Dep/fmt: Upgrade to 10.2.1

13 files changed, 3904 insertions, 3465 deletions

diff --git a/dep/fmt/include/fmt/args.h b/dep/fmt/include/fmt/args.h
index a3966d14071..ad1654bbb60 100644
--- a/dep/fmt/include/fmt/args.h
+++ b/dep/fmt/include/fmt/args.h
@@ -1,4 +1,4 @@
-// Formatting library for C++ - dynamic format arguments
+// Formatting library for C++ - dynamic argument lists
 //
 // Copyright (c) 2012 - present, Victor Zverovich
 // All rights reserved.
@@ -22,8 +22,9 @@ template <typename T> struct is_reference_wrapper : std::false_type {};
 template <typename T>
 struct is_reference_wrapper<std::reference_wrapper<T>> : std::true_type {};
 
-template <typename T> const T& unwrap(const T& v) { return v; }
-template <typename T> const T& unwrap(const std::reference_wrapper<T>& v) {
+template <typename T> auto unwrap(const T& v) -> const T& { return v; }
+template <typename T>
+auto unwrap(const std::reference_wrapper<T>& v) -> const T& {
   return static_cast<const T&>(v);
 }
 
@@ -50,7 +51,7 @@ class dynamic_arg_list {
   std::unique_ptr<node<>> head_;
 
  public:
-  template <typename T, typename Arg> const T& push(const Arg& arg) {
+  template <typename T, typename Arg> auto push(const Arg& arg) -> const T& {
     auto new_node = std::unique_ptr<typed_node<T>>(new typed_node<T>(arg));
     auto& value = new_node->value;
     new_node->next = std::move(head_);
@@ -110,14 +111,14 @@ class dynamic_format_arg_store
 
   friend class basic_format_args<Context>;
 
-  unsigned long long get_types() const {
+  auto get_types() const -> unsigned long long {
     return detail::is_unpacked_bit | data_.size() |
            (named_info_.empty()
                 ? 0ULL
                 : static_cast<unsigned long long>(detail::has_named_args_bit));
   }
 
-  const basic_format_arg<Context>* data() const {
+  auto data() const -> const basic_format_arg<Context>* {
     return named_info_.empty() ? data_.data() : data_.data() + 1;
   }
 
diff --git a/dep/fmt/include/fmt/chrono.h b/dep/fmt/include/fmt/chrono.h
index b112f76e991..9d54574e168 100644
--- a/dep/fmt/include/fmt/chrono.h
+++ b/dep/fmt/include/fmt/chrono.h
@@ -18,10 +18,28 @@
 #include <ostream>
 #include <type_traits>
 
-#include "format.h"
+#include "ostream.h"  // formatbuf
 
 FMT_BEGIN_NAMESPACE
 
+// Check if std::chrono::local_t is available.
+#ifndef FMT_USE_LOCAL_TIME
+#  ifdef __cpp_lib_chrono
+#    define FMT_USE_LOCAL_TIME (__cpp_lib_chrono >= 201907L)
+#  else
+#    define FMT_USE_LOCAL_TIME 0
+#  endif
+#endif
+
+// Check if std::chrono::utc_timestamp is available.
+#ifndef FMT_USE_UTC_TIME
+#  ifdef __cpp_lib_chrono
+#    define FMT_USE_UTC_TIME (__cpp_lib_chrono >= 201907L)
+#  else
+#    define FMT_USE_UTC_TIME 0
+#  endif
+#endif
+
 // Enable tzset.
 #ifndef FMT_USE_TZSET
 // UWP doesn't provide _tzset.
@@ -54,7 +72,8 @@ template <typename To, typename From,
           FMT_ENABLE_IF(!std::is_same<From, To>::value &&
                         std::numeric_limits<From>::is_signed ==
                             std::numeric_limits<To>::is_signed)>
-FMT_CONSTEXPR To lossless_integral_conversion(const From from, int& ec) {
+FMT_CONSTEXPR auto lossless_integral_conversion(const From from, int& ec)
+    -> To {
   ec = 0;
   using F = std::numeric_limits<From>;
   using T = std::numeric_limits<To>;
@@ -83,7 +102,8 @@ template <typename To, typename From,
           FMT_ENABLE_IF(!std::is_same<From, To>::value &&
                         std::numeric_limits<From>::is_signed !=
                             std::numeric_limits<To>::is_signed)>
-FMT_CONSTEXPR To lossless_integral_conversion(const From from, int& ec) {
+FMT_CONSTEXPR auto lossless_integral_conversion(const From from, int& ec)
+    -> To {
   ec = 0;
   using F = std::numeric_limits<From>;
   using T = std::numeric_limits<To>;
@@ -115,7 +135,8 @@ FMT_CONSTEXPR To lossless_integral_conversion(const From from, int& ec) {
 
 template <typename To, typename From,
           FMT_ENABLE_IF(std::is_same<From, To>::value)>
-FMT_CONSTEXPR To lossless_integral_conversion(const From from, int& ec) {
+FMT_CONSTEXPR auto lossless_integral_conversion(const From from, int& ec)
+    -> To {
   ec = 0;
   return from;
 }  // function
@@ -136,7 +157,7 @@ FMT_CONSTEXPR To lossless_integral_conversion(const From from, int& ec) {
 // clang-format on
 template <typename To, typename From,
           FMT_ENABLE_IF(!std::is_same<From, To>::value)>
-FMT_CONSTEXPR To safe_float_conversion(const From from, int& ec) {
+FMT_CONSTEXPR auto safe_float_conversion(const From from, int& ec) -> To {
   ec = 0;
   using T = std::numeric_limits<To>;
   static_assert(std::is_floating_point<From>::value, "From must be floating");
@@ -158,7 +179,7 @@ FMT_CONSTEXPR To safe_float_conversion(const From from, int& ec) {
 
 template <typename To, typename From,
           FMT_ENABLE_IF(std::is_same<From, To>::value)>
-FMT_CONSTEXPR To safe_float_conversion(const From from, int& ec) {
+FMT_CONSTEXPR auto safe_float_conversion(const From from, int& ec) -> To {
   ec = 0;
   static_assert(std::is_floating_point<From>::value, "From must be floating");
   return from;
@@ -170,8 +191,8 @@ FMT_CONSTEXPR To safe_float_conversion(const From from, int& ec) {
 template <typename To, typename FromRep, typename FromPeriod,
           FMT_ENABLE_IF(std::is_integral<FromRep>::value),
           FMT_ENABLE_IF(std::is_integral<typename To::rep>::value)>
-To safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
-                      int& ec) {
+auto safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
+                        int& ec) -> To {
   using From = std::chrono::duration<FromRep, FromPeriod>;
   ec = 0;
   // the basic idea is that we need to convert from count() in the from type
@@ -203,7 +224,8 @@ To safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
     }
     const auto min1 =
         (std::numeric_limits<IntermediateRep>::min)() / Factor::num;
-    if (!std::is_unsigned<IntermediateRep>::value && count < min1) {
+    if (detail::const_check(!std::is_unsigned<IntermediateRep>::value) &&
+        count < min1) {
       ec = 1;
       return {};
     }
@@ -221,8 +243,8 @@ To safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
 template <typename To, typename FromRep, typename FromPeriod,
           FMT_ENABLE_IF(std::is_floating_point<FromRep>::value),
           FMT_ENABLE_IF(std::is_floating_point<typename To::rep>::value)>
-To safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
-                      int& ec) {
+auto safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
+                        int& ec) -> To {
   using From = std::chrono::duration<FromRep, FromPeriod>;
   ec = 0;
   if (std::isnan(from.count())) {
@@ -302,12 +324,12 @@ To safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from,
 
 namespace detail {
 template <typename T = void> struct null {};
-inline null<> localtime_r FMT_NOMACRO(...) { return null<>(); }
-inline null<> localtime_s(...) { return null<>(); }
-inline null<> gmtime_r(...) { return null<>(); }
-inline null<> gmtime_s(...) { return null<>(); }
+inline auto localtime_r FMT_NOMACRO(...) -> null<> { return null<>(); }
+inline auto localtime_s(...) -> null<> { return null<>(); }
+inline auto gmtime_r(...) -> null<> { return null<>(); }
+inline auto gmtime_s(...) -> null<> { return null<>(); }
 
-inline const std::locale& get_classic_locale() {
+inline auto get_classic_locale() -> const std::locale& {
   static const auto& locale = std::locale::classic();
   return locale;
 }
@@ -317,8 +339,6 @@ template <typename CodeUnit> struct codecvt_result {
   CodeUnit buf[max_size];
   CodeUnit* end;
 };
-template <typename CodeUnit>
-constexpr const size_t codecvt_result<CodeUnit>::max_size;
 
 template <typename CodeUnit>
 void write_codecvt(codecvt_result<CodeUnit>& out, string_view in_buf,
@@ -358,37 +378,11 @@ auto write_encoded_tm_str(OutputIt out, string_view in, const std::locale& loc)
     unit_t unit;
     write_codecvt(unit, in, loc);
     // In UTF-8 is used one to four one-byte code units.
-    auto&& buf = basic_memory_buffer<char, unit_t::max_size * 4>();
-    for (code_unit* p = unit.buf; p != unit.end; ++p) {
-      uint32_t c = static_cast<uint32_t>(*p);
-      if (sizeof(code_unit) == 2 && c >= 0xd800 && c <= 0xdfff) {
-        // surrogate pair
-        ++p;
-        if (p == unit.end || (c & 0xfc00) != 0xd800 ||
-            (*p & 0xfc00) != 0xdc00) {
-          FMT_THROW(format_error("failed to format time"));
-        }
-        c = (c << 10) + static_cast<uint32_t>(*p) - 0x35fdc00;
-      }
-      if (c < 0x80) {
-        buf.push_back(static_cast<char>(c));
-      } else if (c < 0x800) {
-        buf.push_back(static_cast<char>(0xc0 | (c >> 6)));
-        buf.push_back(static_cast<char>(0x80 | (c & 0x3f)));
-      } else if ((c >= 0x800 && c <= 0xd7ff) || (c >= 0xe000 && c <= 0xffff)) {
-        buf.push_back(static_cast<char>(0xe0 | (c >> 12)));
-        buf.push_back(static_cast<char>(0x80 | ((c & 0xfff) >> 6)));
-        buf.push_back(static_cast<char>(0x80 | (c & 0x3f)));
-      } else if (c >= 0x10000 && c <= 0x10ffff) {
-        buf.push_back(static_cast<char>(0xf0 | (c >> 18)));
-        buf.push_back(static_cast<char>(0x80 | ((c & 0x3ffff) >> 12)));
-        buf.push_back(static_cast<char>(0x80 | ((c & 0xfff) >> 6)));
-        buf.push_back(static_cast<char>(0x80 | (c & 0x3f)));
-      } else {
-        FMT_THROW(format_error("failed to format time"));
-      }
-    }
-    return copy_str<char>(buf.data(), buf.data() + buf.size(), out);
+    auto u =
+        to_utf8<code_unit, basic_memory_buffer<char, unit_t::max_size * 4>>();
+    if (!u.convert({unit.buf, to_unsigned(unit.end - unit.buf)}))
+      FMT_THROW(format_error("failed to format time"));
+    return copy_str<char>(u.c_str(), u.c_str() + u.size(), out);
   }
   return copy_str<char>(in.data(), in.data() + in.size(), out);
 }
@@ -415,8 +409,7 @@ inline void do_write(buffer<Char>& buf, const std::tm& time,
   auto&& format_buf = formatbuf<std::basic_streambuf<Char>>(buf);
   auto&& os = std::basic_ostream<Char>(&format_buf);
   os.imbue(loc);
-  using iterator = std::ostreambuf_iterator<Char>;
-  const auto& facet = std::use_facet<std::time_put<Char, iterator>>(loc);
+  const auto& facet = std::use_facet<std::time_put<Char>>(loc);
   auto end = facet.put(os, os, Char(' '), &time, format, modifier);
   if (end.failed()) FMT_THROW(format_error("failed to format time"));
 }
@@ -427,7 +420,7 @@ auto write(OutputIt out, const std::tm& time, const std::locale& loc,
            char format, char modifier = 0) -> OutputIt {
   auto&& buf = get_buffer<Char>(out);
   do_write<Char>(buf, time, loc, format, modifier);
-  return buf.out();
+  return get_iterator(buf, out);
 }
 
 template <typename Char, typename OutputIt,
@@ -439,38 +432,83 @@ auto write(OutputIt out, const std::tm& time, const std::locale& loc,
   return write_encoded_tm_str(out, string_view(buf.data(), buf.size()), loc);
 }
 
+template <typename Rep1, typename Rep2>
+struct is_same_arithmetic_type
+    : public std::integral_constant<bool,
+                                    (std::is_integral<Rep1>::value &&
+                                     std::is_integral<Rep2>::value) ||
+                                        (std::is_floating_point<Rep1>::value &&
+                                         std::is_floating_point<Rep2>::value)> {
+};
+
+template <
+    typename To, typename FromRep, typename FromPeriod,
+    FMT_ENABLE_IF(is_same_arithmetic_type<FromRep, typename To::rep>::value)>
+auto fmt_duration_cast(std::chrono::duration<FromRep, FromPeriod> from) -> To {
+#if FMT_SAFE_DURATION_CAST
+  // Throwing version of safe_duration_cast is only available for
+  // integer to integer or float to float casts.
+  int ec;
+  To to = safe_duration_cast::safe_duration_cast<To>(from, ec);
+  if (ec) FMT_THROW(format_error("cannot format duration"));
+  return to;
+#else
+  // Standard duration cast, may overflow.
+  return std::chrono::duration_cast<To>(from);
+#endif
+}
+
+template <
+    typename To, typename FromRep, typename FromPeriod,
+    FMT_ENABLE_IF(!is_same_arithmetic_type<FromRep, typename To::rep>::value)>
+auto fmt_duration_cast(std::chrono::duration<FromRep, FromPeriod> from) -> To {
+  // Mixed integer <-> float cast is not supported by safe_duration_cast.
+  return std::chrono::duration_cast<To>(from);
+}
+
+template <typename Duration>
+auto to_time_t(
+    std::chrono::time_point<std::chrono::system_clock, Duration> time_point)
+    -> std::time_t {
+  // Cannot use std::chrono::system_clock::to_time_t since this would first
+  // require a cast to std::chrono::system_clock::time_point, which could
+  // overflow.
+  return fmt_duration_cast<std::chrono::duration<std::time_t>>(
+             time_point.time_since_epoch())
+      .count();
+}
 }  // namespace detail
 
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
 /**
   Converts given time since epoch as ``std::time_t`` value into calendar time,
   expressed in local time. Unlike ``std::localtime``, this function is
   thread-safe on most platforms.
  */
-inline std::tm localtime(std::time_t time) {
+inline auto localtime(std::time_t time) -> std::tm {
   struct dispatcher {
     std::time_t time_;
     std::tm tm_;
 
     dispatcher(std::time_t t) : time_(t) {}
 
-    bool run() {
+    auto run() -> bool {
       using namespace fmt::detail;
       return handle(localtime_r(&time_, &tm_));
     }
 
-    bool handle(std::tm* tm) { return tm != nullptr; }
+    auto handle(std::tm* tm) -> bool { return tm != nullptr; }
 
-    bool handle(detail::null<>) {
+    auto handle(detail::null<>) -> bool {
       using namespace fmt::detail;
       return fallback(localtime_s(&tm_, &time_));
     }
 
-    bool fallback(int res) { return res == 0; }
+    auto fallback(int res) -> bool { return res == 0; }
 
 #if !FMT_MSC_VERSION
-    bool fallback(detail::null<>) {
+    auto fallback(detail::null<>) -> bool {
       using namespace fmt::detail;
       std::tm* tm = std::localtime(&time_);
       if (tm) tm_ = *tm;
@@ -484,57 +522,62 @@ inline std::tm localtime(std::time_t time) {
   return lt.tm_;
 }
 
-inline std::tm localtime(
-    std::chrono::time_point<std::chrono::system_clock> time_point) {
-  return localtime(std::chrono::system_clock::to_time_t(time_point));
+#if FMT_USE_LOCAL_TIME
+template <typename Duration>
+inline auto localtime(std::chrono::local_time<Duration> time) -> std::tm {
+  return localtime(
+      detail::to_time_t(std::chrono::current_zone()->to_sys(time)));
 }
+#endif
 
 /**
   Converts given time since epoch as ``std::time_t`` value into calendar time,
   expressed in Coordinated Universal Time (UTC). Unlike ``std::gmtime``, this
   function is thread-safe on most platforms.
  */
-inline std::tm gmtime(std::time_t time) {
+inline auto gmtime(std::time_t time) -> std::tm {
   struct dispatcher {
     std::time_t time_;
     std::tm tm_;
 
     dispatcher(std::time_t t) : time_(t) {}
 
-    bool run() {
+    auto run() -> bool {
       using namespace fmt::detail;
       return handle(gmtime_r(&time_, &tm_));
     }
 
-    bool handle(std::tm* tm) { return tm != nullptr; }
+    auto handle(std::tm* tm) -> bool { return tm != nullptr; }
 
-    bool handle(detail::null<>) {
+    auto handle(detail::null<>) -> bool {
       using namespace fmt::detail;
       return fallback(gmtime_s(&tm_, &time_));
     }
 
-    bool fallback(int res) { return res == 0; }
+    auto fallback(int res) -> bool { return res == 0; }
 
 #if !FMT_MSC_VERSION
-    bool fallback(detail::null<>) {
+    auto fallback(detail::null<>) -> bool {
       std::tm* tm = std::gmtime(&time_);
       if (tm) tm_ = *tm;
       return tm != nullptr;
     }
 #endif
   };
-  dispatcher gt(time);
+  auto gt = dispatcher(time);
   // Too big time values may be unsupported.
   if (!gt.run()) FMT_THROW(format_error("time_t value out of range"));
   return gt.tm_;
 }
 
-inline std::tm gmtime(
-    std::chrono::time_point<std::chrono::system_clock> time_point) {
-  return gmtime(std::chrono::system_clock::to_time_t(time_point));
+template <typename Duration>
+inline auto gmtime(
+    std::chrono::time_point<std::chrono::system_clock, Duration> time_point)
+    -> std::tm {
+  return gmtime(detail::to_time_t(time_point));
 }
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 // Writes two-digit numbers a, b and c separated by sep to buf.
 // The method by Pavel Novikov based on
@@ -570,7 +613,8 @@ inline void write_digit2_separated(char* buf, unsigned a, unsigned b,
   }
 }
 
-template <typename Period> FMT_CONSTEXPR inline const char* get_units() {
+template <typename Period>
+FMT_CONSTEXPR inline auto get_units() -> const char* {
   if (std::is_same<Period, std::atto>::value) return "as";
   if (std::is_same<Period, std::femto>::value) return "fs";
   if (std::is_same<Period, std::pico>::value) return "ps";
@@ -588,8 +632,9 @@ template <typename Period> FMT_CONSTEXPR inline const char* get_units() {
   if (std::is_same<Period, std::tera>::value) return "Ts";
   if (std::is_same<Period, std::peta>::value) return "Ps";
   if (std::is_same<Period, std::exa>::value) return "Es";
-  if (std::is_same<Period, std::ratio<60>>::value) return "m";
+  if (std::is_same<Period, std::ratio<60>>::value) return "min";
   if (std::is_same<Period, std::ratio<3600>>::value) return "h";
+  if (std::is_same<Period, std::ratio<86400>>::value) return "d";
   return nullptr;
 }
 
@@ -599,12 +644,38 @@ enum class numeric_system {
   alternative
 };
 
+// Glibc extensions for formatting numeric values.
+enum class pad_type {
+  unspecified,
+  // Do not pad a numeric result string.
+  none,
+  // Pad a numeric result string with zeros even if the conversion specifier
+  // character uses space-padding by default.
+  zero,
+  // Pad a numeric result string with spaces.
+  space,
+};
+
+template <typename OutputIt>
+auto write_padding(OutputIt out, pad_type pad, int width) -> OutputIt {
+  if (pad == pad_type::none) return out;
+  return std::fill_n(out, width, pad == pad_type::space ? ' ' : '0');
+}
+
+template <typename OutputIt>
+auto write_padding(OutputIt out, pad_type pad) -> OutputIt {
+  if (pad != pad_type::none) *out++ = pad == pad_type::space ? ' ' : '0';
+  return out;
+}
+
 // Parses a put_time-like format string and invokes handler actions.
 template <typename Char, typename Handler>
-FMT_CONSTEXPR const Char* parse_chrono_format(const Char* begin,
-                                              const Char* end,
-                                              Handler&& handler) {
+FMT_CONSTEXPR auto parse_chrono_format(const Char* begin, const Char* end,
+                                       Handler&& handler) -> const Char* {
+  if (begin == end || *begin == '}') return begin;
+  if (*begin != '%') FMT_THROW(format_error("invalid format"));
   auto ptr = begin;
+  pad_type pad = pad_type::unspecified;
   while (ptr != end) {
     auto c = *ptr;
     if (c == '}') break;
@@ -615,6 +686,22 @@ FMT_CONSTEXPR const Char* parse_chrono_format(const Char* begin,
     if (begin != ptr) handler.on_text(begin, ptr);
     ++ptr;  // consume '%'
     if (ptr == end) FMT_THROW(format_error("invalid format"));
+    c = *ptr;
+    switch (c) {
+    case '_':
+      pad = pad_type::space;
+      ++ptr;
+      break;
+    case '-':
+      pad = pad_type::none;
+      ++ptr;
+      break;
+    case '0':
+      pad = pad_type::zero;
+      ++ptr;
+      break;
+    }
+    if (ptr == end) FMT_THROW(format_error("invalid format"));
     c = *ptr++;
     switch (c) {
     case '%':
@@ -691,16 +778,16 @@ FMT_CONSTEXPR const Char* parse_chrono_format(const Char* begin,
       break;
     // Hour, minute, second:
     case 'H':
-      handler.on_24_hour(numeric_system::standard);
+      handler.on_24_hour(numeric_system::standard, pad);
       break;
     case 'I':
-      handler.on_12_hour(numeric_system::standard);
+      handler.on_12_hour(numeric_system::standard, pad);
       break;
     case 'M':
-      handler.on_minute(numeric_system::standard);
+      handler.on_minute(numeric_system::standard, pad);
       break;
     case 'S':
-      handler.on_second(numeric_system::standard);
+      handler.on_second(numeric_system::standard, pad);
       break;
     // Other:
     case 'c':
@@ -737,7 +824,7 @@ FMT_CONSTEXPR const Char* parse_chrono_format(const Char* begin,
       handler.on_duration_unit();
       break;
     case 'z':
-      handler.on_utc_offset();
+      handler.on_utc_offset(numeric_system::standard);
       break;
     case 'Z':
       handler.on_tz_name();
@@ -765,6 +852,9 @@ FMT_CONSTEXPR const Char* parse_chrono_format(const Char* begin,
       case 'X':
         handler.on_loc_time(numeric_system::alternative);
         break;
+      case 'z':
+        handler.on_utc_offset(numeric_system::alternative);
+        break;
       default:
         FMT_THROW(format_error("invalid format"));
       }
@@ -802,16 +892,19 @@ FMT_CONSTEXPR const Char* parse_chrono_format(const Char* begin,
         handler.on_dec1_weekday(numeric_system::alternative);
         break;
       case 'H':
-        handler.on_24_hour(numeric_system::alternative);
+        handler.on_24_hour(numeric_system::alternative, pad);
         break;
       case 'I':
-        handler.on_12_hour(numeric_system::alternative);
+        handler.on_12_hour(numeric_system::alternative, pad);
         break;
       case 'M':
-        handler.on_minute(numeric_system::alternative);
+        handler.on_minute(numeric_system::alternative, pad);
         break;
       case 'S':
-        handler.on_second(numeric_system::alternative);
+        handler.on_second(numeric_system::alternative, pad);
+        break;
+      case 'z':
+        handler.on_utc_offset(numeric_system::alternative);
         break;
       default:
         FMT_THROW(format_error("invalid format"));
@@ -864,7 +957,7 @@ template <typename Derived> struct null_chrono_spec_handler {
   FMT_CONSTEXPR void on_am_pm() { unsupported(); }
   FMT_CONSTEXPR void on_duration_value() { unsupported(); }
   FMT_CONSTEXPR void on_duration_unit() { unsupported(); }
-  FMT_CONSTEXPR void on_utc_offset() { unsupported(); }
+  FMT_CONSTEXPR void on_utc_offset(numeric_system) { unsupported(); }
   FMT_CONSTEXPR void on_tz_name() { unsupported(); }
 };
 
@@ -892,10 +985,10 @@ struct tm_format_checker : null_chrono_spec_handler<tm_format_checker> {
   FMT_CONSTEXPR void on_day_of_year() {}
   FMT_CONSTEXPR void on_day_of_month(numeric_system) {}
   FMT_CONSTEXPR void on_day_of_month_space(numeric_system) {}
-  FMT_CONSTEXPR void on_24_hour(numeric_system) {}
-  FMT_CONSTEXPR void on_12_hour(numeric_system) {}
-  FMT_CONSTEXPR void on_minute(numeric_system) {}
-  FMT_CONSTEXPR void on_second(numeric_system) {}
+  FMT_CONSTEXPR void on_24_hour(numeric_system, pad_type) {}
+  FMT_CONSTEXPR void on_12_hour(numeric_system, pad_type) {}
+  FMT_CONSTEXPR void on_minute(numeric_system, pad_type) {}
+  FMT_CONSTEXPR void on_second(numeric_system, pad_type) {}
   FMT_CONSTEXPR void on_datetime(numeric_system) {}
   FMT_CONSTEXPR void on_loc_date(numeric_system) {}
   FMT_CONSTEXPR void on_loc_time(numeric_system) {}
@@ -905,29 +998,29 @@ struct tm_format_checker : null_chrono_spec_handler<tm_format_checker> {
   FMT_CONSTEXPR void on_24_hour_time() {}
   FMT_CONSTEXPR void on_iso_time() {}
   FMT_CONSTEXPR void on_am_pm() {}
-  FMT_CONSTEXPR void on_utc_offset() {}
+  FMT_CONSTEXPR void on_utc_offset(numeric_system) {}
   FMT_CONSTEXPR void on_tz_name() {}
 };
 
-inline const char* tm_wday_full_name(int wday) {
+inline auto tm_wday_full_name(int wday) -> const char* {
   static constexpr const char* full_name_list[] = {
       "Sunday",   "Monday", "Tuesday", "Wednesday",
       "Thursday", "Friday", "Saturday"};
   return wday >= 0 && wday <= 6 ? full_name_list[wday] : "?";
 }
-inline const char* tm_wday_short_name(int wday) {
+inline auto tm_wday_short_name(int wday) -> const char* {
   static constexpr const char* short_name_list[] = {"Sun", "Mon", "Tue", "Wed",
                                                     "Thu", "Fri", "Sat"};
   return wday >= 0 && wday <= 6 ? short_name_list[wday] : "???";
 }
 
-inline const char* tm_mon_full_name(int mon) {
+inline auto tm_mon_full_name(int mon) -> const char* {
   static constexpr const char* full_name_list[] = {
       "January", "February", "March",     "April",   "May",      "June",
       "July",    "August",   "September", "October", "November", "December"};
   return mon >= 0 && mon <= 11 ? full_name_list[mon] : "?";
 }
-inline const char* tm_mon_short_name(int mon) {
+inline auto tm_mon_short_name(int mon) -> const char* {
   static constexpr const char* short_name_list[] = {
       "Jan", "Feb", "Mar", "Apr", "May", "Jun",
       "Jul", "Aug", "Sep", "Oct", "Nov", "Dec",
@@ -957,13 +1050,129 @@ inline void tzset_once() {
 }
 #endif
 
-template <typename OutputIt, typename Char> class tm_writer {
+// Converts value to Int and checks that it's in the range [0, upper).
+template <typename T, typename Int, FMT_ENABLE_IF(std::is_integral<T>::value)>
+inline auto to_nonnegative_int(T value, Int upper) -> Int {
+  if (!std::is_unsigned<Int>::value &&
+      (value < 0 || to_unsigned(value) > to_unsigned(upper))) {
+    FMT_THROW(fmt::format_error("chrono value is out of range"));
+  }
+  return static_cast<Int>(value);
+}
+template <typename T, typename Int, FMT_ENABLE_IF(!std::is_integral<T>::value)>
+inline auto to_nonnegative_int(T value, Int upper) -> Int {
+  if (value < 0 || value > static_cast<T>(upper))
+    FMT_THROW(format_error("invalid value"));
+  return static_cast<Int>(value);
+}
+
+constexpr auto pow10(std::uint32_t n) -> long long {
+  return n == 0 ? 1 : 10 * pow10(n - 1);
+}
+
+// Counts the number of fractional digits in the range [0, 18] according to the
+// C++20 spec. If more than 18 fractional digits are required then returns 6 for
+// microseconds precision.
+template <long long Num, long long Den, int N = 0,
+          bool Enabled = (N < 19) && (Num <= max_value<long long>() / 10)>
+struct count_fractional_digits {
+  static constexpr int value =
+      Num % Den == 0 ? N : count_fractional_digits<Num * 10, Den, N + 1>::value;
+};
+
+// Base case that doesn't instantiate any more templates
+// in order to avoid overflow.
+template <long long Num, long long Den, int N>
+struct count_fractional_digits<Num, Den, N, false> {
+  static constexpr int value = (Num % Den == 0) ? N : 6;
+};
+
+// Format subseconds which are given as an integer type with an appropriate
+// number of digits.
+template <typename Char, typename OutputIt, typename Duration>
+void write_fractional_seconds(OutputIt& out, Duration d, int precision = -1) {
+  constexpr auto num_fractional_digits =
+      count_fractional_digits<Duration::period::num,
+                              Duration::period::den>::value;
+
+  using subsecond_precision = std::chrono::duration<
+      typename std::common_type<typename Duration::rep,
+                                std::chrono::seconds::rep>::type,
+      std::ratio<1, detail::pow10(num_fractional_digits)>>;
+
+  const auto fractional = d - fmt_duration_cast<std::chrono::seconds>(d);
+  const auto subseconds =
+      std::chrono::treat_as_floating_point<
+          typename subsecond_precision::rep>::value
+          ? fractional.count()
+          : fmt_duration_cast<subsecond_precision>(fractional).count();
+  auto n = static_cast<uint32_or_64_or_128_t<long long>>(subseconds);
+  const int num_digits = detail::count_digits(n);
+
+  int leading_zeroes = (std::max)(0, num_fractional_digits - num_digits);
+  if (precision < 0) {
+    FMT_ASSERT(!std::is_floating_point<typename Duration::rep>::value, "");
+    if (std::ratio_less<typename subsecond_precision::period,
+                        std::chrono::seconds::period>::value) {
+      *out++ = '.';
+      out = std::fill_n(out, leading_zeroes, '0');
+      out = format_decimal<Char>(out, n, num_digits).end;
+    }
+  } else {
+    *out++ = '.';
+    leading_zeroes = (std::min)(leading_zeroes, precision);
+    out = std::fill_n(out, leading_zeroes, '0');
+    int remaining = precision - leading_zeroes;
+    if (remaining != 0 && remaining < num_digits) {
+      n /= to_unsigned(detail::pow10(to_unsigned(num_digits - remaining)));
+      out = format_decimal<Char>(out, n, remaining).end;
+      return;
+    }
+    out = format_decimal<Char>(out, n, num_digits).end;
+    remaining -= num_digits;
+    out = std::fill_n(out, remaining, '0');
+  }
+}
+
+// Format subseconds which are given as a floating point type with an
+// appropriate number of digits. We cannot pass the Duration here, as we
+// explicitly need to pass the Rep value in the chrono_formatter.
+template <typename Duration>
+void write_floating_seconds(memory_buffer& buf, Duration duration,
+                            int num_fractional_digits = -1) {
+  using rep = typename Duration::rep;
+  FMT_ASSERT(std::is_floating_point<rep>::value, "");
+
+  auto val = duration.count();
+
+  if (num_fractional_digits < 0) {
+    // For `std::round` with fallback to `round`:
+    // On some toolchains `std::round` is not available (e.g. GCC 6).
+    using namespace std;
+    num_fractional_digits =
+        count_fractional_digits<Duration::period::num,
+                                Duration::period::den>::value;
+    if (num_fractional_digits < 6 && static_cast<rep>(round(val)) != val)
+      num_fractional_digits = 6;
+  }
+
+  fmt::format_to(std::back_inserter(buf), FMT_STRING("{:.{}f}"),
+                 std::fmod(val * static_cast<rep>(Duration::period::num) /
+                               static_cast<rep>(Duration::period::den),
+                           static_cast<rep>(60)),
+                 num_fractional_digits);
+}
+
+template <typename OutputIt, typename Char,
+          typename Duration = std::chrono::seconds>
+class tm_writer {
  private:
   static constexpr int days_per_week = 7;
 
   const std::locale& loc_;
   const bool is_classic_;
   OutputIt out_;
+  const Duration* subsecs_;
   const std::tm& tm_;
 
   auto tm_sec() const noexcept -> int {
@@ -1012,8 +1221,7 @@ template <typename OutputIt, typename Char> class tm_writer {
     return static_cast<int>(l);
   }
 
-  // Algorithm:
-  // https://en.wikipedia.org/wiki/ISO_week_date#Calculating_the_week_number_from_a_month_and_day_of_the_month_or_ordinal_date
+  // Algorithm: https://en.wikipedia.org/wiki/ISO_week_date.
   auto iso_year_weeks(long long curr_year) const noexcept -> int {
     const auto prev_year = curr_year - 1;
     const auto curr_p =
@@ -1051,6 +1259,17 @@ template <typename OutputIt, typename Char> class tm_writer {
     *out_++ = *d++;
     *out_++ = *d;
   }
+  void write2(int value, pad_type pad) {
+    unsigned int v = to_unsigned(value) % 100;
+    if (v >= 10) {
+      const char* d = digits2(v);
+      *out_++ = *d++;
+      *out_++ = *d;
+    } else {
+      out_ = detail::write_padding(out_, pad);
+      *out_++ = static_cast<char>('0' + v);
+    }
+  }
 
   void write_year_extended(long long year) {
     // At least 4 characters.
@@ -1074,7 +1293,7 @@ template <typename OutputIt, typename Char> class tm_writer {
     }
   }
 
-  void write_utc_offset(long offset) {
+  void write_utc_offset(long offset, numeric_system ns) {
     if (offset < 0) {
       *out_++ = '-';
       offset = -offset;
@@ -1083,14 +1302,15 @@ template <typename OutputIt, typename Char> class tm_writer {
     }
     offset /= 60;
     write2(static_cast<int>(offset / 60));
+    if (ns != numeric_system::standard) *out_++ = ':';
     write2(static_cast<int>(offset % 60));
   }
   template <typename T, FMT_ENABLE_IF(has_member_data_tm_gmtoff<T>::value)>
-  void format_utc_offset_impl(const T& tm) {
-    write_utc_offset(tm.tm_gmtoff);
+  void format_utc_offset_impl(const T& tm, numeric_system ns) {
+    write_utc_offset(tm.tm_gmtoff, ns);
   }
   template <typename T, FMT_ENABLE_IF(!has_member_data_tm_gmtoff<T>::value)>
-  void format_utc_offset_impl(const T& tm) {
+  void format_utc_offset_impl(const T& tm, numeric_system ns) {
 #if defined(_WIN32) && defined(_UCRT)
 #  if FMT_USE_TZSET
     tzset_once();
@@ -1102,10 +1322,17 @@ template <typename OutputIt, typename Char> class tm_writer {
       _get_dstbias(&dstbias);
       offset += dstbias;
     }
-    write_utc_offset(-offset);
+    write_utc_offset(-offset, ns);
 #else
-    ignore_unused(tm);
-    format_localized('z');
+    if (ns == numeric_system::standard) return format_localized('z');
+
+    // Extract timezone offset from timezone conversion functions.
+    std::tm gtm = tm;
+    std::time_t gt = std::mktime(&gtm);
+    std::tm ltm = gmtime(gt);
+    std::time_t lt = std::mktime(&ltm);
+    long offset = gt - lt;
+    write_utc_offset(offset, ns);
 #endif
   }
 
@@ -1126,13 +1353,15 @@ template <typename OutputIt, typename Char> class tm_writer {
   }
 
  public:
-  tm_writer(const std::locale& loc, OutputIt out, const std::tm& tm)
+  tm_writer(const std::locale& loc, OutputIt out, const std::tm& tm,
+            const Duration* subsecs = nullptr)
       : loc_(loc),
         is_classic_(loc_ == get_classic_locale()),
         out_(out),
+        subsecs_(subsecs),
         tm_(tm) {}
 
-  OutputIt out() const { return out_; }
+  auto out() const -> OutputIt { return out_; }
 
   FMT_CONSTEXPR void on_text(const Char* begin, const Char* end) {
     out_ = copy_str<Char>(begin, end, out_);
@@ -1227,7 +1456,7 @@ template <typename OutputIt, typename Char> class tm_writer {
     out_ = copy_str<Char>(std::begin(buf) + offset, std::end(buf), out_);
   }
 
-  void on_utc_offset() { format_utc_offset_impl(tm_); }
+  void on_utc_offset(numeric_system ns) { format_utc_offset_impl(tm_, ns); }
   void on_tz_name() { format_tz_name_impl(tm_); }
 
   void on_year(numeric_system ns) {
@@ -1315,22 +1544,41 @@ template <typename OutputIt, typename Char> class tm_writer {
     }
   }
 
-  void on_24_hour(numeric_system ns) {
-    if (is_classic_ || ns == numeric_system::standard) return write2(tm_hour());
+  void on_24_hour(numeric_system ns, pad_type pad) {
+    if (is_classic_ || ns == numeric_system::standard)
+      return write2(tm_hour(), pad);
     format_localized('H', 'O');
   }
-  void on_12_hour(numeric_system ns) {
+  void on_12_hour(numeric_system ns, pad_type pad) {
     if (is_classic_ || ns == numeric_system::standard)
-      return write2(tm_hour12());
+      return write2(tm_hour12(), pad);
     format_localized('I', 'O');
   }
-  void on_minute(numeric_system ns) {
-    if (is_classic_ || ns == numeric_system::standard) return write2(tm_min());
+  void on_minute(numeric_system ns, pad_type pad) {
+    if (is_classic_ || ns == numeric_system::standard)
+      return write2(tm_min(), pad);
     format_localized('M', 'O');
   }
-  void on_second(numeric_system ns) {
-    if (is_classic_ || ns == numeric_system::standard) return write2(tm_sec());
-    format_localized('S', 'O');
+
+  void on_second(numeric_system ns, pad_type pad) {
+    if (is_classic_ || ns == numeric_system::standard) {
+      write2(tm_sec(), pad);
+      if (subsecs_) {
+        if (std::is_floating_point<typename Duration::rep>::value) {
+          auto buf = memory_buffer();
+          write_floating_seconds(buf, *subsecs_);
+          if (buf.size() > 1) {
+            // Remove the leading "0", write something like ".123".
+            out_ = std::copy(buf.begin() + 1, buf.end(), out_);
+          }
+        } else {
+          write_fractional_seconds<Char>(out_, *subsecs_);
+        }
+      }
+    } else {
+      // Currently no formatting of subseconds when a locale is set.
+      format_localized('S', 'O');
+    }
   }
 
   void on_12_hour_time() {
@@ -1351,10 +1599,9 @@ template <typename OutputIt, typename Char> class tm_writer {
     write2(tm_min());
   }
   void on_iso_time() {
-    char buf[8];
-    write_digit2_separated(buf, to_unsigned(tm_hour()), to_unsigned(tm_min()),
-                           to_unsigned(tm_sec()), ':');
-    out_ = copy_str<Char>(std::begin(buf), std::end(buf), out_);
+    on_24_hour_time();
+    *out_++ = ':';
+    on_second(numeric_system::standard, pad_type::unspecified);
   }
 
   void on_am_pm() {
@@ -1372,49 +1619,41 @@ template <typename OutputIt, typename Char> class tm_writer {
 };
 
 struct chrono_format_checker : null_chrono_spec_handler<chrono_format_checker> {
+  bool has_precision_integral = false;
+
   FMT_NORETURN void unsupported() { FMT_THROW(format_error("no date")); }
 
   template <typename Char>
   FMT_CONSTEXPR void on_text(const Char*, const Char*) {}
-  FMT_CONSTEXPR void on_24_hour(numeric_system) {}
-  FMT_CONSTEXPR void on_12_hour(numeric_system) {}
-  FMT_CONSTEXPR void on_minute(numeric_system) {}
-  FMT_CONSTEXPR void on_second(numeric_system) {}
+  FMT_CONSTEXPR void on_day_of_year() {}
+  FMT_CONSTEXPR void on_24_hour(numeric_system, pad_type) {}
+  FMT_CONSTEXPR void on_12_hour(numeric_system, pad_type) {}
+  FMT_CONSTEXPR void on_minute(numeric_system, pad_type) {}
+  FMT_CONSTEXPR void on_second(numeric_system, pad_type) {}
   FMT_CONSTEXPR void on_12_hour_time() {}
   FMT_CONSTEXPR void on_24_hour_time() {}
   FMT_CONSTEXPR void on_iso_time() {}
   FMT_CONSTEXPR void on_am_pm() {}
-  FMT_CONSTEXPR void on_duration_value() {}
+  FMT_CONSTEXPR void on_duration_value() const {
+    if (has_precision_integral) {
+      FMT_THROW(format_error("precision not allowed for this argument type"));
+    }
+  }
   FMT_CONSTEXPR void on_duration_unit() {}
 };
 
-template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
-inline bool isfinite(T) {
+template <typename T,
+          FMT_ENABLE_IF(std::is_integral<T>::value&& has_isfinite<T>::value)>
+inline auto isfinite(T) -> bool {
   return true;
 }
 
-// Converts value to Int and checks that it's in the range [0, upper).
-template <typename T, typename Int, FMT_ENABLE_IF(std::is_integral<T>::value)>
-inline Int to_nonnegative_int(T value, Int upper) {
-  FMT_ASSERT(std::is_unsigned<Int>::value ||
-             (value >= 0 && to_unsigned(value) <= to_unsigned(upper)),
-             "invalid value");
-  (void)upper;
-  return static_cast<Int>(value);
-}
-template <typename T, typename Int, FMT_ENABLE_IF(!std::is_integral<T>::value)>
-inline Int to_nonnegative_int(T value, Int upper) {
-  if (value < 0 || value > static_cast<T>(upper))
-    FMT_THROW(format_error("invalid value"));
-  return static_cast<Int>(value);
-}
-
 template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
-inline T mod(T x, int y) {
+inline auto mod(T x, int y) -> T {
   return x % static_cast<T>(y);
 }
 template <typename T, FMT_ENABLE_IF(std::is_floating_point<T>::value)>
-inline T mod(T x, int y) {
+inline auto mod(T x, int y) -> T {
   return std::fmod(x, static_cast<T>(y));
 }
 
@@ -1429,91 +1668,39 @@ template <typename T> struct make_unsigned_or_unchanged<T, true> {
   using type = typename std::make_unsigned<T>::type;
 };
 
-#if FMT_SAFE_DURATION_CAST
-// throwing version of safe_duration_cast
-template <typename To, typename FromRep, typename FromPeriod>
-To fmt_safe_duration_cast(std::chrono::duration<FromRep, FromPeriod> from) {
-  int ec;
-  To to = safe_duration_cast::safe_duration_cast<To>(from, ec);
-  if (ec) FMT_THROW(format_error("cannot format duration"));
-  return to;
-}
-#endif
-
 template <typename Rep, typename Period,
           FMT_ENABLE_IF(std::is_integral<Rep>::value)>
-inline std::chrono::duration<Rep, std::milli> get_milliseconds(
-    std::chrono::duration<Rep, Period> d) {
+inline auto get_milliseconds(std::chrono::duration<Rep, Period> d)
+    -> std::chrono::duration<Rep, std::milli> {
   // this may overflow and/or the result may not fit in the
   // target type.
 #if FMT_SAFE_DURATION_CAST
   using CommonSecondsType =
       typename std::common_type<decltype(d), std::chrono::seconds>::type;
-  const auto d_as_common = fmt_safe_duration_cast<CommonSecondsType>(d);
+  const auto d_as_common = fmt_duration_cast<CommonSecondsType>(d);
   const auto d_as_whole_seconds =
-      fmt_safe_duration_cast<std::chrono::seconds>(d_as_common);
+      fmt_duration_cast<std::chrono::seconds>(d_as_common);
   // this conversion should be nonproblematic
   const auto diff = d_as_common - d_as_whole_seconds;
   const auto ms =
-      fmt_safe_duration_cast<std::chrono::duration<Rep, std::milli>>(diff);
+      fmt_duration_cast<std::chrono::duration<Rep, std::milli>>(diff);
   return ms;
 #else
-  auto s = std::chrono::duration_cast<std::chrono::seconds>(d);
-  return std::chrono::duration_cast<std::chrono::milliseconds>(d - s);
+  auto s = fmt_duration_cast<std::chrono::seconds>(d);
+  return fmt_duration_cast<std::chrono::milliseconds>(d - s);
 #endif
 }
 
-// Counts the number of fractional digits in the range [0, 18] according to the
-// C++20 spec. If more than 18 fractional digits are required then returns 6 for
-// microseconds precision.
-template <long long Num, long long Den, int N = 0,
-          bool Enabled = (N < 19) && (Num <= max_value<long long>() / 10)>
-struct count_fractional_digits {
-  static constexpr int value =
-      Num % Den == 0 ? N : count_fractional_digits<Num * 10, Den, N + 1>::value;
-};
-
-// Base case that doesn't instantiate any more templates
-// in order to avoid overflow.
-template <long long Num, long long Den, int N>
-struct count_fractional_digits<Num, Den, N, false> {
-  static constexpr int value = (Num % Den == 0) ? N : 6;
-};
-
-constexpr long long pow10(std::uint32_t n) {
-  return n == 0 ? 1 : 10 * pow10(n - 1);
-}
-
-template <class Rep, class Period,
-          FMT_ENABLE_IF(std::numeric_limits<Rep>::is_signed)>
-constexpr std::chrono::duration<Rep, Period> abs(
-    std::chrono::duration<Rep, Period> d) {
-  // We need to compare the duration using the count() method directly
-  // due to a compiler bug in clang-11 regarding the spaceship operator,
-  // when -Wzero-as-null-pointer-constant is enabled.
-  // In clang-12 the bug has been fixed. See
-  // https://bugs.llvm.org/show_bug.cgi?id=46235 and the reproducible example:
-  // https://www.godbolt.org/z/Knbb5joYx.
-  return d.count() >= d.zero().count() ? d : -d;
-}
-
-template <class Rep, class Period,
-          FMT_ENABLE_IF(!std::numeric_limits<Rep>::is_signed)>
-constexpr std::chrono::duration<Rep, Period> abs(
-    std::chrono::duration<Rep, Period> d) {
-  return d;
-}
-
 template <typename Char, typename Rep, typename OutputIt,
           FMT_ENABLE_IF(std::is_integral<Rep>::value)>
-OutputIt format_duration_value(OutputIt out, Rep val, int) {
+auto format_duration_value(OutputIt out, Rep val, int) -> OutputIt {
   return write<Char>(out, val);
 }
 
 template <typename Char, typename Rep, typename OutputIt,
           FMT_ENABLE_IF(std::is_floating_point<Rep>::value)>
-OutputIt format_duration_value(OutputIt out, Rep val, int precision) {
-  auto specs = basic_format_specs<Char>();
+auto format_duration_value(OutputIt out, Rep val, int precision) -> OutputIt {
+  auto specs = format_specs<Char>();
   specs.precision = precision;
   specs.type = precision >= 0 ? presentation_type::fixed_lower
                               : presentation_type::general_lower;
@@ -1521,12 +1708,12 @@ OutputIt format_duration_value(OutputIt out, Rep val, int precision) {
 }
 
 template <typename Char, typename OutputIt>
-OutputIt copy_unit(string_view unit, OutputIt out, Char) {
+auto copy_unit(string_view unit, OutputIt out, Char) -> OutputIt {
   return std::copy(unit.begin(), unit.end(), out);
 }
 
 template <typename OutputIt>
-OutputIt copy_unit(string_view unit, OutputIt out, wchar_t) {
+auto copy_unit(string_view unit, OutputIt out, wchar_t) -> OutputIt {
   // This works when wchar_t is UTF-32 because units only contain characters
   // that have the same representation in UTF-16 and UTF-32.
   utf8_to_utf16 u(unit);
@@ -1534,7 +1721,7 @@ OutputIt copy_unit(string_view unit, OutputIt out, wchar_t) {
 }
 
 template <typename Char, typename Period, typename OutputIt>
-OutputIt format_duration_unit(OutputIt out) {
+auto format_duration_unit(OutputIt out) -> OutputIt {
   if (const char* unit = get_units<Period>())
     return copy_unit(string_view(unit), out, Char());
   *out++ = '[';
@@ -1601,18 +1788,12 @@ struct chrono_formatter {
 
     // this may overflow and/or the result may not fit in the
     // target type.
-#if FMT_SAFE_DURATION_CAST
     // might need checked conversion (rep!=Rep)
-    auto tmpval = std::chrono::duration<rep, Period>(val);
-    s = fmt_safe_duration_cast<seconds>(tmpval);
-#else
-    s = std::chrono::duration_cast<seconds>(
-        std::chrono::duration<rep, Period>(val));
-#endif
+    s = fmt_duration_cast<seconds>(std::chrono::duration<rep, Period>(val));
   }
 
   // returns true if nan or inf, writes to out.
-  bool handle_nan_inf() {
+  auto handle_nan_inf() -> bool {
     if (isfinite(val)) {
       return false;
     }
@@ -1629,17 +1810,22 @@ struct chrono_formatter {
     return true;
   }
 
-  Rep hour() const { return static_cast<Rep>(mod((s.count() / 3600), 24)); }
+  auto days() const -> Rep { return static_cast<Rep>(s.count() / 86400); }
+  auto hour() const -> Rep {
+    return static_cast<Rep>(mod((s.count() / 3600), 24));
+  }
 
-  Rep hour12() const {
+  auto hour12() const -> Rep {
     Rep hour = static_cast<Rep>(mod((s.count() / 3600), 12));
     return hour <= 0 ? 12 : hour;
   }
 
-  Rep minute() const { return static_cast<Rep>(mod((s.count() / 60), 60)); }
-  Rep second() const { return static_cast<Rep>(mod(s.count(), 60)); }
+  auto minute() const -> Rep {
+    return static_cast<Rep>(mod((s.count() / 60), 60));
+  }
+  auto second() const -> Rep { return static_cast<Rep>(mod(s.count(), 60)); }
 
-  std::tm time() const {
+  auto time() const -> std::tm {
     auto time = std::tm();
     time.tm_hour = to_nonnegative_int(hour(), 24);
     time.tm_min = to_nonnegative_int(minute(), 60);
@@ -1654,44 +1840,16 @@ struct chrono_formatter {
     }
   }
 
-  void write(Rep value, int width) {
+  void write(Rep value, int width, pad_type pad = pad_type::unspecified) {
     write_sign();
     if (isnan(value)) return write_nan();
     uint32_or_64_or_128_t<int> n =
         to_unsigned(to_nonnegative_int(value, max_value<int>()));
     int num_digits = detail::count_digits(n);
-    if (width > num_digits) out = std::fill_n(out, width - num_digits, '0');
-    out = format_decimal<char_type>(out, n, num_digits).end;
-  }
-
-  template <typename Duration> void write_fractional_seconds(Duration d) {
-    FMT_ASSERT(!std::is_floating_point<typename Duration::rep>::value, "");
-    constexpr auto num_fractional_digits =
-        count_fractional_digits<Duration::period::num,
-                                Duration::period::den>::value;
-
-    using subsecond_precision = std::chrono::duration<
-        typename std::common_type<typename Duration::rep,
-                                  std::chrono::seconds::rep>::type,
-        std::ratio<1, detail::pow10(num_fractional_digits)>>;
-    if (std::ratio_less<typename subsecond_precision::period,
-                        std::chrono::seconds::period>::value) {
-      *out++ = '.';
-      auto fractional =
-          detail::abs(d) - std::chrono::duration_cast<std::chrono::seconds>(d);
-      auto subseconds =
-          std::chrono::treat_as_floating_point<
-              typename subsecond_precision::rep>::value
-              ? fractional.count()
-              : std::chrono::duration_cast<subsecond_precision>(fractional)
-                    .count();
-      uint32_or_64_or_128_t<long long> n =
-          to_unsigned(to_nonnegative_int(subseconds, max_value<long long>()));
-      int num_digits = detail::count_digits(n);
-      if (num_fractional_digits > num_digits)
-        out = std::fill_n(out, num_fractional_digits - num_digits, '0');
-      out = format_decimal<char_type>(out, n, num_digits).end;
+    if (width > num_digits) {
+      out = detail::write_padding(out, pad, width - num_digits);
     }
+    out = format_decimal<char_type>(out, n, num_digits).end;
   }
 
   void write_nan() { std::copy_n("nan", 3, out); }
@@ -1723,7 +1881,7 @@ struct chrono_formatter {
   void on_loc_time(numeric_system) {}
   void on_us_date() {}
   void on_iso_date() {}
-  void on_utc_offset() {}
+  void on_utc_offset(numeric_system) {}
   void on_tz_name() {}
   void on_year(numeric_system) {}
   void on_short_year(numeric_system) {}
@@ -1735,62 +1893,64 @@ struct chrono_formatter {
   void on_dec0_week_of_year(numeric_system) {}
   void on_dec1_week_of_year(numeric_system) {}
   void on_iso_week_of_year(numeric_system) {}
-  void on_day_of_year() {}
   void on_day_of_month(numeric_system) {}
   void on_day_of_month_space(numeric_system) {}
 
-  void on_24_hour(numeric_system ns) {
+  void on_day_of_year() {
     if (handle_nan_inf()) return;
+    write(days(), 0);
+  }
 
-    if (ns == numeric_system::standard) return write(hour(), 2);
+  void on_24_hour(numeric_system ns, pad_type pad) {
+    if (handle_nan_inf()) return;
+
+    if (ns == numeric_system::standard) return write(hour(), 2, pad);
     auto time = tm();
     time.tm_hour = to_nonnegative_int(hour(), 24);
-    format_tm(time, &tm_writer_type::on_24_hour, ns);
+    format_tm(time, &tm_writer_type::on_24_hour, ns, pad);
   }
 
-  void on_12_hour(numeric_system ns) {
+  void on_12_hour(numeric_system ns, pad_type pad) {
     if (handle_nan_inf()) return;
 
-    if (ns == numeric_system::standard) return write(hour12(), 2);
+    if (ns == numeric_system::standard) return write(hour12(), 2, pad);
     auto time = tm();
     time.tm_hour = to_nonnegative_int(hour12(), 12);
-    format_tm(time, &tm_writer_type::on_12_hour, ns);
+    format_tm(time, &tm_writer_type::on_12_hour, ns, pad);
   }
 
-  void on_minute(numeric_system ns) {
+  void on_minute(numeric_system ns, pad_type pad) {
     if (handle_nan_inf()) return;
 
-    if (ns == numeric_system::standard) return write(minute(), 2);
+    if (ns == numeric_system::standard) return write(minute(), 2, pad);
     auto time = tm();
     time.tm_min = to_nonnegative_int(minute(), 60);
-    format_tm(time, &tm_writer_type::on_minute, ns);
+    format_tm(time, &tm_writer_type::on_minute, ns, pad);
   }
 
-  void on_second(numeric_system ns) {
+  void on_second(numeric_system ns, pad_type pad) {
     if (handle_nan_inf()) return;
 
     if (ns == numeric_system::standard) {
       if (std::is_floating_point<rep>::value) {
-        constexpr auto num_fractional_digits =
-            count_fractional_digits<Period::num, Period::den>::value;
         auto buf = memory_buffer();
-        format_to(std::back_inserter(buf), runtime("{:.{}f}"),
-                  std::fmod(val * static_cast<rep>(Period::num) /
-                                static_cast<rep>(Period::den),
-                            static_cast<rep>(60)),
-                  num_fractional_digits);
+        write_floating_seconds(buf, std::chrono::duration<rep, Period>(val),
+                               precision);
         if (negative) *out++ = '-';
-        if (buf.size() < 2 || buf[1] == '.') *out++ = '0';
+        if (buf.size() < 2 || buf[1] == '.') {
+          out = detail::write_padding(out, pad);
+        }
         out = std::copy(buf.begin(), buf.end(), out);
       } else {
-        write(second(), 2);
-        write_fractional_seconds(std::chrono::duration<rep, Period>(val));
+        write(second(), 2, pad);
+        write_fractional_seconds<char_type>(
+            out, std::chrono::duration<rep, Period>(val), precision);
       }
       return;
     }
     auto time = tm();
     time.tm_sec = to_nonnegative_int(second(), 60);
-    format_tm(time, &tm_writer_type::on_second, ns);
+    format_tm(time, &tm_writer_type::on_second, ns, pad);
   }
 
   void on_12_hour_time() {
@@ -1814,7 +1974,7 @@ struct chrono_formatter {
     on_24_hour_time();
     *out++ = ':';
     if (handle_nan_inf()) return;
-    on_second(numeric_system::standard);
+    on_second(numeric_system::standard, pad_type::unspecified);
   }
 
   void on_am_pm() {
@@ -1833,7 +1993,7 @@ struct chrono_formatter {
   }
 };
 
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
 #if defined(__cpp_lib_chrono) && __cpp_lib_chrono >= 201907
 using weekday = std::chrono::weekday;
@@ -1847,7 +2007,7 @@ class weekday {
   weekday() = default;
   explicit constexpr weekday(unsigned wd) noexcept
       : value(static_cast<unsigned char>(wd != 7 ? wd : 0)) {}
-  constexpr unsigned c_encoding() const noexcept { return value; }
+  constexpr auto c_encoding() const noexcept -> unsigned { return value; }
 };
 
 class year_month_day {};
@@ -1883,118 +2043,67 @@ template <typename Char> struct formatter<weekday, Char> {
 template <typename Rep, typename Period, typename Char>
 struct formatter<std::chrono::duration<Rep, Period>, Char> {
  private:
-  basic_format_specs<Char> specs;
-  int precision = -1;
-  using arg_ref_type = detail::arg_ref<Char>;
-  arg_ref_type width_ref;
-  arg_ref_type precision_ref;
-  bool localized = false;
-  basic_string_view<Char> format_str;
-  using duration = std::chrono::duration<Rep, Period>;
-
-  struct spec_handler {
-    formatter& f;
-    basic_format_parse_context<Char>& context;
-    basic_string_view<Char> format_str;
-
-    template <typename Id> FMT_CONSTEXPR arg_ref_type make_arg_ref(Id arg_id) {
-      context.check_arg_id(arg_id);
-      return arg_ref_type(arg_id);
-    }
+  format_specs<Char> specs_;
+  detail::arg_ref<Char> width_ref_;
+  detail::arg_ref<Char> precision_ref_;
+  bool localized_ = false;
+  basic_string_view<Char> format_str_;
 
-    FMT_CONSTEXPR arg_ref_type make_arg_ref(basic_string_view<Char> arg_id) {
-      context.check_arg_id(arg_id);
-      return arg_ref_type(arg_id);
-    }
-
-    FMT_CONSTEXPR arg_ref_type make_arg_ref(detail::auto_id) {
-      return arg_ref_type(context.next_arg_id());
-    }
+ public:
+  FMT_CONSTEXPR auto parse(basic_format_parse_context<Char>& ctx)
+      -> decltype(ctx.begin()) {
+    auto it = ctx.begin(), end = ctx.end();
+    if (it == end || *it == '}') return it;
 
-    void on_error(const char* msg) { FMT_THROW(format_error(msg)); }
-    FMT_CONSTEXPR void on_fill(basic_string_view<Char> fill) {
-      f.specs.fill = fill;
-    }
-    FMT_CONSTEXPR void on_align(align_t align) { f.specs.align = align; }
-    FMT_CONSTEXPR void on_width(int width) { f.specs.width = width; }
-    FMT_CONSTEXPR void on_precision(int _precision) {
-      f.precision = _precision;
-    }
-    FMT_CONSTEXPR void end_precision() {}
+    it = detail::parse_align(it, end, specs_);
+    if (it == end) return it;
 
-    template <typename Id> FMT_CONSTEXPR void on_dynamic_width(Id arg_id) {
-      f.width_ref = make_arg_ref(arg_id);
-    }
+    it = detail::parse_dynamic_spec(it, end, specs_.width, width_ref_, ctx);
+    if (it == end) return it;
 
-    template <typename Id> FMT_CONSTEXPR void on_dynamic_precision(Id arg_id) {
-      f.precision_ref = make_arg_ref(arg_id);
+    auto checker = detail::chrono_format_checker();
+    if (*it == '.') {
+      checker.has_precision_integral = !std::is_floating_point<Rep>::value;
+      it = detail::parse_precision(it, end, specs_.precision, precision_ref_,
+                                   ctx);
     }
-  };
-
-  using iterator = typename basic_format_parse_context<Char>::iterator;
-  struct parse_range {
-    iterator begin;
-    iterator end;
-  };
-
-  FMT_CONSTEXPR parse_range do_parse(basic_format_parse_context<Char>& ctx) {
-    auto begin = ctx.begin(), end = ctx.end();
-    if (begin == end || *begin == '}') return {begin, begin};
-    spec_handler handler{*this, ctx, format_str};
-    begin = detail::parse_align(begin, end, handler);
-    if (begin == end) return {begin, begin};
-    begin = detail::parse_width(begin, end, handler);
-    if (begin == end) return {begin, begin};
-    if (*begin == '.') {
-      if (std::is_floating_point<Rep>::value)
-        begin = detail::parse_precision(begin, end, handler);
-      else
-        handler.on_error("precision not allowed for this argument type");
+    if (it != end && *it == 'L') {
+      localized_ = true;
+      ++it;
     }
-    if (begin != end && *begin == 'L') {
-      ++begin;
-      localized = true;
-    }
-    end = detail::parse_chrono_format(begin, end,
-                                      detail::chrono_format_checker());
-    return {begin, end};
-  }
-
- public:
-  FMT_CONSTEXPR auto parse(basic_format_parse_context<Char>& ctx)
-      -> decltype(ctx.begin()) {
-    auto range = do_parse(ctx);
-    format_str = basic_string_view<Char>(
-        &*range.begin, detail::to_unsigned(range.end - range.begin));
-    return range.end;
+    end = detail::parse_chrono_format(it, end, checker);
+    format_str_ = {it, detail::to_unsigned(end - it)};
+    return end;
   }
 
   template <typename FormatContext>
-  auto format(const duration& d, FormatContext& ctx) const
+  auto format(std::chrono::duration<Rep, Period> d, FormatContext& ctx) const
       -> decltype(ctx.out()) {
-    auto specs_copy = specs;
-    auto precision_copy = precision;
-    auto begin = format_str.begin(), end = format_str.end();
+    auto specs = specs_;
+    auto precision = specs.precision;
+    specs.precision = -1;
+    auto begin = format_str_.begin(), end = format_str_.end();
     // As a possible future optimization, we could avoid extra copying if width
     // is not specified.
-    basic_memory_buffer<Char> buf;
+    auto buf = basic_memory_buffer<Char>();
     auto out = std::back_inserter(buf);
-    detail::handle_dynamic_spec<detail::width_checker>(specs_copy.width,
-                                                       width_ref, ctx);
-    detail::handle_dynamic_spec<detail::precision_checker>(precision_copy,
-                                                           precision_ref, ctx);
+    detail::handle_dynamic_spec<detail::width_checker>(specs.width, width_ref_,
+                                                       ctx);
+    detail::handle_dynamic_spec<detail::precision_checker>(precision,
+                                                           precision_ref_, ctx);
     if (begin == end || *begin == '}') {
-      out = detail::format_duration_value<Char>(out, d.count(), precision_copy);
+      out = detail::format_duration_value<Char>(out, d.count(), precision);
       detail::format_duration_unit<Char, Period>(out);
     } else {
-      detail::chrono_formatter<FormatContext, decltype(out), Rep, Period> f(
-          ctx, out, d);
-      f.precision = precision_copy;
-      f.localized = localized;
+      using chrono_formatter =
+          detail::chrono_formatter<FormatContext, decltype(out), Rep, Period>;
+      auto f = chrono_formatter(ctx, out, d);
+      f.precision = precision;
+      f.localized = localized_;
       detail::parse_chrono_format(begin, end, f);
     }
     return detail::write(
-        ctx.out(), basic_string_view<Char>(buf.data(), buf.size()), specs_copy);
+        ctx.out(), basic_string_view<Char>(buf.data(), buf.size()), specs);
   }
 };
 
@@ -2002,68 +2111,130 @@ template <typename Char, typename Duration>
 struct formatter<std::chrono::time_point<std::chrono::system_clock, Duration>,
                  Char> : formatter<std::tm, Char> {
   FMT_CONSTEXPR formatter() {
-    basic_string_view<Char> default_specs =
-        detail::string_literal<Char, '%', 'F', ' ', '%', 'T'>{};
-    this->do_parse(default_specs.begin(), default_specs.end());
+    this->format_str_ = detail::string_literal<Char, '%', 'F', ' ', '%', 'T'>{};
   }
 
   template <typename FormatContext>
-  auto format(std::chrono::time_point<std::chrono::system_clock> val,
+  auto format(std::chrono::time_point<std::chrono::system_clock, Duration> val,
               FormatContext& ctx) const -> decltype(ctx.out()) {
+    using period = typename Duration::period;
+    if (detail::const_check(
+            period::num != 1 || period::den != 1 ||
+            std::is_floating_point<typename Duration::rep>::value)) {
+      const auto epoch = val.time_since_epoch();
+      auto subsecs = detail::fmt_duration_cast<Duration>(
+          epoch - detail::fmt_duration_cast<std::chrono::seconds>(epoch));
+
+      if (subsecs.count() < 0) {
+        auto second =
+            detail::fmt_duration_cast<Duration>(std::chrono::seconds(1));
+        if (epoch.count() < ((Duration::min)() + second).count())
+          FMT_THROW(format_error("duration is too small"));
+        subsecs += second;
+        val -= second;
+      }
+
+      return formatter<std::tm, Char>::do_format(gmtime(val), ctx, &subsecs);
+    }
+
+    return formatter<std::tm, Char>::format(gmtime(val), ctx);
+  }
+};
+
+#if FMT_USE_LOCAL_TIME
+template <typename Char, typename Duration>
+struct formatter<std::chrono::local_time<Duration>, Char>
+    : formatter<std::tm, Char> {
+  FMT_CONSTEXPR formatter() {
+    this->format_str_ = detail::string_literal<Char, '%', 'F', ' ', '%', 'T'>{};
+  }
+
+  template <typename FormatContext>
+  auto format(std::chrono::local_time<Duration> val, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    using period = typename Duration::period;
+    if (period::num != 1 || period::den != 1 ||
+        std::is_floating_point<typename Duration::rep>::value) {
+      const auto epoch = val.time_since_epoch();
+      const auto subsecs = detail::fmt_duration_cast<Duration>(
+          epoch - detail::fmt_duration_cast<std::chrono::seconds>(epoch));
+
+      return formatter<std::tm, Char>::do_format(localtime(val), ctx, &subsecs);
+    }
+
     return formatter<std::tm, Char>::format(localtime(val), ctx);
   }
 };
+#endif
+
+#if FMT_USE_UTC_TIME
+template <typename Char, typename Duration>
+struct formatter<std::chrono::time_point<std::chrono::utc_clock, Duration>,
+                 Char>
+    : formatter<std::chrono::time_point<std::chrono::system_clock, Duration>,
+                Char> {
+  template <typename FormatContext>
+  auto format(std::chrono::time_point<std::chrono::utc_clock, Duration> val,
+              FormatContext& ctx) const -> decltype(ctx.out()) {
+    return formatter<
+        std::chrono::time_point<std::chrono::system_clock, Duration>,
+        Char>::format(std::chrono::utc_clock::to_sys(val), ctx);
+  }
+};
+#endif
 
 template <typename Char> struct formatter<std::tm, Char> {
  private:
-  enum class spec {
-    unknown,
-    year_month_day,
-    hh_mm_ss,
-  };
-  spec spec_ = spec::unknown;
-  basic_string_view<Char> specs;
+  format_specs<Char> specs_;
+  detail::arg_ref<Char> width_ref_;
 
  protected:
-  template <typename It> FMT_CONSTEXPR auto do_parse(It begin, It end) -> It {
-    if (begin != end && *begin == ':') ++begin;
-    end = detail::parse_chrono_format(begin, end, detail::tm_format_checker());
-    // Replace default spec only if the new spec is not empty.
-    if (end != begin) specs = {begin, detail::to_unsigned(end - begin)};
-    return end;
+  basic_string_view<Char> format_str_;
+
+  template <typename FormatContext, typename Duration>
+  auto do_format(const std::tm& tm, FormatContext& ctx,
+                 const Duration* subsecs) const -> decltype(ctx.out()) {
+    auto specs = specs_;
+    auto buf = basic_memory_buffer<Char>();
+    auto out = std::back_inserter(buf);
+    detail::handle_dynamic_spec<detail::width_checker>(specs.width, width_ref_,
+                                                       ctx);
+
+    auto loc_ref = ctx.locale();
+    detail::get_locale loc(static_cast<bool>(loc_ref), loc_ref);
+    auto w =
+        detail::tm_writer<decltype(out), Char, Duration>(loc, out, tm, subsecs);
+    detail::parse_chrono_format(format_str_.begin(), format_str_.end(), w);
+    return detail::write(
+        ctx.out(), basic_string_view<Char>(buf.data(), buf.size()), specs);
   }
 
  public:
   FMT_CONSTEXPR auto parse(basic_format_parse_context<Char>& ctx)
       -> decltype(ctx.begin()) {
-    auto end = this->do_parse(ctx.begin(), ctx.end());
-    // basic_string_view<>::compare isn't constexpr before C++17.
-    if (specs.size() == 2 && specs[0] == Char('%')) {
-      if (specs[1] == Char('F'))
-        spec_ = spec::year_month_day;
-      else if (specs[1] == Char('T'))
-        spec_ = spec::hh_mm_ss;
-    }
+    auto it = ctx.begin(), end = ctx.end();
+    if (it == end || *it == '}') return it;
+
+    it = detail::parse_align(it, end, specs_);
+    if (it == end) return it;
+
+    it = detail::parse_dynamic_spec(it, end, specs_.width, width_ref_, ctx);
+    if (it == end) return it;
+
+    end = detail::parse_chrono_format(it, end, detail::tm_format_checker());
+    // Replace the default format_str only if the new spec is not empty.
+    if (end != it) format_str_ = {it, detail::to_unsigned(end - it)};
     return end;
   }
 
   template <typename FormatContext>
   auto format(const std::tm& tm, FormatContext& ctx) const
       -> decltype(ctx.out()) {
-    const auto loc_ref = ctx.locale();
-    detail::get_locale loc(static_cast<bool>(loc_ref), loc_ref);
-    auto w = detail::tm_writer<decltype(ctx.out()), Char>(loc, ctx.out(), tm);
-    if (spec_ == spec::year_month_day)
-      w.on_iso_date();
-    else if (spec_ == spec::hh_mm_ss)
-      w.on_iso_time();
-    else
-      detail::parse_chrono_format(specs.begin(), specs.end(), w);
-    return w.out();
+    return do_format<FormatContext, std::chrono::seconds>(tm, ctx, nullptr);
   }
 };
 
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_CHRONO_H_
diff --git a/dep/fmt/include/fmt/color.h b/dep/fmt/include/fmt/color.h
index 4c163277ef1..367849a86a7 100644
--- a/dep/fmt/include/fmt/color.h
+++ b/dep/fmt/include/fmt/color.h
@@ -11,7 +11,7 @@
 #include "format.h"
 
 FMT_BEGIN_NAMESPACE
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
 enum class color : uint32_t {
   alice_blue = 0xF0F8FF,               // rgb(240,248,255)
@@ -203,7 +203,7 @@ struct rgb {
   uint8_t b;
 };
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 // color is a struct of either a rgb color or a terminal color.
 struct color_type {
@@ -225,8 +225,7 @@ struct color_type {
     uint32_t rgb_color;
   } value;
 };
-
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
 /** A text style consisting of foreground and background colors and emphasis. */
 class text_style {
@@ -234,7 +233,7 @@ class text_style {
   FMT_CONSTEXPR text_style(emphasis em = emphasis()) noexcept
       : set_foreground_color(), set_background_color(), ems(em) {}
 
-  FMT_CONSTEXPR text_style& operator|=(const text_style& rhs) {
+  FMT_CONSTEXPR auto operator|=(const text_style& rhs) -> text_style& {
     if (!set_foreground_color) {
       set_foreground_color = rhs.set_foreground_color;
       foreground_color = rhs.foreground_color;
@@ -258,29 +257,29 @@ class text_style {
     return *this;
   }
 
-  friend FMT_CONSTEXPR text_style operator|(text_style lhs,
-                                            const text_style& rhs) {
+  friend FMT_CONSTEXPR auto operator|(text_style lhs, const text_style& rhs)
+      -> text_style {
     return lhs |= rhs;
   }
 
-  FMT_CONSTEXPR bool has_foreground() const noexcept {
+  FMT_CONSTEXPR auto has_foreground() const noexcept -> bool {
     return set_foreground_color;
   }
-  FMT_CONSTEXPR bool has_background() const noexcept {
+  FMT_CONSTEXPR auto has_background() const noexcept -> bool {
     return set_background_color;
   }
-  FMT_CONSTEXPR bool has_emphasis() const noexcept {
+  FMT_CONSTEXPR auto has_emphasis() const noexcept -> bool {
     return static_cast<uint8_t>(ems) != 0;
   }
-  FMT_CONSTEXPR detail::color_type get_foreground() const noexcept {
+  FMT_CONSTEXPR auto get_foreground() const noexcept -> detail::color_type {
     FMT_ASSERT(has_foreground(), "no foreground specified for this style");
     return foreground_color;
   }
-  FMT_CONSTEXPR detail::color_type get_background() const noexcept {
+  FMT_CONSTEXPR auto get_background() const noexcept -> detail::color_type {
     FMT_ASSERT(has_background(), "no background specified for this style");
     return background_color;
   }
-  FMT_CONSTEXPR emphasis get_emphasis() const noexcept {
+  FMT_CONSTEXPR auto get_emphasis() const noexcept -> emphasis {
     FMT_ASSERT(has_emphasis(), "no emphasis specified for this style");
     return ems;
   }
@@ -298,9 +297,11 @@ class text_style {
     }
   }
 
-  friend FMT_CONSTEXPR text_style fg(detail::color_type foreground) noexcept;
+  friend FMT_CONSTEXPR auto fg(detail::color_type foreground) noexcept
+      -> text_style;
 
-  friend FMT_CONSTEXPR text_style bg(detail::color_type background) noexcept;
+  friend FMT_CONSTEXPR auto bg(detail::color_type background) noexcept
+      -> text_style;
 
   detail::color_type foreground_color;
   detail::color_type background_color;
@@ -310,20 +311,23 @@ class text_style {
 };
 
 /** Creates a text style from the foreground (text) color. */
-FMT_CONSTEXPR inline text_style fg(detail::color_type foreground) noexcept {
+FMT_CONSTEXPR inline auto fg(detail::color_type foreground) noexcept
+    -> text_style {
   return text_style(true, foreground);
 }
 
 /** Creates a text style from the background color. */
-FMT_CONSTEXPR inline text_style bg(detail::color_type background) noexcept {
+FMT_CONSTEXPR inline auto bg(detail::color_type background) noexcept
+    -> text_style {
   return text_style(false, background);
 }
 
-FMT_CONSTEXPR inline text_style operator|(emphasis lhs, emphasis rhs) noexcept {
+FMT_CONSTEXPR inline auto operator|(emphasis lhs, emphasis rhs) noexcept
+    -> text_style {
   return text_style(lhs) | rhs;
 }
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 template <typename Char> struct ansi_color_escape {
   FMT_CONSTEXPR ansi_color_escape(detail::color_type text_color,
@@ -385,8 +389,8 @@ template <typename Char> struct ansi_color_escape {
   }
   FMT_CONSTEXPR operator const Char*() const noexcept { return buffer; }
 
-  FMT_CONSTEXPR const Char* begin() const noexcept { return buffer; }
-  FMT_CONSTEXPR_CHAR_TRAITS const Char* end() const noexcept {
+  FMT_CONSTEXPR auto begin() const noexcept -> const Char* { return buffer; }
+  FMT_CONSTEXPR_CHAR_TRAITS auto end() const noexcept -> const Char* {
     return buffer + std::char_traits<Char>::length(buffer);
   }
 
@@ -401,56 +405,39 @@ template <typename Char> struct ansi_color_escape {
     out[2] = static_cast<Char>('0' + c % 10);
     out[3] = static_cast<Char>(delimiter);
   }
-  static FMT_CONSTEXPR bool has_emphasis(emphasis em, emphasis mask) noexcept {
+  static FMT_CONSTEXPR auto has_emphasis(emphasis em, emphasis mask) noexcept
+      -> bool {
     return static_cast<uint8_t>(em) & static_cast<uint8_t>(mask);
   }
 };
 
 template <typename Char>
-FMT_CONSTEXPR ansi_color_escape<Char> make_foreground_color(
-    detail::color_type foreground) noexcept {
+FMT_CONSTEXPR auto make_foreground_color(detail::color_type foreground) noexcept
+    -> ansi_color_escape<Char> {
   return ansi_color_escape<Char>(foreground, "\x1b[38;2;");
 }
 
 template <typename Char>
-FMT_CONSTEXPR ansi_color_escape<Char> make_background_color(
-    detail::color_type background) noexcept {
+FMT_CONSTEXPR auto make_background_color(detail::color_type background) noexcept
+    -> ansi_color_escape<Char> {
   return ansi_color_escape<Char>(background, "\x1b[48;2;");
 }
 
 template <typename Char>
-FMT_CONSTEXPR ansi_color_escape<Char> make_emphasis(emphasis em) noexcept {
+FMT_CONSTEXPR auto make_emphasis(emphasis em) noexcept
+    -> ansi_color_escape<Char> {
   return ansi_color_escape<Char>(em);
 }
 
-template <typename Char> inline void fputs(const Char* chars, FILE* stream) {
-  int result = std::fputs(chars, stream);
-  if (result < 0)
-    FMT_THROW(system_error(errno, FMT_STRING("cannot write to file")));
-}
-
-template <> inline void fputs<wchar_t>(const wchar_t* chars, FILE* stream) {
-  int result = std::fputws(chars, stream);
-  if (result < 0)
-    FMT_THROW(system_error(errno, FMT_STRING("cannot write to file")));
-}
-
-template <typename Char> inline void reset_color(FILE* stream) {
-  fputs("\x1b[0m", stream);
-}
-
-template <> inline void reset_color<wchar_t>(FILE* stream) {
-  fputs(L"\x1b[0m", stream);
-}
-
 template <typename Char> inline void reset_color(buffer<Char>& buffer) {
   auto reset_color = string_view("\x1b[0m");
   buffer.append(reset_color.begin(), reset_color.end());
 }
 
-template <typename T> struct styled_arg {
+template <typename T> struct styled_arg : detail::view {
   const T& value;
   text_style style;
+  styled_arg(const T& v, text_style s) : value(v), style(s) {}
 };
 
 template <typename Char>
@@ -477,19 +464,21 @@ void vformat_to(buffer<Char>& buf, const text_style& ts,
   if (has_style) detail::reset_color<Char>(buf);
 }
 
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
-template <typename S, typename Char = char_t<S>>
-void vprint(std::FILE* f, const text_style& ts, const S& format,
-            basic_format_args<buffer_context<type_identity_t<Char>>> args) {
-  basic_memory_buffer<Char> buf;
-  detail::vformat_to(buf, ts, detail::to_string_view(format), args);
+inline void vprint(std::FILE* f, const text_style& ts, string_view fmt,
+                   format_args args) {
+  // Legacy wide streams are not supported.
+  auto buf = memory_buffer();
+  detail::vformat_to(buf, ts, fmt, args);
   if (detail::is_utf8()) {
-    detail::print(f, basic_string_view<Char>(buf.begin(), buf.size()));
-  } else {
-    buf.push_back(Char(0));
-    detail::fputs(buf.data(), f);
+    detail::print(f, string_view(buf.begin(), buf.size()));
+    return;
   }
+  buf.push_back('\0');
+  int result = std::fputs(buf.data(), f);
+  if (result < 0)
+    FMT_THROW(system_error(errno, FMT_STRING("cannot write to file")));
 }
 
 /**
@@ -529,9 +518,10 @@ void print(const text_style& ts, const S& format_str, const Args&... args) {
 }
 
 template <typename S, typename Char = char_t<S>>
-inline std::basic_string<Char> vformat(
+inline auto vformat(
     const text_style& ts, const S& format_str,
-    basic_format_args<buffer_context<type_identity_t<Char>>> args) {
+    basic_format_args<buffer_context<type_identity_t<Char>>> args)
+    -> std::basic_string<Char> {
   basic_memory_buffer<Char> buf;
   detail::vformat_to(buf, ts, detail::to_string_view(format_str), args);
   return fmt::to_string(buf);
@@ -550,8 +540,8 @@ inline std::basic_string<Char> vformat(
   \endrst
 */
 template <typename S, typename... Args, typename Char = char_t<S>>
-inline std::basic_string<Char> format(const text_style& ts, const S& format_str,
-                                      const Args&... args) {
+inline auto format(const text_style& ts, const S& format_str,
+                   const Args&... args) -> std::basic_string<Char> {
   return fmt::vformat(ts, detail::to_string_view(format_str),
                       fmt::make_format_args<buffer_context<Char>>(args...));
 }
@@ -561,12 +551,13 @@ inline std::basic_string<Char> format(const text_style& ts, const S& format_str,
  */
 template <typename OutputIt, typename Char,
           FMT_ENABLE_IF(detail::is_output_iterator<OutputIt, Char>::value)>
-OutputIt vformat_to(
-    OutputIt out, const text_style& ts, basic_string_view<Char> format_str,
-    basic_format_args<buffer_context<type_identity_t<Char>>> args) {
+auto vformat_to(OutputIt out, const text_style& ts,
+                basic_string_view<Char> format_str,
+                basic_format_args<buffer_context<type_identity_t<Char>>> args)
+    -> OutputIt {
   auto&& buf = detail::get_buffer<Char>(out);
   detail::vformat_to(buf, ts, format_str, args);
-  return detail::get_iterator(buf);
+  return detail::get_iterator(buf, out);
 }
 
 /**
@@ -581,9 +572,10 @@ OutputIt vformat_to(
                    fmt::emphasis::bold | fg(fmt::color::red), "{}", 42);
   \endrst
 */
-template <typename OutputIt, typename S, typename... Args,
-          bool enable = detail::is_output_iterator<OutputIt, char_t<S>>::value&&
-              detail::is_string<S>::value>
+template <
+    typename OutputIt, typename S, typename... Args,
+    bool enable = detail::is_output_iterator<OutputIt, char_t<S>>::value &&
+                  detail::is_string<S>::value>
 inline auto format_to(OutputIt out, const text_style& ts, const S& format_str,
                       Args&&... args) ->
     typename std::enable_if<enable, OutputIt>::type {
@@ -645,7 +637,7 @@ FMT_CONSTEXPR auto styled(const T& value, text_style ts)
   return detail::styled_arg<remove_cvref_t<T>>{value, ts};
 }
 
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_COLOR_H_
diff --git a/dep/fmt/include/fmt/compile.h b/dep/fmt/include/fmt/compile.h
index 933668c41c3..3b3f166e0cd 100644
--- a/dep/fmt/include/fmt/compile.h
+++ b/dep/fmt/include/fmt/compile.h
@@ -14,89 +14,11 @@ FMT_BEGIN_NAMESPACE
 namespace detail {
 
 template <typename Char, typename InputIt>
-FMT_CONSTEXPR inline counting_iterator copy_str(InputIt begin, InputIt end,
-                                                counting_iterator it) {
+FMT_CONSTEXPR inline auto copy_str(InputIt begin, InputIt end,
+                                   counting_iterator it) -> counting_iterator {
   return it + (end - begin);
 }
 
-template <typename OutputIt> class truncating_iterator_base {
- protected:
-  OutputIt out_;
-  size_t limit_;
-  size_t count_ = 0;
-
-  truncating_iterator_base() : out_(), limit_(0) {}
-
-  truncating_iterator_base(OutputIt out, size_t limit)
-      : out_(out), limit_(limit) {}
-
- public:
-  using iterator_category = std::output_iterator_tag;
-  using value_type = typename std::iterator_traits<OutputIt>::value_type;
-  using difference_type = std::ptrdiff_t;
-  using pointer = void;
-  using reference = void;
-  FMT_UNCHECKED_ITERATOR(truncating_iterator_base);
-
-  OutputIt base() const { return out_; }
-  size_t count() const { return count_; }
-};
-
-// An output iterator that truncates the output and counts the number of objects
-// written to it.
-template <typename OutputIt,
-          typename Enable = typename std::is_void<
-              typename std::iterator_traits<OutputIt>::value_type>::type>
-class truncating_iterator;
-
-template <typename OutputIt>
-class truncating_iterator<OutputIt, std::false_type>
-    : public truncating_iterator_base<OutputIt> {
-  mutable typename truncating_iterator_base<OutputIt>::value_type blackhole_;
-
- public:
-  using value_type = typename truncating_iterator_base<OutputIt>::value_type;
-
-  truncating_iterator() = default;
-
-  truncating_iterator(OutputIt out, size_t limit)
-      : truncating_iterator_base<OutputIt>(out, limit) {}
-
-  truncating_iterator& operator++() {
-    if (this->count_++ < this->limit_) ++this->out_;
-    return *this;
-  }
-
-  truncating_iterator operator++(int) {
-    auto it = *this;
-    ++*this;
-    return it;
-  }
-
-  value_type& operator*() const {
-    return this->count_ < this->limit_ ? *this->out_ : blackhole_;
-  }
-};
-
-template <typename OutputIt>
-class truncating_iterator<OutputIt, std::true_type>
-    : public truncating_iterator_base<OutputIt> {
- public:
-  truncating_iterator() = default;
-
-  truncating_iterator(OutputIt out, size_t limit)
-      : truncating_iterator_base<OutputIt>(out, limit) {}
-
-  template <typename T> truncating_iterator& operator=(T val) {
-    if (this->count_++ < this->limit_) *this->out_++ = val;
-    return *this;
-  }
-
-  truncating_iterator& operator++() { return *this; }
-  truncating_iterator& operator++(int) { return *this; }
-  truncating_iterator& operator*() { return *this; }
-};
-
 // A compile-time string which is compiled into fast formatting code.
 class compiled_string {};
 
@@ -135,7 +57,7 @@ struct udl_compiled_string : compiled_string {
 #endif
 
 template <typename T, typename... Tail>
-const T& first(const T& value, const Tail&...) {
+auto first(const T& value, const Tail&...) -> const T& {
   return value;
 }
 
@@ -196,7 +118,8 @@ template <typename Char> struct code_unit {
 
   template <typename OutputIt, typename... Args>
   constexpr OutputIt format(OutputIt out, const Args&...) const {
-    return write<Char>(out, value);
+    *out++ = value;
+    return out;
   }
 };
 
@@ -220,7 +143,12 @@ template <typename Char, typename T, int N> struct field {
 
   template <typename OutputIt, typename... Args>
   constexpr OutputIt format(OutputIt out, const Args&... args) const {
-    return write<Char>(out, get_arg_checked<T, N>(args...));
+    const T& arg = get_arg_checked<T, N>(args...);
+    if constexpr (std::is_convertible_v<T, basic_string_view<Char>>) {
+      auto s = basic_string_view<Char>(arg);
+      return copy_str<Char>(s.begin(), s.end(), out);
+    }
+    return write<Char>(out, arg);
   }
 };
 
@@ -331,14 +259,14 @@ template <typename T, typename Char> struct parse_specs_result {
   int next_arg_id;
 };
 
-constexpr int manual_indexing_id = -1;
+enum { manual_indexing_id = -1 };
 
 template <typename T, typename Char>
 constexpr parse_specs_result<T, Char> parse_specs(basic_string_view<Char> str,
                                                   size_t pos, int next_arg_id) {
   str.remove_prefix(pos);
-  auto ctx = compile_parse_context<Char>(str, max_value<int>(), nullptr, {},
-                                         next_arg_id);
+  auto ctx =
+      compile_parse_context<Char>(str, max_value<int>(), nullptr, next_arg_id);
   auto f = formatter<T, Char>();
   auto end = f.parse(ctx);
   return {f, pos + fmt::detail::to_unsigned(end - str.data()),
@@ -348,22 +276,18 @@ constexpr parse_specs_result<T, Char> parse_specs(basic_string_view<Char> str,
 template <typename Char> struct arg_id_handler {
   arg_ref<Char> arg_id;
 
-  constexpr int operator()() {
+  constexpr int on_auto() {
     FMT_ASSERT(false, "handler cannot be used with automatic indexing");
     return 0;
   }
-  constexpr int operator()(int id) {
+  constexpr int on_index(int id) {
     arg_id = arg_ref<Char>(id);
     return 0;
   }
-  constexpr int operator()(basic_string_view<Char> id) {
+  constexpr int on_name(basic_string_view<Char> id) {
     arg_id = arg_ref<Char>(id);
     return 0;
   }
-
-  constexpr void on_error(const char* message) {
-    FMT_THROW(format_error(message));
-  }
 };
 
 template <typename Char> struct parse_arg_id_result {
@@ -452,20 +376,18 @@ constexpr auto compile_format_string(S format_str) {
       } else if constexpr (arg_id_result.arg_id.kind == arg_id_kind::name) {
         constexpr auto arg_index =
             get_arg_index_by_name(arg_id_result.arg_id.val.name, Args{});
-        if constexpr (arg_index != invalid_arg_index) {
+        if constexpr (arg_index >= 0) {
           constexpr auto next_id =
               ID != manual_indexing_id ? ID + 1 : manual_indexing_id;
           return parse_replacement_field_then_tail<
               decltype(get_type<arg_index, Args>::value), Args, arg_id_end_pos,
               arg_index, next_id>(format_str);
-        } else {
-          if constexpr (c == '}') {
-            return parse_tail<Args, arg_id_end_pos + 1, ID>(
-                runtime_named_field<char_type>{arg_id_result.arg_id.val.name},
-                format_str);
-          } else if constexpr (c == ':') {
-            return unknown_format();  // no type info for specs parsing
-          }
+        } else if constexpr (c == '}') {
+          return parse_tail<Args, arg_id_end_pos + 1, ID>(
+              runtime_named_field<char_type>{arg_id_result.arg_id.val.name},
+              format_str);
+        } else if constexpr (c == ':') {
+          return unknown_format();  // no type info for specs parsing
         }
       }
     }
@@ -501,7 +423,7 @@ constexpr auto compile(S format_str) {
 #endif  // defined(__cpp_if_constexpr) && defined(__cpp_return_type_deduction)
 }  // namespace detail
 
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
 #if defined(__cpp_if_constexpr) && defined(__cpp_return_type_deduction)
 
@@ -566,17 +488,19 @@ FMT_CONSTEXPR OutputIt format_to(OutputIt out, const S&, Args&&... args) {
 
 template <typename OutputIt, typename S, typename... Args,
           FMT_ENABLE_IF(detail::is_compiled_string<S>::value)>
-format_to_n_result<OutputIt> format_to_n(OutputIt out, size_t n,
-                                         const S& format_str, Args&&... args) {
-  auto it = fmt::format_to(detail::truncating_iterator<OutputIt>(out, n),
-                           format_str, std::forward<Args>(args)...);
-  return {it.base(), it.count()};
+auto format_to_n(OutputIt out, size_t n, const S& format_str, Args&&... args)
+    -> format_to_n_result<OutputIt> {
+  using traits = detail::fixed_buffer_traits;
+  auto buf = detail::iterator_buffer<OutputIt, char, traits>(out, n);
+  fmt::format_to(std::back_inserter(buf), format_str,
+                 std::forward<Args>(args)...);
+  return {buf.out(), buf.count()};
 }
 
 template <typename S, typename... Args,
           FMT_ENABLE_IF(detail::is_compiled_string<S>::value)>
-FMT_CONSTEXPR20 size_t formatted_size(const S& format_str,
-                                      const Args&... args) {
+FMT_CONSTEXPR20 auto formatted_size(const S& format_str, const Args&... args)
+    -> size_t {
   return fmt::format_to(detail::counting_iterator(), format_str, args...)
       .count();
 }
@@ -605,7 +529,7 @@ template <detail_exported::fixed_string Str> constexpr auto operator""_cf() {
 }  // namespace literals
 #endif
 
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_COMPILE_H_
diff --git a/dep/fmt/include/fmt/core.h b/dep/fmt/include/fmt/core.h
index a4f0d2752f8..b51c1406a99 100644
--- a/dep/fmt/include/fmt/core.h
+++ b/dep/fmt/include/fmt/core.h
@@ -13,11 +13,12 @@
 #include <cstring>  // std::strlen
 #include <iterator>
 #include <limits>
+#include <memory>  // std::addressof
 #include <string>
 #include <type_traits>
 
 // The fmt library version in the form major * 10000 + minor * 100 + patch.
-#define FMT_VERSION 90100
+#define FMT_VERSION 100201
 
 #if defined(__clang__) && !defined(__ibmxl__)
 #  define FMT_CLANG_VERSION (__clang_major__ * 100 + __clang_minor__)
@@ -69,9 +70,7 @@
 #  define FMT_HAS_FEATURE(x) 0
 #endif
 
-#if (defined(__has_include) || FMT_ICC_VERSION >= 1600 || \
-     FMT_MSC_VERSION > 1900) &&                           \
-    !defined(__INTELLISENSE__)
+#if defined(__has_include) || FMT_ICC_VERSION >= 1600 || FMT_MSC_VERSION > 1900
 #  define FMT_HAS_INCLUDE(x) __has_include(x)
 #else
 #  define FMT_HAS_INCLUDE(x) 0
@@ -94,7 +93,7 @@
 #ifndef FMT_USE_CONSTEXPR
 #  if (FMT_HAS_FEATURE(cxx_relaxed_constexpr) || FMT_MSC_VERSION >= 1912 || \
        (FMT_GCC_VERSION >= 600 && FMT_CPLUSPLUS >= 201402L)) &&             \
-      !FMT_ICC_VERSION && !defined(__NVCC__)
+      !FMT_ICC_VERSION && (!defined(__NVCC__) || FMT_CPLUSPLUS >= 202002L)
 #    define FMT_USE_CONSTEXPR 1
 #  else
 #    define FMT_USE_CONSTEXPR 0
@@ -106,9 +105,12 @@
 #  define FMT_CONSTEXPR
 #endif
 
-#if ((FMT_CPLUSPLUS >= 202002L) &&                            \
-     (!defined(_GLIBCXX_RELEASE) || _GLIBCXX_RELEASE > 9)) || \
-    (FMT_CPLUSPLUS >= 201709L && FMT_GCC_VERSION >= 1002)
+#if (FMT_CPLUSPLUS >= 202002L ||                                \
+     (FMT_CPLUSPLUS >= 201709L && FMT_GCC_VERSION >= 1002)) &&  \
+    ((!defined(_GLIBCXX_RELEASE) || _GLIBCXX_RELEASE >= 10) &&  \
+     (!defined(_LIBCPP_VERSION) || _LIBCPP_VERSION >= 10000) && \
+     (!FMT_MSC_VERSION || FMT_MSC_VERSION >= 1928)) &&          \
+    defined(__cpp_lib_is_constant_evaluated)
 #  define FMT_CONSTEXPR20 constexpr
 #else
 #  define FMT_CONSTEXPR20
@@ -140,22 +142,7 @@
 #  endif
 #endif
 
-#ifndef FMT_DEPRECATED
-#  if FMT_HAS_CPP14_ATTRIBUTE(deprecated) || FMT_MSC_VERSION >= 1900
-#    define FMT_DEPRECATED [[deprecated]]
-#  else
-#    if (defined(__GNUC__) && !defined(__LCC__)) || defined(__clang__)
-#      define FMT_DEPRECATED __attribute__((deprecated))
-#    elif FMT_MSC_VERSION
-#      define FMT_DEPRECATED __declspec(deprecated)
-#    else
-#      define FMT_DEPRECATED /* deprecated */
-#    endif
-#  endif
-#endif
-
-// [[noreturn]] is disabled on MSVC and NVCC because of bogus unreachable code
-// warnings.
+// Disable [[noreturn]] on MSVC/NVCC because of bogus unreachable code warnings.
 #if FMT_EXCEPTIONS && FMT_HAS_CPP_ATTRIBUTE(noreturn) && !FMT_MSC_VERSION && \
     !defined(__NVCC__)
 #  define FMT_NORETURN [[noreturn]]
@@ -163,17 +150,6 @@
 #  define FMT_NORETURN
 #endif
 
-#if FMT_HAS_CPP17_ATTRIBUTE(fallthrough)
-#  define FMT_FALLTHROUGH [[fallthrough]]
-#elif defined(__clang__)
-#  define FMT_FALLTHROUGH [[clang::fallthrough]]
-#elif FMT_GCC_VERSION >= 700 && \
-    (!defined(__EDG_VERSION__) || __EDG_VERSION__ >= 520)
-#  define FMT_FALLTHROUGH [[gnu::fallthrough]]
-#else
-#  define FMT_FALLTHROUGH
-#endif
-
 #ifndef FMT_NODISCARD
 #  if FMT_HAS_CPP17_ATTRIBUTE(nodiscard)
 #    define FMT_NODISCARD [[nodiscard]]
@@ -182,16 +158,6 @@
 #  endif
 #endif
 
-#ifndef FMT_USE_FLOAT
-#  define FMT_USE_FLOAT 1
-#endif
-#ifndef FMT_USE_DOUBLE
-#  define FMT_USE_DOUBLE 1
-#endif
-#ifndef FMT_USE_LONG_DOUBLE
-#  define FMT_USE_LONG_DOUBLE 1
-#endif
-
 #ifndef FMT_INLINE
 #  if FMT_GCC_VERSION || FMT_CLANG_VERSION
 #    define FMT_INLINE inline __attribute__((always_inline))
@@ -200,9 +166,6 @@
 #  endif
 #endif
 
-// An inline std::forward replacement.
-#define FMT_FORWARD(...) static_cast<decltype(__VA_ARGS__)&&>(__VA_ARGS__)
-
 #ifdef _MSC_VER
 #  define FMT_UNCHECKED_ITERATOR(It) \
     using _Unchecked_type = It  // Mark iterator as checked.
@@ -213,34 +176,32 @@
 #ifndef FMT_BEGIN_NAMESPACE
 #  define FMT_BEGIN_NAMESPACE \
     namespace fmt {           \
-    inline namespace v9 {
+    inline namespace v10 {
 #  define FMT_END_NAMESPACE \
     }                       \
     }
 #endif
 
-#ifndef FMT_MODULE_EXPORT
-#  define FMT_MODULE_EXPORT
-#  define FMT_MODULE_EXPORT_BEGIN
-#  define FMT_MODULE_EXPORT_END
-#  define FMT_BEGIN_DETAIL_NAMESPACE namespace detail {
-#  define FMT_END_DETAIL_NAMESPACE }
+#ifndef FMT_EXPORT
+#  define FMT_EXPORT
+#  define FMT_BEGIN_EXPORT
+#  define FMT_END_EXPORT
+#endif
+
+#if FMT_GCC_VERSION || FMT_CLANG_VERSION
+#  define FMT_VISIBILITY(value) __attribute__((visibility(value)))
+#else
+#  define FMT_VISIBILITY(value)
 #endif
 
 #if !defined(FMT_HEADER_ONLY) && defined(_WIN32)
-#  define FMT_CLASS_API FMT_MSC_WARNING(suppress : 4275)
-#  ifdef FMT_EXPORT
+#  if defined(FMT_LIB_EXPORT)
 #    define FMT_API __declspec(dllexport)
 #  elif defined(FMT_SHARED)
 #    define FMT_API __declspec(dllimport)
 #  endif
-#else
-#  define FMT_CLASS_API
-#  if defined(FMT_EXPORT) || defined(FMT_SHARED)
-#    if defined(__GNUC__) || defined(__clang__)
-#      define FMT_API __attribute__((visibility("default")))
-#    endif
-#  endif
+#elif defined(FMT_LIB_EXPORT) || defined(FMT_SHARED)
+#  define FMT_API FMT_VISIBILITY("default")
 #endif
 #ifndef FMT_API
 #  define FMT_API
@@ -261,11 +222,14 @@
 #endif
 
 #ifndef FMT_CONSTEVAL
-#  if ((FMT_GCC_VERSION >= 1000 || FMT_CLANG_VERSION >= 1101) &&         \
-       FMT_CPLUSPLUS >= 202002L && !defined(__apple_build_version__)) || \
-      (defined(__cpp_consteval) &&                                       \
-       (!FMT_MSC_VERSION || _MSC_FULL_VER >= 193030704))
-// consteval is broken in MSVC before VS2022 and Apple clang 13.
+#  if ((FMT_GCC_VERSION >= 1000 || FMT_CLANG_VERSION >= 1101) && \
+       (!defined(__apple_build_version__) ||                     \
+        __apple_build_version__ >= 14000029L) &&                 \
+       FMT_CPLUSPLUS >= 202002L) ||                              \
+      (defined(__cpp_consteval) &&                               \
+       (!FMT_MSC_VERSION || FMT_MSC_VERSION >= 1929))
+// consteval is broken in MSVC before VS2019 version 16.10 and Apple clang
+// before 14.
 #    define FMT_CONSTEVAL consteval
 #    define FMT_HAS_CONSTEVAL
 #  else
@@ -277,21 +241,30 @@
 #  if defined(__cpp_nontype_template_args) &&                  \
       ((FMT_GCC_VERSION >= 903 && FMT_CPLUSPLUS >= 201709L) || \
        __cpp_nontype_template_args >= 201911L) &&              \
-      !defined(__NVCOMPILER)
+      !defined(__NVCOMPILER) && !defined(__LCC__)
 #    define FMT_USE_NONTYPE_TEMPLATE_ARGS 1
 #  else
 #    define FMT_USE_NONTYPE_TEMPLATE_ARGS 0
 #  endif
 #endif
 
+// GCC < 5 requires this-> in decltype
+#ifndef FMT_DECLTYPE_THIS
+#  if FMT_GCC_VERSION && FMT_GCC_VERSION < 500
+#    define FMT_DECLTYPE_THIS this->
+#  else
+#    define FMT_DECLTYPE_THIS
+#  endif
+#endif
+
 // Enable minimal optimizations for more compact code in debug mode.
 FMT_GCC_PRAGMA("GCC push_options")
-#if !defined(__OPTIMIZE__) && !defined(__NVCOMPILER)
+#if !defined(__OPTIMIZE__) && !defined(__NVCOMPILER) && !defined(__LCC__) && \
+    !defined(__CUDACC__)
 FMT_GCC_PRAGMA("GCC optimize(\"Og\")")
 #endif
 
 FMT_BEGIN_NAMESPACE
-FMT_MODULE_EXPORT_BEGIN
 
 // Implementations of enable_if_t and other metafunctions for older systems.
 template <bool B, typename T = void>
@@ -305,22 +278,17 @@ template <typename T>
 using remove_const_t = typename std::remove_const<T>::type;
 template <typename T>
 using remove_cvref_t = typename std::remove_cv<remove_reference_t<T>>::type;
-template <typename T> struct type_identity { using type = T; };
+template <typename T> struct type_identity {
+  using type = T;
+};
 template <typename T> using type_identity_t = typename type_identity<T>::type;
 template <typename T>
 using underlying_t = typename std::underlying_type<T>::type;
 
-template <typename...> struct disjunction : std::false_type {};
-template <typename P> struct disjunction<P> : P {};
-template <typename P1, typename... Pn>
-struct disjunction<P1, Pn...>
-    : conditional_t<bool(P1::value), P1, disjunction<Pn...>> {};
-
-template <typename...> struct conjunction : std::true_type {};
-template <typename P> struct conjunction<P> : P {};
-template <typename P1, typename... Pn>
-struct conjunction<P1, Pn...>
-    : conditional_t<bool(P1::value), conjunction<Pn...>, P1> {};
+// Checks whether T is a container with contiguous storage.
+template <typename T> struct is_contiguous : std::false_type {};
+template <typename Char>
+struct is_contiguous<std::basic_string<Char>> : std::true_type {};
 
 struct monostate {
   constexpr monostate() {}
@@ -332,11 +300,19 @@ struct monostate {
 #ifdef FMT_DOC
 #  define FMT_ENABLE_IF(...)
 #else
-#  define FMT_ENABLE_IF(...) enable_if_t<(__VA_ARGS__), int> = 0
+#  define FMT_ENABLE_IF(...) fmt::enable_if_t<(__VA_ARGS__), int> = 0
 #endif
 
-FMT_BEGIN_DETAIL_NAMESPACE
+// This is defined in core.h instead of format.h to avoid injecting in std.
+// It is a template to avoid undesirable implicit conversions to std::byte.
+#ifdef __cpp_lib_byte
+template <typename T, FMT_ENABLE_IF(std::is_same<T, std::byte>::value)>
+inline auto format_as(T b) -> unsigned char {
+  return static_cast<unsigned char>(b);
+}
+#endif
 
+namespace detail {
 // Suppresses "unused variable" warnings with the method described in
 // https://herbsutter.com/2009/10/18/mailbag-shutting-up-compiler-warnings/.
 // (void)var does not work on many Intel compilers.
@@ -344,7 +320,15 @@ template <typename... T> FMT_CONSTEXPR void ignore_unused(const T&...) {}
 
 constexpr FMT_INLINE auto is_constant_evaluated(
     bool default_value = false) noexcept -> bool {
-#ifdef __cpp_lib_is_constant_evaluated
+// Workaround for incompatibility between libstdc++ consteval-based
+// std::is_constant_evaluated() implementation and clang-14.
+// https://github.com/fmtlib/fmt/issues/3247
+#if FMT_CPLUSPLUS >= 202002L && defined(_GLIBCXX_RELEASE) && \
+    _GLIBCXX_RELEASE >= 12 &&                                \
+    (FMT_CLANG_VERSION >= 1400 && FMT_CLANG_VERSION < 1500)
+  ignore_unused(default_value);
+  return __builtin_is_constant_evaluated();
+#elif defined(__cpp_lib_is_constant_evaluated)
   ignore_unused(default_value);
   return std::is_constant_evaluated();
 #else
@@ -364,12 +348,12 @@ FMT_NORETURN FMT_API void assert_fail(const char* file, int line,
 #  ifdef NDEBUG
 // FMT_ASSERT is not empty to avoid -Wempty-body.
 #    define FMT_ASSERT(condition, message) \
-      ::fmt::detail::ignore_unused((condition), (message))
+      fmt::detail::ignore_unused((condition), (message))
 #  else
 #    define FMT_ASSERT(condition, message)                                    \
       ((condition) /* void() fails with -Winvalid-constexpr on clang 4.0.1 */ \
            ? (void)0                                                          \
-           : ::fmt::detail::assert_fail(__FILE__, __LINE__, (message)))
+           : fmt::detail::assert_fail(__FILE__, __LINE__, (message)))
 #  endif
 #endif
 
@@ -410,15 +394,15 @@ FMT_CONSTEXPR auto to_unsigned(Int value) ->
   return static_cast<typename std::make_unsigned<Int>::type>(value);
 }
 
-FMT_MSC_WARNING(suppress : 4566) constexpr unsigned char micro[] = "\u00B5";
+FMT_CONSTEXPR inline auto is_utf8() -> bool {
+  FMT_MSC_WARNING(suppress : 4566) constexpr unsigned char section[] = "\u00A7";
 
-constexpr auto is_utf8() -> bool {
   // Avoid buggy sign extensions in MSVC's constant evaluation mode (#2297).
   using uchar = unsigned char;
-  return FMT_UNICODE || (sizeof(micro) == 3 && uchar(micro[0]) == 0xC2 &&
-                         uchar(micro[1]) == 0xB5);
+  return FMT_UNICODE || (sizeof(section) == 3 && uchar(section[0]) == 0xC2 &&
+                         uchar(section[1]) == 0xA7);
 }
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
 /**
   An implementation of ``std::basic_string_view`` for pre-C++17. It provides a
@@ -427,6 +411,7 @@ FMT_END_DETAIL_NAMESPACE
   compiled with a different ``-std`` option than the client code (which is not
   recommended).
  */
+FMT_EXPORT
 template <typename Char> class basic_string_view {
  private:
   const Char* data_;
@@ -486,6 +471,18 @@ template <typename Char> class basic_string_view {
     size_ -= n;
   }
 
+  FMT_CONSTEXPR_CHAR_TRAITS auto starts_with(
+      basic_string_view<Char> sv) const noexcept -> bool {
+    return size_ >= sv.size_ &&
+           std::char_traits<Char>::compare(data_, sv.data_, sv.size_) == 0;
+  }
+  FMT_CONSTEXPR_CHAR_TRAITS auto starts_with(Char c) const noexcept -> bool {
+    return size_ >= 1 && std::char_traits<Char>::eq(*data_, c);
+  }
+  FMT_CONSTEXPR_CHAR_TRAITS auto starts_with(const Char* s) const -> bool {
+    return starts_with(basic_string_view<Char>(s));
+  }
+
   // Lexicographically compare this string reference to other.
   FMT_CONSTEXPR_CHAR_TRAITS auto compare(basic_string_view other) const -> int {
     size_t str_size = size_ < other.size_ ? size_ : other.size_;
@@ -517,13 +514,15 @@ template <typename Char> class basic_string_view {
   }
 };
 
+FMT_EXPORT
 using string_view = basic_string_view<char>;
 
 /** Specifies if ``T`` is a character type. Can be specialized by users. */
+FMT_EXPORT
 template <typename T> struct is_char : std::false_type {};
 template <> struct is_char<char> : std::true_type {};
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 // A base class for compile-time strings.
 struct compile_string {};
@@ -531,7 +530,6 @@ struct compile_string {};
 template <typename S>
 struct is_compile_string : std::is_base_of<compile_string, S> {};
 
-// Returns a string view of `s`.
 template <typename Char, FMT_ENABLE_IF(is_char<Char>::value)>
 FMT_INLINE auto to_string_view(const Char* s) -> basic_string_view<Char> {
   return s;
@@ -561,10 +559,10 @@ void to_string_view(...);
 // Specifies whether S is a string type convertible to fmt::basic_string_view.
 // It should be a constexpr function but MSVC 2017 fails to compile it in
 // enable_if and MSVC 2015 fails to compile it as an alias template.
-// ADL invocation of to_string_view is DEPRECATED!
+// ADL is intentionally disabled as to_string_view is not an extension point.
 template <typename S>
-struct is_string : std::is_class<decltype(to_string_view(std::declval<S>()))> {
-};
+struct is_string
+    : std::is_class<decltype(detail::to_string_view(std::declval<S>()))> {};
 
 template <typename S, typename = void> struct char_t_impl {};
 template <typename S> struct char_t_impl<S, enable_if_t<is_string<S>::value>> {
@@ -619,26 +617,48 @@ FMT_TYPE_CONSTANT(const Char*, cstring_type);
 FMT_TYPE_CONSTANT(basic_string_view<Char>, string_type);
 FMT_TYPE_CONSTANT(const void*, pointer_type);
 
-constexpr bool is_integral_type(type t) {
+constexpr auto is_integral_type(type t) -> bool {
   return t > type::none_type && t <= type::last_integer_type;
 }
-
-constexpr bool is_arithmetic_type(type t) {
+constexpr auto is_arithmetic_type(type t) -> bool {
   return t > type::none_type && t <= type::last_numeric_type;
 }
 
+constexpr auto set(type rhs) -> int { return 1 << static_cast<int>(rhs); }
+constexpr auto in(type t, int set) -> bool {
+  return ((set >> static_cast<int>(t)) & 1) != 0;
+}
+
+// Bitsets of types.
+enum {
+  sint_set =
+      set(type::int_type) | set(type::long_long_type) | set(type::int128_type),
+  uint_set = set(type::uint_type) | set(type::ulong_long_type) |
+             set(type::uint128_type),
+  bool_set = set(type::bool_type),
+  char_set = set(type::char_type),
+  float_set = set(type::float_type) | set(type::double_type) |
+              set(type::long_double_type),
+  string_set = set(type::string_type),
+  cstring_set = set(type::cstring_type),
+  pointer_set = set(type::pointer_type)
+};
+
+// DEPRECATED!
 FMT_NORETURN FMT_API void throw_format_error(const char* message);
 
 struct error_handler {
   constexpr error_handler() = default;
-  constexpr error_handler(const error_handler&) = default;
 
   // This function is intentionally not constexpr to give a compile-time error.
   FMT_NORETURN void on_error(const char* message) {
     throw_format_error(message);
   }
 };
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
+
+/** Throws ``format_error`` with a given message. */
+using detail::throw_format_error;
 
 /** String's character type. */
 template <typename S> using char_t = typename detail::char_t_impl<S>::type;
@@ -650,8 +670,8 @@ template <typename S> using char_t = typename detail::char_t_impl<S>::type;
   You can use the ``format_parse_context`` type alias for ``char`` instead.
   \endrst
  */
-template <typename Char, typename ErrorHandler = detail::error_handler>
-class basic_format_parse_context : private ErrorHandler {
+FMT_EXPORT
+template <typename Char> class basic_format_parse_context {
  private:
   basic_string_view<Char> format_str_;
   int next_arg_id_;
@@ -660,12 +680,11 @@ class basic_format_parse_context : private ErrorHandler {
 
  public:
   using char_type = Char;
-  using iterator = typename basic_string_view<Char>::iterator;
+  using iterator = const Char*;
 
   explicit constexpr basic_format_parse_context(
-      basic_string_view<Char> format_str, ErrorHandler eh = {},
-      int next_arg_id = 0)
-      : ErrorHandler(eh), format_str_(format_str), next_arg_id_(next_arg_id) {}
+      basic_string_view<Char> format_str, int next_arg_id = 0)
+      : format_str_(format_str), next_arg_id_(next_arg_id) {}
 
   /**
     Returns an iterator to the beginning of the format string range being
@@ -691,7 +710,8 @@ class basic_format_parse_context : private ErrorHandler {
    */
   FMT_CONSTEXPR auto next_arg_id() -> int {
     if (next_arg_id_ < 0) {
-      on_error("cannot switch from manual to automatic argument indexing");
+      detail::throw_format_error(
+          "cannot switch from manual to automatic argument indexing");
       return 0;
     }
     int id = next_arg_id_++;
@@ -705,7 +725,8 @@ class basic_format_parse_context : private ErrorHandler {
    */
   FMT_CONSTEXPR void check_arg_id(int id) {
     if (next_arg_id_ > 0) {
-      on_error("cannot switch from automatic to manual argument indexing");
+      detail::throw_format_error(
+          "cannot switch from automatic to manual argument indexing");
       return;
     }
     next_arg_id_ = -1;
@@ -713,117 +734,50 @@ class basic_format_parse_context : private ErrorHandler {
   }
   FMT_CONSTEXPR void check_arg_id(basic_string_view<Char>) {}
   FMT_CONSTEXPR void check_dynamic_spec(int arg_id);
-
-  FMT_CONSTEXPR void on_error(const char* message) {
-    ErrorHandler::on_error(message);
-  }
-
-  constexpr auto error_handler() const -> ErrorHandler { return *this; }
 };
 
+FMT_EXPORT
 using format_parse_context = basic_format_parse_context<char>;
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 // A parse context with extra data used only in compile-time checks.
-template <typename Char, typename ErrorHandler = detail::error_handler>
-class compile_parse_context
-    : public basic_format_parse_context<Char, ErrorHandler> {
+template <typename Char>
+class compile_parse_context : public basic_format_parse_context<Char> {
  private:
   int num_args_;
   const type* types_;
-  using base = basic_format_parse_context<Char, ErrorHandler>;
+  using base = basic_format_parse_context<Char>;
 
  public:
   explicit FMT_CONSTEXPR compile_parse_context(
       basic_string_view<Char> format_str, int num_args, const type* types,
-      ErrorHandler eh = {}, int next_arg_id = 0)
-      : base(format_str, eh, next_arg_id), num_args_(num_args), types_(types) {}
+      int next_arg_id = 0)
+      : base(format_str, next_arg_id), num_args_(num_args), types_(types) {}
 
   constexpr auto num_args() const -> int { return num_args_; }
   constexpr auto arg_type(int id) const -> type { return types_[id]; }
 
   FMT_CONSTEXPR auto next_arg_id() -> int {
     int id = base::next_arg_id();
-    if (id >= num_args_) this->on_error("argument not found");
+    if (id >= num_args_) throw_format_error("argument not found");
     return id;
   }
 
   FMT_CONSTEXPR void check_arg_id(int id) {
     base::check_arg_id(id);
-    if (id >= num_args_) this->on_error("argument not found");
+    if (id >= num_args_) throw_format_error("argument not found");
   }
   using base::check_arg_id;
 
   FMT_CONSTEXPR void check_dynamic_spec(int arg_id) {
+    detail::ignore_unused(arg_id);
+#if !defined(__LCC__)
     if (arg_id < num_args_ && types_ && !is_integral_type(types_[arg_id]))
-      this->on_error("width/precision is not integer");
-  }
-};
-FMT_END_DETAIL_NAMESPACE
-
-template <typename Char, typename ErrorHandler>
-FMT_CONSTEXPR void
-basic_format_parse_context<Char, ErrorHandler>::do_check_arg_id(int id) {
-  // Argument id is only checked at compile-time during parsing because
-  // formatting has its own validation.
-  if (detail::is_constant_evaluated() && FMT_GCC_VERSION >= 1200) {
-    using context = detail::compile_parse_context<Char, ErrorHandler>;
-    if (id >= static_cast<context*>(this)->num_args())
-      on_error("argument not found");
-  }
-}
-
-template <typename Char, typename ErrorHandler>
-FMT_CONSTEXPR void
-basic_format_parse_context<Char, ErrorHandler>::check_dynamic_spec(int arg_id) {
-  if (detail::is_constant_evaluated()) {
-    using context = detail::compile_parse_context<Char, ErrorHandler>;
-    static_cast<context*>(this)->check_dynamic_spec(arg_id);
+      throw_format_error("width/precision is not integer");
+#endif
   }
-}
-
-template <typename Context> class basic_format_arg;
-template <typename Context> class basic_format_args;
-template <typename Context> class dynamic_format_arg_store;
-
-// A formatter for objects of type T.
-template <typename T, typename Char = char, typename Enable = void>
-struct formatter {
-  // A deleted default constructor indicates a disabled formatter.
-  formatter() = delete;
 };
 
-// Specifies if T has an enabled formatter specialization. A type can be
-// formattable even if it doesn't have a formatter e.g. via a conversion.
-template <typename T, typename Context>
-using has_formatter =
-    std::is_constructible<typename Context::template formatter_type<T>>;
-
-// Checks whether T is a container with contiguous storage.
-template <typename T> struct is_contiguous : std::false_type {};
-template <typename Char>
-struct is_contiguous<std::basic_string<Char>> : std::true_type {};
-
-class appender;
-
-FMT_BEGIN_DETAIL_NAMESPACE
-
-template <typename Context, typename T>
-constexpr auto has_const_formatter_impl(T*)
-    -> decltype(typename Context::template formatter_type<T>().format(
-                    std::declval<const T&>(), std::declval<Context&>()),
-                true) {
-  return true;
-}
-template <typename Context>
-constexpr auto has_const_formatter_impl(...) -> bool {
-  return false;
-}
-template <typename T, typename Context>
-constexpr auto has_const_formatter() -> bool {
-  return has_const_formatter_impl<Context>(static_cast<T*>(nullptr));
-}
-
 // Extracts a reference to the container from back_insert_iterator.
 template <typename Container>
 inline auto get_container(std::back_insert_iterator<Container> it)
@@ -849,7 +803,7 @@ template <typename Char, typename T, typename U,
 FMT_CONSTEXPR auto copy_str(T* begin, T* end, U* out) -> U* {
   if (is_constant_evaluated()) return copy_str<Char, T*, U*>(begin, end, out);
   auto size = to_unsigned(end - begin);
-  memcpy(out, begin, size * sizeof(U));
+  if (size > 0) memcpy(out, begin, size * sizeof(U));
   return out + size;
 }
 
@@ -868,7 +822,7 @@ template <typename T> class buffer {
  protected:
   // Don't initialize ptr_ since it is not accessed to save a few cycles.
   FMT_MSC_WARNING(suppress : 26495)
-  buffer(size_t sz) noexcept : size_(sz), capacity_(sz) {}
+  FMT_CONSTEXPR buffer(size_t sz) noexcept : size_(sz), capacity_(sz) {}
 
   FMT_CONSTEXPR20 buffer(T* p = nullptr, size_t sz = 0, size_t cap = 0) noexcept
       : ptr_(p), size_(sz), capacity_(cap) {}
@@ -883,6 +837,7 @@ template <typename T> class buffer {
   }
 
   /** Increases the buffer capacity to hold at least *capacity* elements. */
+  // DEPRECATED!
   virtual FMT_CONSTEXPR20 void grow(size_t capacity) = 0;
 
  public:
@@ -892,11 +847,11 @@ template <typename T> class buffer {
   buffer(const buffer&) = delete;
   void operator=(const buffer&) = delete;
 
-  auto begin() noexcept -> T* { return ptr_; }
-  auto end() noexcept -> T* { return ptr_ + size_; }
+  FMT_INLINE auto begin() noexcept -> T* { return ptr_; }
+  FMT_INLINE auto end() noexcept -> T* { return ptr_ + size_; }
 
-  auto begin() const noexcept -> const T* { return ptr_; }
-  auto end() const noexcept -> const T* { return ptr_ + size_; }
+  FMT_INLINE auto begin() const noexcept -> const T* { return ptr_; }
+  FMT_INLINE auto end() const noexcept -> const T* { return ptr_ + size_; }
 
   /** Returns the size of this buffer. */
   constexpr auto size() const noexcept -> size_t { return size_; }
@@ -904,10 +859,8 @@ template <typename T> class buffer {
   /** Returns the capacity of this buffer. */
   constexpr auto capacity() const noexcept -> size_t { return capacity_; }
 
-  /** Returns a pointer to the buffer data. */
+  /** Returns a pointer to the buffer data (not null-terminated). */
   FMT_CONSTEXPR auto data() noexcept -> T* { return ptr_; }
-
-  /** Returns a pointer to the buffer data. */
   FMT_CONSTEXPR auto data() const noexcept -> const T* { return ptr_; }
 
   /** Clears this buffer. */
@@ -1100,6 +1053,79 @@ template <typename T = char> class counting_buffer final : public buffer<T> {
 
   auto count() -> size_t { return count_ + this->size(); }
 };
+}  // namespace detail
+
+template <typename Char>
+FMT_CONSTEXPR void basic_format_parse_context<Char>::do_check_arg_id(int id) {
+  // Argument id is only checked at compile-time during parsing because
+  // formatting has its own validation.
+  if (detail::is_constant_evaluated() &&
+      (!FMT_GCC_VERSION || FMT_GCC_VERSION >= 1200)) {
+    using context = detail::compile_parse_context<Char>;
+    if (id >= static_cast<context*>(this)->num_args())
+      detail::throw_format_error("argument not found");
+  }
+}
+
+template <typename Char>
+FMT_CONSTEXPR void basic_format_parse_context<Char>::check_dynamic_spec(
+    int arg_id) {
+  if (detail::is_constant_evaluated() &&
+      (!FMT_GCC_VERSION || FMT_GCC_VERSION >= 1200)) {
+    using context = detail::compile_parse_context<Char>;
+    static_cast<context*>(this)->check_dynamic_spec(arg_id);
+  }
+}
+
+FMT_EXPORT template <typename Context> class basic_format_arg;
+FMT_EXPORT template <typename Context> class basic_format_args;
+FMT_EXPORT template <typename Context> class dynamic_format_arg_store;
+
+// A formatter for objects of type T.
+FMT_EXPORT
+template <typename T, typename Char = char, typename Enable = void>
+struct formatter {
+  // A deleted default constructor indicates a disabled formatter.
+  formatter() = delete;
+};
+
+// Specifies if T has an enabled formatter specialization. A type can be
+// formattable even if it doesn't have a formatter e.g. via a conversion.
+template <typename T, typename Context>
+using has_formatter =
+    std::is_constructible<typename Context::template formatter_type<T>>;
+
+// An output iterator that appends to a buffer.
+// It is used to reduce symbol sizes for the common case.
+class appender : public std::back_insert_iterator<detail::buffer<char>> {
+  using base = std::back_insert_iterator<detail::buffer<char>>;
+
+ public:
+  using std::back_insert_iterator<detail::buffer<char>>::back_insert_iterator;
+  appender(base it) noexcept : base(it) {}
+  FMT_UNCHECKED_ITERATOR(appender);
+
+  auto operator++() noexcept -> appender& { return *this; }
+  auto operator++(int) noexcept -> appender { return *this; }
+};
+
+namespace detail {
+
+template <typename Context, typename T>
+constexpr auto has_const_formatter_impl(T*)
+    -> decltype(typename Context::template formatter_type<T>().format(
+                    std::declval<const T&>(), std::declval<Context&>()),
+                true) {
+  return true;
+}
+template <typename Context>
+constexpr auto has_const_formatter_impl(...) -> bool {
+  return false;
+}
+template <typename T, typename Context>
+constexpr auto has_const_formatter() -> bool {
+  return has_const_formatter_impl<Context>(static_cast<T*>(nullptr));
+}
 
 template <typename T>
 using buffer_appender = conditional_t<std::is_same<T, char>::value, appender,
@@ -1110,29 +1136,21 @@ template <typename T, typename OutputIt>
 auto get_buffer(OutputIt out) -> iterator_buffer<OutputIt, T> {
   return iterator_buffer<OutputIt, T>(out);
 }
+template <typename T, typename Buf,
+          FMT_ENABLE_IF(std::is_base_of<buffer<char>, Buf>::value)>
+auto get_buffer(std::back_insert_iterator<Buf> out) -> buffer<char>& {
+  return get_container(out);
+}
 
-template <typename Buffer>
-auto get_iterator(Buffer& buf) -> decltype(buf.out()) {
+template <typename Buf, typename OutputIt>
+FMT_INLINE auto get_iterator(Buf& buf, OutputIt) -> decltype(buf.out()) {
   return buf.out();
 }
-template <typename T> auto get_iterator(buffer<T>& buf) -> buffer_appender<T> {
-  return buffer_appender<T>(buf);
+template <typename T, typename OutputIt>
+auto get_iterator(buffer<T>&, OutputIt out) -> OutputIt {
+  return out;
 }
 
-template <typename T, typename Char = char, typename Enable = void>
-struct fallback_formatter {
-  fallback_formatter() = delete;
-};
-
-// Specifies if T has an enabled fallback_formatter specialization.
-template <typename T, typename Char>
-using has_fallback_formatter =
-#ifdef FMT_DEPRECATED_OSTREAM
-    std::is_constructible<fallback_formatter<T, Char>>;
-#else
-    std::false_type;
-#endif
-
 struct view {};
 
 template <typename Char, typename T> struct named_arg : view {
@@ -1217,7 +1235,6 @@ constexpr auto count_statically_named_args() -> size_t {
 
 struct unformattable {};
 struct unformattable_char : unformattable {};
-struct unformattable_const : unformattable {};
 struct unformattable_pointer : unformattable {};
 
 template <typename Char> struct string_value {
@@ -1284,21 +1301,17 @@ template <typename Context> class value {
   FMT_INLINE value(const named_arg_info<char_type>* args, size_t size)
       : named_args{args, size} {}
 
-  template <typename T> FMT_CONSTEXPR FMT_INLINE value(T& val) {
-    using value_type = remove_cvref_t<T>;
-    custom.value = const_cast<value_type*>(&val);
+  template <typename T> FMT_CONSTEXPR20 FMT_INLINE value(T& val) {
+    using value_type = remove_const_t<T>;
+    custom.value = const_cast<value_type*>(std::addressof(val));
     // Get the formatter type through the context to allow different contexts
     // have different extension points, e.g. `formatter<T>` for `format` and
     // `printf_formatter<T>` for `printf`.
     custom.format = format_custom_arg<
-        value_type,
-        conditional_t<has_formatter<value_type, Context>::value,
-                      typename Context::template formatter_type<value_type>,
-                      fallback_formatter<value_type, char_type>>>;
+        value_type, typename Context::template formatter_type<value_type>>;
   }
   value(unformattable);
   value(unformattable_char);
-  value(unformattable_const);
   value(unformattable_pointer);
 
  private:
@@ -1311,33 +1324,30 @@ template <typename Context> class value {
     parse_ctx.advance_to(f.parse(parse_ctx));
     using qualified_type =
         conditional_t<has_const_formatter<T, Context>(), const T, T>;
+    // Calling format through a mutable reference is deprecated.
     ctx.advance_to(f.format(*static_cast<qualified_type*>(arg), ctx));
   }
 };
 
-template <typename Context, typename T>
-FMT_CONSTEXPR auto make_arg(T&& value) -> basic_format_arg<Context>;
-
 // To minimize the number of types we need to deal with, long is translated
 // either to int or to long long depending on its size.
 enum { long_short = sizeof(long) == sizeof(int) };
 using long_type = conditional_t<long_short, int, long long>;
 using ulong_type = conditional_t<long_short, unsigned, unsigned long long>;
 
-#ifdef __cpp_lib_byte
-inline auto format_as(std::byte b) -> unsigned char {
-  return static_cast<unsigned char>(b);
-}
-#endif
-
-template <typename T> struct has_format_as {
-  template <typename U, typename V = decltype(format_as(U())),
-            FMT_ENABLE_IF(std::is_enum<U>::value&& std::is_integral<V>::value)>
-  static auto check(U*) -> std::true_type;
-  static auto check(...) -> std::false_type;
+template <typename T> struct format_as_result {
+  template <typename U,
+            FMT_ENABLE_IF(std::is_enum<U>::value || std::is_class<U>::value)>
+  static auto map(U*) -> remove_cvref_t<decltype(format_as(std::declval<U>()))>;
+  static auto map(...) -> void;
 
-  enum { value = decltype(check(static_cast<T*>(nullptr)))::value };
+  using type = decltype(map(static_cast<T*>(nullptr)));
 };
+template <typename T> using format_as_t = typename format_as_result<T>::type;
+
+template <typename T>
+struct has_format_as
+    : bool_constant<!std::is_same<format_as_t<T>, void>::value> {};
 
 // Maps formatting arguments to core types.
 // arg_mapper reports errors by returning unformattable instead of using
@@ -1414,25 +1424,6 @@ template <typename Context> struct arg_mapper {
   FMT_CONSTEXPR FMT_INLINE auto map(const T&) -> unformattable_char {
     return {};
   }
-  template <typename T,
-            FMT_ENABLE_IF(
-                std::is_convertible<T, basic_string_view<char_type>>::value &&
-                !is_string<T>::value && !has_formatter<T, Context>::value &&
-                !has_fallback_formatter<T, char_type>::value)>
-  FMT_CONSTEXPR FMT_INLINE auto map(const T& val)
-      -> basic_string_view<char_type> {
-    return basic_string_view<char_type>(val);
-  }
-  template <typename T,
-            FMT_ENABLE_IF(
-                std::is_convertible<T, std_string_view<char_type>>::value &&
-                !std::is_convertible<T, basic_string_view<char_type>>::value &&
-                !is_string<T>::value && !has_formatter<T, Context>::value &&
-                !has_fallback_formatter<T, char_type>::value)>
-  FMT_CONSTEXPR FMT_INLINE auto map(const T& val)
-      -> basic_string_view<char_type> {
-    return std_string_view<char_type>(val);
-  }
 
   FMT_CONSTEXPR FMT_INLINE auto map(void* val) -> const void* { return val; }
   FMT_CONSTEXPR FMT_INLINE auto map(const void* val) -> const void* {
@@ -1442,16 +1433,15 @@ template <typename Context> struct arg_mapper {
     return val;
   }
 
-  // We use SFINAE instead of a const T* parameter to avoid conflicting with
-  // the C array overload.
+  // Use SFINAE instead of a const T* parameter to avoid a conflict with the
+  // array overload.
   template <
       typename T,
       FMT_ENABLE_IF(
           std::is_pointer<T>::value || std::is_member_pointer<T>::value ||
           std::is_function<typename std::remove_pointer<T>::type>::value ||
-          (std::is_convertible<const T&, const void*>::value &&
-           !std::is_convertible<const T&, const char_type*>::value &&
-           !has_formatter<T, Context>::value))>
+          (std::is_array<T>::value &&
+           !std::is_convertible<T, const char_type*>::value))>
   FMT_CONSTEXPR auto map(const T&) -> unformattable_pointer {
     return {};
   }
@@ -1462,62 +1452,42 @@ template <typename Context> struct arg_mapper {
     return values;
   }
 
-  template <typename T,
-            FMT_ENABLE_IF(
-                std::is_enum<T>::value&& std::is_convertible<T, int>::value &&
-                !has_format_as<T>::value && !has_formatter<T, Context>::value &&
-                !has_fallback_formatter<T, char_type>::value)>
-  FMT_CONSTEXPR FMT_INLINE auto map(const T& val)
-      -> decltype(std::declval<arg_mapper>().map(
-          static_cast<underlying_t<T>>(val))) {
-    return map(static_cast<underlying_t<T>>(val));
-  }
-
-  template <typename T, FMT_ENABLE_IF(has_format_as<T>::value &&
-                                      !has_formatter<T, Context>::value)>
+  // Only map owning types because mapping views can be unsafe.
+  template <typename T, typename U = format_as_t<T>,
+            FMT_ENABLE_IF(std::is_arithmetic<U>::value)>
   FMT_CONSTEXPR FMT_INLINE auto map(const T& val)
-      -> decltype(std::declval<arg_mapper>().map(format_as(T()))) {
+      -> decltype(FMT_DECLTYPE_THIS map(U())) {
     return map(format_as(val));
   }
 
-  template <typename T, typename U = remove_cvref_t<T>>
-  struct formattable
-      : bool_constant<has_const_formatter<U, Context>() ||
-                      !std::is_const<remove_reference_t<T>>::value ||
-                      has_fallback_formatter<U, char_type>::value> {};
+  template <typename T, typename U = remove_const_t<T>>
+  struct formattable : bool_constant<has_const_formatter<U, Context>() ||
+                                     (has_formatter<U, Context>::value &&
+                                      !std::is_const<T>::value)> {};
 
-#if (FMT_MSC_VERSION != 0 && FMT_MSC_VERSION < 1910) || \
-    FMT_ICC_VERSION != 0 || defined(__NVCC__)
-  // Workaround a bug in MSVC and Intel (Issue 2746).
-  template <typename T> FMT_CONSTEXPR FMT_INLINE auto do_map(T&& val) -> T& {
-    return val;
-  }
-#else
   template <typename T, FMT_ENABLE_IF(formattable<T>::value)>
-  FMT_CONSTEXPR FMT_INLINE auto do_map(T&& val) -> T& {
+  FMT_CONSTEXPR FMT_INLINE auto do_map(T& val) -> T& {
     return val;
   }
   template <typename T, FMT_ENABLE_IF(!formattable<T>::value)>
-  FMT_CONSTEXPR FMT_INLINE auto do_map(T&&) -> unformattable_const {
+  FMT_CONSTEXPR FMT_INLINE auto do_map(T&) -> unformattable {
     return {};
   }
-#endif
 
-  template <typename T, typename U = remove_cvref_t<T>,
-            FMT_ENABLE_IF(!is_string<U>::value && !is_char<U>::value &&
-                          !std::is_array<U>::value &&
-                          !std::is_pointer<U>::value &&
-                          !has_format_as<U>::value &&
-                          (has_formatter<U, Context>::value ||
-                           has_fallback_formatter<U, char_type>::value))>
-  FMT_CONSTEXPR FMT_INLINE auto map(T&& val)
-      -> decltype(this->do_map(std::forward<T>(val))) {
-    return do_map(std::forward<T>(val));
+  template <typename T, typename U = remove_const_t<T>,
+            FMT_ENABLE_IF((std::is_class<U>::value || std::is_enum<U>::value ||
+                           std::is_union<U>::value) &&
+                          !is_string<U>::value && !is_char<U>::value &&
+                          !is_named_arg<U>::value &&
+                          !std::is_arithmetic<format_as_t<U>>::value)>
+  FMT_CONSTEXPR FMT_INLINE auto map(T& val)
+      -> decltype(FMT_DECLTYPE_THIS do_map(val)) {
+    return do_map(val);
   }
 
   template <typename T, FMT_ENABLE_IF(is_named_arg<T>::value)>
   FMT_CONSTEXPR FMT_INLINE auto map(const T& named_arg)
-      -> decltype(std::declval<arg_mapper>().map(named_arg.value)) {
+      -> decltype(FMT_DECLTYPE_THIS map(named_arg.value)) {
     return map(named_arg.value);
   }
 
@@ -1536,36 +1506,131 @@ enum { max_packed_args = 62 / packed_arg_bits };
 enum : unsigned long long { is_unpacked_bit = 1ULL << 63 };
 enum : unsigned long long { has_named_args_bit = 1ULL << 62 };
 
-FMT_END_DETAIL_NAMESPACE
+template <typename Char, typename InputIt>
+auto copy_str(InputIt begin, InputIt end, appender out) -> appender {
+  get_container(out).append(begin, end);
+  return out;
+}
+template <typename Char, typename InputIt>
+auto copy_str(InputIt begin, InputIt end,
+              std::back_insert_iterator<std::string> out)
+    -> std::back_insert_iterator<std::string> {
+  get_container(out).append(begin, end);
+  return out;
+}
 
-// An output iterator that appends to a buffer.
-// It is used to reduce symbol sizes for the common case.
-class appender : public std::back_insert_iterator<detail::buffer<char>> {
-  using base = std::back_insert_iterator<detail::buffer<char>>;
+template <typename Char, typename R, typename OutputIt>
+FMT_CONSTEXPR auto copy_str(R&& rng, OutputIt out) -> OutputIt {
+  return detail::copy_str<Char>(rng.begin(), rng.end(), out);
+}
 
-  template <typename T>
-  friend auto get_buffer(appender out) -> detail::buffer<char>& {
-    return detail::get_container(out);
-  }
+#if FMT_GCC_VERSION && FMT_GCC_VERSION < 500
+// A workaround for gcc 4.8 to make void_t work in a SFINAE context.
+template <typename...> struct void_t_impl {
+  using type = void;
+};
+template <typename... T> using void_t = typename void_t_impl<T...>::type;
+#else
+template <typename...> using void_t = void;
+#endif
+
+template <typename It, typename T, typename Enable = void>
+struct is_output_iterator : std::false_type {};
+
+template <typename It, typename T>
+struct is_output_iterator<
+    It, T,
+    void_t<typename std::iterator_traits<It>::iterator_category,
+           decltype(*std::declval<It>() = std::declval<T>())>>
+    : std::true_type {};
+
+template <typename It> struct is_back_insert_iterator : std::false_type {};
+template <typename Container>
+struct is_back_insert_iterator<std::back_insert_iterator<Container>>
+    : std::true_type {};
+
+// A type-erased reference to an std::locale to avoid a heavy <locale> include.
+class locale_ref {
+ private:
+  const void* locale_;  // A type-erased pointer to std::locale.
 
  public:
-  using std::back_insert_iterator<detail::buffer<char>>::back_insert_iterator;
-  appender(base it) noexcept : base(it) {}
-  FMT_UNCHECKED_ITERATOR(appender);
+  constexpr FMT_INLINE locale_ref() : locale_(nullptr) {}
+  template <typename Locale> explicit locale_ref(const Locale& loc);
 
-  auto operator++() noexcept -> appender& { return *this; }
-  auto operator++(int) noexcept -> appender { return *this; }
+  explicit operator bool() const noexcept { return locale_ != nullptr; }
+
+  template <typename Locale> auto get() const -> Locale;
 };
 
-// A formatting argument. It is a trivially copyable/constructible type to
-// allow storage in basic_memory_buffer.
+template <typename> constexpr auto encode_types() -> unsigned long long {
+  return 0;
+}
+
+template <typename Context, typename Arg, typename... Args>
+constexpr auto encode_types() -> unsigned long long {
+  return static_cast<unsigned>(mapped_type_constant<Arg, Context>::value) |
+         (encode_types<Context, Args...>() << packed_arg_bits);
+}
+
+#if defined(__cpp_if_constexpr)
+// This type is intentionally undefined, only used for errors
+template <typename T, typename Char> struct type_is_unformattable_for;
+#endif
+
+template <bool PACKED, typename Context, typename T, FMT_ENABLE_IF(PACKED)>
+FMT_CONSTEXPR FMT_INLINE auto make_arg(T& val) -> value<Context> {
+  using arg_type = remove_cvref_t<decltype(arg_mapper<Context>().map(val))>;
+
+  constexpr bool formattable_char =
+      !std::is_same<arg_type, unformattable_char>::value;
+  static_assert(formattable_char, "Mixing character types is disallowed.");
+
+  // Formatting of arbitrary pointers is disallowed. If you want to format a
+  // pointer cast it to `void*` or `const void*`. In particular, this forbids
+  // formatting of `[const] volatile char*` printed as bool by iostreams.
+  constexpr bool formattable_pointer =
+      !std::is_same<arg_type, unformattable_pointer>::value;
+  static_assert(formattable_pointer,
+                "Formatting of non-void pointers is disallowed.");
+
+  constexpr bool formattable = !std::is_same<arg_type, unformattable>::value;
+#if defined(__cpp_if_constexpr)
+  if constexpr (!formattable) {
+    type_is_unformattable_for<T, typename Context::char_type> _;
+  }
+#endif
+  static_assert(
+      formattable,
+      "Cannot format an argument. To make type T formattable provide a "
+      "formatter<T> specialization: https://fmt.dev/latest/api.html#udt");
+  return {arg_mapper<Context>().map(val)};
+}
+
+template <typename Context, typename T>
+FMT_CONSTEXPR auto make_arg(T& val) -> basic_format_arg<Context> {
+  auto arg = basic_format_arg<Context>();
+  arg.type_ = mapped_type_constant<T, Context>::value;
+  arg.value_ = make_arg<true, Context>(val);
+  return arg;
+}
+
+template <bool PACKED, typename Context, typename T, FMT_ENABLE_IF(!PACKED)>
+FMT_CONSTEXPR inline auto make_arg(T& val) -> basic_format_arg<Context> {
+  return make_arg<Context>(val);
+}
+}  // namespace detail
+FMT_BEGIN_EXPORT
+
+// A formatting argument. Context is a template parameter for the compiled API
+// where output can be unbuffered.
 template <typename Context> class basic_format_arg {
  private:
   detail::value<Context> value_;
   detail::type type_;
 
   template <typename ContextType, typename T>
-  friend FMT_CONSTEXPR auto detail::make_arg(T&& value)
+  friend FMT_CONSTEXPR auto detail::make_arg(T& value)
       -> basic_format_arg<ContextType>;
 
   template <typename Visitor, typename Ctx>
@@ -1610,6 +1675,15 @@ template <typename Context> class basic_format_arg {
   auto is_arithmetic() const -> bool {
     return detail::is_arithmetic_type(type_);
   }
+
+  FMT_INLINE auto format_custom(const char_type* parse_begin,
+                                typename Context::parse_context_type& parse_ctx,
+                                Context& ctx) -> bool {
+    if (type_ != detail::type::custom_type) return false;
+    parse_ctx.advance_to(parse_begin);
+    value_.custom.format(value_.custom.value, parse_ctx, ctx);
+    return true;
+  }
 };
 
 /**
@@ -1619,6 +1693,7 @@ template <typename Context> class basic_format_arg {
   ``vis(value)`` will be called with the value of type ``double``.
   \endrst
  */
+// DEPRECATED!
 template <typename Visitor, typename Context>
 FMT_CONSTEXPR FMT_INLINE auto visit_format_arg(
     Visitor&& vis, const basic_format_arg<Context>& arg) -> decltype(vis(0)) {
@@ -1660,136 +1735,8 @@ FMT_CONSTEXPR FMT_INLINE auto visit_format_arg(
   return vis(monostate());
 }
 
-FMT_BEGIN_DETAIL_NAMESPACE
-
-template <typename Char, typename InputIt>
-auto copy_str(InputIt begin, InputIt end, appender out) -> appender {
-  get_container(out).append(begin, end);
-  return out;
-}
-
-template <typename Char, typename R, typename OutputIt>
-FMT_CONSTEXPR auto copy_str(R&& rng, OutputIt out) -> OutputIt {
-  return detail::copy_str<Char>(rng.begin(), rng.end(), out);
-}
-
-#if FMT_GCC_VERSION && FMT_GCC_VERSION < 500
-// A workaround for gcc 4.8 to make void_t work in a SFINAE context.
-template <typename... Ts> struct void_t_impl { using type = void; };
-template <typename... Ts>
-using void_t = typename detail::void_t_impl<Ts...>::type;
-#else
-template <typename...> using void_t = void;
-#endif
-
-template <typename It, typename T, typename Enable = void>
-struct is_output_iterator : std::false_type {};
-
-template <typename It, typename T>
-struct is_output_iterator<
-    It, T,
-    void_t<typename std::iterator_traits<It>::iterator_category,
-           decltype(*std::declval<It>() = std::declval<T>())>>
-    : std::true_type {};
-
-template <typename OutputIt>
-struct is_back_insert_iterator : std::false_type {};
-template <typename Container>
-struct is_back_insert_iterator<std::back_insert_iterator<Container>>
-    : std::true_type {};
-
-template <typename OutputIt>
-struct is_contiguous_back_insert_iterator : std::false_type {};
-template <typename Container>
-struct is_contiguous_back_insert_iterator<std::back_insert_iterator<Container>>
-    : is_contiguous<Container> {};
-template <>
-struct is_contiguous_back_insert_iterator<appender> : std::true_type {};
-
-// A type-erased reference to an std::locale to avoid a heavy <locale> include.
-class locale_ref {
- private:
-  const void* locale_;  // A type-erased pointer to std::locale.
-
- public:
-  constexpr locale_ref() : locale_(nullptr) {}
-  template <typename Locale> explicit locale_ref(const Locale& loc);
-
-  explicit operator bool() const noexcept { return locale_ != nullptr; }
-
-  template <typename Locale> auto get() const -> Locale;
-};
-
-template <typename> constexpr auto encode_types() -> unsigned long long {
-  return 0;
-}
-
-template <typename Context, typename Arg, typename... Args>
-constexpr auto encode_types() -> unsigned long long {
-  return static_cast<unsigned>(mapped_type_constant<Arg, Context>::value) |
-         (encode_types<Context, Args...>() << packed_arg_bits);
-}
-
-template <typename Context, typename T>
-FMT_CONSTEXPR FMT_INLINE auto make_value(T&& val) -> value<Context> {
-  const auto& arg = arg_mapper<Context>().map(FMT_FORWARD(val));
-
-  constexpr bool formattable_char =
-      !std::is_same<decltype(arg), const unformattable_char&>::value;
-  static_assert(formattable_char, "Mixing character types is disallowed.");
-
-  constexpr bool formattable_const =
-      !std::is_same<decltype(arg), const unformattable_const&>::value;
-  static_assert(formattable_const, "Cannot format a const argument.");
-
-  // Formatting of arbitrary pointers is disallowed. If you want to output
-  // a pointer cast it to "void *" or "const void *". In particular, this
-  // forbids formatting of "[const] volatile char *" which is printed as bool
-  // by iostreams.
-  constexpr bool formattable_pointer =
-      !std::is_same<decltype(arg), const unformattable_pointer&>::value;
-  static_assert(formattable_pointer,
-                "Formatting of non-void pointers is disallowed.");
-
-  constexpr bool formattable =
-      !std::is_same<decltype(arg), const unformattable&>::value;
-  static_assert(
-      formattable,
-      "Cannot format an argument. To make type T formattable provide a "
-      "formatter<T> specialization: https://fmt.dev/latest/api.html#udt");
-  return {arg};
-}
-
-template <typename Context, typename T>
-FMT_CONSTEXPR auto make_arg(T&& value) -> basic_format_arg<Context> {
-  basic_format_arg<Context> arg;
-  arg.type_ = mapped_type_constant<T, Context>::value;
-  arg.value_ = make_value<Context>(value);
-  return arg;
-}
-
-// The type template parameter is there to avoid an ODR violation when using
-// a fallback formatter in one translation unit and an implicit conversion in
-// another (not recommended).
-template <bool IS_PACKED, typename Context, type, typename T,
-          FMT_ENABLE_IF(IS_PACKED)>
-FMT_CONSTEXPR FMT_INLINE auto make_arg(T&& val) -> value<Context> {
-  return make_value<Context>(val);
-}
-
-template <bool IS_PACKED, typename Context, type, typename T,
-          FMT_ENABLE_IF(!IS_PACKED)>
-FMT_CONSTEXPR inline auto make_arg(T&& value) -> basic_format_arg<Context> {
-  return make_arg<Context>(value);
-}
-FMT_END_DETAIL_NAMESPACE
-
 // Formatting context.
 template <typename OutputIt, typename Char> class basic_format_context {
- public:
-  /** The character type for the output. */
-  using char_type = Char;
-
  private:
   OutputIt out_;
   basic_format_args<basic_format_context> args_;
@@ -1798,32 +1745,34 @@ template <typename OutputIt, typename Char> class basic_format_context {
  public:
   using iterator = OutputIt;
   using format_arg = basic_format_arg<basic_format_context>;
+  using format_args = basic_format_args<basic_format_context>;
   using parse_context_type = basic_format_parse_context<Char>;
-  template <typename T> using formatter_type = formatter<T, char_type>;
+  template <typename T> using formatter_type = formatter<T, Char>;
+
+  /** The character type for the output. */
+  using char_type = Char;
 
   basic_format_context(basic_format_context&&) = default;
   basic_format_context(const basic_format_context&) = delete;
   void operator=(const basic_format_context&) = delete;
   /**
-   Constructs a ``basic_format_context`` object. References to the arguments are
-   stored in the object so make sure they have appropriate lifetimes.
+    Constructs a ``basic_format_context`` object. References to the arguments
+    are stored in the object so make sure they have appropriate lifetimes.
    */
-  constexpr basic_format_context(
-      OutputIt out, basic_format_args<basic_format_context> ctx_args,
-      detail::locale_ref loc = detail::locale_ref())
+  constexpr basic_format_context(OutputIt out, format_args ctx_args,
+                                 detail::locale_ref loc = {})
       : out_(out), args_(ctx_args), loc_(loc) {}
 
   constexpr auto arg(int id) const -> format_arg { return args_.get(id); }
-  FMT_CONSTEXPR auto arg(basic_string_view<char_type> name) -> format_arg {
+  FMT_CONSTEXPR auto arg(basic_string_view<Char> name) -> format_arg {
     return args_.get(name);
   }
-  FMT_CONSTEXPR auto arg_id(basic_string_view<char_type> name) -> int {
+  FMT_CONSTEXPR auto arg_id(basic_string_view<Char> name) -> int {
     return args_.get_id(name);
   }
-  auto args() const -> const basic_format_args<basic_format_context>& {
-    return args_;
-  }
+  auto args() const -> const format_args& { return args_; }
 
+  // DEPRECATED!
   FMT_CONSTEXPR auto error_handler() -> detail::error_handler { return {}; }
   void on_error(const char* message) { error_handler().on_error(message); }
 
@@ -1843,16 +1792,10 @@ using buffer_context =
     basic_format_context<detail::buffer_appender<Char>, Char>;
 using format_context = buffer_context<char>;
 
-// Workaround an alias issue: https://stackoverflow.com/q/62767544/471164.
-#define FMT_BUFFER_CONTEXT(Char) \
-  basic_format_context<detail::buffer_appender<Char>, Char>
-
 template <typename T, typename Char = char>
-using is_formattable = bool_constant<
-    !std::is_base_of<detail::unformattable,
-                     decltype(detail::arg_mapper<buffer_context<Char>>().map(
-                         std::declval<T>()))>::value &&
-    !detail::has_fallback_formatter<T, Char>::value>;
+using is_formattable = bool_constant<!std::is_base_of<
+    detail::unformattable, decltype(detail::arg_mapper<buffer_context<Char>>()
+                                        .map(std::declval<T&>()))>::value>;
 
 /**
   \rst
@@ -1870,7 +1813,7 @@ class format_arg_store
 {
  private:
   static const size_t num_args = sizeof...(Args);
-  static const size_t num_named_args = detail::count_named_args<Args...>();
+  static constexpr size_t num_named_args = detail::count_named_args<Args...>();
   static const bool is_packed = num_args <= detail::max_packed_args;
 
   using value_type = conditional_t<is_packed, detail::value<Context>,
@@ -1891,16 +1834,14 @@ class format_arg_store
 
  public:
   template <typename... T>
-  FMT_CONSTEXPR FMT_INLINE format_arg_store(T&&... args)
+  FMT_CONSTEXPR FMT_INLINE format_arg_store(T&... args)
       :
 #if FMT_GCC_VERSION && FMT_GCC_VERSION < 409
         basic_format_args<Context>(*this),
 #endif
-        data_{detail::make_arg<
-            is_packed, Context,
-            detail::mapped_type_constant<remove_cvref_t<T>, Context>::value>(
-            FMT_FORWARD(args))...} {
-    detail::init_named_args(data_.named_args(), 0, 0, args...);
+        data_{detail::make_arg<is_packed, Context>(args)...} {
+    if (detail::const_check(num_named_args != 0))
+      detail::init_named_args(data_.named_args(), 0, 0, args...);
   }
 };
 
@@ -1908,14 +1849,15 @@ class format_arg_store
   \rst
   Constructs a `~fmt::format_arg_store` object that contains references to
   arguments and can be implicitly converted to `~fmt::format_args`. `Context`
-  can be omitted in which case it defaults to `~fmt::context`.
+  can be omitted in which case it defaults to `~fmt::format_context`.
   See `~fmt::arg` for lifetime considerations.
   \endrst
  */
-template <typename Context = format_context, typename... Args>
-constexpr auto make_format_args(Args&&... args)
-    -> format_arg_store<Context, remove_cvref_t<Args>...> {
-  return {FMT_FORWARD(args)...};
+// Arguments are taken by lvalue references to avoid some lifetime issues.
+template <typename Context = format_context, typename... T>
+constexpr auto make_format_args(T&... args)
+    -> format_arg_store<Context, remove_cvref_t<T>...> {
+  return {args...};
 }
 
 /**
@@ -1934,6 +1876,7 @@ inline auto arg(const Char* name, const T& arg) -> detail::named_arg<Char, T> {
   static_assert(!detail::is_named_arg<T>(), "nested named arguments");
   return {name, arg};
 }
+FMT_END_EXPORT
 
 /**
   \rst
@@ -1942,7 +1885,7 @@ inline auto arg(const Char* name, const T& arg) -> detail::named_arg<Char, T> {
   ``vformat``::
 
     void vlog(string_view format_str, format_args args);  // OK
-    format_args args = make_format_args(42);  // Error: dangling reference
+    format_args args = make_format_args();  // Error: dangling reference
   \endrst
  */
 template <typename Context> class basic_format_args {
@@ -2059,7 +2002,7 @@ template <typename Context> class basic_format_args {
 /** An alias to ``basic_format_args<format_context>``. */
 // A separate type would result in shorter symbols but break ABI compatibility
 // between clang and gcc on ARM (#1919).
-using format_args = basic_format_args<format_context>;
+FMT_EXPORT using format_args = basic_format_args<format_context>;
 
 // We cannot use enum classes as bit fields because of a gcc bug, so we put them
 // in namespaces instead (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61414).
@@ -2080,7 +2023,7 @@ enum type FMT_ENUM_UNDERLYING_TYPE(unsigned char){none, minus, plus, space};
 }
 using sign_t = sign::type;
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 // Workaround an array initialization issue in gcc 4.8.
 template <typename Char> struct fill_t {
@@ -2092,7 +2035,7 @@ template <typename Char> struct fill_t {
  public:
   FMT_CONSTEXPR void operator=(basic_string_view<Char> s) {
     auto size = s.size();
-    if (size > max_size) return throw_format_error("invalid fill");
+    FMT_ASSERT(size <= max_size, "invalid fill");
     for (size_t i = 0; i < size; ++i) data_[i] = s[i];
     size_ = static_cast<unsigned char>(size);
   }
@@ -2105,11 +2048,10 @@ template <typename Char> struct fill_t {
     return data_[index];
   }
 };
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
 enum class presentation_type : unsigned char {
   none,
-  // Integer types should go first,
   dec,             // 'd'
   oct,             // 'o'
   hex_lower,       // 'x'
@@ -2131,7 +2073,7 @@ enum class presentation_type : unsigned char {
 };
 
 // Format specifiers for built-in and string types.
-template <typename Char> struct basic_format_specs {
+template <typename Char = char> struct format_specs {
   int width;
   int precision;
   presentation_type type;
@@ -2141,7 +2083,7 @@ template <typename Char> struct basic_format_specs {
   bool localized : 1;
   detail::fill_t<Char> fill;
 
-  constexpr basic_format_specs()
+  constexpr format_specs()
       : width(0),
         precision(-1),
         type(presentation_type::none),
@@ -2151,9 +2093,7 @@ template <typename Char> struct basic_format_specs {
         localized(false) {}
 };
 
-using format_specs = basic_format_specs<char>;
-
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 enum class arg_id_kind { none, index, name };
 
@@ -2174,7 +2114,7 @@ template <typename Char> struct arg_ref {
 
   arg_id_kind kind;
   union value {
-    FMT_CONSTEXPR value(int id = 0) : index{id} {}
+    FMT_CONSTEXPR value(int idx = 0) : index(idx) {}
     FMT_CONSTEXPR value(basic_string_view<Char> n) : name(n) {}
 
     int index;
@@ -2183,134 +2123,30 @@ template <typename Char> struct arg_ref {
 };
 
 // Format specifiers with width and precision resolved at formatting rather
-// than parsing time to allow re-using the same parsed specifiers with
+// than parsing time to allow reusing the same parsed specifiers with
 // different sets of arguments (precompilation of format strings).
-template <typename Char>
-struct dynamic_format_specs : basic_format_specs<Char> {
+template <typename Char = char>
+struct dynamic_format_specs : format_specs<Char> {
   arg_ref<Char> width_ref;
   arg_ref<Char> precision_ref;
 };
 
-struct auto_id {};
-
-// A format specifier handler that sets fields in basic_format_specs.
-template <typename Char> class specs_setter {
- protected:
-  basic_format_specs<Char>& specs_;
-
- public:
-  explicit FMT_CONSTEXPR specs_setter(basic_format_specs<Char>& specs)
-      : specs_(specs) {}
-
-  FMT_CONSTEXPR specs_setter(const specs_setter& other)
-      : specs_(other.specs_) {}
-
-  FMT_CONSTEXPR void on_align(align_t align) { specs_.align = align; }
-  FMT_CONSTEXPR void on_fill(basic_string_view<Char> fill) {
-    specs_.fill = fill;
-  }
-  FMT_CONSTEXPR void on_sign(sign_t s) { specs_.sign = s; }
-  FMT_CONSTEXPR void on_hash() { specs_.alt = true; }
-  FMT_CONSTEXPR void on_localized() { specs_.localized = true; }
-
-  FMT_CONSTEXPR void on_zero() {
-    if (specs_.align == align::none) specs_.align = align::numeric;
-    specs_.fill[0] = Char('0');
-  }
-
-  FMT_CONSTEXPR void on_width(int width) { specs_.width = width; }
-  FMT_CONSTEXPR void on_precision(int precision) {
-    specs_.precision = precision;
-  }
-  FMT_CONSTEXPR void end_precision() {}
-
-  FMT_CONSTEXPR void on_type(presentation_type type) { specs_.type = type; }
-};
-
-// Format spec handler that saves references to arguments representing dynamic
-// width and precision to be resolved at formatting time.
-template <typename ParseContext>
-class dynamic_specs_handler
-    : public specs_setter<typename ParseContext::char_type> {
- public:
-  using char_type = typename ParseContext::char_type;
-
-  FMT_CONSTEXPR dynamic_specs_handler(dynamic_format_specs<char_type>& specs,
-                                      ParseContext& ctx)
-      : specs_setter<char_type>(specs), specs_(specs), context_(ctx) {}
-
-  FMT_CONSTEXPR dynamic_specs_handler(const dynamic_specs_handler& other)
-      : specs_setter<char_type>(other),
-        specs_(other.specs_),
-        context_(other.context_) {}
-
-  template <typename Id> FMT_CONSTEXPR void on_dynamic_width(Id arg_id) {
-    specs_.width_ref = make_arg_ref(arg_id);
-  }
-
-  template <typename Id> FMT_CONSTEXPR void on_dynamic_precision(Id arg_id) {
-    specs_.precision_ref = make_arg_ref(arg_id);
-  }
-
-  FMT_CONSTEXPR void on_error(const char* message) {
-    context_.on_error(message);
-  }
-
- private:
-  dynamic_format_specs<char_type>& specs_;
-  ParseContext& context_;
-
-  using arg_ref_type = arg_ref<char_type>;
-
-  FMT_CONSTEXPR auto make_arg_ref(int arg_id) -> arg_ref_type {
-    context_.check_arg_id(arg_id);
-    context_.check_dynamic_spec(arg_id);
-    return arg_ref_type(arg_id);
-  }
-
-  FMT_CONSTEXPR auto make_arg_ref(auto_id) -> arg_ref_type {
-    int arg_id = context_.next_arg_id();
-    context_.check_dynamic_spec(arg_id);
-    return arg_ref_type(arg_id);
-  }
-
-  FMT_CONSTEXPR auto make_arg_ref(basic_string_view<char_type> arg_id)
-      -> arg_ref_type {
-    context_.check_arg_id(arg_id);
-    basic_string_view<char_type> format_str(
-        context_.begin(), to_unsigned(context_.end() - context_.begin()));
-    return arg_ref_type(arg_id);
-  }
-};
-
-template <typename Char> constexpr bool is_ascii_letter(Char c) {
-  return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z');
-}
-
-// Converts a character to ASCII. Returns a number > 127 on conversion failure.
+// Converts a character to ASCII. Returns '\0' on conversion failure.
 template <typename Char, FMT_ENABLE_IF(std::is_integral<Char>::value)>
-constexpr auto to_ascii(Char c) -> Char {
-  return c;
+constexpr auto to_ascii(Char c) -> char {
+  return c <= 0xff ? static_cast<char>(c) : '\0';
 }
 template <typename Char, FMT_ENABLE_IF(std::is_enum<Char>::value)>
-constexpr auto to_ascii(Char c) -> underlying_t<Char> {
-  return c;
-}
-
-FMT_CONSTEXPR inline auto code_point_length_impl(char c) -> int {
-  return "\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\0\0\0\0\0\0\0\0\2\2\2\2\3\3\4"
-      [static_cast<unsigned char>(c) >> 3];
+constexpr auto to_ascii(Char c) -> char {
+  return c <= 0xff ? static_cast<char>(c) : '\0';
 }
 
+// Returns the number of code units in a code point or 1 on error.
 template <typename Char>
 FMT_CONSTEXPR auto code_point_length(const Char* begin) -> int {
   if (const_check(sizeof(Char) != 1)) return 1;
-  int len = code_point_length_impl(static_cast<char>(*begin));
-
-  // Compute the pointer to the next character early so that the next
-  // iteration can start working on the next character. Neither Clang
-  // nor GCC figure out this reordering on their own.
-  return len + !len;
+  auto c = static_cast<unsigned char>(*begin);
+  return static_cast<int>((0x3a55000000000000ull >> (2 * (c >> 3))) & 0x3) + 1;
 }
 
 // Return the result via the out param to workaround gcc bug 77539.
@@ -2355,279 +2191,295 @@ FMT_CONSTEXPR auto parse_nonnegative_int(const Char*& begin, const Char* end,
              : error_value;
 }
 
-// Parses fill and alignment.
-template <typename Char, typename Handler>
-FMT_CONSTEXPR auto parse_align(const Char* begin, const Char* end,
-                               Handler&& handler) -> const Char* {
-  FMT_ASSERT(begin != end, "");
-  auto align = align::none;
-  auto p = begin + code_point_length(begin);
-  if (end - p <= 0) p = begin;
-  for (;;) {
-    switch (to_ascii(*p)) {
-    case '<':
-      align = align::left;
-      break;
-    case '>':
-      align = align::right;
-      break;
-    case '^':
-      align = align::center;
-      break;
-    default:
-      break;
-    }
-    if (align != align::none) {
-      if (p != begin) {
-        auto c = *begin;
-        if (c == '{')
-          return handler.on_error("invalid fill character '{'"), begin;
-        handler.on_fill(basic_string_view<Char>(begin, to_unsigned(p - begin)));
-        begin = p + 1;
-      } else
-        ++begin;
-      handler.on_align(align);
-      break;
-    } else if (p == begin) {
-      break;
-    }
-    p = begin;
+FMT_CONSTEXPR inline auto parse_align(char c) -> align_t {
+  switch (c) {
+  case '<':
+    return align::left;
+  case '>':
+    return align::right;
+  case '^':
+    return align::center;
   }
-  return begin;
+  return align::none;
 }
 
-template <typename Char> FMT_CONSTEXPR bool is_name_start(Char c) {
-  return ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || '_' == c;
+template <typename Char> constexpr auto is_name_start(Char c) -> bool {
+  return ('a' <= c && c <= 'z') || ('A' <= c && c <= 'Z') || c == '_';
 }
 
-template <typename Char, typename IDHandler>
+template <typename Char, typename Handler>
 FMT_CONSTEXPR auto do_parse_arg_id(const Char* begin, const Char* end,
-                                   IDHandler&& handler) -> const Char* {
-  FMT_ASSERT(begin != end, "");
+                                   Handler&& handler) -> const Char* {
   Char c = *begin;
   if (c >= '0' && c <= '9') {
     int index = 0;
+    constexpr int max = (std::numeric_limits<int>::max)();
     if (c != '0')
-      index =
-          parse_nonnegative_int(begin, end, (std::numeric_limits<int>::max)());
+      index = parse_nonnegative_int(begin, end, max);
     else
       ++begin;
     if (begin == end || (*begin != '}' && *begin != ':'))
-      handler.on_error("invalid format string");
+      throw_format_error("invalid format string");
     else
-      handler(index);
+      handler.on_index(index);
     return begin;
   }
   if (!is_name_start(c)) {
-    handler.on_error("invalid format string");
+    throw_format_error("invalid format string");
     return begin;
   }
   auto it = begin;
   do {
     ++it;
-  } while (it != end && (is_name_start(c = *it) || ('0' <= c && c <= '9')));
-  handler(basic_string_view<Char>(begin, to_unsigned(it - begin)));
+  } while (it != end && (is_name_start(*it) || ('0' <= *it && *it <= '9')));
+  handler.on_name({begin, to_unsigned(it - begin)});
   return it;
 }
 
-template <typename Char, typename IDHandler>
+template <typename Char, typename Handler>
 FMT_CONSTEXPR FMT_INLINE auto parse_arg_id(const Char* begin, const Char* end,
-                                           IDHandler&& handler) -> const Char* {
+                                           Handler&& handler) -> const Char* {
+  FMT_ASSERT(begin != end, "");
   Char c = *begin;
   if (c != '}' && c != ':') return do_parse_arg_id(begin, end, handler);
-  handler();
+  handler.on_auto();
   return begin;
 }
 
-template <typename Char, typename Handler>
-FMT_CONSTEXPR auto parse_width(const Char* begin, const Char* end,
-                               Handler&& handler) -> const Char* {
-  using detail::auto_id;
-  struct width_adapter {
-    Handler& handler;
+template <typename Char> struct dynamic_spec_id_handler {
+  basic_format_parse_context<Char>& ctx;
+  arg_ref<Char>& ref;
 
-    FMT_CONSTEXPR void operator()() { handler.on_dynamic_width(auto_id()); }
-    FMT_CONSTEXPR void operator()(int id) { handler.on_dynamic_width(id); }
-    FMT_CONSTEXPR void operator()(basic_string_view<Char> id) {
-      handler.on_dynamic_width(id);
-    }
-    FMT_CONSTEXPR void on_error(const char* message) {
-      if (message) handler.on_error(message);
-    }
-  };
+  FMT_CONSTEXPR void on_auto() {
+    int id = ctx.next_arg_id();
+    ref = arg_ref<Char>(id);
+    ctx.check_dynamic_spec(id);
+  }
+  FMT_CONSTEXPR void on_index(int id) {
+    ref = arg_ref<Char>(id);
+    ctx.check_arg_id(id);
+    ctx.check_dynamic_spec(id);
+  }
+  FMT_CONSTEXPR void on_name(basic_string_view<Char> id) {
+    ref = arg_ref<Char>(id);
+    ctx.check_arg_id(id);
+  }
+};
 
+// Parses [integer | "{" [arg_id] "}"].
+template <typename Char>
+FMT_CONSTEXPR auto parse_dynamic_spec(const Char* begin, const Char* end,
+                                      int& value, arg_ref<Char>& ref,
+                                      basic_format_parse_context<Char>& ctx)
+    -> const Char* {
   FMT_ASSERT(begin != end, "");
   if ('0' <= *begin && *begin <= '9') {
-    int width = parse_nonnegative_int(begin, end, -1);
-    if (width != -1)
-      handler.on_width(width);
+    int val = parse_nonnegative_int(begin, end, -1);
+    if (val != -1)
+      value = val;
     else
-      handler.on_error("number is too big");
+      throw_format_error("number is too big");
   } else if (*begin == '{') {
     ++begin;
-    if (begin != end) begin = parse_arg_id(begin, end, width_adapter{handler});
-    if (begin == end || *begin != '}')
-      return handler.on_error("invalid format string"), begin;
-    ++begin;
+    auto handler = dynamic_spec_id_handler<Char>{ctx, ref};
+    if (begin != end) begin = parse_arg_id(begin, end, handler);
+    if (begin != end && *begin == '}') return ++begin;
+    throw_format_error("invalid format string");
   }
   return begin;
 }
 
-template <typename Char, typename Handler>
-FMT_CONSTEXPR auto parse_precision(const Char* begin, const Char* end,
-                                   Handler&& handler) -> const Char* {
-  using detail::auto_id;
-  struct precision_adapter {
-    Handler& handler;
-
-    FMT_CONSTEXPR void operator()() { handler.on_dynamic_precision(auto_id()); }
-    FMT_CONSTEXPR void operator()(int id) { handler.on_dynamic_precision(id); }
-    FMT_CONSTEXPR void operator()(basic_string_view<Char> id) {
-      handler.on_dynamic_precision(id);
-    }
-    FMT_CONSTEXPR void on_error(const char* message) {
-      if (message) handler.on_error(message);
-    }
-  };
-
-  ++begin;
-  auto c = begin != end ? *begin : Char();
-  if ('0' <= c && c <= '9') {
-    auto precision = parse_nonnegative_int(begin, end, -1);
-    if (precision != -1)
-      handler.on_precision(precision);
-    else
-      handler.on_error("number is too big");
-  } else if (c == '{') {
-    ++begin;
-    if (begin != end)
-      begin = parse_arg_id(begin, end, precision_adapter{handler});
-    if (begin == end || *begin++ != '}')
-      return handler.on_error("invalid format string"), begin;
-  } else {
-    return handler.on_error("missing precision specifier"), begin;
-  }
-  handler.end_precision();
-  return begin;
-}
-
 template <typename Char>
-FMT_CONSTEXPR auto parse_presentation_type(Char type) -> presentation_type {
-  switch (to_ascii(type)) {
-  case 'd':
-    return presentation_type::dec;
-  case 'o':
-    return presentation_type::oct;
-  case 'x':
-    return presentation_type::hex_lower;
-  case 'X':
-    return presentation_type::hex_upper;
-  case 'b':
-    return presentation_type::bin_lower;
-  case 'B':
-    return presentation_type::bin_upper;
-  case 'a':
-    return presentation_type::hexfloat_lower;
-  case 'A':
-    return presentation_type::hexfloat_upper;
-  case 'e':
-    return presentation_type::exp_lower;
-  case 'E':
-    return presentation_type::exp_upper;
-  case 'f':
-    return presentation_type::fixed_lower;
-  case 'F':
-    return presentation_type::fixed_upper;
-  case 'g':
-    return presentation_type::general_lower;
-  case 'G':
-    return presentation_type::general_upper;
-  case 'c':
-    return presentation_type::chr;
-  case 's':
-    return presentation_type::string;
-  case 'p':
-    return presentation_type::pointer;
-  case '?':
-    return presentation_type::debug;
-  default:
-    return presentation_type::none;
-  }
-}
-
-// Parses standard format specifiers and sends notifications about parsed
-// components to handler.
-template <typename Char, typename SpecHandler>
-FMT_CONSTEXPR FMT_INLINE auto parse_format_specs(const Char* begin,
-                                                 const Char* end,
-                                                 SpecHandler&& handler)
+FMT_CONSTEXPR auto parse_precision(const Char* begin, const Char* end,
+                                   int& value, arg_ref<Char>& ref,
+                                   basic_format_parse_context<Char>& ctx)
     -> const Char* {
-  if (1 < end - begin && begin[1] == '}' && is_ascii_letter(*begin) &&
-      *begin != 'L') {
-    presentation_type type = parse_presentation_type(*begin++);
-    if (type == presentation_type::none)
-      handler.on_error("invalid type specifier");
-    handler.on_type(type);
+  ++begin;
+  if (begin == end || *begin == '}') {
+    throw_format_error("invalid precision");
     return begin;
   }
+  return parse_dynamic_spec(begin, end, value, ref, ctx);
+}
 
-  if (begin == end) return begin;
-
-  begin = parse_align(begin, end, handler);
-  if (begin == end) return begin;
-
-  // Parse sign.
-  switch (to_ascii(*begin)) {
-  case '+':
-    handler.on_sign(sign::plus);
-    ++begin;
-    break;
-  case '-':
-    handler.on_sign(sign::minus);
-    ++begin;
-    break;
-  case ' ':
-    handler.on_sign(sign::space);
-    ++begin;
-    break;
-  default:
-    break;
-  }
-  if (begin == end) return begin;
-
-  if (*begin == '#') {
-    handler.on_hash();
-    if (++begin == end) return begin;
-  }
-
-  // Parse zero flag.
-  if (*begin == '0') {
-    handler.on_zero();
-    if (++begin == end) return begin;
-  }
-
-  begin = parse_width(begin, end, handler);
-  if (begin == end) return begin;
+enum class state { start, align, sign, hash, zero, width, precision, locale };
 
-  // Parse precision.
-  if (*begin == '.') {
-    begin = parse_precision(begin, end, handler);
+// Parses standard format specifiers.
+template <typename Char>
+FMT_CONSTEXPR FMT_INLINE auto parse_format_specs(
+    const Char* begin, const Char* end, dynamic_format_specs<Char>& specs,
+    basic_format_parse_context<Char>& ctx, type arg_type) -> const Char* {
+  auto c = '\0';
+  if (end - begin > 1) {
+    auto next = to_ascii(begin[1]);
+    c = parse_align(next) == align::none ? to_ascii(*begin) : '\0';
+  } else {
     if (begin == end) return begin;
+    c = to_ascii(*begin);
   }
 
-  if (*begin == 'L') {
-    handler.on_localized();
-    ++begin;
-  }
+  struct {
+    state current_state = state::start;
+    FMT_CONSTEXPR void operator()(state s, bool valid = true) {
+      if (current_state >= s || !valid)
+        throw_format_error("invalid format specifier");
+      current_state = s;
+    }
+  } enter_state;
+
+  using pres = presentation_type;
+  constexpr auto integral_set = sint_set | uint_set | bool_set | char_set;
+  struct {
+    const Char*& begin;
+    dynamic_format_specs<Char>& specs;
+    type arg_type;
+
+    FMT_CONSTEXPR auto operator()(pres pres_type, int set) -> const Char* {
+      if (!in(arg_type, set)) {
+        if (arg_type == type::none_type) return begin;
+        throw_format_error("invalid format specifier");
+      }
+      specs.type = pres_type;
+      return begin + 1;
+    }
+  } parse_presentation_type{begin, specs, arg_type};
 
-  // Parse type.
-  if (begin != end && *begin != '}') {
-    presentation_type type = parse_presentation_type(*begin++);
-    if (type == presentation_type::none)
-      handler.on_error("invalid type specifier");
-    handler.on_type(type);
+  for (;;) {
+    switch (c) {
+    case '<':
+    case '>':
+    case '^':
+      enter_state(state::align);
+      specs.align = parse_align(c);
+      ++begin;
+      break;
+    case '+':
+    case '-':
+    case ' ':
+      if (arg_type == type::none_type) return begin;
+      enter_state(state::sign, in(arg_type, sint_set | float_set));
+      switch (c) {
+      case '+':
+        specs.sign = sign::plus;
+        break;
+      case '-':
+        specs.sign = sign::minus;
+        break;
+      case ' ':
+        specs.sign = sign::space;
+        break;
+      }
+      ++begin;
+      break;
+    case '#':
+      if (arg_type == type::none_type) return begin;
+      enter_state(state::hash, is_arithmetic_type(arg_type));
+      specs.alt = true;
+      ++begin;
+      break;
+    case '0':
+      enter_state(state::zero);
+      if (!is_arithmetic_type(arg_type)) {
+        if (arg_type == type::none_type) return begin;
+        throw_format_error("format specifier requires numeric argument");
+      }
+      if (specs.align == align::none) {
+        // Ignore 0 if align is specified for compatibility with std::format.
+        specs.align = align::numeric;
+        specs.fill[0] = Char('0');
+      }
+      ++begin;
+      break;
+    case '1':
+    case '2':
+    case '3':
+    case '4':
+    case '5':
+    case '6':
+    case '7':
+    case '8':
+    case '9':
+    case '{':
+      enter_state(state::width);
+      begin = parse_dynamic_spec(begin, end, specs.width, specs.width_ref, ctx);
+      break;
+    case '.':
+      if (arg_type == type::none_type) return begin;
+      enter_state(state::precision,
+                  in(arg_type, float_set | string_set | cstring_set));
+      begin = parse_precision(begin, end, specs.precision, specs.precision_ref,
+                              ctx);
+      break;
+    case 'L':
+      if (arg_type == type::none_type) return begin;
+      enter_state(state::locale, is_arithmetic_type(arg_type));
+      specs.localized = true;
+      ++begin;
+      break;
+    case 'd':
+      return parse_presentation_type(pres::dec, integral_set);
+    case 'o':
+      return parse_presentation_type(pres::oct, integral_set);
+    case 'x':
+      return parse_presentation_type(pres::hex_lower, integral_set);
+    case 'X':
+      return parse_presentation_type(pres::hex_upper, integral_set);
+    case 'b':
+      return parse_presentation_type(pres::bin_lower, integral_set);
+    case 'B':
+      return parse_presentation_type(pres::bin_upper, integral_set);
+    case 'a':
+      return parse_presentation_type(pres::hexfloat_lower, float_set);
+    case 'A':
+      return parse_presentation_type(pres::hexfloat_upper, float_set);
+    case 'e':
+      return parse_presentation_type(pres::exp_lower, float_set);
+    case 'E':
+      return parse_presentation_type(pres::exp_upper, float_set);
+    case 'f':
+      return parse_presentation_type(pres::fixed_lower, float_set);
+    case 'F':
+      return parse_presentation_type(pres::fixed_upper, float_set);
+    case 'g':
+      return parse_presentation_type(pres::general_lower, float_set);
+    case 'G':
+      return parse_presentation_type(pres::general_upper, float_set);
+    case 'c':
+      if (arg_type == type::bool_type)
+        throw_format_error("invalid format specifier");
+      return parse_presentation_type(pres::chr, integral_set);
+    case 's':
+      return parse_presentation_type(pres::string,
+                                     bool_set | string_set | cstring_set);
+    case 'p':
+      return parse_presentation_type(pres::pointer, pointer_set | cstring_set);
+    case '?':
+      return parse_presentation_type(pres::debug,
+                                     char_set | string_set | cstring_set);
+    case '}':
+      return begin;
+    default: {
+      if (*begin == '}') return begin;
+      // Parse fill and alignment.
+      auto fill_end = begin + code_point_length(begin);
+      if (end - fill_end <= 0) {
+        throw_format_error("invalid format specifier");
+        return begin;
+      }
+      if (*begin == '{') {
+        throw_format_error("invalid fill character '{'");
+        return begin;
+      }
+      auto align = parse_align(to_ascii(*fill_end));
+      enter_state(state::align, align != align::none);
+      specs.fill = {begin, to_unsigned(fill_end - begin)};
+      specs.align = align;
+      begin = fill_end + 1;
+    }
+    }
+    if (begin == end) return begin;
+    c = to_ascii(*begin);
   }
-  return begin;
 }
 
 template <typename Char, typename Handler>
@@ -2637,14 +2489,11 @@ FMT_CONSTEXPR auto parse_replacement_field(const Char* begin, const Char* end,
     Handler& handler;
     int arg_id;
 
-    FMT_CONSTEXPR void operator()() { arg_id = handler.on_arg_id(); }
-    FMT_CONSTEXPR void operator()(int id) { arg_id = handler.on_arg_id(id); }
-    FMT_CONSTEXPR void operator()(basic_string_view<Char> id) {
+    FMT_CONSTEXPR void on_auto() { arg_id = handler.on_arg_id(); }
+    FMT_CONSTEXPR void on_index(int id) { arg_id = handler.on_arg_id(id); }
+    FMT_CONSTEXPR void on_name(basic_string_view<Char> id) {
       arg_id = handler.on_arg_id(id);
     }
-    FMT_CONSTEXPR void on_error(const char* message) {
-      if (message) handler.on_error(message);
-    }
   };
 
   ++begin;
@@ -2673,9 +2522,6 @@ FMT_CONSTEXPR auto parse_replacement_field(const Char* begin, const Char* end,
 template <bool IS_CONSTEXPR, typename Char, typename Handler>
 FMT_CONSTEXPR FMT_INLINE void parse_format_string(
     basic_string_view<Char> format_str, Handler&& handler) {
-  // Workaround a name-lookup bug in MSVC's modules implementation.
-  using detail::find;
-
   auto begin = format_str.data();
   auto end = begin + format_str.size();
   if (end - begin < 32) {
@@ -2735,189 +2581,46 @@ FMT_CONSTEXPR auto parse_format_specs(ParseContext& ctx)
     -> decltype(ctx.begin()) {
   using char_type = typename ParseContext::char_type;
   using context = buffer_context<char_type>;
-  using stripped_type = typename strip_named_arg<T>::type;
   using mapped_type = conditional_t<
       mapped_type_constant<T, context>::value != type::custom_type,
       decltype(arg_mapper<context>().map(std::declval<const T&>())),
-      stripped_type>;
-  auto f = conditional_t<has_formatter<mapped_type, context>::value,
-                         formatter<mapped_type, char_type>,
-                         fallback_formatter<stripped_type, char_type>>();
-  return f.parse(ctx);
-}
-
-template <typename ErrorHandler>
-FMT_CONSTEXPR void check_int_type_spec(presentation_type type,
-                                       ErrorHandler&& eh) {
-  if (type > presentation_type::bin_upper && type != presentation_type::chr)
-    eh.on_error("invalid type specifier");
+      typename strip_named_arg<T>::type>;
+#if defined(__cpp_if_constexpr)
+  if constexpr (std::is_default_constructible<
+                    formatter<mapped_type, char_type>>::value) {
+    return formatter<mapped_type, char_type>().parse(ctx);
+  } else {
+    type_is_unformattable_for<T, char_type> _;
+    return ctx.begin();
+  }
+#else
+  return formatter<mapped_type, char_type>().parse(ctx);
+#endif
 }
 
-// Checks char specs and returns true if the type spec is char (and not int).
-template <typename Char, typename ErrorHandler = error_handler>
-FMT_CONSTEXPR auto check_char_specs(const basic_format_specs<Char>& specs,
-                                    ErrorHandler&& eh = {}) -> bool {
+// Checks char specs and returns true iff the presentation type is char-like.
+template <typename Char>
+FMT_CONSTEXPR auto check_char_specs(const format_specs<Char>& specs) -> bool {
   if (specs.type != presentation_type::none &&
       specs.type != presentation_type::chr &&
       specs.type != presentation_type::debug) {
-    check_int_type_spec(specs.type, eh);
     return false;
   }
   if (specs.align == align::numeric || specs.sign != sign::none || specs.alt)
-    eh.on_error("invalid format specifier for char");
+    throw_format_error("invalid format specifier for char");
   return true;
 }
 
-// A floating-point presentation format.
-enum class float_format : unsigned char {
-  general,  // General: exponent notation or fixed point based on magnitude.
-  exp,      // Exponent notation with the default precision of 6, e.g. 1.2e-3.
-  fixed,    // Fixed point with the default precision of 6, e.g. 0.0012.
-  hex
-};
-
-struct float_specs {
-  int precision;
-  float_format format : 8;
-  sign_t sign : 8;
-  bool upper : 1;
-  bool locale : 1;
-  bool binary32 : 1;
-  bool showpoint : 1;
-};
-
-template <typename ErrorHandler = error_handler, typename Char>
-FMT_CONSTEXPR auto parse_float_type_spec(const basic_format_specs<Char>& specs,
-                                         ErrorHandler&& eh = {})
-    -> float_specs {
-  auto result = float_specs();
-  result.showpoint = specs.alt;
-  result.locale = specs.localized;
-  switch (specs.type) {
-  case presentation_type::none:
-    result.format = float_format::general;
-    break;
-  case presentation_type::general_upper:
-    result.upper = true;
-    FMT_FALLTHROUGH;
-  case presentation_type::general_lower:
-    result.format = float_format::general;
-    break;
-  case presentation_type::exp_upper:
-    result.upper = true;
-    FMT_FALLTHROUGH;
-  case presentation_type::exp_lower:
-    result.format = float_format::exp;
-    result.showpoint |= specs.precision != 0;
-    break;
-  case presentation_type::fixed_upper:
-    result.upper = true;
-    FMT_FALLTHROUGH;
-  case presentation_type::fixed_lower:
-    result.format = float_format::fixed;
-    result.showpoint |= specs.precision != 0;
-    break;
-  case presentation_type::hexfloat_upper:
-    result.upper = true;
-    FMT_FALLTHROUGH;
-  case presentation_type::hexfloat_lower:
-    result.format = float_format::hex;
-    break;
-  default:
-    eh.on_error("invalid type specifier");
-    break;
-  }
-  return result;
-}
-
-template <typename ErrorHandler = error_handler>
-FMT_CONSTEXPR auto check_cstring_type_spec(presentation_type type,
-                                           ErrorHandler&& eh = {}) -> bool {
-  if (type == presentation_type::none || type == presentation_type::string ||
-      type == presentation_type::debug)
-    return true;
-  if (type != presentation_type::pointer) eh.on_error("invalid type specifier");
-  return false;
-}
-
-template <typename ErrorHandler = error_handler>
-FMT_CONSTEXPR void check_string_type_spec(presentation_type type,
-                                          ErrorHandler&& eh = {}) {
-  if (type != presentation_type::none && type != presentation_type::string &&
-      type != presentation_type::debug)
-    eh.on_error("invalid type specifier");
-}
-
-template <typename ErrorHandler>
-FMT_CONSTEXPR void check_pointer_type_spec(presentation_type type,
-                                           ErrorHandler&& eh) {
-  if (type != presentation_type::none && type != presentation_type::pointer)
-    eh.on_error("invalid type specifier");
-}
-
-// A parse_format_specs handler that checks if specifiers are consistent with
-// the argument type.
-template <typename Handler> class specs_checker : public Handler {
- private:
-  detail::type arg_type_;
-
-  FMT_CONSTEXPR void require_numeric_argument() {
-    if (!is_arithmetic_type(arg_type_))
-      this->on_error("format specifier requires numeric argument");
-  }
-
- public:
-  FMT_CONSTEXPR specs_checker(const Handler& handler, detail::type arg_type)
-      : Handler(handler), arg_type_(arg_type) {}
-
-  FMT_CONSTEXPR void on_align(align_t align) {
-    if (align == align::numeric) require_numeric_argument();
-    Handler::on_align(align);
-  }
-
-  FMT_CONSTEXPR void on_sign(sign_t s) {
-    require_numeric_argument();
-    if (is_integral_type(arg_type_) && arg_type_ != type::int_type &&
-        arg_type_ != type::long_long_type && arg_type_ != type::int128_type &&
-        arg_type_ != type::char_type) {
-      this->on_error("format specifier requires signed argument");
-    }
-    Handler::on_sign(s);
-  }
-
-  FMT_CONSTEXPR void on_hash() {
-    require_numeric_argument();
-    Handler::on_hash();
-  }
-
-  FMT_CONSTEXPR void on_localized() {
-    require_numeric_argument();
-    Handler::on_localized();
-  }
-
-  FMT_CONSTEXPR void on_zero() {
-    require_numeric_argument();
-    Handler::on_zero();
-  }
-
-  FMT_CONSTEXPR void end_precision() {
-    if (is_integral_type(arg_type_) || arg_type_ == type::pointer_type)
-      this->on_error("precision not allowed for this argument type");
-  }
-};
-
-constexpr int invalid_arg_index = -1;
-
 #if FMT_USE_NONTYPE_TEMPLATE_ARGS
 template <int N, typename T, typename... Args, typename Char>
 constexpr auto get_arg_index_by_name(basic_string_view<Char> name) -> int {
-  if constexpr (detail::is_statically_named_arg<T>()) {
+  if constexpr (is_statically_named_arg<T>()) {
     if (name == T::name) return N;
   }
   if constexpr (sizeof...(Args) > 0)
     return get_arg_index_by_name<N + 1, Args...>(name);
   (void)name;  // Workaround an MSVC bug about "unused" parameter.
-  return invalid_arg_index;
+  return -1;
 }
 #endif
 
@@ -2928,34 +2631,29 @@ FMT_CONSTEXPR auto get_arg_index_by_name(basic_string_view<Char> name) -> int {
     return get_arg_index_by_name<0, Args...>(name);
 #endif
   (void)name;
-  return invalid_arg_index;
+  return -1;
 }
 
-template <typename Char, typename ErrorHandler, typename... Args>
-class format_string_checker {
+template <typename Char, typename... Args> class format_string_checker {
  private:
-  // In the future basic_format_parse_context will replace compile_parse_context
-  // here and will use is_constant_evaluated and downcasting to access the data
-  // needed for compile-time checks: https://godbolt.org/z/GvWzcTjh1.
-  using parse_context_type = compile_parse_context<Char, ErrorHandler>;
+  using parse_context_type = compile_parse_context<Char>;
   static constexpr int num_args = sizeof...(Args);
 
   // Format specifier parsing function.
+  // In the future basic_format_parse_context will replace compile_parse_context
+  // here and will use is_constant_evaluated and downcasting to access the data
+  // needed for compile-time checks: https://godbolt.org/z/GvWzcTjh1.
   using parse_func = const Char* (*)(parse_context_type&);
 
+  type types_[num_args > 0 ? static_cast<size_t>(num_args) : 1];
   parse_context_type context_;
   parse_func parse_funcs_[num_args > 0 ? static_cast<size_t>(num_args) : 1];
-  type types_[num_args > 0 ? static_cast<size_t>(num_args) : 1];
 
  public:
-  explicit FMT_CONSTEXPR format_string_checker(
-      basic_string_view<Char> format_str, ErrorHandler eh)
-      : context_(format_str, num_args, types_, eh),
-        parse_funcs_{&parse_format_specs<Args, parse_context_type>...},
-        types_{
-            mapped_type_constant<Args,
-                                 basic_format_context<Char*, Char>>::value...} {
-  }
+  explicit FMT_CONSTEXPR format_string_checker(basic_string_view<Char> fmt)
+      : types_{mapped_type_constant<Args, buffer_context<Char>>::value...},
+        context_(fmt, num_args, types_),
+        parse_funcs_{&parse_format_specs<Args, parse_context_type>...} {}
 
   FMT_CONSTEXPR void on_text(const Char*, const Char*) {}
 
@@ -2966,8 +2664,8 @@ class format_string_checker {
   FMT_CONSTEXPR auto on_arg_id(basic_string_view<Char> id) -> int {
 #if FMT_USE_NONTYPE_TEMPLATE_ARGS
     auto index = get_arg_index_by_name<Args...>(id);
-    if (index == invalid_arg_index) on_error("named argument is not found");
-    return context_.check_arg_id(index), index;
+    if (index < 0) on_error("named argument is not found");
+    return index;
 #else
     (void)id;
     on_error("compile-time checks for named arguments require C++20 support");
@@ -2975,17 +2673,19 @@ class format_string_checker {
 #endif
   }
 
-  FMT_CONSTEXPR void on_replacement_field(int, const Char*) {}
+  FMT_CONSTEXPR void on_replacement_field(int id, const Char* begin) {
+    on_format_specs(id, begin, begin);  // Call parse() on empty specs.
+  }
 
   FMT_CONSTEXPR auto on_format_specs(int id, const Char* begin, const Char*)
       -> const Char* {
-    context_.advance_to(context_.begin() + (begin - &*context_.begin()));
+    context_.advance_to(begin);
     // id >= 0 check is a workaround for gcc 10 bug (#2065).
     return id >= 0 && id < num_args ? parse_funcs_[id](context_) : begin;
   }
 
   FMT_CONSTEXPR void on_error(const char* message) {
-    context_.on_error(message);
+    throw_format_error(message);
   }
 };
 
@@ -3001,28 +2701,35 @@ FMT_INLINE void check_format_string(const S&) {
 template <typename... Args, typename S,
           FMT_ENABLE_IF(is_compile_string<S>::value)>
 void check_format_string(S format_str) {
-  FMT_CONSTEXPR auto s = basic_string_view<typename S::char_type>(format_str);
-  using checker = format_string_checker<typename S::char_type, error_handler,
-                                        remove_cvref_t<Args>...>;
-  FMT_CONSTEXPR bool invalid_format =
-      (parse_format_string<true>(s, checker(s, {})), true);
-  ignore_unused(invalid_format);
+  using char_t = typename S::char_type;
+  FMT_CONSTEXPR auto s = basic_string_view<char_t>(format_str);
+  using checker = format_string_checker<char_t, remove_cvref_t<Args>...>;
+  FMT_CONSTEXPR bool error = (parse_format_string<true>(s, checker(s)), true);
+  ignore_unused(error);
 }
 
+template <typename Char = char> struct vformat_args {
+  using type = basic_format_args<
+      basic_format_context<std::back_insert_iterator<buffer<Char>>, Char>>;
+};
+template <> struct vformat_args<char> {
+  using type = format_args;
+};
+
+// Use vformat_args and avoid type_identity to keep symbols short.
 template <typename Char>
-void vformat_to(
-    buffer<Char>& buf, basic_string_view<Char> fmt,
-    basic_format_args<FMT_BUFFER_CONTEXT(type_identity_t<Char>)> args,
-    locale_ref loc = {});
+void vformat_to(buffer<Char>& buf, basic_string_view<Char> fmt,
+                typename vformat_args<Char>::type args, locale_ref loc = {});
 
 FMT_API void vprint_mojibake(std::FILE*, string_view, format_args);
 #ifndef _WIN32
 inline void vprint_mojibake(std::FILE*, string_view, format_args) {}
 #endif
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
-// A formatter specialization for the core types corresponding to detail::type
-// constants.
+FMT_BEGIN_EXPORT
+
+// A formatter specialization for natively supported types.
 template <typename T, typename Char>
 struct formatter<T, Char,
                  enable_if_t<detail::type_constant<T, Char>::value !=
@@ -3031,81 +2738,21 @@ struct formatter<T, Char,
   detail::dynamic_format_specs<Char> specs_;
 
  public:
-  // Parses format specifiers stopping either at the end of the range or at the
-  // terminating '}'.
   template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
-    auto begin = ctx.begin(), end = ctx.end();
-    if (begin == end) return begin;
-    using handler_type = detail::dynamic_specs_handler<ParseContext>;
+  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> const Char* {
     auto type = detail::type_constant<T, Char>::value;
-    auto checker =
-        detail::specs_checker<handler_type>(handler_type(specs_, ctx), type);
-    auto it = detail::parse_format_specs(begin, end, checker);
-    auto eh = ctx.error_handler();
-    switch (type) {
-    case detail::type::none_type:
-      FMT_ASSERT(false, "invalid argument type");
-      break;
-    case detail::type::bool_type:
-      if (specs_.type == presentation_type::none ||
-          specs_.type == presentation_type::string) {
-        break;
-      }
-      FMT_FALLTHROUGH;
-    case detail::type::int_type:
-    case detail::type::uint_type:
-    case detail::type::long_long_type:
-    case detail::type::ulong_long_type:
-    case detail::type::int128_type:
-    case detail::type::uint128_type:
-      detail::check_int_type_spec(specs_.type, eh);
-      break;
-    case detail::type::char_type:
-      detail::check_char_specs(specs_, eh);
-      break;
-    case detail::type::float_type:
-      if (detail::const_check(FMT_USE_FLOAT))
-        detail::parse_float_type_spec(specs_, eh);
-      else
-        FMT_ASSERT(false, "float support disabled");
-      break;
-    case detail::type::double_type:
-      if (detail::const_check(FMT_USE_DOUBLE))
-        detail::parse_float_type_spec(specs_, eh);
-      else
-        FMT_ASSERT(false, "double support disabled");
-      break;
-    case detail::type::long_double_type:
-      if (detail::const_check(FMT_USE_LONG_DOUBLE))
-        detail::parse_float_type_spec(specs_, eh);
-      else
-        FMT_ASSERT(false, "long double support disabled");
-      break;
-    case detail::type::cstring_type:
-      detail::check_cstring_type_spec(specs_.type, eh);
-      break;
-    case detail::type::string_type:
-      detail::check_string_type_spec(specs_.type, eh);
-      break;
-    case detail::type::pointer_type:
-      detail::check_pointer_type_spec(specs_.type, eh);
-      break;
-    case detail::type::custom_type:
-      // Custom format specifiers are checked in parse functions of
-      // formatter specializations.
-      break;
-    }
-    return it;
+    auto end =
+        detail::parse_format_specs(ctx.begin(), ctx.end(), specs_, ctx, type);
+    if (type == detail::type::char_type) detail::check_char_specs(specs_);
+    return end;
   }
 
   template <detail::type U = detail::type_constant<T, Char>::value,
-            enable_if_t<(U == detail::type::string_type ||
-                         U == detail::type::cstring_type ||
-                         U == detail::type::char_type),
-                        int> = 0>
-  FMT_CONSTEXPR void set_debug_format() {
-    specs_.type = presentation_type::debug;
+            FMT_ENABLE_IF(U == detail::type::string_type ||
+                          U == detail::type::cstring_type ||
+                          U == detail::type::char_type)>
+  FMT_CONSTEXPR void set_debug_format(bool set = true) {
+    specs_.type = set ? presentation_type::debug : presentation_type::none;
   }
 
   template <typename FormatContext>
@@ -3113,28 +2760,9 @@ struct formatter<T, Char,
       -> decltype(ctx.out());
 };
 
-#define FMT_FORMAT_AS(Type, Base)                                        \
-  template <typename Char>                                               \
-  struct formatter<Type, Char> : formatter<Base, Char> {                 \
-    template <typename FormatContext>                                    \
-    auto format(Type const& val, FormatContext& ctx) const               \
-        -> decltype(ctx.out()) {                                         \
-      return formatter<Base, Char>::format(static_cast<Base>(val), ctx); \
-    }                                                                    \
-  }
-
-FMT_FORMAT_AS(signed char, int);
-FMT_FORMAT_AS(unsigned char, unsigned);
-FMT_FORMAT_AS(short, int);
-FMT_FORMAT_AS(unsigned short, unsigned);
-FMT_FORMAT_AS(long, long long);
-FMT_FORMAT_AS(unsigned long, unsigned long long);
-FMT_FORMAT_AS(Char*, const Char*);
-FMT_FORMAT_AS(std::basic_string<Char>, basic_string_view<Char>);
-FMT_FORMAT_AS(std::nullptr_t, const void*);
-FMT_FORMAT_AS(detail::std_string_view<Char>, basic_string_view<Char>);
-
-template <typename Char> struct basic_runtime { basic_string_view<Char> str; };
+template <typename Char = char> struct runtime_format_string {
+  basic_string_view<Char> str;
+};
 
 /** A compile-time format string. */
 template <typename Char, typename... Args> class basic_format_string {
@@ -3154,17 +2782,18 @@ template <typename Char, typename... Args> class basic_format_string {
 #ifdef FMT_HAS_CONSTEVAL
     if constexpr (detail::count_named_args<Args...>() ==
                   detail::count_statically_named_args<Args...>()) {
-      using checker = detail::format_string_checker<Char, detail::error_handler,
-                                                    remove_cvref_t<Args>...>;
-      detail::parse_format_string<true>(str_, checker(s, {}));
+      using checker =
+          detail::format_string_checker<Char, remove_cvref_t<Args>...>;
+      detail::parse_format_string<true>(str_, checker(s));
     }
 #else
     detail::check_format_string<Args...>(s);
 #endif
   }
-  basic_format_string(basic_runtime<Char> r) : str_(r.str) {}
+  basic_format_string(runtime_format_string<Char> fmt) : str_(fmt.str) {}
 
   FMT_INLINE operator basic_string_view<Char>() const { return str_; }
+  FMT_INLINE auto get() const -> basic_string_view<Char> { return str_; }
 };
 
 #if FMT_GCC_VERSION && FMT_GCC_VERSION < 409
@@ -3184,7 +2813,7 @@ using format_string = basic_format_string<char, type_identity_t<Args>...>;
     fmt::print(fmt::runtime("{:d}"), "I am not a number");
   \endrst
  */
-inline auto runtime(string_view s) -> basic_runtime<char> { return {{s}}; }
+inline auto runtime(string_view s) -> runtime_format_string<> { return {{s}}; }
 #endif
 
 FMT_API auto vformat(string_view fmt, format_args args) -> std::string;
@@ -3210,10 +2839,9 @@ FMT_NODISCARD FMT_INLINE auto format(format_string<T...> fmt, T&&... args)
 template <typename OutputIt,
           FMT_ENABLE_IF(detail::is_output_iterator<OutputIt, char>::value)>
 auto vformat_to(OutputIt out, string_view fmt, format_args args) -> OutputIt {
-  using detail::get_buffer;
-  auto&& buf = get_buffer<char>(out);
+  auto&& buf = detail::get_buffer<char>(out);
   detail::vformat_to(buf, fmt, args, {});
-  return detail::get_iterator(buf);
+  return detail::get_iterator(buf, out);
 }
 
 /**
@@ -3272,7 +2900,7 @@ template <typename... T>
 FMT_NODISCARD FMT_INLINE auto formatted_size(format_string<T...> fmt,
                                              T&&... args) -> size_t {
   auto buf = detail::counting_buffer<>();
-  detail::vformat_to(buf, string_view(fmt), fmt::make_format_args(args...), {});
+  detail::vformat_to<char>(buf, fmt, fmt::make_format_args(args...), {});
   return buf.count();
 }
 
@@ -3313,7 +2941,25 @@ FMT_INLINE void print(std::FILE* f, format_string<T...> fmt, T&&... args) {
                            : detail::vprint_mojibake(f, fmt, vargs);
 }
 
-FMT_MODULE_EXPORT_END
+/**
+  Formats ``args`` according to specifications in ``fmt`` and writes the
+  output to the file ``f`` followed by a newline.
+ */
+template <typename... T>
+FMT_INLINE void println(std::FILE* f, format_string<T...> fmt, T&&... args) {
+  return fmt::print(f, "{}\n", fmt::format(fmt, std::forward<T>(args)...));
+}
+
+/**
+  Formats ``args`` according to specifications in ``fmt`` and writes the output
+  to ``stdout`` followed by a newline.
+ */
+template <typename... T>
+FMT_INLINE void println(format_string<T...> fmt, T&&... args) {
+  return fmt::println(stdout, fmt, std::forward<T>(args)...);
+}
+
+FMT_END_EXPORT
 FMT_GCC_PRAGMA("GCC pop_options")
 FMT_END_NAMESPACE
 
diff --git a/dep/fmt/include/fmt/format-inl.h b/dep/fmt/include/fmt/format-inl.h
index 22b1ec8df0e..efac5d1f88f 100644
--- a/dep/fmt/include/fmt/format-inl.h
+++ b/dep/fmt/include/fmt/format-inl.h
@@ -9,20 +9,16 @@
 #define FMT_FORMAT_INL_H_
 
 #include <algorithm>
-#include <cctype>
 #include <cerrno>  // errno
 #include <climits>
 #include <cmath>
-#include <cstdarg>
-#include <cstring>  // std::memmove
-#include <cwchar>
 #include <exception>
 
 #ifndef FMT_STATIC_THOUSANDS_SEPARATOR
 #  include <locale>
 #endif
 
-#ifdef _WIN32
+#if defined(_WIN32) && !defined(FMT_WINDOWS_NO_WCHAR)
 #  include <io.h>  // _isatty
 #endif
 
@@ -62,8 +58,8 @@ FMT_FUNC void format_error_code(detail::buffer<char>& out, int error_code,
   error_code_size += detail::to_unsigned(detail::count_digits(abs_value));
   auto it = buffer_appender<char>(out);
   if (message.size() <= inline_buffer_size - error_code_size)
-    format_to(it, FMT_STRING("{}{}"), message, SEP);
-  format_to(it, FMT_STRING("{}{}"), ERROR_STR, error_code);
+    fmt::format_to(it, FMT_STRING("{}{}"), message, SEP);
+  fmt::format_to(it, FMT_STRING("{}{}"), ERROR_STR, error_code);
   FMT_ASSERT(out.size() <= inline_buffer_size, "");
 }
 
@@ -77,9 +73,8 @@ FMT_FUNC void report_error(format_func func, int error_code,
 }
 
 // A wrapper around fwrite that throws on error.
-inline void fwrite_fully(const void* ptr, size_t size, size_t count,
-                         FILE* stream) {
-  size_t written = std::fwrite(ptr, size, count, stream);
+inline void fwrite_fully(const void* ptr, size_t count, FILE* stream) {
+  size_t written = std::fwrite(ptr, 1, count, stream);
   if (written < count)
     FMT_THROW(system_error(errno, FMT_STRING("cannot write to file")));
 }
@@ -90,7 +85,7 @@ locale_ref::locale_ref(const Locale& loc) : locale_(&loc) {
   static_assert(std::is_same<Locale, std::locale>::value, "");
 }
 
-template <typename Locale> Locale locale_ref::get() const {
+template <typename Locale> auto locale_ref::get() const -> Locale {
   static_assert(std::is_same<Locale, std::locale>::value, "");
   return locale_ ? *static_cast<const std::locale*>(locale_) : std::locale();
 }
@@ -102,7 +97,8 @@ FMT_FUNC auto thousands_sep_impl(locale_ref loc) -> thousands_sep_result<Char> {
   auto thousands_sep = grouping.empty() ? Char() : facet.thousands_sep();
   return {std::move(grouping), thousands_sep};
 }
-template <typename Char> FMT_FUNC Char decimal_point_impl(locale_ref loc) {
+template <typename Char>
+FMT_FUNC auto decimal_point_impl(locale_ref loc) -> Char {
   return std::use_facet<std::numpunct<Char>>(loc.get<std::locale>())
       .decimal_point();
 }
@@ -115,96 +111,74 @@ template <typename Char> FMT_FUNC Char decimal_point_impl(locale_ref) {
   return '.';
 }
 #endif
+
+FMT_FUNC auto write_loc(appender out, loc_value value,
+                        const format_specs<>& specs, locale_ref loc) -> bool {
+#ifndef FMT_STATIC_THOUSANDS_SEPARATOR
+  auto locale = loc.get<std::locale>();
+  // We cannot use the num_put<char> facet because it may produce output in
+  // a wrong encoding.
+  using facet = format_facet<std::locale>;
+  if (std::has_facet<facet>(locale))
+    return std::use_facet<facet>(locale).put(out, value, specs);
+  return facet(locale).put(out, value, specs);
+#endif
+  return false;
+}
 }  // namespace detail
 
-#if !FMT_MSC_VERSION
-FMT_API FMT_FUNC format_error::~format_error() noexcept = default;
+template <typename Locale> typename Locale::id format_facet<Locale>::id;
+
+#ifndef FMT_STATIC_THOUSANDS_SEPARATOR
+template <typename Locale> format_facet<Locale>::format_facet(Locale& loc) {
+  auto& numpunct = std::use_facet<std::numpunct<char>>(loc);
+  grouping_ = numpunct.grouping();
+  if (!grouping_.empty()) separator_ = std::string(1, numpunct.thousands_sep());
+}
+
+template <>
+FMT_API FMT_FUNC auto format_facet<std::locale>::do_put(
+    appender out, loc_value val, const format_specs<>& specs) const -> bool {
+  return val.visit(
+      detail::loc_writer<>{out, specs, separator_, grouping_, decimal_point_});
+}
 #endif
 
-FMT_FUNC std::system_error vsystem_error(int error_code, string_view format_str,
-                                         format_args args) {
+FMT_FUNC auto vsystem_error(int error_code, string_view fmt, format_args args)
+    -> std::system_error {
   auto ec = std::error_code(error_code, std::generic_category());
-  return std::system_error(ec, vformat(format_str, args));
+  return std::system_error(ec, vformat(fmt, args));
 }
 
 namespace detail {
 
-template <typename F> inline bool operator==(basic_fp<F> x, basic_fp<F> y) {
+template <typename F>
+inline auto operator==(basic_fp<F> x, basic_fp<F> y) -> bool {
   return x.f == y.f && x.e == y.e;
 }
 
 // Compilers should be able to optimize this into the ror instruction.
-FMT_CONSTEXPR inline uint32_t rotr(uint32_t n, uint32_t r) noexcept {
+FMT_CONSTEXPR inline auto rotr(uint32_t n, uint32_t r) noexcept -> uint32_t {
   r &= 31;
   return (n >> r) | (n << (32 - r));
 }
-FMT_CONSTEXPR inline uint64_t rotr(uint64_t n, uint32_t r) noexcept {
+FMT_CONSTEXPR inline auto rotr(uint64_t n, uint32_t r) noexcept -> uint64_t {
   r &= 63;
   return (n >> r) | (n << (64 - r));
 }
 
-// Computes 128-bit result of multiplication of two 64-bit unsigned integers.
-inline uint128_fallback umul128(uint64_t x, uint64_t y) noexcept {
-#if FMT_USE_INT128
-  auto p = static_cast<uint128_opt>(x) * static_cast<uint128_opt>(y);
-  return {static_cast<uint64_t>(p >> 64), static_cast<uint64_t>(p)};
-#elif defined(_MSC_VER) && defined(_M_X64)
-  auto result = uint128_fallback();
-  result.lo_ = _umul128(x, y, &result.hi_);
-  return result;
-#else
-  const uint64_t mask = static_cast<uint64_t>(max_value<uint32_t>());
-
-  uint64_t a = x >> 32;
-  uint64_t b = x & mask;
-  uint64_t c = y >> 32;
-  uint64_t d = y & mask;
-
-  uint64_t ac = a * c;
-  uint64_t bc = b * c;
-  uint64_t ad = a * d;
-  uint64_t bd = b * d;
-
-  uint64_t intermediate = (bd >> 32) + (ad & mask) + (bc & mask);
-
-  return {ac + (intermediate >> 32) + (ad >> 32) + (bc >> 32),
-          (intermediate << 32) + (bd & mask)};
-#endif
-}
-
 // Implementation of Dragonbox algorithm: https://github.com/jk-jeon/dragonbox.
 namespace dragonbox {
-// Computes upper 64 bits of multiplication of two 64-bit unsigned integers.
-inline uint64_t umul128_upper64(uint64_t x, uint64_t y) noexcept {
-#if FMT_USE_INT128
-  auto p = static_cast<uint128_opt>(x) * static_cast<uint128_opt>(y);
-  return static_cast<uint64_t>(p >> 64);
-#elif defined(_MSC_VER) && defined(_M_X64)
-  return __umulh(x, y);
-#else
-  return umul128(x, y).high();
-#endif
-}
-
-// Computes upper 128 bits of multiplication of a 64-bit unsigned integer and a
-// 128-bit unsigned integer.
-inline uint128_fallback umul192_upper128(uint64_t x,
-                                         uint128_fallback y) noexcept {
-  uint128_fallback r = umul128(x, y.high());
-  r += umul128_upper64(x, y.low());
-  return r;
-}
-
 // Computes upper 64 bits of multiplication of a 32-bit unsigned integer and a
 // 64-bit unsigned integer.
-inline uint64_t umul96_upper64(uint32_t x, uint64_t y) noexcept {
+inline auto umul96_upper64(uint32_t x, uint64_t y) noexcept -> uint64_t {
   return umul128_upper64(static_cast<uint64_t>(x) << 32, y);
 }
 
 // Computes lower 128 bits of multiplication of a 64-bit unsigned integer and a
 // 128-bit unsigned integer.
-inline uint128_fallback umul192_lower128(uint64_t x,
-                                         uint128_fallback y) noexcept {
+inline auto umul192_lower128(uint64_t x, uint128_fallback y) noexcept
+    -> uint128_fallback {
   uint64_t high = x * y.high();
   uint128_fallback high_low = umul128(x, y.low());
   return {high + high_low.high(), high_low.low()};
@@ -212,29 +186,17 @@ inline uint128_fallback umul192_lower128(uint64_t x,
 
 // Computes lower 64 bits of multiplication of a 32-bit unsigned integer and a
 // 64-bit unsigned integer.
-inline uint64_t umul96_lower64(uint32_t x, uint64_t y) noexcept {
+inline auto umul96_lower64(uint32_t x, uint64_t y) noexcept -> uint64_t {
   return x * y;
 }
 
-// Computes floor(log10(pow(2, e))) for e in [-2620, 2620] using the method from
-// https://fmt.dev/papers/Dragonbox.pdf#page=28, section 6.1.
-inline int floor_log10_pow2(int e) noexcept {
-  FMT_ASSERT(e <= 2620 && e >= -2620, "too large exponent");
-  static_assert((-1 >> 1) == -1, "right shift is not arithmetic");
-  return (e * 315653) >> 20;
-}
-
 // Various fast log computations.
-inline int floor_log2_pow10(int e) noexcept {
-  FMT_ASSERT(e <= 1233 && e >= -1233, "too large exponent");
-  return (e * 1741647) >> 19;
-}
-inline int floor_log10_pow2_minus_log10_4_over_3(int e) noexcept {
+inline auto floor_log10_pow2_minus_log10_4_over_3(int e) noexcept -> int {
   FMT_ASSERT(e <= 2936 && e >= -2985, "too large exponent");
   return (e * 631305 - 261663) >> 21;
 }
 
-static constexpr struct {
+FMT_INLINE_VARIABLE constexpr struct {
   uint32_t divisor;
   int shift_amount;
 } div_small_pow10_infos[] = {{10, 16}, {100, 16}};
@@ -243,7 +205,7 @@ static constexpr struct {
 // divisible by pow(10, N).
 // Precondition: n <= pow(10, N + 1).
 template <int N>
-bool check_divisibility_and_divide_by_pow10(uint32_t& n) noexcept {
+auto check_divisibility_and_divide_by_pow10(uint32_t& n) noexcept -> bool {
   // The numbers below are chosen such that:
   //   1. floor(n/d) = floor(nm / 2^k) where d=10 or d=100,
   //   2. nm mod 2^k < m if and only if n is divisible by d,
@@ -268,7 +230,7 @@ bool check_divisibility_and_divide_by_pow10(uint32_t& n) noexcept {
 
 // Computes floor(n / pow(10, N)) for small n and N.
 // Precondition: n <= pow(10, N + 1).
-template <int N> uint32_t small_division_by_pow10(uint32_t n) noexcept {
+template <int N> auto small_division_by_pow10(uint32_t n) noexcept -> uint32_t {
   constexpr auto info = div_small_pow10_infos[N - 1];
   FMT_ASSERT(n <= info.divisor * 10, "n is too large");
   constexpr uint32_t magic_number =
@@ -277,24 +239,24 @@ template <int N> uint32_t small_division_by_pow10(uint32_t n) noexcept {
 }
 
 // Computes floor(n / 10^(kappa + 1)) (float)
-inline uint32_t divide_by_10_to_kappa_plus_1(uint32_t n) noexcept {
+inline auto divide_by_10_to_kappa_plus_1(uint32_t n) noexcept -> uint32_t {
   // 1374389535 = ceil(2^37/100)
   return static_cast<uint32_t>((static_cast<uint64_t>(n) * 1374389535) >> 37);
 }
 // Computes floor(n / 10^(kappa + 1)) (double)
-inline uint64_t divide_by_10_to_kappa_plus_1(uint64_t n) noexcept {
+inline auto divide_by_10_to_kappa_plus_1(uint64_t n) noexcept -> uint64_t {
   // 2361183241434822607 = ceil(2^(64+7)/1000)
   return umul128_upper64(n, 2361183241434822607ull) >> 7;
 }
 
 // Various subroutines using pow10 cache
-template <class T> struct cache_accessor;
+template <typename T> struct cache_accessor;
 
 template <> struct cache_accessor<float> {
   using carrier_uint = float_info<float>::carrier_uint;
   using cache_entry_type = uint64_t;
 
-  static uint64_t get_cached_power(int k) noexcept {
+  static auto get_cached_power(int k) noexcept -> uint64_t {
     FMT_ASSERT(k >= float_info<float>::min_k && k <= float_info<float>::max_k,
                "k is out of range");
     static constexpr const uint64_t pow10_significands[] = {
@@ -336,20 +298,23 @@ template <> struct cache_accessor<float> {
     bool is_integer;
   };
 
-  static compute_mul_result compute_mul(
-      carrier_uint u, const cache_entry_type& cache) noexcept {
+  static auto compute_mul(carrier_uint u,
+                          const cache_entry_type& cache) noexcept
+      -> compute_mul_result {
     auto r = umul96_upper64(u, cache);
     return {static_cast<carrier_uint>(r >> 32),
             static_cast<carrier_uint>(r) == 0};
   }
 
-  static uint32_t compute_delta(const cache_entry_type& cache,
-                                int beta) noexcept {
+  static auto compute_delta(const cache_entry_type& cache, int beta) noexcept
+      -> uint32_t {
     return static_cast<uint32_t>(cache >> (64 - 1 - beta));
   }
 
-  static compute_mul_parity_result compute_mul_parity(
-      carrier_uint two_f, const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_mul_parity(carrier_uint two_f,
+                                 const cache_entry_type& cache,
+                                 int beta) noexcept
+      -> compute_mul_parity_result {
     FMT_ASSERT(beta >= 1, "");
     FMT_ASSERT(beta < 64, "");
 
@@ -358,22 +323,22 @@ template <> struct cache_accessor<float> {
             static_cast<uint32_t>(r >> (32 - beta)) == 0};
   }
 
-  static carrier_uint compute_left_endpoint_for_shorter_interval_case(
-      const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_left_endpoint_for_shorter_interval_case(
+      const cache_entry_type& cache, int beta) noexcept -> carrier_uint {
     return static_cast<carrier_uint>(
         (cache - (cache >> (num_significand_bits<float>() + 2))) >>
         (64 - num_significand_bits<float>() - 1 - beta));
   }
 
-  static carrier_uint compute_right_endpoint_for_shorter_interval_case(
-      const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_right_endpoint_for_shorter_interval_case(
+      const cache_entry_type& cache, int beta) noexcept -> carrier_uint {
     return static_cast<carrier_uint>(
         (cache + (cache >> (num_significand_bits<float>() + 1))) >>
         (64 - num_significand_bits<float>() - 1 - beta));
   }
 
-  static carrier_uint compute_round_up_for_shorter_interval_case(
-      const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_round_up_for_shorter_interval_case(
+      const cache_entry_type& cache, int beta) noexcept -> carrier_uint {
     return (static_cast<carrier_uint>(
                 cache >> (64 - num_significand_bits<float>() - 2 - beta)) +
             1) /
@@ -385,7 +350,7 @@ template <> struct cache_accessor<double> {
   using carrier_uint = float_info<double>::carrier_uint;
   using cache_entry_type = uint128_fallback;
 
-  static uint128_fallback get_cached_power(int k) noexcept {
+  static auto get_cached_power(int k) noexcept -> uint128_fallback {
     FMT_ASSERT(k >= float_info<double>::min_k && k <= float_info<double>::max_k,
                "k is out of range");
 
@@ -1009,8 +974,22 @@ template <> struct cache_accessor<double> {
       {0xfcf62c1dee382c42, 0x46729e03dd9ed7b6},
       {0x9e19db92b4e31ba9, 0x6c07a2c26a8346d2},
       {0xc5a05277621be293, 0xc7098b7305241886},
-      { 0xf70867153aa2db38,
-        0xb8cbee4fc66d1ea8 }
+      {0xf70867153aa2db38, 0xb8cbee4fc66d1ea8},
+      {0x9a65406d44a5c903, 0x737f74f1dc043329},
+      {0xc0fe908895cf3b44, 0x505f522e53053ff3},
+      {0xf13e34aabb430a15, 0x647726b9e7c68ff0},
+      {0x96c6e0eab509e64d, 0x5eca783430dc19f6},
+      {0xbc789925624c5fe0, 0xb67d16413d132073},
+      {0xeb96bf6ebadf77d8, 0xe41c5bd18c57e890},
+      {0x933e37a534cbaae7, 0x8e91b962f7b6f15a},
+      {0xb80dc58e81fe95a1, 0x723627bbb5a4adb1},
+      {0xe61136f2227e3b09, 0xcec3b1aaa30dd91d},
+      {0x8fcac257558ee4e6, 0x213a4f0aa5e8a7b2},
+      {0xb3bd72ed2af29e1f, 0xa988e2cd4f62d19e},
+      {0xe0accfa875af45a7, 0x93eb1b80a33b8606},
+      {0x8c6c01c9498d8b88, 0xbc72f130660533c4},
+      {0xaf87023b9bf0ee6a, 0xeb8fad7c7f8680b5},
+      {0xdb68c2ca82ed2a05, 0xa67398db9f6820e2},
 #else
       {0xff77b1fcbebcdc4f, 0x25e8e89c13bb0f7b},
       {0xce5d73ff402d98e3, 0xfb0a3d212dc81290},
@@ -1034,8 +1013,8 @@ template <> struct cache_accessor<double> {
       {0x8da471a9de737e24, 0x5ceaecfed289e5d3},
       {0xe4d5e82392a40515, 0x0fabaf3feaa5334b},
       {0xb8da1662e7b00a17, 0x3d6a751f3b936244},
-      { 0x95527a5202df0ccb,
-        0x0f37801e0c43ebc9 }
+      {0x95527a5202df0ccb, 0x0f37801e0c43ebc9},
+      {0xf13e34aabb430a15, 0x647726b9e7c68ff0}
 #endif
     };
 
@@ -1095,19 +1074,22 @@ template <> struct cache_accessor<double> {
     bool is_integer;
   };
 
-  static compute_mul_result compute_mul(
-      carrier_uint u, const cache_entry_type& cache) noexcept {
+  static auto compute_mul(carrier_uint u,
+                          const cache_entry_type& cache) noexcept
+      -> compute_mul_result {
     auto r = umul192_upper128(u, cache);
     return {r.high(), r.low() == 0};
   }
 
-  static uint32_t compute_delta(cache_entry_type const& cache,
-                                int beta) noexcept {
+  static auto compute_delta(cache_entry_type const& cache, int beta) noexcept
+      -> uint32_t {
     return static_cast<uint32_t>(cache.high() >> (64 - 1 - beta));
   }
 
-  static compute_mul_parity_result compute_mul_parity(
-      carrier_uint two_f, const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_mul_parity(carrier_uint two_f,
+                                 const cache_entry_type& cache,
+                                 int beta) noexcept
+      -> compute_mul_parity_result {
     FMT_ASSERT(beta >= 1, "");
     FMT_ASSERT(beta < 64, "");
 
@@ -1116,31 +1098,35 @@ template <> struct cache_accessor<double> {
             ((r.high() << beta) | (r.low() >> (64 - beta))) == 0};
   }
 
-  static carrier_uint compute_left_endpoint_for_shorter_interval_case(
-      const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_left_endpoint_for_shorter_interval_case(
+      const cache_entry_type& cache, int beta) noexcept -> carrier_uint {
     return (cache.high() -
             (cache.high() >> (num_significand_bits<double>() + 2))) >>
            (64 - num_significand_bits<double>() - 1 - beta);
   }
 
-  static carrier_uint compute_right_endpoint_for_shorter_interval_case(
-      const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_right_endpoint_for_shorter_interval_case(
+      const cache_entry_type& cache, int beta) noexcept -> carrier_uint {
     return (cache.high() +
             (cache.high() >> (num_significand_bits<double>() + 1))) >>
            (64 - num_significand_bits<double>() - 1 - beta);
   }
 
-  static carrier_uint compute_round_up_for_shorter_interval_case(
-      const cache_entry_type& cache, int beta) noexcept {
+  static auto compute_round_up_for_shorter_interval_case(
+      const cache_entry_type& cache, int beta) noexcept -> carrier_uint {
     return ((cache.high() >> (64 - num_significand_bits<double>() - 2 - beta)) +
             1) /
            2;
   }
 };
 
+FMT_FUNC auto get_cached_power(int k) noexcept -> uint128_fallback {
+  return cache_accessor<double>::get_cached_power(k);
+}
+
 // Various integer checks
-template <class T>
-bool is_left_endpoint_integer_shorter_interval(int exponent) noexcept {
+template <typename T>
+auto is_left_endpoint_integer_shorter_interval(int exponent) noexcept -> bool {
   const int case_shorter_interval_left_endpoint_lower_threshold = 2;
   const int case_shorter_interval_left_endpoint_upper_threshold = 3;
   return exponent >= case_shorter_interval_left_endpoint_lower_threshold &&
@@ -1148,12 +1134,12 @@ bool is_left_endpoint_integer_shorter_interval(int exponent) noexcept {
 }
 
 // Remove trailing zeros from n and return the number of zeros removed (float)
-FMT_INLINE int remove_trailing_zeros(uint32_t& n) noexcept {
+FMT_INLINE int remove_trailing_zeros(uint32_t& n, int s = 0) noexcept {
   FMT_ASSERT(n != 0, "");
-  const uint32_t mod_inv_5 = 0xcccccccd;
-  const uint32_t mod_inv_25 = mod_inv_5 * mod_inv_5;
+  // Modular inverse of 5 (mod 2^32): (mod_inv_5 * 5) mod 2^32 = 1.
+  constexpr uint32_t mod_inv_5 = 0xcccccccd;
+  constexpr uint32_t mod_inv_25 = 0xc28f5c29;  // = mod_inv_5 * mod_inv_5
 
-  int s = 0;
   while (true) {
     auto q = rotr(n * mod_inv_25, 2);
     if (q > max_value<uint32_t>() / 100) break;
@@ -1165,7 +1151,6 @@ FMT_INLINE int remove_trailing_zeros(uint32_t& n) noexcept {
     n = q;
     s |= 1;
   }
-
   return s;
 }
 
@@ -1179,32 +1164,17 @@ FMT_INLINE int remove_trailing_zeros(uint64_t& n) noexcept {
 
   // Is n is divisible by 10^8?
   if ((nm.high() & ((1ull << (90 - 64)) - 1)) == 0 && nm.low() < magic_number) {
-    // If yes, work with the quotient.
+    // If yes, work with the quotient...
     auto n32 = static_cast<uint32_t>(nm.high() >> (90 - 64));
-
-    const uint32_t mod_inv_5 = 0xcccccccd;
-    const uint32_t mod_inv_25 = mod_inv_5 * mod_inv_5;
-
-    int s = 8;
-    while (true) {
-      auto q = rotr(n32 * mod_inv_25, 2);
-      if (q > max_value<uint32_t>() / 100) break;
-      n32 = q;
-      s += 2;
-    }
-    auto q = rotr(n32 * mod_inv_5, 1);
-    if (q <= max_value<uint32_t>() / 10) {
-      n32 = q;
-      s |= 1;
-    }
-
+    // ... and use the 32 bit variant of the function
+    int s = remove_trailing_zeros(n32, 8);
     n = n32;
     return s;
   }
 
   // If n is not divisible by 10^8, work with n itself.
-  const uint64_t mod_inv_5 = 0xcccccccccccccccd;
-  const uint64_t mod_inv_25 = mod_inv_5 * mod_inv_5;
+  constexpr uint64_t mod_inv_5 = 0xcccccccccccccccd;
+  constexpr uint64_t mod_inv_25 = 0x8f5c28f5c28f5c29;  // mod_inv_5 * mod_inv_5
 
   int s = 0;
   while (true) {
@@ -1223,7 +1193,7 @@ FMT_INLINE int remove_trailing_zeros(uint64_t& n) noexcept {
 }
 
 // The main algorithm for shorter interval case
-template <class T>
+template <typename T>
 FMT_INLINE decimal_fp<T> shorter_interval_case(int exponent) noexcept {
   decimal_fp<T> ret_value;
   // Compute k and beta
@@ -1270,7 +1240,7 @@ FMT_INLINE decimal_fp<T> shorter_interval_case(int exponent) noexcept {
   return ret_value;
 }
 
-template <typename T> decimal_fp<T> to_decimal(T x) noexcept {
+template <typename T> auto to_decimal(T x) noexcept -> decimal_fp<T> {
   // Step 1: integer promotion & Schubfach multiplier calculation.
 
   using carrier_uint = typename float_info<T>::carrier_uint;
@@ -1394,17 +1364,6 @@ small_divisor_case_label:
   return ret_value;
 }
 }  // namespace dragonbox
-
-#ifdef _MSC_VER
-FMT_FUNC auto fmt_snprintf(char* buf, size_t size, const char* fmt, ...)
-    -> int {
-  auto args = va_list();
-  va_start(args, fmt);
-  int result = vsnprintf_s(buf, size, _TRUNCATE, fmt, args);
-  va_end(args);
-  return result;
-}
-#endif
 }  // namespace detail
 
 template <> struct formatter<detail::bigint> {
@@ -1413,23 +1372,22 @@ template <> struct formatter<detail::bigint> {
     return ctx.begin();
   }
 
-  template <typename FormatContext>
-  auto format(const detail::bigint& n, FormatContext& ctx) const ->
-      typename FormatContext::iterator {
+  auto format(const detail::bigint& n, format_context& ctx) const
+      -> format_context::iterator {
     auto out = ctx.out();
     bool first = true;
     for (auto i = n.bigits_.size(); i > 0; --i) {
       auto value = n.bigits_[i - 1u];
       if (first) {
-        out = format_to(out, FMT_STRING("{:x}"), value);
+        out = fmt::format_to(out, FMT_STRING("{:x}"), value);
         first = false;
         continue;
       }
-      out = format_to(out, FMT_STRING("{:08x}"), value);
+      out = fmt::format_to(out, FMT_STRING("{:08x}"), value);
     }
     if (n.exp_ > 0)
-      out = format_to(out, FMT_STRING("p{}"),
-                      n.exp_ * detail::bigint::bigit_bits);
+      out = fmt::format_to(out, FMT_STRING("p{}"),
+                           n.exp_ * detail::bigint::bigit_bits);
     return out;
   }
 };
@@ -1465,7 +1423,7 @@ FMT_FUNC void report_system_error(int error_code,
   report_error(format_system_error, error_code, message);
 }
 
-FMT_FUNC std::string vformat(string_view fmt, format_args args) {
+FMT_FUNC auto vformat(string_view fmt, format_args args) -> std::string {
   // Don't optimize the "{}" case to keep the binary size small and because it
   // can be better optimized in fmt::format anyway.
   auto buffer = memory_buffer();
@@ -1474,57 +1432,54 @@ FMT_FUNC std::string vformat(string_view fmt, format_args args) {
 }
 
 namespace detail {
-#ifdef _WIN32
+#if !defined(_WIN32) || defined(FMT_WINDOWS_NO_WCHAR)
+FMT_FUNC auto write_console(int, string_view) -> bool { return false; }
+FMT_FUNC auto write_console(std::FILE*, string_view) -> bool { return false; }
+#else
 using dword = conditional_t<sizeof(long) == 4, unsigned long, unsigned>;
 extern "C" __declspec(dllimport) int __stdcall WriteConsoleW(  //
     void*, const void*, dword, dword*, void*);
 
-FMT_FUNC bool write_console(std::FILE* f, string_view text) {
-  auto fd = _fileno(f);
-  if (_isatty(fd)) {
-    detail::utf8_to_utf16 u16(string_view(text.data(), text.size()));
-    auto written = detail::dword();
-    if (detail::WriteConsoleW(reinterpret_cast<void*>(_get_osfhandle(fd)),
-                              u16.c_str(), static_cast<uint32_t>(u16.size()),
-                              &written, nullptr)) {
-      return true;
-    }
-  }
-  // We return false if the file descriptor was not TTY, or it was but
-  // SetConsoleW failed which can happen if the output has been redirected to
-  // NUL. In both cases when we return false, we should attempt to do regular
-  // write via fwrite or std::ostream::write.
-  return false;
+FMT_FUNC bool write_console(int fd, string_view text) {
+  auto u16 = utf8_to_utf16(text);
+  return WriteConsoleW(reinterpret_cast<void*>(_get_osfhandle(fd)), u16.c_str(),
+                       static_cast<dword>(u16.size()), nullptr, nullptr) != 0;
+}
+
+FMT_FUNC auto write_console(std::FILE* f, string_view text) -> bool {
+  return write_console(_fileno(f), text);
+}
+#endif
+
+#ifdef _WIN32
+// Print assuming legacy (non-Unicode) encoding.
+FMT_FUNC void vprint_mojibake(std::FILE* f, string_view fmt, format_args args) {
+  auto buffer = memory_buffer();
+  detail::vformat_to(buffer, fmt, args);
+  fwrite_fully(buffer.data(), buffer.size(), f);
 }
 #endif
 
 FMT_FUNC void print(std::FILE* f, string_view text) {
 #ifdef _WIN32
-  if (write_console(f, text)) return;
+  int fd = _fileno(f);
+  if (_isatty(fd)) {
+    std::fflush(f);
+    if (write_console(fd, text)) return;
+  }
 #endif
-  detail::fwrite_fully(text.data(), 1, text.size(), f);
+  fwrite_fully(text.data(), text.size(), f);
 }
 }  // namespace detail
 
-FMT_FUNC void vprint(std::FILE* f, string_view format_str, format_args args) {
-  memory_buffer buffer;
-  detail::vformat_to(buffer, format_str, args);
+FMT_FUNC void vprint(std::FILE* f, string_view fmt, format_args args) {
+  auto buffer = memory_buffer();
+  detail::vformat_to(buffer, fmt, args);
   detail::print(f, {buffer.data(), buffer.size()});
 }
 
-#ifdef _WIN32
-// Print assuming legacy (non-Unicode) encoding.
-FMT_FUNC void detail::vprint_mojibake(std::FILE* f, string_view format_str,
-                                      format_args args) {
-  memory_buffer buffer;
-  detail::vformat_to(buffer, format_str,
-                     basic_format_args<buffer_context<char>>(args));
-  fwrite_fully(buffer.data(), 1, buffer.size(), f);
-}
-#endif
-
-FMT_FUNC void vprint(string_view format_str, format_args args) {
-  vprint(stdout, format_str, args);
+FMT_FUNC void vprint(string_view fmt, format_args args) {
+  vprint(stdout, fmt, args);
 }
 
 namespace detail {
diff --git a/dep/fmt/include/fmt/format.h b/dep/fmt/include/fmt/format.h
index 7c607dbd304..7637c8a0d06 100644
--- a/dep/fmt/include/fmt/format.h
+++ b/dep/fmt/include/fmt/format.h
@@ -33,30 +33,71 @@
 #ifndef FMT_FORMAT_H_
 #define FMT_FORMAT_H_
 
-#include <cmath>         // std::signbit
-#include <cstdint>       // uint32_t
-#include <cstring>       // std::memcpy
-#include <limits>        // std::numeric_limits
-#include <memory>        // std::uninitialized_copy
-#include <stdexcept>     // std::runtime_error
-#include <system_error>  // std::system_error
+#include <cmath>             // std::signbit
+#include <cstdint>           // uint32_t
+#include <cstring>           // std::memcpy
+#include <initializer_list>  // std::initializer_list
+#include <limits>            // std::numeric_limits
+#include <memory>            // std::uninitialized_copy
+#include <stdexcept>         // std::runtime_error
+#include <system_error>      // std::system_error
 
 #ifdef __cpp_lib_bit_cast
-#  include <bit>  // std::bitcast
+#  include <bit>  // std::bit_cast
 #endif
 
 #include "core.h"
 
-#if FMT_GCC_VERSION
-#  define FMT_GCC_VISIBILITY_HIDDEN __attribute__((visibility("hidden")))
+#if defined __cpp_inline_variables && __cpp_inline_variables >= 201606L
+#  define FMT_INLINE_VARIABLE inline
 #else
-#  define FMT_GCC_VISIBILITY_HIDDEN
+#  define FMT_INLINE_VARIABLE
 #endif
 
-#ifdef __NVCC__
-#  define FMT_CUDA_VERSION (__CUDACC_VER_MAJOR__ * 100 + __CUDACC_VER_MINOR__)
+#if FMT_HAS_CPP17_ATTRIBUTE(fallthrough)
+#  define FMT_FALLTHROUGH [[fallthrough]]
+#elif defined(__clang__)
+#  define FMT_FALLTHROUGH [[clang::fallthrough]]
+#elif FMT_GCC_VERSION >= 700 && \
+    (!defined(__EDG_VERSION__) || __EDG_VERSION__ >= 520)
+#  define FMT_FALLTHROUGH [[gnu::fallthrough]]
 #else
-#  define FMT_CUDA_VERSION 0
+#  define FMT_FALLTHROUGH
+#endif
+
+#ifndef FMT_DEPRECATED
+#  if FMT_HAS_CPP14_ATTRIBUTE(deprecated) || FMT_MSC_VERSION >= 1900
+#    define FMT_DEPRECATED [[deprecated]]
+#  else
+#    if (defined(__GNUC__) && !defined(__LCC__)) || defined(__clang__)
+#      define FMT_DEPRECATED __attribute__((deprecated))
+#    elif FMT_MSC_VERSION
+#      define FMT_DEPRECATED __declspec(deprecated)
+#    else
+#      define FMT_DEPRECATED /* deprecated */
+#    endif
+#  endif
+#endif
+
+#ifndef FMT_NO_UNIQUE_ADDRESS
+#  if FMT_CPLUSPLUS >= 202002L
+#    if FMT_HAS_CPP_ATTRIBUTE(no_unique_address)
+#      define FMT_NO_UNIQUE_ADDRESS [[no_unique_address]]
+// VS2019 v16.10 and later except clang-cl (https://reviews.llvm.org/D110485)
+#    elif (FMT_MSC_VERSION >= 1929) && !FMT_CLANG_VERSION
+#      define FMT_NO_UNIQUE_ADDRESS [[msvc::no_unique_address]]
+#    endif
+#  endif
+#endif
+#ifndef FMT_NO_UNIQUE_ADDRESS
+#  define FMT_NO_UNIQUE_ADDRESS
+#endif
+
+// Visibility when compiled as a shared library/object.
+#if defined(FMT_LIB_EXPORT) || defined(FMT_SHARED)
+#  define FMT_SO_VISIBILITY(value) FMT_VISIBILITY(value)
+#else
+#  define FMT_SO_VISIBILITY(value)
 #endif
 
 #ifdef __has_builtin
@@ -71,12 +112,6 @@
 #  define FMT_NOINLINE
 #endif
 
-#if FMT_MSC_VERSION
-#  define FMT_MSC_DEFAULT = default
-#else
-#  define FMT_MSC_DEFAULT
-#endif
-
 #ifndef FMT_THROW
 #  if FMT_EXCEPTIONS
 #    if FMT_MSC_VERSION || defined(__NVCC__)
@@ -95,10 +130,8 @@ FMT_END_NAMESPACE
 #      define FMT_THROW(x) throw x
 #    endif
 #  else
-#    define FMT_THROW(x)               \
-      do {                             \
-        FMT_ASSERT(false, (x).what()); \
-      } while (false)
+#    define FMT_THROW(x) \
+      ::fmt::detail::assert_fail(__FILE__, __LINE__, (x).what())
 #  endif
 #endif
 
@@ -120,7 +153,10 @@ FMT_END_NAMESPACE
 
 #ifndef FMT_USE_USER_DEFINED_LITERALS
 // EDG based compilers (Intel, NVIDIA, Elbrus, etc), GCC and MSVC support UDLs.
-#  if (FMT_HAS_FEATURE(cxx_user_literals) || FMT_GCC_VERSION >= 407 || \
+//
+// GCC before 4.9 requires a space in `operator"" _a` which is invalid in later
+// compiler versions.
+#  if (FMT_HAS_FEATURE(cxx_user_literals) || FMT_GCC_VERSION >= 409 || \
        FMT_MSC_VERSION >= 1900) &&                                     \
       (!defined(__EDG_VERSION__) || __EDG_VERSION__ >= /* UDL feature */ 480)
 #    define FMT_USE_USER_DEFINED_LITERALS 1
@@ -200,7 +236,8 @@ inline auto clzll(uint64_t x) -> int {
   _BitScanReverse64(&r, x);
 #  else
   // Scan the high 32 bits.
-  if (_BitScanReverse(&r, static_cast<uint32_t>(x >> 32))) return 63 ^ (r + 32);
+  if (_BitScanReverse(&r, static_cast<uint32_t>(x >> 32)))
+    return 63 ^ static_cast<int>(r + 32);
   // Scan the low 32 bits.
   _BitScanReverse(&r, static_cast<uint32_t>(x));
 #  endif
@@ -261,37 +298,6 @@ template <typename CharT, CharT... C>
 constexpr CharT string_literal<CharT, C...>::value[sizeof...(C)];
 #endif
 
-template <typename Streambuf> class formatbuf : public Streambuf {
- private:
-  using char_type = typename Streambuf::char_type;
-  using streamsize = decltype(std::declval<Streambuf>().sputn(nullptr, 0));
-  using int_type = typename Streambuf::int_type;
-  using traits_type = typename Streambuf::traits_type;
-
-  buffer<char_type>& buffer_;
-
- public:
-  explicit formatbuf(buffer<char_type>& buf) : buffer_(buf) {}
-
- protected:
-  // The put area is always empty. This makes the implementation simpler and has
-  // the advantage that the streambuf and the buffer are always in sync and
-  // sputc never writes into uninitialized memory. A disadvantage is that each
-  // call to sputc always results in a (virtual) call to overflow. There is no
-  // disadvantage here for sputn since this always results in a call to xsputn.
-
-  auto overflow(int_type ch) -> int_type override {
-    if (!traits_type::eq_int_type(ch, traits_type::eof()))
-      buffer_.push_back(static_cast<char_type>(ch));
-    return ch;
-  }
-
-  auto xsputn(const char_type* s, streamsize count) -> streamsize override {
-    buffer_.append(s, s + count);
-    return count;
-  }
-};
-
 // Implementation of std::bit_cast for pre-C++20.
 template <typename To, typename From, FMT_ENABLE_IF(sizeof(To) == sizeof(From))>
 FMT_CONSTEXPR20 auto bit_cast(const From& from) -> To {
@@ -323,14 +329,12 @@ class uint128_fallback {
  private:
   uint64_t lo_, hi_;
 
-  friend uint128_fallback umul128(uint64_t x, uint64_t y) noexcept;
-
  public:
   constexpr uint128_fallback(uint64_t hi, uint64_t lo) : lo_(lo), hi_(hi) {}
   constexpr uint128_fallback(uint64_t value = 0) : lo_(value), hi_(0) {}
 
-  constexpr uint64_t high() const noexcept { return hi_; }
-  constexpr uint64_t low() const noexcept { return lo_; }
+  constexpr auto high() const noexcept -> uint64_t { return hi_; }
+  constexpr auto low() const noexcept -> uint64_t { return lo_; }
 
   template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
   constexpr explicit operator T() const {
@@ -359,6 +363,10 @@ class uint128_fallback {
       -> uint128_fallback {
     return {lhs.hi_ & rhs.hi_, lhs.lo_ & rhs.lo_};
   }
+  friend constexpr auto operator~(const uint128_fallback& n)
+      -> uint128_fallback {
+    return {~n.hi_, ~n.lo_};
+  }
   friend auto operator+(const uint128_fallback& lhs,
                         const uint128_fallback& rhs) -> uint128_fallback {
     auto result = uint128_fallback(lhs);
@@ -397,8 +405,12 @@ class uint128_fallback {
     lo_ = new_lo;
     hi_ = new_hi;
   }
+  FMT_CONSTEXPR void operator&=(uint128_fallback n) {
+    lo_ &= n.lo_;
+    hi_ &= n.hi_;
+  }
 
-  FMT_CONSTEXPR20 uint128_fallback& operator+=(uint64_t n) noexcept {
+  FMT_CONSTEXPR20 auto operator+=(uint64_t n) noexcept -> uint128_fallback& {
     if (is_constant_evaluated()) {
       lo_ += n;
       hi_ += (lo_ < n ? 1 : 0);
@@ -463,10 +475,34 @@ inline auto bit_cast(const From& from) -> To {
   return result;
 }
 
+template <typename UInt>
+FMT_CONSTEXPR20 inline auto countl_zero_fallback(UInt n) -> int {
+  int lz = 0;
+  constexpr UInt msb_mask = static_cast<UInt>(1) << (num_bits<UInt>() - 1);
+  for (; (n & msb_mask) == 0; n <<= 1) lz++;
+  return lz;
+}
+
+FMT_CONSTEXPR20 inline auto countl_zero(uint32_t n) -> int {
+#ifdef FMT_BUILTIN_CLZ
+  if (!is_constant_evaluated()) return FMT_BUILTIN_CLZ(n);
+#endif
+  return countl_zero_fallback(n);
+}
+
+FMT_CONSTEXPR20 inline auto countl_zero(uint64_t n) -> int {
+#ifdef FMT_BUILTIN_CLZLL
+  if (!is_constant_evaluated()) return FMT_BUILTIN_CLZLL(n);
+#endif
+  return countl_zero_fallback(n);
+}
+
 FMT_INLINE void assume(bool condition) {
   (void)condition;
 #if FMT_HAS_BUILTIN(__builtin_assume) && !FMT_ICC_VERSION
   __builtin_assume(condition);
+#elif FMT_GCC_VERSION
+  if (!condition) __builtin_unreachable();
 #endif
 }
 
@@ -485,20 +521,6 @@ inline auto get_data(Container& c) -> typename Container::value_type* {
   return c.data();
 }
 
-#if defined(_SECURE_SCL) && _SECURE_SCL
-// Make a checked iterator to avoid MSVC warnings.
-template <typename T> using checked_ptr = stdext::checked_array_iterator<T*>;
-template <typename T>
-constexpr auto make_checked(T* p, size_t size) -> checked_ptr<T> {
-  return {p, size};
-}
-#else
-template <typename T> using checked_ptr = T*;
-template <typename T> constexpr auto make_checked(T* p, size_t) -> T* {
-  return p;
-}
-#endif
-
 // Attempts to reserve space for n extra characters in the output range.
 // Returns a pointer to the reserved range or a reference to it.
 template <typename Container, FMT_ENABLE_IF(is_contiguous<Container>::value)>
@@ -506,12 +528,12 @@ template <typename Container, FMT_ENABLE_IF(is_contiguous<Container>::value)>
 __attribute__((no_sanitize("undefined")))
 #endif
 inline auto
-reserve(std::back_insert_iterator<Container> it, size_t n)
-    -> checked_ptr<typename Container::value_type> {
+reserve(std::back_insert_iterator<Container> it, size_t n) ->
+    typename Container::value_type* {
   Container& c = get_container(it);
   size_t size = c.size();
   c.resize(size + n);
-  return make_checked(get_data(c) + size, n);
+  return get_data(c) + size;
 }
 
 template <typename T>
@@ -543,8 +565,8 @@ template <typename T> auto to_pointer(buffer_appender<T> it, size_t n) -> T* {
 }
 
 template <typename Container, FMT_ENABLE_IF(is_contiguous<Container>::value)>
-inline auto base_iterator(std::back_insert_iterator<Container>& it,
-                          checked_ptr<typename Container::value_type>)
+inline auto base_iterator(std::back_insert_iterator<Container> it,
+                          typename Container::value_type*)
     -> std::back_insert_iterator<Container> {
   return it;
 }
@@ -607,7 +629,8 @@ FMT_CONSTEXPR inline auto utf8_decode(const char* s, uint32_t* c, int* e)
   constexpr const int shiftc[] = {0, 18, 12, 6, 0};
   constexpr const int shifte[] = {0, 6, 4, 2, 0};
 
-  int len = code_point_length_impl(*s);
+  int len = "\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\1\0\0\0\0\0\0\0\0\2\2\2\2\3\3\4"
+      [static_cast<unsigned char>(*s) >> 3];
   // Compute the pointer to the next character early so that the next
   // iteration can start working on the next character. Neither Clang
   // nor GCC figure out this reordering on their own.
@@ -636,7 +659,7 @@ FMT_CONSTEXPR inline auto utf8_decode(const char* s, uint32_t* c, int* e)
   return next;
 }
 
-constexpr uint32_t invalid_code_point = ~uint32_t();
+constexpr FMT_INLINE_VARIABLE uint32_t invalid_code_point = ~uint32_t();
 
 // Invokes f(cp, sv) for every code point cp in s with sv being the string view
 // corresponding to the code point. cp is invalid_code_point on error.
@@ -677,7 +700,7 @@ inline auto compute_width(basic_string_view<Char> s) -> size_t {
 }
 
 // Computes approximate display width of a UTF-8 string.
-FMT_CONSTEXPR inline size_t compute_width(string_view s) {
+FMT_CONSTEXPR inline auto compute_width(string_view s) -> size_t {
   size_t num_code_points = 0;
   // It is not a lambda for compatibility with C++14.
   struct count_code_points {
@@ -706,6 +729,7 @@ FMT_CONSTEXPR inline size_t compute_width(string_view s) {
       return true;
     }
   };
+  // We could avoid branches by using utf8_decode directly.
   for_each_codepoint(s, count_code_points{&num_code_points});
   return num_code_points;
 }
@@ -723,12 +747,17 @@ inline auto code_point_index(basic_string_view<Char> s, size_t n) -> size_t {
 
 // Calculates the index of the nth code point in a UTF-8 string.
 inline auto code_point_index(string_view s, size_t n) -> size_t {
-  const char* data = s.data();
-  size_t num_code_points = 0;
-  for (size_t i = 0, size = s.size(); i != size; ++i) {
-    if ((data[i] & 0xc0) != 0x80 && ++num_code_points > n) return i;
-  }
-  return s.size();
+  size_t result = s.size();
+  const char* begin = s.begin();
+  for_each_codepoint(s, [begin, &n, &result](uint32_t, string_view sv) {
+    if (n != 0) {
+      --n;
+      return true;
+    }
+    result = to_unsigned(sv.begin() - begin);
+    return false;
+  });
+  return result;
 }
 
 inline auto code_point_index(basic_string_view<char8_type> s, size_t n)
@@ -737,13 +766,48 @@ inline auto code_point_index(basic_string_view<char8_type> s, size_t n)
       string_view(reinterpret_cast<const char*>(s.data()), s.size()), n);
 }
 
+template <typename T> struct is_integral : std::is_integral<T> {};
+template <> struct is_integral<int128_opt> : std::true_type {};
+template <> struct is_integral<uint128_t> : std::true_type {};
+
+template <typename T>
+using is_signed =
+    std::integral_constant<bool, std::numeric_limits<T>::is_signed ||
+                                     std::is_same<T, int128_opt>::value>;
+
+template <typename T>
+using is_integer =
+    bool_constant<is_integral<T>::value && !std::is_same<T, bool>::value &&
+                  !std::is_same<T, char>::value &&
+                  !std::is_same<T, wchar_t>::value>;
+
+#ifndef FMT_USE_FLOAT
+#  define FMT_USE_FLOAT 1
+#endif
+#ifndef FMT_USE_DOUBLE
+#  define FMT_USE_DOUBLE 1
+#endif
+#ifndef FMT_USE_LONG_DOUBLE
+#  define FMT_USE_LONG_DOUBLE 1
+#endif
+
 #ifndef FMT_USE_FLOAT128
-#  ifdef __SIZEOF_FLOAT128__
-#    define FMT_USE_FLOAT128 1
-#  else
+#  ifdef __clang__
+// Clang emulates GCC, so it has to appear early.
+#    if FMT_HAS_INCLUDE(<quadmath.h>)
+#      define FMT_USE_FLOAT128 1
+#    endif
+#  elif defined(__GNUC__)
+// GNU C++:
+#    if defined(_GLIBCXX_USE_FLOAT128) && !defined(__STRICT_ANSI__)
+#      define FMT_USE_FLOAT128 1
+#    endif
+#  endif
+#  ifndef FMT_USE_FLOAT128
 #    define FMT_USE_FLOAT128 0
 #  endif
 #endif
+
 #if FMT_USE_FLOAT128
 using float128 = __float128;
 #else
@@ -775,7 +839,7 @@ void buffer<T>::append(const U* begin, const U* end) {
     try_reserve(size_ + count);
     auto free_cap = capacity_ - size_;
     if (free_cap < count) count = free_cap;
-    std::uninitialized_copy_n(begin, count, make_checked(ptr_ + size_, count));
+    std::uninitialized_copy_n(begin, count, ptr_ + size_);
     size_ += count;
     begin += count;
   }
@@ -787,7 +851,7 @@ template <typename T>
 struct is_locale<T, void_t<decltype(T::classic())>> : std::true_type {};
 }  // namespace detail
 
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
 // The number of characters to store in the basic_memory_buffer object itself
 // to avoid dynamic memory allocation.
@@ -803,7 +867,7 @@ enum { inline_buffer_size = 500 };
   **Example**::
 
      auto out = fmt::memory_buffer();
-     format_to(std::back_inserter(out), "The answer is {}.", 42);
+     fmt::format_to(std::back_inserter(out), "The answer is {}.", 42);
 
   This will append the following output to the ``out`` object:
 
@@ -820,8 +884,8 @@ class basic_memory_buffer final : public detail::buffer<T> {
  private:
   T store_[SIZE];
 
-  // Don't inherit from Allocator avoid generating type_info for it.
-  Allocator alloc_;
+  // Don't inherit from Allocator to avoid generating type_info for it.
+  FMT_NO_UNIQUE_ADDRESS Allocator alloc_;
 
   // Deallocate memory allocated by the buffer.
   FMT_CONSTEXPR20 void deallocate() {
@@ -830,7 +894,28 @@ class basic_memory_buffer final : public detail::buffer<T> {
   }
 
  protected:
-  FMT_CONSTEXPR20 void grow(size_t size) override;
+  FMT_CONSTEXPR20 void grow(size_t size) override {
+    detail::abort_fuzzing_if(size > 5000);
+    const size_t max_size = std::allocator_traits<Allocator>::max_size(alloc_);
+    size_t old_capacity = this->capacity();
+    size_t new_capacity = old_capacity + old_capacity / 2;
+    if (size > new_capacity)
+      new_capacity = size;
+    else if (new_capacity > max_size)
+      new_capacity = size > max_size ? size : max_size;
+    T* old_data = this->data();
+    T* new_data =
+        std::allocator_traits<Allocator>::allocate(alloc_, new_capacity);
+    // Suppress a bogus -Wstringop-overflow in gcc 13.1 (#3481).
+    detail::assume(this->size() <= new_capacity);
+    // The following code doesn't throw, so the raw pointer above doesn't leak.
+    std::uninitialized_copy_n(old_data, this->size(), new_data);
+    this->set(new_data, new_capacity);
+    // deallocate must not throw according to the standard, but even if it does,
+    // the buffer already uses the new storage and will deallocate it in
+    // destructor.
+    if (old_data != store_) alloc_.deallocate(old_data, old_capacity);
+  }
 
  public:
   using value_type = T;
@@ -852,8 +937,7 @@ class basic_memory_buffer final : public detail::buffer<T> {
     size_t size = other.size(), capacity = other.capacity();
     if (data == other.store_) {
       this->set(store_, capacity);
-      detail::copy_str<T>(other.store_, other.store_ + size,
-                          detail::make_checked(store_, capacity));
+      detail::copy_str<T>(other.store_, other.store_ + size, store_);
     } else {
       this->set(data, capacity);
       // Set pointer to the inline array so that delete is not called
@@ -899,7 +983,6 @@ class basic_memory_buffer final : public detail::buffer<T> {
   /** Increases the buffer capacity to *new_capacity*. */
   void reserve(size_t new_capacity) { this->try_reserve(new_capacity); }
 
-  // Directly append data into the buffer
   using detail::buffer<T>::append;
   template <typename ContiguousRange>
   void append(const ContiguousRange& range) {
@@ -907,55 +990,30 @@ class basic_memory_buffer final : public detail::buffer<T> {
   }
 };
 
-template <typename T, size_t SIZE, typename Allocator>
-FMT_CONSTEXPR20 void basic_memory_buffer<T, SIZE, Allocator>::grow(
-    size_t size) {
-  detail::abort_fuzzing_if(size > 5000);
-  const size_t max_size = std::allocator_traits<Allocator>::max_size(alloc_);
-  size_t old_capacity = this->capacity();
-  size_t new_capacity = old_capacity + old_capacity / 2;
-  if (size > new_capacity)
-    new_capacity = size;
-  else if (new_capacity > max_size)
-    new_capacity = size > max_size ? size : max_size;
-  T* old_data = this->data();
-  T* new_data =
-      std::allocator_traits<Allocator>::allocate(alloc_, new_capacity);
-  // The following code doesn't throw, so the raw pointer above doesn't leak.
-  std::uninitialized_copy(old_data, old_data + this->size(),
-                          detail::make_checked(new_data, new_capacity));
-  this->set(new_data, new_capacity);
-  // deallocate must not throw according to the standard, but even if it does,
-  // the buffer already uses the new storage and will deallocate it in
-  // destructor.
-  if (old_data != store_) alloc_.deallocate(old_data, old_capacity);
-}
-
 using memory_buffer = basic_memory_buffer<char>;
 
 template <typename T, size_t SIZE, typename Allocator>
 struct is_contiguous<basic_memory_buffer<T, SIZE, Allocator>> : std::true_type {
 };
 
+FMT_END_EXPORT
 namespace detail {
-#ifdef _WIN32
-FMT_API bool write_console(std::FILE* f, string_view text);
-#endif
+FMT_API auto write_console(int fd, string_view text) -> bool;
+FMT_API auto write_console(std::FILE* f, string_view text) -> bool;
 FMT_API void print(std::FILE*, string_view);
 }  // namespace detail
 
-/** A formatting error such as invalid format string. */
-FMT_CLASS_API
-class FMT_API format_error : public std::runtime_error {
+FMT_BEGIN_EXPORT
+
+// Suppress a misleading warning in older versions of clang.
+#if FMT_CLANG_VERSION
+#  pragma clang diagnostic ignored "-Wweak-vtables"
+#endif
+
+/** An error reported from a formatting function. */
+class FMT_SO_VISIBILITY("default") format_error : public std::runtime_error {
  public:
-  explicit format_error(const char* message) : std::runtime_error(message) {}
-  explicit format_error(const std::string& message)
-      : std::runtime_error(message) {}
-  format_error(const format_error&) = default;
-  format_error& operator=(const format_error&) = default;
-  format_error(format_error&&) = default;
-  format_error& operator=(format_error&&) = default;
-  ~format_error() noexcept override FMT_MSC_DEFAULT;
+  using std::runtime_error::runtime_error;
 };
 
 namespace detail_exported {
@@ -984,16 +1042,52 @@ constexpr auto compile_string_to_view(detail::std_string_view<Char> s)
 }
 }  // namespace detail_exported
 
-FMT_BEGIN_DETAIL_NAMESPACE
+class loc_value {
+ private:
+  basic_format_arg<format_context> value_;
 
-template <typename T> struct is_integral : std::is_integral<T> {};
-template <> struct is_integral<int128_opt> : std::true_type {};
-template <> struct is_integral<uint128_t> : std::true_type {};
+ public:
+  template <typename T, FMT_ENABLE_IF(!detail::is_float128<T>::value)>
+  loc_value(T value) : value_(detail::make_arg<format_context>(value)) {}
 
-template <typename T>
-using is_signed =
-    std::integral_constant<bool, std::numeric_limits<T>::is_signed ||
-                                     std::is_same<T, int128_opt>::value>;
+  template <typename T, FMT_ENABLE_IF(detail::is_float128<T>::value)>
+  loc_value(T) {}
+
+  template <typename Visitor> auto visit(Visitor&& vis) -> decltype(vis(0)) {
+    return visit_format_arg(vis, value_);
+  }
+};
+
+// A locale facet that formats values in UTF-8.
+// It is parameterized on the locale to avoid the heavy <locale> include.
+template <typename Locale> class format_facet : public Locale::facet {
+ private:
+  std::string separator_;
+  std::string grouping_;
+  std::string decimal_point_;
+
+ protected:
+  virtual auto do_put(appender out, loc_value val,
+                      const format_specs<>& specs) const -> bool;
+
+ public:
+  static FMT_API typename Locale::id id;
+
+  explicit format_facet(Locale& loc);
+  explicit format_facet(string_view sep = "",
+                        std::initializer_list<unsigned char> g = {3},
+                        std::string decimal_point = ".")
+      : separator_(sep.data(), sep.size()),
+        grouping_(g.begin(), g.end()),
+        decimal_point_(decimal_point) {}
+
+  auto put(appender out, loc_value val, const format_specs<>& specs) const
+      -> bool {
+    return do_put(out, val, specs);
+  }
+};
+
+namespace detail {
 
 // Returns true if value is negative, false otherwise.
 // Same as `value < 0` but doesn't produce warnings if T is an unsigned type.
@@ -1024,13 +1118,13 @@ using uint32_or_64_or_128_t =
 template <typename T>
 using uint64_or_128_t = conditional_t<num_bits<T>() <= 64, uint64_t, uint128_t>;
 
-#define FMT_POWERS_OF_10(factor)                                             \
-  factor * 10, (factor)*100, (factor)*1000, (factor)*10000, (factor)*100000, \
-      (factor)*1000000, (factor)*10000000, (factor)*100000000,               \
-      (factor)*1000000000
+#define FMT_POWERS_OF_10(factor)                                  \
+  factor * 10, (factor) * 100, (factor) * 1000, (factor) * 10000, \
+      (factor) * 100000, (factor) * 1000000, (factor) * 10000000, \
+      (factor) * 100000000, (factor) * 1000000000
 
 // Converts value in the range [0, 100) to a string.
-constexpr const char* digits2(size_t value) {
+constexpr auto digits2(size_t value) -> const char* {
   // GCC generates slightly better code when value is pointer-size.
   return &"0001020304050607080910111213141516171819"
          "2021222324252627282930313233343536373839"
@@ -1040,7 +1134,7 @@ constexpr const char* digits2(size_t value) {
 }
 
 // Sign is a template parameter to workaround a bug in gcc 4.8.
-template <typename Char, typename Sign> constexpr Char sign(Sign s) {
+template <typename Char, typename Sign> constexpr auto sign(Sign s) -> Char {
 #if !FMT_GCC_VERSION || FMT_GCC_VERSION >= 604
   static_assert(std::is_same<Sign, sign_t>::value, "");
 #endif
@@ -1122,7 +1216,7 @@ FMT_CONSTEXPR auto count_digits(UInt n) -> int {
 FMT_INLINE auto do_count_digits(uint32_t n) -> int {
 // An optimization by Kendall Willets from https://bit.ly/3uOIQrB.
 // This increments the upper 32 bits (log10(T) - 1) when >= T is added.
-#  define FMT_INC(T) (((sizeof(#  T) - 1ull) << 32) - T)
+#  define FMT_INC(T) (((sizeof(#T) - 1ull) << 32) - T)
   static constexpr uint64_t table[] = {
       FMT_INC(0),          FMT_INC(0),          FMT_INC(0),           // 8
       FMT_INC(10),         FMT_INC(10),         FMT_INC(10),          // 64
@@ -1238,7 +1332,7 @@ template <typename Char, typename UInt, typename Iterator,
 FMT_CONSTEXPR inline auto format_decimal(Iterator out, UInt value, int size)
     -> format_decimal_result<Iterator> {
   // Buffer is large enough to hold all digits (digits10 + 1).
-  Char buffer[digits10<UInt>() + 1];
+  Char buffer[digits10<UInt>() + 1] = {};
   auto end = format_decimal(buffer, value, size).end;
   return {out, detail::copy_str_noinline<Char>(buffer, end, out)};
 }
@@ -1258,14 +1352,14 @@ FMT_CONSTEXPR auto format_uint(Char* buffer, UInt value, int num_digits,
 }
 
 template <unsigned BASE_BITS, typename Char, typename It, typename UInt>
-inline auto format_uint(It out, UInt value, int num_digits, bool upper = false)
-    -> It {
+FMT_CONSTEXPR inline auto format_uint(It out, UInt value, int num_digits,
+                                      bool upper = false) -> It {
   if (auto ptr = to_pointer<Char>(out, to_unsigned(num_digits))) {
     format_uint<BASE_BITS>(ptr, value, num_digits, upper);
     return out;
   }
   // Buffer should be large enough to hold all digits (digits / BASE_BITS + 1).
-  char buffer[num_bits<UInt>() / BASE_BITS + 1];
+  char buffer[num_bits<UInt>() / BASE_BITS + 1] = {};
   format_uint<BASE_BITS>(buffer, value, num_digits, upper);
   return detail::copy_str_noinline<Char>(buffer, buffer + num_digits, out);
 }
@@ -1283,7 +1377,140 @@ class utf8_to_utf16 {
   auto str() const -> std::wstring { return {&buffer_[0], size()}; }
 };
 
+enum class to_utf8_error_policy { abort, replace };
+
+// A converter from UTF-16/UTF-32 (host endian) to UTF-8.
+template <typename WChar, typename Buffer = memory_buffer> class to_utf8 {
+ private:
+  Buffer buffer_;
+
+ public:
+  to_utf8() {}
+  explicit to_utf8(basic_string_view<WChar> s,
+                   to_utf8_error_policy policy = to_utf8_error_policy::abort) {
+    static_assert(sizeof(WChar) == 2 || sizeof(WChar) == 4,
+                  "Expect utf16 or utf32");
+    if (!convert(s, policy))
+      FMT_THROW(std::runtime_error(sizeof(WChar) == 2 ? "invalid utf16"
+                                                      : "invalid utf32"));
+  }
+  operator string_view() const { return string_view(&buffer_[0], size()); }
+  auto size() const -> size_t { return buffer_.size() - 1; }
+  auto c_str() const -> const char* { return &buffer_[0]; }
+  auto str() const -> std::string { return std::string(&buffer_[0], size()); }
+
+  // Performs conversion returning a bool instead of throwing exception on
+  // conversion error. This method may still throw in case of memory allocation
+  // error.
+  auto convert(basic_string_view<WChar> s,
+               to_utf8_error_policy policy = to_utf8_error_policy::abort)
+      -> bool {
+    if (!convert(buffer_, s, policy)) return false;
+    buffer_.push_back(0);
+    return true;
+  }
+  static auto convert(Buffer& buf, basic_string_view<WChar> s,
+                      to_utf8_error_policy policy = to_utf8_error_policy::abort)
+      -> bool {
+    for (auto p = s.begin(); p != s.end(); ++p) {
+      uint32_t c = static_cast<uint32_t>(*p);
+      if (sizeof(WChar) == 2 && c >= 0xd800 && c <= 0xdfff) {
+        // Handle a surrogate pair.
+        ++p;
+        if (p == s.end() || (c & 0xfc00) != 0xd800 || (*p & 0xfc00) != 0xdc00) {
+          if (policy == to_utf8_error_policy::abort) return false;
+          buf.append(string_view("\xEF\xBF\xBD"));
+          --p;
+        } else {
+          c = (c << 10) + static_cast<uint32_t>(*p) - 0x35fdc00;
+        }
+      } else if (c < 0x80) {
+        buf.push_back(static_cast<char>(c));
+      } else if (c < 0x800) {
+        buf.push_back(static_cast<char>(0xc0 | (c >> 6)));
+        buf.push_back(static_cast<char>(0x80 | (c & 0x3f)));
+      } else if ((c >= 0x800 && c <= 0xd7ff) || (c >= 0xe000 && c <= 0xffff)) {
+        buf.push_back(static_cast<char>(0xe0 | (c >> 12)));
+        buf.push_back(static_cast<char>(0x80 | ((c & 0xfff) >> 6)));
+        buf.push_back(static_cast<char>(0x80 | (c & 0x3f)));
+      } else if (c >= 0x10000 && c <= 0x10ffff) {
+        buf.push_back(static_cast<char>(0xf0 | (c >> 18)));
+        buf.push_back(static_cast<char>(0x80 | ((c & 0x3ffff) >> 12)));
+        buf.push_back(static_cast<char>(0x80 | ((c & 0xfff) >> 6)));
+        buf.push_back(static_cast<char>(0x80 | (c & 0x3f)));
+      } else {
+        return false;
+      }
+    }
+    return true;
+  }
+};
+
+// Computes 128-bit result of multiplication of two 64-bit unsigned integers.
+inline auto umul128(uint64_t x, uint64_t y) noexcept -> uint128_fallback {
+#if FMT_USE_INT128
+  auto p = static_cast<uint128_opt>(x) * static_cast<uint128_opt>(y);
+  return {static_cast<uint64_t>(p >> 64), static_cast<uint64_t>(p)};
+#elif defined(_MSC_VER) && defined(_M_X64)
+  auto hi = uint64_t();
+  auto lo = _umul128(x, y, &hi);
+  return {hi, lo};
+#else
+  const uint64_t mask = static_cast<uint64_t>(max_value<uint32_t>());
+
+  uint64_t a = x >> 32;
+  uint64_t b = x & mask;
+  uint64_t c = y >> 32;
+  uint64_t d = y & mask;
+
+  uint64_t ac = a * c;
+  uint64_t bc = b * c;
+  uint64_t ad = a * d;
+  uint64_t bd = b * d;
+
+  uint64_t intermediate = (bd >> 32) + (ad & mask) + (bc & mask);
+
+  return {ac + (intermediate >> 32) + (ad >> 32) + (bc >> 32),
+          (intermediate << 32) + (bd & mask)};
+#endif
+}
+
 namespace dragonbox {
+// Computes floor(log10(pow(2, e))) for e in [-2620, 2620] using the method from
+// https://fmt.dev/papers/Dragonbox.pdf#page=28, section 6.1.
+inline auto floor_log10_pow2(int e) noexcept -> int {
+  FMT_ASSERT(e <= 2620 && e >= -2620, "too large exponent");
+  static_assert((-1 >> 1) == -1, "right shift is not arithmetic");
+  return (e * 315653) >> 20;
+}
+
+inline auto floor_log2_pow10(int e) noexcept -> int {
+  FMT_ASSERT(e <= 1233 && e >= -1233, "too large exponent");
+  return (e * 1741647) >> 19;
+}
+
+// Computes upper 64 bits of multiplication of two 64-bit unsigned integers.
+inline auto umul128_upper64(uint64_t x, uint64_t y) noexcept -> uint64_t {
+#if FMT_USE_INT128
+  auto p = static_cast<uint128_opt>(x) * static_cast<uint128_opt>(y);
+  return static_cast<uint64_t>(p >> 64);
+#elif defined(_MSC_VER) && defined(_M_X64)
+  return __umulh(x, y);
+#else
+  return umul128(x, y).high();
+#endif
+}
+
+// Computes upper 128 bits of multiplication of a 64-bit unsigned integer and a
+// 128-bit unsigned integer.
+inline auto umul192_upper128(uint64_t x, uint128_fallback y) noexcept
+    -> uint128_fallback {
+  uint128_fallback r = umul128(x, y.high());
+  r += umul128_upper64(x, y.low());
+  return r;
+}
+
+FMT_API auto get_cached_power(int k) noexcept -> uint128_fallback;
 
 // Type-specific information that Dragonbox uses.
 template <typename T, typename Enable = void> struct float_info;
@@ -1307,7 +1534,7 @@ template <> struct float_info<double> {
   static const int big_divisor = 1000;
   static const int small_divisor = 100;
   static const int min_k = -292;
-  static const int max_k = 326;
+  static const int max_k = 341;
   static const int shorter_interval_tie_lower_threshold = -77;
   static const int shorter_interval_tie_upper_threshold = -77;
 };
@@ -1337,14 +1564,14 @@ template <typename T> FMT_API auto to_decimal(T x) noexcept -> decimal_fp<T>;
 }  // namespace dragonbox
 
 // Returns true iff Float has the implicit bit which is not stored.
-template <typename Float> constexpr bool has_implicit_bit() {
+template <typename Float> constexpr auto has_implicit_bit() -> bool {
   // An 80-bit FP number has a 64-bit significand an no implicit bit.
   return std::numeric_limits<Float>::digits != 64;
 }
 
 // Returns the number of significand bits stored in Float. The implicit bit is
 // not counted since it is not stored.
-template <typename Float> constexpr int num_significand_bits() {
+template <typename Float> constexpr auto num_significand_bits() -> int {
   // std::numeric_limits may not support __float128.
   return is_float128<Float>() ? 112
                               : (std::numeric_limits<Float>::digits -
@@ -1354,8 +1581,8 @@ template <typename Float> constexpr int num_significand_bits() {
 template <typename Float>
 constexpr auto exponent_mask() ->
     typename dragonbox::float_info<Float>::carrier_uint {
-  using uint = typename dragonbox::float_info<Float>::carrier_uint;
-  return ((uint(1) << dragonbox::float_info<Float>::exponent_bits) - 1)
+  using float_uint = typename dragonbox::float_info<Float>::carrier_uint;
+  return ((float_uint(1) << dragonbox::float_info<Float>::exponent_bits) - 1)
          << num_significand_bits<Float>();
 }
 template <typename Float> constexpr auto exponent_bias() -> int {
@@ -1437,7 +1664,7 @@ using fp = basic_fp<unsigned long long>;
 
 // Normalizes the value converted from double and multiplied by (1 << SHIFT).
 template <int SHIFT = 0, typename F>
-FMT_CONSTEXPR basic_fp<F> normalize(basic_fp<F> value) {
+FMT_CONSTEXPR auto normalize(basic_fp<F> value) -> basic_fp<F> {
   // Handle subnormals.
   const auto implicit_bit = F(1) << num_significand_bits<double>();
   const auto shifted_implicit_bit = implicit_bit << SHIFT;
@@ -1454,7 +1681,7 @@ FMT_CONSTEXPR basic_fp<F> normalize(basic_fp<F> value) {
 }
 
 // Computes lhs * rhs / pow(2, 64) rounded to nearest with half-up tie breaking.
-FMT_CONSTEXPR inline uint64_t multiply(uint64_t lhs, uint64_t rhs) {
+FMT_CONSTEXPR inline auto multiply(uint64_t lhs, uint64_t rhs) -> uint64_t {
 #if FMT_USE_INT128
   auto product = static_cast<__uint128_t>(lhs) * rhs;
   auto f = static_cast<uint64_t>(product >> 64);
@@ -1471,162 +1698,13 @@ FMT_CONSTEXPR inline uint64_t multiply(uint64_t lhs, uint64_t rhs) {
 #endif
 }
 
-FMT_CONSTEXPR inline fp operator*(fp x, fp y) {
+FMT_CONSTEXPR inline auto operator*(fp x, fp y) -> fp {
   return {multiply(x.f, y.f), x.e + y.e + 64};
 }
 
-template <typename T = void> struct basic_data {
-  // Normalized 64-bit significands of pow(10, k), for k = -348, -340, ..., 340.
-  // These are generated by support/compute-powers.py.
-  static constexpr uint64_t pow10_significands[87] = {
-      0xfa8fd5a0081c0288, 0xbaaee17fa23ebf76, 0x8b16fb203055ac76,
-      0xcf42894a5dce35ea, 0x9a6bb0aa55653b2d, 0xe61acf033d1a45df,
-      0xab70fe17c79ac6ca, 0xff77b1fcbebcdc4f, 0xbe5691ef416bd60c,
-      0x8dd01fad907ffc3c, 0xd3515c2831559a83, 0x9d71ac8fada6c9b5,
-      0xea9c227723ee8bcb, 0xaecc49914078536d, 0x823c12795db6ce57,
-      0xc21094364dfb5637, 0x9096ea6f3848984f, 0xd77485cb25823ac7,
-      0xa086cfcd97bf97f4, 0xef340a98172aace5, 0xb23867fb2a35b28e,
-      0x84c8d4dfd2c63f3b, 0xc5dd44271ad3cdba, 0x936b9fcebb25c996,
-      0xdbac6c247d62a584, 0xa3ab66580d5fdaf6, 0xf3e2f893dec3f126,
-      0xb5b5ada8aaff80b8, 0x87625f056c7c4a8b, 0xc9bcff6034c13053,
-      0x964e858c91ba2655, 0xdff9772470297ebd, 0xa6dfbd9fb8e5b88f,
-      0xf8a95fcf88747d94, 0xb94470938fa89bcf, 0x8a08f0f8bf0f156b,
-      0xcdb02555653131b6, 0x993fe2c6d07b7fac, 0xe45c10c42a2b3b06,
-      0xaa242499697392d3, 0xfd87b5f28300ca0e, 0xbce5086492111aeb,
-      0x8cbccc096f5088cc, 0xd1b71758e219652c, 0x9c40000000000000,
-      0xe8d4a51000000000, 0xad78ebc5ac620000, 0x813f3978f8940984,
-      0xc097ce7bc90715b3, 0x8f7e32ce7bea5c70, 0xd5d238a4abe98068,
-      0x9f4f2726179a2245, 0xed63a231d4c4fb27, 0xb0de65388cc8ada8,
-      0x83c7088e1aab65db, 0xc45d1df942711d9a, 0x924d692ca61be758,
-      0xda01ee641a708dea, 0xa26da3999aef774a, 0xf209787bb47d6b85,
-      0xb454e4a179dd1877, 0x865b86925b9bc5c2, 0xc83553c5c8965d3d,
-      0x952ab45cfa97a0b3, 0xde469fbd99a05fe3, 0xa59bc234db398c25,
-      0xf6c69a72a3989f5c, 0xb7dcbf5354e9bece, 0x88fcf317f22241e2,
-      0xcc20ce9bd35c78a5, 0x98165af37b2153df, 0xe2a0b5dc971f303a,
-      0xa8d9d1535ce3b396, 0xfb9b7cd9a4a7443c, 0xbb764c4ca7a44410,
-      0x8bab8eefb6409c1a, 0xd01fef10a657842c, 0x9b10a4e5e9913129,
-      0xe7109bfba19c0c9d, 0xac2820d9623bf429, 0x80444b5e7aa7cf85,
-      0xbf21e44003acdd2d, 0x8e679c2f5e44ff8f, 0xd433179d9c8cb841,
-      0x9e19db92b4e31ba9, 0xeb96bf6ebadf77d9, 0xaf87023b9bf0ee6b,
-  };
-
-#if FMT_GCC_VERSION && FMT_GCC_VERSION < 409
-#  pragma GCC diagnostic push
-#  pragma GCC diagnostic ignored "-Wnarrowing"
-#endif
-  // Binary exponents of pow(10, k), for k = -348, -340, ..., 340, corresponding
-  // to significands above.
-  static constexpr int16_t pow10_exponents[87] = {
-      -1220, -1193, -1166, -1140, -1113, -1087, -1060, -1034, -1007, -980, -954,
-      -927,  -901,  -874,  -847,  -821,  -794,  -768,  -741,  -715,  -688, -661,
-      -635,  -608,  -582,  -555,  -529,  -502,  -475,  -449,  -422,  -396, -369,
-      -343,  -316,  -289,  -263,  -236,  -210,  -183,  -157,  -130,  -103, -77,
-      -50,   -24,   3,     30,    56,    83,    109,   136,   162,   189,  216,
-      242,   269,   295,   322,   348,   375,   402,   428,   455,   481,  508,
-      534,   561,   588,   614,   641,   667,   694,   720,   747,   774,  800,
-      827,   853,   880,   907,   933,   960,   986,   1013,  1039,  1066};
-#if FMT_GCC_VERSION && FMT_GCC_VERSION < 409
-#  pragma GCC diagnostic pop
-#endif
-
-  static constexpr uint64_t power_of_10_64[20] = {
-      1, FMT_POWERS_OF_10(1ULL), FMT_POWERS_OF_10(1000000000ULL),
-      10000000000000000000ULL};
-};
-
-#if FMT_CPLUSPLUS < 201703L
-template <typename T> constexpr uint64_t basic_data<T>::pow10_significands[];
-template <typename T> constexpr int16_t basic_data<T>::pow10_exponents[];
-template <typename T> constexpr uint64_t basic_data<T>::power_of_10_64[];
-#endif
-
-// This is a struct rather than an alias to avoid shadowing warnings in gcc.
-struct data : basic_data<> {};
-
-// Returns a cached power of 10 `c_k = c_k.f * pow(2, c_k.e)` such that its
-// (binary) exponent satisfies `min_exponent <= c_k.e <= min_exponent + 28`.
-FMT_CONSTEXPR inline fp get_cached_power(int min_exponent,
-                                         int& pow10_exponent) {
-  const int shift = 32;
-  // log10(2) = 0x0.4d104d427de7fbcc...
-  const int64_t significand = 0x4d104d427de7fbcc;
-  int index = static_cast<int>(
-      ((min_exponent + fp::num_significand_bits - 1) * (significand >> shift) +
-       ((int64_t(1) << shift) - 1))  // ceil
-      >> 32                          // arithmetic shift
-  );
-  // Decimal exponent of the first (smallest) cached power of 10.
-  const int first_dec_exp = -348;
-  // Difference between 2 consecutive decimal exponents in cached powers of 10.
-  const int dec_exp_step = 8;
-  index = (index - first_dec_exp - 1) / dec_exp_step + 1;
-  pow10_exponent = first_dec_exp + index * dec_exp_step;
-  // Using *(x + index) instead of x[index] avoids an issue with some compilers
-  // using the EDG frontend (e.g. nvhpc/22.3 in C++17 mode).
-  return {*(data::pow10_significands + index),
-          *(data::pow10_exponents + index)};
-}
-
-#ifndef _MSC_VER
-#  define FMT_SNPRINTF snprintf
-#else
-FMT_API auto fmt_snprintf(char* buf, size_t size, const char* fmt, ...) -> int;
-#  define FMT_SNPRINTF fmt_snprintf
-#endif  // _MSC_VER
-
-// Formats a floating-point number with snprintf using the hexfloat format.
-template <typename T>
-auto snprintf_float(T value, int precision, float_specs specs,
-                    buffer<char>& buf) -> int {
-  // Buffer capacity must be non-zero, otherwise MSVC's vsnprintf_s will fail.
-  FMT_ASSERT(buf.capacity() > buf.size(), "empty buffer");
-  FMT_ASSERT(specs.format == float_format::hex, "");
-  static_assert(!std::is_same<T, float>::value, "");
-
-  // Build the format string.
-  char format[7];  // The longest format is "%#.*Le".
-  char* format_ptr = format;
-  *format_ptr++ = '%';
-  if (specs.showpoint) *format_ptr++ = '#';
-  if (precision >= 0) {
-    *format_ptr++ = '.';
-    *format_ptr++ = '*';
-  }
-  if (std::is_same<T, long double>()) *format_ptr++ = 'L';
-  *format_ptr++ = specs.upper ? 'A' : 'a';
-  *format_ptr = '\0';
-
-  // Format using snprintf.
-  auto offset = buf.size();
-  for (;;) {
-    auto begin = buf.data() + offset;
-    auto capacity = buf.capacity() - offset;
-    abort_fuzzing_if(precision > 100000);
-    // Suppress the warning about a nonliteral format string.
-    // Cannot use auto because of a bug in MinGW (#1532).
-    int (*snprintf_ptr)(char*, size_t, const char*, ...) = FMT_SNPRINTF;
-    int result = precision >= 0
-                     ? snprintf_ptr(begin, capacity, format, precision, value)
-                     : snprintf_ptr(begin, capacity, format, value);
-    if (result < 0) {
-      // The buffer will grow exponentially.
-      buf.try_reserve(buf.capacity() + 1);
-      continue;
-    }
-    auto size = to_unsigned(result);
-    // Size equal to capacity means that the last character was truncated.
-    if (size < capacity) {
-      buf.try_resize(size + offset);
-      return 0;
-    }
-    buf.try_reserve(size + offset + 1);  // Add 1 for the terminating '\0'.
-  }
-}
-
-template <typename T>
+template <typename T, bool doublish = num_bits<T>() == num_bits<double>()>
 using convert_float_result =
-    conditional_t<std::is_same<T, float>::value || sizeof(T) == sizeof(double),
-                  double, T>;
+    conditional_t<std::is_same<T, float>::value || doublish, double, T>;
 
 template <typename T>
 constexpr auto convert_float(T value) -> convert_float_result<T> {
@@ -1649,8 +1727,7 @@ FMT_NOINLINE FMT_CONSTEXPR auto fill(OutputIt it, size_t n,
 // width: output display width in (terminal) column positions.
 template <align::type align = align::left, typename OutputIt, typename Char,
           typename F>
-FMT_CONSTEXPR auto write_padded(OutputIt out,
-                                const basic_format_specs<Char>& specs,
+FMT_CONSTEXPR auto write_padded(OutputIt out, const format_specs<Char>& specs,
                                 size_t size, size_t width, F&& f) -> OutputIt {
   static_assert(align == align::left || align == align::right, "");
   unsigned spec_width = to_unsigned(specs.width);
@@ -1669,15 +1746,14 @@ FMT_CONSTEXPR auto write_padded(OutputIt out,
 
 template <align::type align = align::left, typename OutputIt, typename Char,
           typename F>
-constexpr auto write_padded(OutputIt out, const basic_format_specs<Char>& specs,
+constexpr auto write_padded(OutputIt out, const format_specs<Char>& specs,
                             size_t size, F&& f) -> OutputIt {
   return write_padded<align>(out, specs, size, size, f);
 }
 
 template <align::type align = align::left, typename Char, typename OutputIt>
 FMT_CONSTEXPR auto write_bytes(OutputIt out, string_view bytes,
-                               const basic_format_specs<Char>& specs)
-    -> OutputIt {
+                               const format_specs<Char>& specs) -> OutputIt {
   return write_padded<align>(
       out, specs, bytes.size(), [bytes](reserve_iterator<OutputIt> it) {
         const char* data = bytes.data();
@@ -1686,8 +1762,8 @@ FMT_CONSTEXPR auto write_bytes(OutputIt out, string_view bytes,
 }
 
 template <typename Char, typename OutputIt, typename UIntPtr>
-auto write_ptr(OutputIt out, UIntPtr value,
-               const basic_format_specs<Char>* specs) -> OutputIt {
+auto write_ptr(OutputIt out, UIntPtr value, const format_specs<Char>* specs)
+    -> OutputIt {
   int num_digits = count_digits<4>(value);
   auto size = to_unsigned(num_digits) + size_t(2);
   auto write = [=](reserve_iterator<OutputIt> it) {
@@ -1749,7 +1825,7 @@ inline auto find_escape(const char* begin, const char* end)
   [] {                                                                        \
     /* Use the hidden visibility as a workaround for a GCC bug (#1973). */    \
     /* Use a macro-like name to avoid shadowing warnings. */                  \
-    struct FMT_GCC_VISIBILITY_HIDDEN FMT_COMPILE_STRING : base {              \
+    struct FMT_VISIBILITY("hidden") FMT_COMPILE_STRING : base {               \
       using char_type FMT_MAYBE_UNUSED = fmt::remove_cvref_t<decltype(s[0])>; \
       FMT_MAYBE_UNUSED FMT_CONSTEXPR explicit                                 \
       operator fmt::basic_string_view<char_type>() const {                    \
@@ -1806,16 +1882,14 @@ auto write_escaped_cp(OutputIt out, const find_escape_result<Char>& escape)
     *out++ = static_cast<Char>('\\');
     break;
   default:
-    if (is_utf8()) {
-      if (escape.cp < 0x100) {
-        return write_codepoint<2, Char>(out, 'x', escape.cp);
-      }
-      if (escape.cp < 0x10000) {
-        return write_codepoint<4, Char>(out, 'u', escape.cp);
-      }
-      if (escape.cp < 0x110000) {
-        return write_codepoint<8, Char>(out, 'U', escape.cp);
-      }
+    if (escape.cp < 0x100) {
+      return write_codepoint<2, Char>(out, 'x', escape.cp);
+    }
+    if (escape.cp < 0x10000) {
+      return write_codepoint<4, Char>(out, 'u', escape.cp);
+    }
+    if (escape.cp < 0x110000) {
+      return write_codepoint<8, Char>(out, 'U', escape.cp);
     }
     for (Char escape_char : basic_string_view<Char>(
              escape.begin, to_unsigned(escape.end - escape.begin))) {
@@ -1846,11 +1920,13 @@ auto write_escaped_string(OutputIt out, basic_string_view<Char> str)
 
 template <typename Char, typename OutputIt>
 auto write_escaped_char(OutputIt out, Char v) -> OutputIt {
+  Char v_array[1] = {v};
   *out++ = static_cast<Char>('\'');
   if ((needs_escape(static_cast<uint32_t>(v)) && v != static_cast<Char>('"')) ||
       v == static_cast<Char>('\'')) {
-    out = write_escaped_cp(
-        out, find_escape_result<Char>{&v, &v + 1, static_cast<uint32_t>(v)});
+    out = write_escaped_cp(out,
+                           find_escape_result<Char>{v_array, v_array + 1,
+                                                    static_cast<uint32_t>(v)});
   } else {
     *out++ = v;
   }
@@ -1860,8 +1936,7 @@ auto write_escaped_char(OutputIt out, Char v) -> OutputIt {
 
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR auto write_char(OutputIt out, Char value,
-                              const basic_format_specs<Char>& specs)
-    -> OutputIt {
+                              const format_specs<Char>& specs) -> OutputIt {
   bool is_debug = specs.type == presentation_type::debug;
   return write_padded(out, specs, 1, [=](reserve_iterator<OutputIt> it) {
     if (is_debug) return write_escaped_char(it, value);
@@ -1871,11 +1946,14 @@ FMT_CONSTEXPR auto write_char(OutputIt out, Char value,
 }
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR auto write(OutputIt out, Char value,
-                         const basic_format_specs<Char>& specs,
-                         locale_ref loc = {}) -> OutputIt {
+                         const format_specs<Char>& specs, locale_ref loc = {})
+    -> OutputIt {
+  // char is formatted as unsigned char for consistency across platforms.
+  using unsigned_type =
+      conditional_t<std::is_same<Char, char>::value, unsigned char, unsigned>;
   return check_char_specs(specs)
              ? write_char(out, value, specs)
-             : write(out, static_cast<int>(value), specs, loc);
+             : write(out, static_cast<unsigned_type>(value), specs, loc);
 }
 
 // Data for write_int that doesn't depend on output iterator type. It is used to
@@ -1885,7 +1963,7 @@ template <typename Char> struct write_int_data {
   size_t padding;
 
   FMT_CONSTEXPR write_int_data(int num_digits, unsigned prefix,
-                               const basic_format_specs<Char>& specs)
+                               const format_specs<Char>& specs)
       : size((prefix >> 24) + to_unsigned(num_digits)), padding(0) {
     if (specs.align == align::numeric) {
       auto width = to_unsigned(specs.width);
@@ -1907,7 +1985,7 @@ template <typename Char> struct write_int_data {
 template <typename OutputIt, typename Char, typename W>
 FMT_CONSTEXPR FMT_INLINE auto write_int(OutputIt out, int num_digits,
                                         unsigned prefix,
-                                        const basic_format_specs<Char>& specs,
+                                        const format_specs<Char>& specs,
                                         W write_digits) -> OutputIt {
   // Slightly faster check for specs.width == 0 && specs.precision == -1.
   if ((specs.width | (specs.precision + 1)) == 0) {
@@ -1930,19 +2008,19 @@ FMT_CONSTEXPR FMT_INLINE auto write_int(OutputIt out, int num_digits,
 
 template <typename Char> class digit_grouping {
  private:
-  thousands_sep_result<Char> sep_;
+  std::string grouping_;
+  std::basic_string<Char> thousands_sep_;
 
   struct next_state {
     std::string::const_iterator group;
     int pos;
   };
-  next_state initial_state() const { return {sep_.grouping.begin(), 0}; }
+  auto initial_state() const -> next_state { return {grouping_.begin(), 0}; }
 
   // Returns the next digit group separator position.
-  int next(next_state& state) const {
-    if (!sep_.thousands_sep) return max_value<int>();
-    if (state.group == sep_.grouping.end())
-      return state.pos += sep_.grouping.back();
+  auto next(next_state& state) const -> int {
+    if (thousands_sep_.empty()) return max_value<int>();
+    if (state.group == grouping_.end()) return state.pos += grouping_.back();
     if (*state.group <= 0 || *state.group == max_value<char>())
       return max_value<int>();
     state.pos += *state.group++;
@@ -1951,16 +2029,17 @@ template <typename Char> class digit_grouping {
 
  public:
   explicit digit_grouping(locale_ref loc, bool localized = true) {
-    if (localized)
-      sep_ = thousands_sep<Char>(loc);
-    else
-      sep_.thousands_sep = Char();
+    if (!localized) return;
+    auto sep = thousands_sep<Char>(loc);
+    grouping_ = sep.grouping;
+    if (sep.thousands_sep) thousands_sep_.assign(1, sep.thousands_sep);
   }
-  explicit digit_grouping(thousands_sep_result<Char> sep) : sep_(sep) {}
+  digit_grouping(std::string grouping, std::basic_string<Char> sep)
+      : grouping_(std::move(grouping)), thousands_sep_(std::move(sep)) {}
 
-  Char separator() const { return sep_.thousands_sep; }
+  auto has_separator() const -> bool { return !thousands_sep_.empty(); }
 
-  int count_separators(int num_digits) const {
+  auto count_separators(int num_digits) const -> int {
     int count = 0;
     auto state = initial_state();
     while (num_digits > next(state)) ++count;
@@ -1969,7 +2048,7 @@ template <typename Char> class digit_grouping {
 
   // Applies grouping to digits and write the output to out.
   template <typename Out, typename C>
-  Out apply(Out out, basic_string_view<C> digits) const {
+  auto apply(Out out, basic_string_view<C> digits) const -> Out {
     auto num_digits = static_cast<int>(digits.size());
     auto separators = basic_memory_buffer<int>();
     separators.push_back(0);
@@ -1981,7 +2060,9 @@ template <typename Char> class digit_grouping {
     for (int i = 0, sep_index = static_cast<int>(separators.size() - 1);
          i < num_digits; ++i) {
       if (num_digits - i == separators[sep_index]) {
-        *out++ = separator();
+        out =
+            copy_str<Char>(thousands_sep_.data(),
+                           thousands_sep_.data() + thousands_sep_.size(), out);
         --sep_index;
       }
       *out++ = static_cast<Char>(digits[to_unsigned(i)]);
@@ -1990,38 +2071,76 @@ template <typename Char> class digit_grouping {
   }
 };
 
+FMT_CONSTEXPR inline void prefix_append(unsigned& prefix, unsigned value) {
+  prefix |= prefix != 0 ? value << 8 : value;
+  prefix += (1u + (value > 0xff ? 1 : 0)) << 24;
+}
+
+// Writes a decimal integer with digit grouping.
 template <typename OutputIt, typename UInt, typename Char>
-auto write_int_localized(OutputIt out, UInt value, unsigned prefix,
-                         const basic_format_specs<Char>& specs,
-                         const digit_grouping<Char>& grouping) -> OutputIt {
+auto write_int(OutputIt out, UInt value, unsigned prefix,
+               const format_specs<Char>& specs,
+               const digit_grouping<Char>& grouping) -> OutputIt {
   static_assert(std::is_same<uint64_or_128_t<UInt>, UInt>::value, "");
-  int num_digits = count_digits(value);
-  char digits[40];
-  format_decimal(digits, value, num_digits);
-  unsigned size = to_unsigned((prefix != 0 ? 1 : 0) + num_digits +
-                              grouping.count_separators(num_digits));
+  int num_digits = 0;
+  auto buffer = memory_buffer();
+  switch (specs.type) {
+  case presentation_type::none:
+  case presentation_type::dec: {
+    num_digits = count_digits(value);
+    format_decimal<char>(appender(buffer), value, num_digits);
+    break;
+  }
+  case presentation_type::hex_lower:
+  case presentation_type::hex_upper: {
+    bool upper = specs.type == presentation_type::hex_upper;
+    if (specs.alt)
+      prefix_append(prefix, unsigned(upper ? 'X' : 'x') << 8 | '0');
+    num_digits = count_digits<4>(value);
+    format_uint<4, char>(appender(buffer), value, num_digits, upper);
+    break;
+  }
+  case presentation_type::bin_lower:
+  case presentation_type::bin_upper: {
+    bool upper = specs.type == presentation_type::bin_upper;
+    if (specs.alt)
+      prefix_append(prefix, unsigned(upper ? 'B' : 'b') << 8 | '0');
+    num_digits = count_digits<1>(value);
+    format_uint<1, char>(appender(buffer), value, num_digits);
+    break;
+  }
+  case presentation_type::oct: {
+    num_digits = count_digits<3>(value);
+    // Octal prefix '0' is counted as a digit, so only add it if precision
+    // is not greater than the number of digits.
+    if (specs.alt && specs.precision <= num_digits && value != 0)
+      prefix_append(prefix, '0');
+    format_uint<3, char>(appender(buffer), value, num_digits);
+    break;
+  }
+  case presentation_type::chr:
+    return write_char(out, static_cast<Char>(value), specs);
+  default:
+    throw_format_error("invalid format specifier");
+  }
+
+  unsigned size = (prefix != 0 ? prefix >> 24 : 0) + to_unsigned(num_digits) +
+                  to_unsigned(grouping.count_separators(num_digits));
   return write_padded<align::right>(
       out, specs, size, size, [&](reserve_iterator<OutputIt> it) {
-        if (prefix != 0) {
-          char sign = static_cast<char>(prefix);
-          *it++ = static_cast<Char>(sign);
-        }
-        return grouping.apply(it, string_view(digits, to_unsigned(num_digits)));
+        for (unsigned p = prefix & 0xffffff; p != 0; p >>= 8)
+          *it++ = static_cast<Char>(p & 0xff);
+        return grouping.apply(it, string_view(buffer.data(), buffer.size()));
       });
 }
 
-template <typename OutputIt, typename UInt, typename Char>
-auto write_int_localized(OutputIt& out, UInt value, unsigned prefix,
-                         const basic_format_specs<Char>& specs, locale_ref loc)
-    -> bool {
-  auto grouping = digit_grouping<Char>(loc);
-  out = write_int_localized(out, value, prefix, specs, grouping);
-  return true;
-}
-
-FMT_CONSTEXPR inline void prefix_append(unsigned& prefix, unsigned value) {
-  prefix |= prefix != 0 ? value << 8 : value;
-  prefix += (1u + (value > 0xff ? 1 : 0)) << 24;
+// Writes a localized value.
+FMT_API auto write_loc(appender out, loc_value value,
+                       const format_specs<>& specs, locale_ref loc) -> bool;
+template <typename OutputIt, typename Char>
+inline auto write_loc(OutputIt, loc_value, const format_specs<Char>&,
+                      locale_ref) -> bool {
+  return false;
 }
 
 template <typename UInt> struct write_int_arg {
@@ -2045,21 +2164,37 @@ FMT_CONSTEXPR auto make_write_int_arg(T value, sign_t sign)
   return {abs_value, prefix};
 }
 
+template <typename Char = char> struct loc_writer {
+  buffer_appender<Char> out;
+  const format_specs<Char>& specs;
+  std::basic_string<Char> sep;
+  std::string grouping;
+  std::basic_string<Char> decimal_point;
+
+  template <typename T, FMT_ENABLE_IF(is_integer<T>::value)>
+  auto operator()(T value) -> bool {
+    auto arg = make_write_int_arg(value, specs.sign);
+    write_int(out, static_cast<uint64_or_128_t<T>>(arg.abs_value), arg.prefix,
+              specs, digit_grouping<Char>(grouping, sep));
+    return true;
+  }
+
+  template <typename T, FMT_ENABLE_IF(!is_integer<T>::value)>
+  auto operator()(T) -> bool {
+    return false;
+  }
+};
+
 template <typename Char, typename OutputIt, typename T>
 FMT_CONSTEXPR FMT_INLINE auto write_int(OutputIt out, write_int_arg<T> arg,
-                                        const basic_format_specs<Char>& specs,
-                                        locale_ref loc) -> OutputIt {
+                                        const format_specs<Char>& specs,
+                                        locale_ref) -> OutputIt {
   static_assert(std::is_same<T, uint32_or_64_or_128_t<T>>::value, "");
   auto abs_value = arg.abs_value;
   auto prefix = arg.prefix;
   switch (specs.type) {
   case presentation_type::none:
   case presentation_type::dec: {
-    if (specs.localized &&
-        write_int_localized(out, static_cast<uint64_or_128_t<T>>(abs_value),
-                            prefix, specs, loc)) {
-      return out;
-    }
     auto num_digits = count_digits(abs_value);
     return write_int(
         out, num_digits, prefix, specs, [=](reserve_iterator<OutputIt> it) {
@@ -2102,13 +2237,13 @@ FMT_CONSTEXPR FMT_INLINE auto write_int(OutputIt out, write_int_arg<T> arg,
   case presentation_type::chr:
     return write_char(out, static_cast<Char>(abs_value), specs);
   default:
-    throw_format_error("invalid type specifier");
+    throw_format_error("invalid format specifier");
   }
   return out;
 }
 template <typename Char, typename OutputIt, typename T>
 FMT_CONSTEXPR FMT_NOINLINE auto write_int_noinline(
-    OutputIt out, write_int_arg<T> arg, const basic_format_specs<Char>& specs,
+    OutputIt out, write_int_arg<T> arg, const format_specs<Char>& specs,
     locale_ref loc) -> OutputIt {
   return write_int(out, arg, specs, loc);
 }
@@ -2117,8 +2252,9 @@ template <typename Char, typename OutputIt, typename T,
                         !std::is_same<T, bool>::value &&
                         std::is_same<OutputIt, buffer_appender<Char>>::value)>
 FMT_CONSTEXPR FMT_INLINE auto write(OutputIt out, T value,
-                                    const basic_format_specs<Char>& specs,
+                                    const format_specs<Char>& specs,
                                     locale_ref loc) -> OutputIt {
+  if (specs.localized && write_loc(out, value, specs, loc)) return out;
   return write_int_noinline(out, make_write_int_arg(value, specs.sign), specs,
                             loc);
 }
@@ -2128,8 +2264,9 @@ template <typename Char, typename OutputIt, typename T,
                         !std::is_same<T, bool>::value &&
                         !std::is_same<OutputIt, buffer_appender<Char>>::value)>
 FMT_CONSTEXPR FMT_INLINE auto write(OutputIt out, T value,
-                                    const basic_format_specs<Char>& specs,
+                                    const format_specs<Char>& specs,
                                     locale_ref loc) -> OutputIt {
+  if (specs.localized && write_loc(out, value, specs, loc)) return out;
   return write_int(out, make_write_int_arg(value, specs.sign), specs, loc);
 }
 
@@ -2152,30 +2289,30 @@ class counting_iterator {
 
   FMT_CONSTEXPR counting_iterator() : count_(0) {}
 
-  FMT_CONSTEXPR size_t count() const { return count_; }
+  FMT_CONSTEXPR auto count() const -> size_t { return count_; }
 
-  FMT_CONSTEXPR counting_iterator& operator++() {
+  FMT_CONSTEXPR auto operator++() -> counting_iterator& {
     ++count_;
     return *this;
   }
-  FMT_CONSTEXPR counting_iterator operator++(int) {
+  FMT_CONSTEXPR auto operator++(int) -> counting_iterator {
     auto it = *this;
     ++*this;
     return it;
   }
 
-  FMT_CONSTEXPR friend counting_iterator operator+(counting_iterator it,
-                                                   difference_type n) {
+  FMT_CONSTEXPR friend auto operator+(counting_iterator it, difference_type n)
+      -> counting_iterator {
     it.count_ += static_cast<size_t>(n);
     return it;
   }
 
-  FMT_CONSTEXPR value_type operator*() const { return {}; }
+  FMT_CONSTEXPR auto operator*() const -> value_type { return {}; }
 };
 
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR auto write(OutputIt out, basic_string_view<Char> s,
-                         const basic_format_specs<Char>& specs) -> OutputIt {
+                         const format_specs<Char>& specs) -> OutputIt {
   auto data = s.data();
   auto size = s.size();
   if (specs.precision >= 0 && to_unsigned(specs.precision) < size)
@@ -2197,18 +2334,18 @@ FMT_CONSTEXPR auto write(OutputIt out, basic_string_view<Char> s,
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR auto write(OutputIt out,
                          basic_string_view<type_identity_t<Char>> s,
-                         const basic_format_specs<Char>& specs, locale_ref)
+                         const format_specs<Char>& specs, locale_ref)
     -> OutputIt {
-  check_string_type_spec(specs.type);
   return write(out, s, specs);
 }
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR auto write(OutputIt out, const Char* s,
-                         const basic_format_specs<Char>& specs, locale_ref)
+                         const format_specs<Char>& specs, locale_ref)
     -> OutputIt {
-  return check_cstring_type_spec(specs.type)
-             ? write(out, basic_string_view<Char>(s), specs, {})
-             : write_ptr<Char>(out, bit_cast<uintptr_t>(s), &specs);
+  if (specs.type == presentation_type::pointer)
+    return write_ptr<Char>(out, bit_cast<uintptr_t>(s), &specs);
+  if (!s) throw_format_error("string pointer is null");
+  return write(out, basic_string_view<Char>(s), specs, {});
 }
 
 template <typename Char, typename OutputIt, typename T,
@@ -2233,9 +2370,113 @@ FMT_CONSTEXPR auto write(OutputIt out, T value) -> OutputIt {
   return base_iterator(out, it);
 }
 
+// DEPRECATED!
+template <typename Char>
+FMT_CONSTEXPR auto parse_align(const Char* begin, const Char* end,
+                               format_specs<Char>& specs) -> const Char* {
+  FMT_ASSERT(begin != end, "");
+  auto align = align::none;
+  auto p = begin + code_point_length(begin);
+  if (end - p <= 0) p = begin;
+  for (;;) {
+    switch (to_ascii(*p)) {
+    case '<':
+      align = align::left;
+      break;
+    case '>':
+      align = align::right;
+      break;
+    case '^':
+      align = align::center;
+      break;
+    }
+    if (align != align::none) {
+      if (p != begin) {
+        auto c = *begin;
+        if (c == '}') return begin;
+        if (c == '{') {
+          throw_format_error("invalid fill character '{'");
+          return begin;
+        }
+        specs.fill = {begin, to_unsigned(p - begin)};
+        begin = p + 1;
+      } else {
+        ++begin;
+      }
+      break;
+    } else if (p == begin) {
+      break;
+    }
+    p = begin;
+  }
+  specs.align = align;
+  return begin;
+}
+
+// A floating-point presentation format.
+enum class float_format : unsigned char {
+  general,  // General: exponent notation or fixed point based on magnitude.
+  exp,      // Exponent notation with the default precision of 6, e.g. 1.2e-3.
+  fixed,    // Fixed point with the default precision of 6, e.g. 0.0012.
+  hex
+};
+
+struct float_specs {
+  int precision;
+  float_format format : 8;
+  sign_t sign : 8;
+  bool upper : 1;
+  bool locale : 1;
+  bool binary32 : 1;
+  bool showpoint : 1;
+};
+
+template <typename Char>
+FMT_CONSTEXPR auto parse_float_type_spec(const format_specs<Char>& specs)
+    -> float_specs {
+  auto result = float_specs();
+  result.showpoint = specs.alt;
+  result.locale = specs.localized;
+  switch (specs.type) {
+  case presentation_type::none:
+    result.format = float_format::general;
+    break;
+  case presentation_type::general_upper:
+    result.upper = true;
+    FMT_FALLTHROUGH;
+  case presentation_type::general_lower:
+    result.format = float_format::general;
+    break;
+  case presentation_type::exp_upper:
+    result.upper = true;
+    FMT_FALLTHROUGH;
+  case presentation_type::exp_lower:
+    result.format = float_format::exp;
+    result.showpoint |= specs.precision != 0;
+    break;
+  case presentation_type::fixed_upper:
+    result.upper = true;
+    FMT_FALLTHROUGH;
+  case presentation_type::fixed_lower:
+    result.format = float_format::fixed;
+    result.showpoint |= specs.precision != 0;
+    break;
+  case presentation_type::hexfloat_upper:
+    result.upper = true;
+    FMT_FALLTHROUGH;
+  case presentation_type::hexfloat_lower:
+    result.format = float_format::hex;
+    break;
+  default:
+    throw_format_error("invalid format specifier");
+    break;
+  }
+  return result;
+}
+
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR20 auto write_nonfinite(OutputIt out, bool isnan,
-                                     basic_format_specs<Char> specs,
+                                     format_specs<Char> specs,
                                      const float_specs& fspecs) -> OutputIt {
   auto str =
       isnan ? (fspecs.upper ? "NAN" : "nan") : (fspecs.upper ? "INF" : "inf");
@@ -2281,7 +2522,7 @@ template <typename Char, typename OutputIt, typename T, typename Grouping>
 FMT_CONSTEXPR20 auto write_significand(OutputIt out, T significand,
                                        int significand_size, int exponent,
                                        const Grouping& grouping) -> OutputIt {
-  if (!grouping.separator()) {
+  if (!grouping.has_separator()) {
     out = write_significand<Char>(out, significand, significand_size);
     return detail::fill_n(out, exponent, static_cast<Char>('0'));
   }
@@ -2343,7 +2584,7 @@ FMT_CONSTEXPR20 auto write_significand(OutputIt out, T significand,
                                        int significand_size, int integral_size,
                                        Char decimal_point,
                                        const Grouping& grouping) -> OutputIt {
-  if (!grouping.separator()) {
+  if (!grouping.has_separator()) {
     return write_significand(out, significand, significand_size, integral_size,
                              decimal_point);
   }
@@ -2359,7 +2600,7 @@ FMT_CONSTEXPR20 auto write_significand(OutputIt out, T significand,
 template <typename OutputIt, typename DecimalFP, typename Char,
           typename Grouping = digit_grouping<Char>>
 FMT_CONSTEXPR20 auto do_write_float(OutputIt out, const DecimalFP& f,
-                                    const basic_format_specs<Char>& specs,
+                                    const format_specs<Char>& specs,
                                     float_specs fspecs, locale_ref loc)
     -> OutputIt {
   auto significand = f.significand;
@@ -2418,7 +2659,7 @@ FMT_CONSTEXPR20 auto do_write_float(OutputIt out, const DecimalFP& f,
     abort_fuzzing_if(num_zeros > 5000);
     if (fspecs.showpoint) {
       ++size;
-      if (num_zeros <= 0 && fspecs.format != float_format::fixed) num_zeros = 1;
+      if (num_zeros <= 0 && fspecs.format != float_format::fixed) num_zeros = 0;
       if (num_zeros > 0) size += to_unsigned(num_zeros);
     }
     auto grouping = Grouping(loc, fspecs.locale);
@@ -2436,7 +2677,7 @@ FMT_CONSTEXPR20 auto do_write_float(OutputIt out, const DecimalFP& f,
     int num_zeros = fspecs.showpoint ? fspecs.precision - significand_size : 0;
     size += 1 + to_unsigned(num_zeros > 0 ? num_zeros : 0);
     auto grouping = Grouping(loc, fspecs.locale);
-    size += to_unsigned(grouping.count_separators(significand_size));
+    size += to_unsigned(grouping.count_separators(exp));
     return write_padded<align::right>(out, specs, size, [&](iterator it) {
       if (sign) *it++ = detail::sign<Char>(sign);
       it = write_significand(it, significand, significand_size, exp,
@@ -2466,19 +2707,19 @@ template <typename Char> class fallback_digit_grouping {
  public:
   constexpr fallback_digit_grouping(locale_ref, bool) {}
 
-  constexpr Char separator() const { return Char(); }
+  constexpr auto has_separator() const -> bool { return false; }
 
-  constexpr int count_separators(int) const { return 0; }
+  constexpr auto count_separators(int) const -> int { return 0; }
 
   template <typename Out, typename C>
-  constexpr Out apply(Out out, basic_string_view<C>) const {
+  constexpr auto apply(Out out, basic_string_view<C>) const -> Out {
     return out;
   }
 };
 
 template <typename OutputIt, typename DecimalFP, typename Char>
 FMT_CONSTEXPR20 auto write_float(OutputIt out, const DecimalFP& f,
-                                 const basic_format_specs<Char>& specs,
+                                 const format_specs<Char>& specs,
                                  float_specs fspecs, locale_ref loc)
     -> OutputIt {
   if (is_constant_evaluated()) {
@@ -2490,7 +2731,7 @@ FMT_CONSTEXPR20 auto write_float(OutputIt out, const DecimalFP& f,
   }
 }
 
-template <typename T> constexpr bool isnan(T value) {
+template <typename T> constexpr auto isnan(T value) -> bool {
   return !(value >= value);  // std::isnan doesn't support __float128.
 }
 
@@ -2503,17 +2744,17 @@ struct has_isfinite<T, enable_if_t<sizeof(std::isfinite(T())) != 0>>
 
 template <typename T, FMT_ENABLE_IF(std::is_floating_point<T>::value&&
                                         has_isfinite<T>::value)>
-FMT_CONSTEXPR20 bool isfinite(T value) {
+FMT_CONSTEXPR20 auto isfinite(T value) -> bool {
   constexpr T inf = T(std::numeric_limits<double>::infinity());
   if (is_constant_evaluated())
-    return !detail::isnan(value) && value != inf && value != -inf;
+    return !detail::isnan(value) && value < inf && value > -inf;
   return std::isfinite(value);
 }
 template <typename T, FMT_ENABLE_IF(!has_isfinite<T>::value)>
-FMT_CONSTEXPR bool isfinite(T value) {
+FMT_CONSTEXPR auto isfinite(T value) -> bool {
   T inf = T(std::numeric_limits<double>::infinity());
   // std::isfinite doesn't support __float128.
-  return !detail::isnan(value) && value != inf && value != -inf;
+  return !detail::isnan(value) && value < inf && value > -inf;
 }
 
 template <typename T, FMT_ENABLE_IF(is_floating_point<T>::value)>
@@ -2529,78 +2770,6 @@ FMT_INLINE FMT_CONSTEXPR bool signbit(T value) {
   return std::signbit(static_cast<double>(value));
 }
 
-enum class round_direction { unknown, up, down };
-
-// Given the divisor (normally a power of 10), the remainder = v % divisor for
-// some number v and the error, returns whether v should be rounded up, down, or
-// whether the rounding direction can't be determined due to error.
-// error should be less than divisor / 2.
-FMT_CONSTEXPR inline round_direction get_round_direction(uint64_t divisor,
-                                                         uint64_t remainder,
-                                                         uint64_t error) {
-  FMT_ASSERT(remainder < divisor, "");  // divisor - remainder won't overflow.
-  FMT_ASSERT(error < divisor, "");      // divisor - error won't overflow.
-  FMT_ASSERT(error < divisor - error, "");  // error * 2 won't overflow.
-  // Round down if (remainder + error) * 2 <= divisor.
-  if (remainder <= divisor - remainder && error * 2 <= divisor - remainder * 2)
-    return round_direction::down;
-  // Round up if (remainder - error) * 2 >= divisor.
-  if (remainder >= error &&
-      remainder - error >= divisor - (remainder - error)) {
-    return round_direction::up;
-  }
-  return round_direction::unknown;
-}
-
-namespace digits {
-enum result {
-  more,  // Generate more digits.
-  done,  // Done generating digits.
-  error  // Digit generation cancelled due to an error.
-};
-}
-
-struct gen_digits_handler {
-  char* buf;
-  int size;
-  int precision;
-  int exp10;
-  bool fixed;
-
-  FMT_CONSTEXPR digits::result on_digit(char digit, uint64_t divisor,
-                                        uint64_t remainder, uint64_t error,
-                                        bool integral) {
-    FMT_ASSERT(remainder < divisor, "");
-    buf[size++] = digit;
-    if (!integral && error >= remainder) return digits::error;
-    if (size < precision) return digits::more;
-    if (!integral) {
-      // Check if error * 2 < divisor with overflow prevention.
-      // The check is not needed for the integral part because error = 1
-      // and divisor > (1 << 32) there.
-      if (error >= divisor || error >= divisor - error) return digits::error;
-    } else {
-      FMT_ASSERT(error == 1 && divisor > 2, "");
-    }
-    auto dir = get_round_direction(divisor, remainder, error);
-    if (dir != round_direction::up)
-      return dir == round_direction::down ? digits::done : digits::error;
-    ++buf[size - 1];
-    for (int i = size - 1; i > 0 && buf[i] > '9'; --i) {
-      buf[i] = '0';
-      ++buf[i - 1];
-    }
-    if (buf[0] > '9') {
-      buf[0] = '1';
-      if (fixed)
-        buf[size++] = '0';
-      else
-        ++exp10;
-    }
-    return digits::done;
-  }
-};
-
 inline FMT_CONSTEXPR20 void adjust_precision(int& precision, int exp10) {
   // Adjust fixed precision by exponent because it is relative to decimal
   // point.
@@ -2609,101 +2778,6 @@ inline FMT_CONSTEXPR20 void adjust_precision(int& precision, int exp10) {
   precision += exp10;
 }
 
-// Generates output using the Grisu digit-gen algorithm.
-// error: the size of the region (lower, upper) outside of which numbers
-// definitely do not round to value (Delta in Grisu3).
-FMT_INLINE FMT_CONSTEXPR20 auto grisu_gen_digits(fp value, uint64_t error,
-                                                 int& exp,
-                                                 gen_digits_handler& handler)
-    -> digits::result {
-  const fp one(1ULL << -value.e, value.e);
-  // The integral part of scaled value (p1 in Grisu) = value / one. It cannot be
-  // zero because it contains a product of two 64-bit numbers with MSB set (due
-  // to normalization) - 1, shifted right by at most 60 bits.
-  auto integral = static_cast<uint32_t>(value.f >> -one.e);
-  FMT_ASSERT(integral != 0, "");
-  FMT_ASSERT(integral == value.f >> -one.e, "");
-  // The fractional part of scaled value (p2 in Grisu) c = value % one.
-  uint64_t fractional = value.f & (one.f - 1);
-  exp = count_digits(integral);  // kappa in Grisu.
-  // Non-fixed formats require at least one digit and no precision adjustment.
-  if (handler.fixed) {
-    adjust_precision(handler.precision, exp + handler.exp10);
-    // Check if precision is satisfied just by leading zeros, e.g.
-    // format("{:.2f}", 0.001) gives "0.00" without generating any digits.
-    if (handler.precision <= 0) {
-      if (handler.precision < 0) return digits::done;
-      // Divide by 10 to prevent overflow.
-      uint64_t divisor = data::power_of_10_64[exp - 1] << -one.e;
-      auto dir = get_round_direction(divisor, value.f / 10, error * 10);
-      if (dir == round_direction::unknown) return digits::error;
-      handler.buf[handler.size++] = dir == round_direction::up ? '1' : '0';
-      return digits::done;
-    }
-  }
-  // Generate digits for the integral part. This can produce up to 10 digits.
-  do {
-    uint32_t digit = 0;
-    auto divmod_integral = [&](uint32_t divisor) {
-      digit = integral / divisor;
-      integral %= divisor;
-    };
-    // This optimization by Milo Yip reduces the number of integer divisions by
-    // one per iteration.
-    switch (exp) {
-    case 10:
-      divmod_integral(1000000000);
-      break;
-    case 9:
-      divmod_integral(100000000);
-      break;
-    case 8:
-      divmod_integral(10000000);
-      break;
-    case 7:
-      divmod_integral(1000000);
-      break;
-    case 6:
-      divmod_integral(100000);
-      break;
-    case 5:
-      divmod_integral(10000);
-      break;
-    case 4:
-      divmod_integral(1000);
-      break;
-    case 3:
-      divmod_integral(100);
-      break;
-    case 2:
-      divmod_integral(10);
-      break;
-    case 1:
-      digit = integral;
-      integral = 0;
-      break;
-    default:
-      FMT_ASSERT(false, "invalid number of digits");
-    }
-    --exp;
-    auto remainder = (static_cast<uint64_t>(integral) << -one.e) + fractional;
-    auto result = handler.on_digit(static_cast<char>('0' + digit),
-                                   data::power_of_10_64[exp] << -one.e,
-                                   remainder, error, true);
-    if (result != digits::more) return result;
-  } while (exp > 0);
-  // Generate digits for the fractional part.
-  for (;;) {
-    fractional *= 10;
-    error *= 10;
-    char digit = static_cast<char>('0' + (fractional >> -one.e));
-    fractional &= one.f - 1;
-    --exp;
-    auto result = handler.on_digit(digit, one.f, fractional, error, false);
-    if (result != digits::more) return result;
-  }
-}
-
 class bigint {
  private:
   // A bigint is stored as an array of bigits (big digits), with bigit at index
@@ -2714,10 +2788,10 @@ class bigint {
   basic_memory_buffer<bigit, bigits_capacity> bigits_;
   int exp_;
 
-  FMT_CONSTEXPR20 bigit operator[](int index) const {
+  FMT_CONSTEXPR20 auto operator[](int index) const -> bigit {
     return bigits_[to_unsigned(index)];
   }
-  FMT_CONSTEXPR20 bigit& operator[](int index) {
+  FMT_CONSTEXPR20 auto operator[](int index) -> bigit& {
     return bigits_[to_unsigned(index)];
   }
 
@@ -2804,7 +2878,7 @@ class bigint {
     auto size = other.bigits_.size();
     bigits_.resize(size);
     auto data = other.bigits_.data();
-    std::copy(data, data + size, make_checked(bigits_.data(), size));
+    copy_str<bigit>(data, data + size, bigits_.data());
     exp_ = other.exp_;
   }
 
@@ -2813,11 +2887,11 @@ class bigint {
     assign(uint64_or_128_t<Int>(n));
   }
 
-  FMT_CONSTEXPR20 int num_bigits() const {
+  FMT_CONSTEXPR20 auto num_bigits() const -> int {
     return static_cast<int>(bigits_.size()) + exp_;
   }
 
-  FMT_NOINLINE FMT_CONSTEXPR20 bigint& operator<<=(int shift) {
+  FMT_NOINLINE FMT_CONSTEXPR20 auto operator<<=(int shift) -> bigint& {
     FMT_ASSERT(shift >= 0, "");
     exp_ += shift / bigit_bits;
     shift %= bigit_bits;
@@ -2832,13 +2906,15 @@ class bigint {
     return *this;
   }
 
-  template <typename Int> FMT_CONSTEXPR20 bigint& operator*=(Int value) {
+  template <typename Int>
+  FMT_CONSTEXPR20 auto operator*=(Int value) -> bigint& {
     FMT_ASSERT(value > 0, "");
     multiply(uint32_or_64_or_128_t<Int>(value));
     return *this;
   }
 
-  friend FMT_CONSTEXPR20 int compare(const bigint& lhs, const bigint& rhs) {
+  friend FMT_CONSTEXPR20 auto compare(const bigint& lhs, const bigint& rhs)
+      -> int {
     int num_lhs_bigits = lhs.num_bigits(), num_rhs_bigits = rhs.num_bigits();
     if (num_lhs_bigits != num_rhs_bigits)
       return num_lhs_bigits > num_rhs_bigits ? 1 : -1;
@@ -2855,8 +2931,9 @@ class bigint {
   }
 
   // Returns compare(lhs1 + lhs2, rhs).
-  friend FMT_CONSTEXPR20 int add_compare(const bigint& lhs1, const bigint& lhs2,
-                                         const bigint& rhs) {
+  friend FMT_CONSTEXPR20 auto add_compare(const bigint& lhs1,
+                                          const bigint& lhs2, const bigint& rhs)
+      -> int {
     auto minimum = [](int a, int b) { return a < b ? a : b; };
     auto maximum = [](int a, int b) { return a > b ? a : b; };
     int max_lhs_bigits = maximum(lhs1.num_bigits(), lhs2.num_bigits());
@@ -2937,13 +3014,13 @@ class bigint {
     bigits_.resize(to_unsigned(num_bigits + exp_difference));
     for (int i = num_bigits - 1, j = i + exp_difference; i >= 0; --i, --j)
       bigits_[j] = bigits_[i];
-    std::uninitialized_fill_n(bigits_.data(), exp_difference, 0);
+    std::uninitialized_fill_n(bigits_.data(), exp_difference, 0u);
     exp_ -= exp_difference;
   }
 
   // Divides this bignum by divisor, assigning the remainder to this and
   // returning the quotient.
-  FMT_CONSTEXPR20 int divmod_assign(const bigint& divisor) {
+  FMT_CONSTEXPR20 auto divmod_assign(const bigint& divisor) -> int {
     FMT_ASSERT(this != &divisor, "");
     if (compare(*this, divisor) < 0) return 0;
     FMT_ASSERT(divisor.bigits_[divisor.bigits_.size() - 1u] != 0, "");
@@ -3018,6 +3095,7 @@ FMT_CONSTEXPR20 inline void format_dragon(basic_fp<uint128_t> value,
   }
   int even = static_cast<int>((value.f & 1) == 0);
   if (!upper) upper = &lower;
+  bool shortest = num_digits < 0;
   if ((flags & dragon::fixup) != 0) {
     if (add_compare(numerator, *upper, denominator) + even <= 0) {
       --exp10;
@@ -3030,7 +3108,7 @@ FMT_CONSTEXPR20 inline void format_dragon(basic_fp<uint128_t> value,
     if ((flags & dragon::fixed) != 0) adjust_precision(num_digits, exp10 + 1);
   }
   // Invariant: value == (numerator / denominator) * pow(10, exp10).
-  if (num_digits < 0) {
+  if (shortest) {
     // Generate the shortest representation.
     num_digits = 0;
     char* data = buf.data();
@@ -3060,7 +3138,7 @@ FMT_CONSTEXPR20 inline void format_dragon(basic_fp<uint128_t> value,
   }
   // Generate the given number of digits.
   exp10 -= num_digits - 1;
-  if (num_digits == 0) {
+  if (num_digits <= 0) {
     denominator *= 10;
     auto digit = add_compare(numerator, numerator, denominator) > 0 ? '1' : '0';
     buf.push_back(digit);
@@ -3085,7 +3163,10 @@ FMT_CONSTEXPR20 inline void format_dragon(basic_fp<uint128_t> value,
       }
       if (buf[0] == overflow) {
         buf[0] = '1';
-        ++exp10;
+        if ((flags & dragon::fixed) != 0)
+          buf.push_back('0');
+        else
+          ++exp10;
       }
       return;
     }
@@ -3094,6 +3175,105 @@ FMT_CONSTEXPR20 inline void format_dragon(basic_fp<uint128_t> value,
   buf[num_digits - 1] = static_cast<char>('0' + digit);
 }
 
+// Formats a floating-point number using the hexfloat format.
+template <typename Float, FMT_ENABLE_IF(!is_double_double<Float>::value)>
+FMT_CONSTEXPR20 void format_hexfloat(Float value, int precision,
+                                     float_specs specs, buffer<char>& buf) {
+  // float is passed as double to reduce the number of instantiations and to
+  // simplify implementation.
+  static_assert(!std::is_same<Float, float>::value, "");
+
+  using info = dragonbox::float_info<Float>;
+
+  // Assume Float is in the format [sign][exponent][significand].
+  using carrier_uint = typename info::carrier_uint;
+
+  constexpr auto num_float_significand_bits =
+      detail::num_significand_bits<Float>();
+
+  basic_fp<carrier_uint> f(value);
+  f.e += num_float_significand_bits;
+  if (!has_implicit_bit<Float>()) --f.e;
+
+  constexpr auto num_fraction_bits =
+      num_float_significand_bits + (has_implicit_bit<Float>() ? 1 : 0);
+  constexpr auto num_xdigits = (num_fraction_bits + 3) / 4;
+
+  constexpr auto leading_shift = ((num_xdigits - 1) * 4);
+  const auto leading_mask = carrier_uint(0xF) << leading_shift;
+  const auto leading_xdigit =
+      static_cast<uint32_t>((f.f & leading_mask) >> leading_shift);
+  if (leading_xdigit > 1) f.e -= (32 - countl_zero(leading_xdigit) - 1);
+
+  int print_xdigits = num_xdigits - 1;
+  if (precision >= 0 && print_xdigits > precision) {
+    const int shift = ((print_xdigits - precision - 1) * 4);
+    const auto mask = carrier_uint(0xF) << shift;
+    const auto v = static_cast<uint32_t>((f.f & mask) >> shift);
+
+    if (v >= 8) {
+      const auto inc = carrier_uint(1) << (shift + 4);
+      f.f += inc;
+      f.f &= ~(inc - 1);
+    }
+
+    // Check long double overflow
+    if (!has_implicit_bit<Float>()) {
+      const auto implicit_bit = carrier_uint(1) << num_float_significand_bits;
+      if ((f.f & implicit_bit) == implicit_bit) {
+        f.f >>= 4;
+        f.e += 4;
+      }
+    }
+
+    print_xdigits = precision;
+  }
+
+  char xdigits[num_bits<carrier_uint>() / 4];
+  detail::fill_n(xdigits, sizeof(xdigits), '0');
+  format_uint<4>(xdigits, f.f, num_xdigits, specs.upper);
+
+  // Remove zero tail
+  while (print_xdigits > 0 && xdigits[print_xdigits] == '0') --print_xdigits;
+
+  buf.push_back('0');
+  buf.push_back(specs.upper ? 'X' : 'x');
+  buf.push_back(xdigits[0]);
+  if (specs.showpoint || print_xdigits > 0 || print_xdigits < precision)
+    buf.push_back('.');
+  buf.append(xdigits + 1, xdigits + 1 + print_xdigits);
+  for (; print_xdigits < precision; ++print_xdigits) buf.push_back('0');
+
+  buf.push_back(specs.upper ? 'P' : 'p');
+
+  uint32_t abs_e;
+  if (f.e < 0) {
+    buf.push_back('-');
+    abs_e = static_cast<uint32_t>(-f.e);
+  } else {
+    buf.push_back('+');
+    abs_e = static_cast<uint32_t>(f.e);
+  }
+  format_decimal<char>(appender(buf), abs_e, detail::count_digits(abs_e));
+}
+
+template <typename Float, FMT_ENABLE_IF(is_double_double<Float>::value)>
+FMT_CONSTEXPR20 void format_hexfloat(Float value, int precision,
+                                     float_specs specs, buffer<char>& buf) {
+  format_hexfloat(static_cast<double>(value), precision, specs, buf);
+}
+
+constexpr auto fractional_part_rounding_thresholds(int index) -> uint32_t {
+  // For checking rounding thresholds.
+  // The kth entry is chosen to be the smallest integer such that the
+  // upper 32-bits of 10^(k+1) times it is strictly bigger than 5 * 10^k.
+  // It is equal to ceil(2^31 + 2^32/10^(k + 1)).
+  // These are stored in a string literal because we cannot have static arrays
+  // in constexpr functions and non-static ones are poorly optimized.
+  return U"\x9999999a\x828f5c29\x80418938\x80068db9\x8000a7c6\x800010c7"
+         U"\x800001ae\x8000002b"[index];
+}
+
 template <typename Float>
 FMT_CONSTEXPR20 auto format_float(Float value, int precision, float_specs specs,
                                   buffer<char>& buf) -> int {
@@ -3116,7 +3296,7 @@ FMT_CONSTEXPR20 auto format_float(Float value, int precision, float_specs specs,
   int exp = 0;
   bool use_dragon = true;
   unsigned dragon_flags = 0;
-  if (!is_fast_float<Float>()) {
+  if (!is_fast_float<Float>() || is_constant_evaluated()) {
     const auto inv_log2_10 = 0.3010299956639812;  // 1 / log2(10)
     using info = dragonbox::float_info<decltype(converted_value)>;
     const auto f = basic_fp<typename info::carrier_uint>(converted_value);
@@ -3124,10 +3304,11 @@ FMT_CONSTEXPR20 auto format_float(Float value, int precision, float_specs specs,
     //   10^(exp - 1) <= value < 10^exp or 10^exp <= value < 10^(exp + 1).
     // This is based on log10(value) == log2(value) / log2(10) and approximation
     // of log2(value) by e + num_fraction_bits idea from double-conversion.
-    exp = static_cast<int>(
-        std::ceil((f.e + count_digits<1>(f.f) - 1) * inv_log2_10 - 1e-10));
+    auto e = (f.e + count_digits<1>(f.f) - 1) * inv_log2_10 - 1e-10;
+    exp = static_cast<int>(e);
+    if (e > exp) ++exp;  // Compute ceil.
     dragon_flags = dragon::fixup;
-  } else if (!is_constant_evaluated() && precision < 0) {
+  } else if (precision < 0) {
     // Use Dragonbox for the shortest format.
     if (specs.binary32) {
       auto dec = dragonbox::to_decimal(static_cast<float>(value));
@@ -3138,23 +3319,244 @@ FMT_CONSTEXPR20 auto format_float(Float value, int precision, float_specs specs,
     write<char>(buffer_appender<char>(buf), dec.significand);
     return dec.exponent;
   } else {
-    // Use Grisu + Dragon4 for the given precision:
-    // https://www.cs.tufts.edu/~nr/cs257/archive/florian-loitsch/printf.pdf.
-    const int min_exp = -60;  // alpha in Grisu.
-    int cached_exp10 = 0;     // K in Grisu.
-    fp normalized = normalize(fp(converted_value));
-    const auto cached_pow = get_cached_power(
-        min_exp - (normalized.e + fp::num_significand_bits), cached_exp10);
-    normalized = normalized * cached_pow;
-    gen_digits_handler handler{buf.data(), 0, precision, -cached_exp10, fixed};
-    if (grisu_gen_digits(normalized, 1, exp, handler) != digits::error &&
-        !is_constant_evaluated()) {
-      exp += handler.exp10;
-      buf.try_resize(to_unsigned(handler.size));
-      use_dragon = false;
+    // Extract significand bits and exponent bits.
+    using info = dragonbox::float_info<double>;
+    auto br = bit_cast<uint64_t>(static_cast<double>(value));
+
+    const uint64_t significand_mask =
+        (static_cast<uint64_t>(1) << num_significand_bits<double>()) - 1;
+    uint64_t significand = (br & significand_mask);
+    int exponent = static_cast<int>((br & exponent_mask<double>()) >>
+                                    num_significand_bits<double>());
+
+    if (exponent != 0) {  // Check if normal.
+      exponent -= exponent_bias<double>() + num_significand_bits<double>();
+      significand |=
+          (static_cast<uint64_t>(1) << num_significand_bits<double>());
+      significand <<= 1;
     } else {
-      exp += handler.size - cached_exp10 - 1;
-      precision = handler.precision;
+      // Normalize subnormal inputs.
+      FMT_ASSERT(significand != 0, "zeros should not appear here");
+      int shift = countl_zero(significand);
+      FMT_ASSERT(shift >= num_bits<uint64_t>() - num_significand_bits<double>(),
+                 "");
+      shift -= (num_bits<uint64_t>() - num_significand_bits<double>() - 2);
+      exponent = (std::numeric_limits<double>::min_exponent -
+                  num_significand_bits<double>()) -
+                 shift;
+      significand <<= shift;
+    }
+
+    // Compute the first several nonzero decimal significand digits.
+    // We call the number we get the first segment.
+    const int k = info::kappa - dragonbox::floor_log10_pow2(exponent);
+    exp = -k;
+    const int beta = exponent + dragonbox::floor_log2_pow10(k);
+    uint64_t first_segment;
+    bool has_more_segments;
+    int digits_in_the_first_segment;
+    {
+      const auto r = dragonbox::umul192_upper128(
+          significand << beta, dragonbox::get_cached_power(k));
+      first_segment = r.high();
+      has_more_segments = r.low() != 0;
+
+      // The first segment can have 18 ~ 19 digits.
+      if (first_segment >= 1000000000000000000ULL) {
+        digits_in_the_first_segment = 19;
+      } else {
+        // When it is of 18-digits, we align it to 19-digits by adding a bogus
+        // zero at the end.
+        digits_in_the_first_segment = 18;
+        first_segment *= 10;
+      }
+    }
+
+    // Compute the actual number of decimal digits to print.
+    if (fixed) adjust_precision(precision, exp + digits_in_the_first_segment);
+
+    // Use Dragon4 only when there might be not enough digits in the first
+    // segment.
+    if (digits_in_the_first_segment > precision) {
+      use_dragon = false;
+
+      if (precision <= 0) {
+        exp += digits_in_the_first_segment;
+
+        if (precision < 0) {
+          // Nothing to do, since all we have are just leading zeros.
+          buf.try_resize(0);
+        } else {
+          // We may need to round-up.
+          buf.try_resize(1);
+          if ((first_segment | static_cast<uint64_t>(has_more_segments)) >
+              5000000000000000000ULL) {
+            buf[0] = '1';
+          } else {
+            buf[0] = '0';
+          }
+        }
+      }  // precision <= 0
+      else {
+        exp += digits_in_the_first_segment - precision;
+
+        // When precision > 0, we divide the first segment into three
+        // subsegments, each with 9, 9, and 0 ~ 1 digits so that each fits
+        // in 32-bits which usually allows faster calculation than in
+        // 64-bits. Since some compiler (e.g. MSVC) doesn't know how to optimize
+        // division-by-constant for large 64-bit divisors, we do it here
+        // manually. The magic number 7922816251426433760 below is equal to
+        // ceil(2^(64+32) / 10^10).
+        const uint32_t first_subsegment = static_cast<uint32_t>(
+            dragonbox::umul128_upper64(first_segment, 7922816251426433760ULL) >>
+            32);
+        const uint64_t second_third_subsegments =
+            first_segment - first_subsegment * 10000000000ULL;
+
+        uint64_t prod;
+        uint32_t digits;
+        bool should_round_up;
+        int number_of_digits_to_print = precision > 9 ? 9 : precision;
+
+        // Print a 9-digits subsegment, either the first or the second.
+        auto print_subsegment = [&](uint32_t subsegment, char* buffer) {
+          int number_of_digits_printed = 0;
+
+          // If we want to print an odd number of digits from the subsegment,
+          if ((number_of_digits_to_print & 1) != 0) {
+            // Convert to 64-bit fixed-point fractional form with 1-digit
+            // integer part. The magic number 720575941 is a good enough
+            // approximation of 2^(32 + 24) / 10^8; see
+            // https://jk-jeon.github.io/posts/2022/12/fixed-precision-formatting/#fixed-length-case
+            // for details.
+            prod = ((subsegment * static_cast<uint64_t>(720575941)) >> 24) + 1;
+            digits = static_cast<uint32_t>(prod >> 32);
+            *buffer = static_cast<char>('0' + digits);
+            number_of_digits_printed++;
+          }
+          // If we want to print an even number of digits from the
+          // first_subsegment,
+          else {
+            // Convert to 64-bit fixed-point fractional form with 2-digits
+            // integer part. The magic number 450359963 is a good enough
+            // approximation of 2^(32 + 20) / 10^7; see
+            // https://jk-jeon.github.io/posts/2022/12/fixed-precision-formatting/#fixed-length-case
+            // for details.
+            prod = ((subsegment * static_cast<uint64_t>(450359963)) >> 20) + 1;
+            digits = static_cast<uint32_t>(prod >> 32);
+            copy2(buffer, digits2(digits));
+            number_of_digits_printed += 2;
+          }
+
+          // Print all digit pairs.
+          while (number_of_digits_printed < number_of_digits_to_print) {
+            prod = static_cast<uint32_t>(prod) * static_cast<uint64_t>(100);
+            digits = static_cast<uint32_t>(prod >> 32);
+            copy2(buffer + number_of_digits_printed, digits2(digits));
+            number_of_digits_printed += 2;
+          }
+        };
+
+        // Print first subsegment.
+        print_subsegment(first_subsegment, buf.data());
+
+        // Perform rounding if the first subsegment is the last subsegment to
+        // print.
+        if (precision <= 9) {
+          // Rounding inside the subsegment.
+          // We round-up if:
+          //  - either the fractional part is strictly larger than 1/2, or
+          //  - the fractional part is exactly 1/2 and the last digit is odd.
+          // We rely on the following observations:
+          //  - If fractional_part >= threshold, then the fractional part is
+          //    strictly larger than 1/2.
+          //  - If the MSB of fractional_part is set, then the fractional part
+          //    must be at least 1/2.
+          //  - When the MSB of fractional_part is set, either
+          //    second_third_subsegments being nonzero or has_more_segments
+          //    being true means there are further digits not printed, so the
+          //    fractional part is strictly larger than 1/2.
+          if (precision < 9) {
+            uint32_t fractional_part = static_cast<uint32_t>(prod);
+            should_round_up =
+                fractional_part >= fractional_part_rounding_thresholds(
+                                       8 - number_of_digits_to_print) ||
+                ((fractional_part >> 31) &
+                 ((digits & 1) | (second_third_subsegments != 0) |
+                  has_more_segments)) != 0;
+          }
+          // Rounding at the subsegment boundary.
+          // In this case, the fractional part is at least 1/2 if and only if
+          // second_third_subsegments >= 5000000000ULL, and is strictly larger
+          // than 1/2 if we further have either second_third_subsegments >
+          // 5000000000ULL or has_more_segments == true.
+          else {
+            should_round_up = second_third_subsegments > 5000000000ULL ||
+                              (second_third_subsegments == 5000000000ULL &&
+                               ((digits & 1) != 0 || has_more_segments));
+          }
+        }
+        // Otherwise, print the second subsegment.
+        else {
+          // Compilers are not aware of how to leverage the maximum value of
+          // second_third_subsegments to find out a better magic number which
+          // allows us to eliminate an additional shift. 1844674407370955162 =
+          // ceil(2^64/10) < ceil(2^64*(10^9/(10^10 - 1))).
+          const uint32_t second_subsegment =
+              static_cast<uint32_t>(dragonbox::umul128_upper64(
+                  second_third_subsegments, 1844674407370955162ULL));
+          const uint32_t third_subsegment =
+              static_cast<uint32_t>(second_third_subsegments) -
+              second_subsegment * 10;
+
+          number_of_digits_to_print = precision - 9;
+          print_subsegment(second_subsegment, buf.data() + 9);
+
+          // Rounding inside the subsegment.
+          if (precision < 18) {
+            // The condition third_subsegment != 0 implies that the segment was
+            // of 19 digits, so in this case the third segment should be
+            // consisting of a genuine digit from the input.
+            uint32_t fractional_part = static_cast<uint32_t>(prod);
+            should_round_up =
+                fractional_part >= fractional_part_rounding_thresholds(
+                                       8 - number_of_digits_to_print) ||
+                ((fractional_part >> 31) &
+                 ((digits & 1) | (third_subsegment != 0) |
+                  has_more_segments)) != 0;
+          }
+          // Rounding at the subsegment boundary.
+          else {
+            // In this case, the segment must be of 19 digits, thus
+            // the third subsegment should be consisting of a genuine digit from
+            // the input.
+            should_round_up = third_subsegment > 5 ||
+                              (third_subsegment == 5 &&
+                               ((digits & 1) != 0 || has_more_segments));
+          }
+        }
+
+        // Round-up if necessary.
+        if (should_round_up) {
+          ++buf[precision - 1];
+          for (int i = precision - 1; i > 0 && buf[i] > '9'; --i) {
+            buf[i] = '0';
+            ++buf[i - 1];
+          }
+          if (buf[0] > '9') {
+            buf[0] = '1';
+            if (fixed)
+              buf[precision++] = '0';
+            else
+              ++exp;
+          }
+        }
+        buf.try_resize(to_unsigned(precision));
+      }
+    }  // if (digits_in_the_first_segment > precision)
+    else {
+      // Adjust the exponent for its use in Dragon4.
+      exp += digits_in_the_first_segment - 1;
     }
   }
   if (use_dragon) {
@@ -3181,13 +3583,10 @@ FMT_CONSTEXPR20 auto format_float(Float value, int precision, float_specs specs,
   }
   return exp;
 }
-
-template <typename Char, typename OutputIt, typename T,
-          FMT_ENABLE_IF(is_floating_point<T>::value)>
-FMT_CONSTEXPR20 auto write(OutputIt out, T value,
-                           basic_format_specs<Char> specs, locale_ref loc = {})
+template <typename Char, typename OutputIt, typename T>
+FMT_CONSTEXPR20 auto write_float(OutputIt out, T value,
+                                 format_specs<Char> specs, locale_ref loc)
     -> OutputIt {
-  if (const_check(!is_supported_floating_point(value))) return out;
   float_specs fspecs = parse_float_type_spec(specs);
   fspecs.sign = specs.sign;
   if (detail::signbit(value)) {  // value < 0 is false for NaN so use signbit.
@@ -3211,7 +3610,7 @@ FMT_CONSTEXPR20 auto write(OutputIt out, T value,
   memory_buffer buffer;
   if (fspecs.format == float_format::hex) {
     if (fspecs.sign) buffer.push_back(detail::sign<char>(fspecs.sign));
-    snprintf_float(convert_float(value), specs.precision, fspecs, buffer);
+    format_hexfloat(convert_float(value), specs.precision, fspecs, buffer);
     return write_bytes<align::right>(out, {buffer.data(), buffer.size()},
                                      specs);
   }
@@ -3234,10 +3633,19 @@ FMT_CONSTEXPR20 auto write(OutputIt out, T value,
 }
 
 template <typename Char, typename OutputIt, typename T,
+          FMT_ENABLE_IF(is_floating_point<T>::value)>
+FMT_CONSTEXPR20 auto write(OutputIt out, T value, format_specs<Char> specs,
+                           locale_ref loc = {}) -> OutputIt {
+  if (const_check(!is_supported_floating_point(value))) return out;
+  return specs.localized && write_loc(out, value, specs, loc)
+             ? out
+             : write_float(out, value, specs, loc);
+}
+
+template <typename Char, typename OutputIt, typename T,
           FMT_ENABLE_IF(is_fast_float<T>::value)>
 FMT_CONSTEXPR20 auto write(OutputIt out, T value) -> OutputIt {
-  if (is_constant_evaluated())
-    return write(out, value, basic_format_specs<Char>());
+  if (is_constant_evaluated()) return write(out, value, format_specs<Char>());
   if (const_check(!is_supported_floating_point(value))) return out;
 
   auto fspecs = float_specs();
@@ -3246,11 +3654,11 @@ FMT_CONSTEXPR20 auto write(OutputIt out, T value) -> OutputIt {
     value = -value;
   }
 
-  constexpr auto specs = basic_format_specs<Char>();
+  constexpr auto specs = format_specs<Char>();
   using floaty = conditional_t<std::is_same<T, long double>::value, double, T>;
-  using uint = typename dragonbox::float_info<floaty>::carrier_uint;
-  uint mask = exponent_mask<floaty>();
-  if ((bit_cast<uint>(value) & mask) == mask)
+  using floaty_uint = typename dragonbox::float_info<floaty>::carrier_uint;
+  floaty_uint mask = exponent_mask<floaty>();
+  if ((bit_cast<floaty_uint>(value) & mask) == mask)
     return write_nonfinite(out, std::isnan(value), specs, fspecs);
 
   auto dec = dragonbox::to_decimal(static_cast<floaty>(value));
@@ -3261,12 +3669,12 @@ template <typename Char, typename OutputIt, typename T,
           FMT_ENABLE_IF(is_floating_point<T>::value &&
                         !is_fast_float<T>::value)>
 inline auto write(OutputIt out, T value) -> OutputIt {
-  return write(out, value, basic_format_specs<Char>());
+  return write(out, value, format_specs<Char>());
 }
 
 template <typename Char, typename OutputIt>
-auto write(OutputIt out, monostate, basic_format_specs<Char> = {},
-           locale_ref = {}) -> OutputIt {
+auto write(OutputIt out, monostate, format_specs<Char> = {}, locale_ref = {})
+    -> OutputIt {
   FMT_ASSERT(false, "");
   return out;
 }
@@ -3300,8 +3708,8 @@ FMT_CONSTEXPR auto write(OutputIt out, T value) -> OutputIt {
 template <typename Char, typename OutputIt, typename T,
           FMT_ENABLE_IF(std::is_same<T, bool>::value)>
 FMT_CONSTEXPR auto write(OutputIt out, T value,
-                         const basic_format_specs<Char>& specs = {},
-                         locale_ref = {}) -> OutputIt {
+                         const format_specs<Char>& specs = {}, locale_ref = {})
+    -> OutputIt {
   return specs.type != presentation_type::none &&
                  specs.type != presentation_type::string
              ? write(out, value ? 1 : 0, specs, {})
@@ -3318,20 +3726,15 @@ FMT_CONSTEXPR auto write(OutputIt out, Char value) -> OutputIt {
 template <typename Char, typename OutputIt>
 FMT_CONSTEXPR_CHAR_TRAITS auto write(OutputIt out, const Char* value)
     -> OutputIt {
-  if (!value) {
-    throw_format_error("string pointer is null");
-  } else {
-    out = write(out, basic_string_view<Char>(value));
-  }
+  if (value) return write(out, basic_string_view<Char>(value));
+  throw_format_error("string pointer is null");
   return out;
 }
 
 template <typename Char, typename OutputIt, typename T,
           FMT_ENABLE_IF(std::is_same<T, void>::value)>
-auto write(OutputIt out, const T* value,
-           const basic_format_specs<Char>& specs = {}, locale_ref = {})
-    -> OutputIt {
-  check_pointer_type_spec(specs.type, error_handler());
+auto write(OutputIt out, const T* value, const format_specs<Char>& specs = {},
+           locale_ref = {}) -> OutputIt {
   return write_ptr<Char>(out, bit_cast<uintptr_t>(value), &specs);
 }
 
@@ -3341,8 +3744,8 @@ template <typename Char, typename OutputIt, typename T,
 FMT_CONSTEXPR auto write(OutputIt out, const T& value) -> enable_if_t<
     std::is_class<T>::value && !is_string<T>::value &&
         !is_floating_point<T>::value && !std::is_same<T, Char>::value &&
-        !std::is_same<const T&,
-                      decltype(arg_mapper<Context>().map(value))>::value,
+        !std::is_same<T, remove_cvref_t<decltype(arg_mapper<Context>().map(
+                             value))>>::value,
     OutputIt> {
   return write<Char>(out, arg_mapper<Context>().map(value));
 }
@@ -3352,12 +3755,11 @@ template <typename Char, typename OutputIt, typename T,
 FMT_CONSTEXPR auto write(OutputIt out, const T& value)
     -> enable_if_t<mapped_type_constant<T, Context>::value == type::custom_type,
                    OutputIt> {
-  using formatter_type =
-      conditional_t<has_formatter<T, Context>::value,
-                    typename Context::template formatter_type<T>,
-                    fallback_formatter<T, Char>>;
+  auto formatter = typename Context::template formatter_type<T>();
+  auto parse_ctx = typename Context::parse_context_type({});
+  formatter.parse(parse_ctx);
   auto ctx = Context(out, {}, {});
-  return formatter_type().format(value, ctx);
+  return formatter.format(value, ctx);
 }
 
 // An argument visitor that formats the argument and writes it via the output
@@ -3386,7 +3788,7 @@ template <typename Char> struct arg_formatter {
   using context = buffer_context<Char>;
 
   iterator out;
-  const basic_format_specs<Char>& specs;
+  const format_specs<Char>& specs;
   locale_ref locale;
 
   template <typename T>
@@ -3400,122 +3802,50 @@ template <typename Char> struct arg_formatter {
   }
 };
 
-template <typename Char> struct custom_formatter {
-  basic_format_parse_context<Char>& parse_ctx;
-  buffer_context<Char>& ctx;
-
-  void operator()(
-      typename basic_format_arg<buffer_context<Char>>::handle h) const {
-    h.format(parse_ctx, ctx);
-  }
-  template <typename T> void operator()(T) const {}
-};
-
-template <typename T>
-using is_integer =
-    bool_constant<is_integral<T>::value && !std::is_same<T, bool>::value &&
-                  !std::is_same<T, char>::value &&
-                  !std::is_same<T, wchar_t>::value>;
-
-template <typename ErrorHandler> class width_checker {
- public:
-  explicit FMT_CONSTEXPR width_checker(ErrorHandler& eh) : handler_(eh) {}
-
+struct width_checker {
   template <typename T, FMT_ENABLE_IF(is_integer<T>::value)>
   FMT_CONSTEXPR auto operator()(T value) -> unsigned long long {
-    if (is_negative(value)) handler_.on_error("negative width");
+    if (is_negative(value)) throw_format_error("negative width");
     return static_cast<unsigned long long>(value);
   }
 
   template <typename T, FMT_ENABLE_IF(!is_integer<T>::value)>
   FMT_CONSTEXPR auto operator()(T) -> unsigned long long {
-    handler_.on_error("width is not integer");
+    throw_format_error("width is not integer");
     return 0;
   }
-
- private:
-  ErrorHandler& handler_;
 };
 
-template <typename ErrorHandler> class precision_checker {
- public:
-  explicit FMT_CONSTEXPR precision_checker(ErrorHandler& eh) : handler_(eh) {}
-
+struct precision_checker {
   template <typename T, FMT_ENABLE_IF(is_integer<T>::value)>
   FMT_CONSTEXPR auto operator()(T value) -> unsigned long long {
-    if (is_negative(value)) handler_.on_error("negative precision");
+    if (is_negative(value)) throw_format_error("negative precision");
     return static_cast<unsigned long long>(value);
   }
 
   template <typename T, FMT_ENABLE_IF(!is_integer<T>::value)>
   FMT_CONSTEXPR auto operator()(T) -> unsigned long long {
-    handler_.on_error("precision is not integer");
+    throw_format_error("precision is not integer");
     return 0;
   }
-
- private:
-  ErrorHandler& handler_;
 };
 
-template <template <typename> class Handler, typename FormatArg,
-          typename ErrorHandler>
-FMT_CONSTEXPR auto get_dynamic_spec(FormatArg arg, ErrorHandler eh) -> int {
-  unsigned long long value = visit_format_arg(Handler<ErrorHandler>(eh), arg);
-  if (value > to_unsigned(max_value<int>())) eh.on_error("number is too big");
+template <typename Handler, typename FormatArg>
+FMT_CONSTEXPR auto get_dynamic_spec(FormatArg arg) -> int {
+  unsigned long long value = visit_format_arg(Handler(), arg);
+  if (value > to_unsigned(max_value<int>()))
+    throw_format_error("number is too big");
   return static_cast<int>(value);
 }
 
 template <typename Context, typename ID>
-FMT_CONSTEXPR auto get_arg(Context& ctx, ID id) ->
-    typename Context::format_arg {
+FMT_CONSTEXPR auto get_arg(Context& ctx, ID id) -> decltype(ctx.arg(id)) {
   auto arg = ctx.arg(id);
   if (!arg) ctx.on_error("argument not found");
   return arg;
 }
 
-// The standard format specifier handler with checking.
-template <typename Char> class specs_handler : public specs_setter<Char> {
- private:
-  basic_format_parse_context<Char>& parse_context_;
-  buffer_context<Char>& context_;
-
-  // This is only needed for compatibility with gcc 4.4.
-  using format_arg = basic_format_arg<buffer_context<Char>>;
-
-  FMT_CONSTEXPR auto get_arg(auto_id) -> format_arg {
-    return detail::get_arg(context_, parse_context_.next_arg_id());
-  }
-
-  FMT_CONSTEXPR auto get_arg(int arg_id) -> format_arg {
-    parse_context_.check_arg_id(arg_id);
-    return detail::get_arg(context_, arg_id);
-  }
-
-  FMT_CONSTEXPR auto get_arg(basic_string_view<Char> arg_id) -> format_arg {
-    parse_context_.check_arg_id(arg_id);
-    return detail::get_arg(context_, arg_id);
-  }
-
- public:
-  FMT_CONSTEXPR specs_handler(basic_format_specs<Char>& specs,
-                              basic_format_parse_context<Char>& parse_ctx,
-                              buffer_context<Char>& ctx)
-      : specs_setter<Char>(specs), parse_context_(parse_ctx), context_(ctx) {}
-
-  template <typename Id> FMT_CONSTEXPR void on_dynamic_width(Id arg_id) {
-    this->specs_.width = get_dynamic_spec<width_checker>(
-        get_arg(arg_id), context_.error_handler());
-  }
-
-  template <typename Id> FMT_CONSTEXPR void on_dynamic_precision(Id arg_id) {
-    this->specs_.precision = get_dynamic_spec<precision_checker>(
-        get_arg(arg_id), context_.error_handler());
-  }
-
-  void on_error(const char* message) { context_.on_error(message); }
-};
-
-template <template <typename> class Handler, typename Context>
+template <typename Handler, typename Context>
 FMT_CONSTEXPR void handle_dynamic_spec(int& value,
                                        arg_ref<typename Context::char_type> ref,
                                        Context& ctx) {
@@ -3523,26 +3853,15 @@ FMT_CONSTEXPR void handle_dynamic_spec(int& value,
   case arg_id_kind::none:
     break;
   case arg_id_kind::index:
-    value = detail::get_dynamic_spec<Handler>(ctx.arg(ref.val.index),
-                                              ctx.error_handler());
+    value = detail::get_dynamic_spec<Handler>(get_arg(ctx, ref.val.index));
     break;
   case arg_id_kind::name:
-    value = detail::get_dynamic_spec<Handler>(ctx.arg(ref.val.name),
-                                              ctx.error_handler());
+    value = detail::get_dynamic_spec<Handler>(get_arg(ctx, ref.val.name));
     break;
   }
 }
 
 #if FMT_USE_USER_DEFINED_LITERALS
-template <typename Char> struct udl_formatter {
-  basic_string_view<Char> str;
-
-  template <typename... T>
-  auto operator()(T&&... args) const -> std::basic_string<Char> {
-    return vformat(str, fmt::make_format_args<buffer_context<Char>>(args...));
-  }
-};
-
 #  if FMT_USE_NONTYPE_TEMPLATE_ARGS
 template <typename T, typename Char, size_t N,
           fmt::detail_exported::fixed_string<Char, N> Str>
@@ -3581,12 +3900,12 @@ template <typename Char> struct udl_arg {
 #endif  // FMT_USE_USER_DEFINED_LITERALS
 
 template <typename Locale, typename Char>
-auto vformat(const Locale& loc, basic_string_view<Char> format_str,
+auto vformat(const Locale& loc, basic_string_view<Char> fmt,
              basic_format_args<buffer_context<type_identity_t<Char>>> args)
     -> std::basic_string<Char> {
-  basic_memory_buffer<Char> buffer;
-  detail::vformat_to(buffer, format_str, args, detail::locale_ref(loc));
-  return {buffer.data(), buffer.size()};
+  auto buf = basic_memory_buffer<Char>();
+  detail::vformat_to(buf, fmt, args, detail::locale_ref(loc));
+  return {buf.data(), buf.size()};
 }
 
 using format_func = void (*)(detail::buffer<char>&, int, const char*);
@@ -3596,28 +3915,28 @@ FMT_API void format_error_code(buffer<char>& out, int error_code,
 
 FMT_API void report_error(format_func func, int error_code,
                           const char* message) noexcept;
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
 FMT_API auto vsystem_error(int error_code, string_view format_str,
                            format_args args) -> std::system_error;
 
 /**
- \rst
- Constructs :class:`std::system_error` with a message formatted with
- ``fmt::format(fmt, args...)``.
+  \rst
+  Constructs :class:`std::system_error` with a message formatted with
+  ``fmt::format(fmt, args...)``.
   *error_code* is a system error code as given by ``errno``.
 
- **Example**::
-
-   // This throws std::system_error with the description
-   //   cannot open file 'madeup': No such file or directory
-   // or similar (system message may vary).
-   const char* filename = "madeup";
-   std::FILE* file = std::fopen(filename, "r");
-   if (!file)
-     throw fmt::system_error(errno, "cannot open file '{}'", filename);
- \endrst
-*/
+  **Example**::
+
+    // This throws std::system_error with the description
+    //   cannot open file 'madeup': No such file or directory
+    // or similar (system message may vary).
+    const char* filename = "madeup";
+    std::FILE* file = std::fopen(filename, "r");
+    if (!file)
+      throw fmt::system_error(errno, "cannot open file '{}'", filename);
+  \endrst
+ */
 template <typename... T>
 auto system_error(int error_code, format_string<T...> fmt, T&&... args)
     -> std::system_error {
@@ -3708,93 +4027,33 @@ class format_int {
 };
 
 template <typename T, typename Char>
-template <typename FormatContext>
-FMT_CONSTEXPR FMT_INLINE auto
-formatter<T, Char,
-          enable_if_t<detail::type_constant<T, Char>::value !=
-                      detail::type::custom_type>>::format(const T& val,
-                                                          FormatContext& ctx)
-    const -> decltype(ctx.out()) {
-  if (specs_.width_ref.kind != detail::arg_id_kind::none ||
-      specs_.precision_ref.kind != detail::arg_id_kind::none) {
-    auto specs = specs_;
-    detail::handle_dynamic_spec<detail::width_checker>(specs.width,
-                                                       specs.width_ref, ctx);
-    detail::handle_dynamic_spec<detail::precision_checker>(
-        specs.precision, specs.precision_ref, ctx);
-    return detail::write<Char>(ctx.out(), val, specs, ctx.locale());
-  }
-  return detail::write<Char>(ctx.out(), val, specs_, ctx.locale());
-}
-
-template <typename Char>
-struct formatter<void*, Char> : formatter<const void*, Char> {
+struct formatter<T, Char, enable_if_t<detail::has_format_as<T>::value>>
+    : formatter<detail::format_as_t<T>, Char> {
   template <typename FormatContext>
-  auto format(void* val, FormatContext& ctx) const -> decltype(ctx.out()) {
-    return formatter<const void*, Char>::format(val, ctx);
+  auto format(const T& value, FormatContext& ctx) const -> decltype(ctx.out()) {
+    using base = formatter<detail::format_as_t<T>, Char>;
+    return base::format(format_as(value), ctx);
   }
 };
 
-template <typename Char, size_t N>
-struct formatter<Char[N], Char> : formatter<basic_string_view<Char>, Char> {
-  template <typename FormatContext>
-  FMT_CONSTEXPR auto format(const Char* val, FormatContext& ctx) const
-      -> decltype(ctx.out()) {
-    return formatter<basic_string_view<Char>, Char>::format(val, ctx);
-  }
-};
-
-// A formatter for types known only at run time such as variant alternatives.
-//
-// Usage:
-//   using variant = std::variant<int, std::string>;
-//   template <>
-//   struct formatter<variant>: dynamic_formatter<> {
-//     auto format(const variant& v, format_context& ctx) {
-//       return visit([&](const auto& val) {
-//           return dynamic_formatter<>::format(val, ctx);
-//       }, v);
-//     }
-//   };
-template <typename Char = char> class dynamic_formatter {
- private:
-  detail::dynamic_format_specs<Char> specs_;
-  const Char* format_str_;
-
-  struct null_handler : detail::error_handler {
-    void on_align(align_t) {}
-    void on_sign(sign_t) {}
-    void on_hash() {}
-  };
-
-  template <typename Context> void handle_specs(Context& ctx) {
-    detail::handle_dynamic_spec<detail::width_checker>(specs_.width,
-                                                       specs_.width_ref, ctx);
-    detail::handle_dynamic_spec<detail::precision_checker>(
-        specs_.precision, specs_.precision_ref, ctx);
-  }
+#define FMT_FORMAT_AS(Type, Base) \
+  template <typename Char>        \
+  struct formatter<Type, Char> : formatter<Base, Char> {}
+
+FMT_FORMAT_AS(signed char, int);
+FMT_FORMAT_AS(unsigned char, unsigned);
+FMT_FORMAT_AS(short, int);
+FMT_FORMAT_AS(unsigned short, unsigned);
+FMT_FORMAT_AS(long, detail::long_type);
+FMT_FORMAT_AS(unsigned long, detail::ulong_type);
+FMT_FORMAT_AS(Char*, const Char*);
+FMT_FORMAT_AS(std::basic_string<Char>, basic_string_view<Char>);
+FMT_FORMAT_AS(std::nullptr_t, const void*);
+FMT_FORMAT_AS(detail::std_string_view<Char>, basic_string_view<Char>);
+FMT_FORMAT_AS(void*, const void*);
 
- public:
-  template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
-    format_str_ = ctx.begin();
-    // Checks are deferred to formatting time when the argument type is known.
-    detail::dynamic_specs_handler<ParseContext> handler(specs_, ctx);
-    return detail::parse_format_specs(ctx.begin(), ctx.end(), handler);
-  }
-
-  template <typename T, typename FormatContext>
-  auto format(const T& val, FormatContext& ctx) -> decltype(ctx.out()) {
-    handle_specs(ctx);
-    detail::specs_checker<null_handler> checker(
-        null_handler(), detail::mapped_type_constant<T, FormatContext>::value);
-    checker.on_align(specs_.align);
-    if (specs_.sign != sign::none) checker.on_sign(specs_.sign);
-    if (specs_.alt) checker.on_hash();
-    if (specs_.precision >= 0) checker.end_precision();
-    return detail::write<Char>(ctx.out(), val, specs_, ctx.locale());
-  }
-};
+template <typename Char, size_t N>
+struct formatter<Char[N], Char> : formatter<basic_string_view<Char>, Char> {};
 
 /**
   \rst
@@ -3809,7 +4068,8 @@ template <typename T> auto ptr(T p) -> const void* {
   static_assert(std::is_pointer<T>::value, "");
   return detail::bit_cast<const void*>(p);
 }
-template <typename T> auto ptr(const std::unique_ptr<T>& p) -> const void* {
+template <typename T, typename Deleter>
+auto ptr(const std::unique_ptr<T, Deleter>& p) -> const void* {
   return p.get();
 }
 template <typename T> auto ptr(const std::shared_ptr<T>& p) -> const void* {
@@ -3849,17 +4109,13 @@ class bytes {
 
 template <> struct formatter<bytes> {
  private:
-  detail::dynamic_format_specs<char> specs_;
+  detail::dynamic_format_specs<> specs_;
 
  public:
   template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
-    using handler_type = detail::dynamic_specs_handler<ParseContext>;
-    detail::specs_checker<handler_type> handler(handler_type(specs_, ctx),
-                                                detail::type::string_type);
-    auto it = parse_format_specs(ctx.begin(), ctx.end(), handler);
-    detail::check_string_type_spec(specs_.type, ctx.error_handler());
-    return it;
+  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> const char* {
+    return parse_format_specs(ctx.begin(), ctx.end(), specs_, ctx,
+                              detail::type::string_type);
   }
 
   template <typename FormatContext>
@@ -3873,7 +4129,9 @@ template <> struct formatter<bytes> {
 };
 
 // group_digits_view is not derived from view because it copies the argument.
-template <typename T> struct group_digits_view { T value; };
+template <typename T> struct group_digits_view {
+  T value;
+};
 
 /**
   \rst
@@ -3892,17 +4150,13 @@ template <typename T> auto group_digits(T value) -> group_digits_view<T> {
 
 template <typename T> struct formatter<group_digits_view<T>> : formatter<T> {
  private:
-  detail::dynamic_format_specs<char> specs_;
+  detail::dynamic_format_specs<> specs_;
 
  public:
   template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
-    using handler_type = detail::dynamic_specs_handler<ParseContext>;
-    detail::specs_checker<handler_type> handler(handler_type(specs_, ctx),
-                                                detail::type::int_type);
-    auto it = parse_format_specs(ctx.begin(), ctx.end(), handler);
-    detail::check_string_type_spec(specs_.type, ctx.error_handler());
-    return it;
+  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> const char* {
+    return parse_format_specs(ctx.begin(), ctx.end(), specs_, ctx,
+                              detail::type::int_type);
   }
 
   template <typename FormatContext>
@@ -3912,12 +4166,66 @@ template <typename T> struct formatter<group_digits_view<T>> : formatter<T> {
                                                        specs_.width_ref, ctx);
     detail::handle_dynamic_spec<detail::precision_checker>(
         specs_.precision, specs_.precision_ref, ctx);
-    return detail::write_int_localized(
+    return detail::write_int(
         ctx.out(), static_cast<detail::uint64_or_128_t<T>>(t.value), 0, specs_,
-        detail::digit_grouping<char>({"\3", ','}));
+        detail::digit_grouping<char>("\3", ","));
   }
 };
 
+template <typename T> struct nested_view {
+  const formatter<T>* fmt;
+  const T* value;
+};
+
+template <typename T> struct formatter<nested_view<T>> {
+  FMT_CONSTEXPR auto parse(format_parse_context& ctx) -> const char* {
+    return ctx.begin();
+  }
+  auto format(nested_view<T> view, format_context& ctx) const
+      -> decltype(ctx.out()) {
+    return view.fmt->format(*view.value, ctx);
+  }
+};
+
+template <typename T> struct nested_formatter {
+ private:
+  int width_;
+  detail::fill_t<char> fill_;
+  align_t align_ : 4;
+  formatter<T> formatter_;
+
+ public:
+  constexpr nested_formatter() : width_(0), align_(align_t::none) {}
+
+  FMT_CONSTEXPR auto parse(format_parse_context& ctx) -> const char* {
+    auto specs = detail::dynamic_format_specs<char>();
+    auto it = parse_format_specs(ctx.begin(), ctx.end(), specs, ctx,
+                                 detail::type::none_type);
+    width_ = specs.width;
+    fill_ = specs.fill;
+    align_ = specs.align;
+    ctx.advance_to(it);
+    return formatter_.parse(ctx);
+  }
+
+  template <typename F>
+  auto write_padded(format_context& ctx, F write) const -> decltype(ctx.out()) {
+    if (width_ == 0) return write(ctx.out());
+    auto buf = memory_buffer();
+    write(std::back_inserter(buf));
+    auto specs = format_specs<>();
+    specs.width = width_;
+    specs.fill = fill_;
+    specs.align = align_;
+    return detail::write(ctx.out(), string_view(buf.data(), buf.size()), specs);
+  }
+
+  auto nested(const T& value) const -> nested_view<T> {
+    return nested_view<T>{&formatter_, &value};
+  }
+};
+
+// DEPRECATED! join_view will be moved to ranges.h.
 template <typename It, typename Sentinel, typename Char = char>
 struct join_view : detail::view {
   It begin;
@@ -3937,30 +4245,11 @@ struct formatter<join_view<It, Sentinel, Char>, Char> {
 #else
       typename std::iterator_traits<It>::value_type;
 #endif
-  using context = buffer_context<Char>;
-  using mapper = detail::arg_mapper<context>;
-
-  template <typename T, FMT_ENABLE_IF(has_formatter<T, context>::value)>
-  static auto map(const T& value) -> const T& {
-    return value;
-  }
-  template <typename T, FMT_ENABLE_IF(!has_formatter<T, context>::value)>
-  static auto map(const T& value) -> decltype(mapper().map(value)) {
-    return mapper().map(value);
-  }
-
-  using formatter_type =
-      conditional_t<is_formattable<value_type, Char>::value,
-                    formatter<remove_cvref_t<decltype(map(
-                                  std::declval<const value_type&>()))>,
-                              Char>,
-                    detail::fallback_formatter<value_type, Char>>;
-
-  formatter_type value_formatter_;
+  formatter<remove_cvref_t<value_type>, Char> value_formatter_;
 
  public:
   template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
+  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> const Char* {
     return value_formatter_.parse(ctx);
   }
 
@@ -3970,12 +4259,12 @@ struct formatter<join_view<It, Sentinel, Char>, Char> {
     auto it = value.begin;
     auto out = ctx.out();
     if (it != value.end) {
-      out = value_formatter_.format(map(*it), ctx);
+      out = value_formatter_.format(*it, ctx);
       ++it;
       while (it != value.end) {
         out = detail::copy_str<Char>(value.sep.begin(), value.sep.end(), out);
         ctx.advance_to(out);
-        out = value_formatter_.format(map(*it), ctx);
+        out = value_formatter_.format(*it, ctx);
         ++it;
       }
     }
@@ -4025,11 +4314,12 @@ auto join(Range&& range, string_view sep)
     std::string answer = fmt::to_string(42);
   \endrst
  */
-template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value)>
+template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value &&
+                                    !detail::has_format_as<T>::value)>
 inline auto to_string(const T& value) -> std::string {
-  auto result = std::string();
-  detail::write<char>(std::back_inserter(result), value);
-  return result;
+  auto buffer = memory_buffer();
+  detail::write<char>(appender(buffer), value);
+  return {buffer.data(), buffer.size()};
 }
 
 template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
@@ -4050,31 +4340,23 @@ FMT_NODISCARD auto to_string(const basic_memory_buffer<Char, SIZE>& buf)
   return std::basic_string<Char>(buf.data(), size);
 }
 
-FMT_BEGIN_DETAIL_NAMESPACE
+template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value &&
+                                    detail::has_format_as<T>::value)>
+inline auto to_string(const T& value) -> std::string {
+  return to_string(format_as(value));
+}
+
+FMT_END_EXPORT
+
+namespace detail {
 
 template <typename Char>
-void vformat_to(
-    buffer<Char>& buf, basic_string_view<Char> fmt,
-    basic_format_args<FMT_BUFFER_CONTEXT(type_identity_t<Char>)> args,
-    locale_ref loc) {
-  // workaround for msvc bug regarding name-lookup in module
-  // link names into function scope
-  using detail::arg_formatter;
-  using detail::buffer_appender;
-  using detail::custom_formatter;
-  using detail::default_arg_formatter;
-  using detail::get_arg;
-  using detail::locale_ref;
-  using detail::parse_format_specs;
-  using detail::specs_checker;
-  using detail::specs_handler;
-  using detail::to_unsigned;
-  using detail::type;
-  using detail::write;
+void vformat_to(buffer<Char>& buf, basic_string_view<Char> fmt,
+                typename vformat_args<Char>::type args, locale_ref loc) {
   auto out = buffer_appender<Char>(buf);
   if (fmt.size() == 2 && equal2(fmt.data(), "{}")) {
     auto arg = args.get(0);
-    if (!arg) error_handler().on_error("argument not found");
+    if (!arg) throw_format_error("argument not found");
     visit_format_arg(default_arg_formatter<Char>{out, args, loc}, arg);
     return;
   }
@@ -4101,7 +4383,7 @@ void vformat_to(
     }
     FMT_CONSTEXPR auto on_arg_id(basic_string_view<Char> id) -> int {
       int arg_id = context.arg_id(id);
-      if (arg_id < 0) on_error("argument not found");
+      if (arg_id < 0) throw_format_error("argument not found");
       return arg_id;
     }
 
@@ -4116,18 +4398,17 @@ void vformat_to(
     auto on_format_specs(int id, const Char* begin, const Char* end)
         -> const Char* {
       auto arg = get_arg(context, id);
-      if (arg.type() == type::custom_type) {
-        parse_context.advance_to(parse_context.begin() +
-                                 (begin - &*parse_context.begin()));
-        visit_format_arg(custom_formatter<Char>{parse_context, context}, arg);
+      // Not using a visitor for custom types gives better codegen.
+      if (arg.format_custom(begin, parse_context, context))
         return parse_context.begin();
-      }
-      auto specs = basic_format_specs<Char>();
-      specs_checker<specs_handler<Char>> handler(
-          specs_handler<Char>(specs, parse_context, context), arg.type());
-      begin = parse_format_specs(begin, end, handler);
+      auto specs = detail::dynamic_format_specs<Char>();
+      begin = parse_format_specs(begin, end, specs, parse_context, arg.type());
+      detail::handle_dynamic_spec<detail::width_checker>(
+          specs.width, specs.width_ref, context);
+      detail::handle_dynamic_spec<detail::precision_checker>(
+          specs.precision, specs.precision_ref, context);
       if (begin == end || *begin != '}')
-        on_error("missing '}' in format string");
+        throw_format_error("missing '}' in format string");
       auto f = arg_formatter<Char>{context.out(), specs, context.locale()};
       context.advance_to(visit_format_arg(f, arg));
       return begin;
@@ -4136,7 +4417,12 @@ void vformat_to(
   detail::parse_format_string<false>(fmt, format_handler(out, fmt, args, loc));
 }
 
+FMT_BEGIN_EXPORT
+
 #ifndef FMT_HEADER_ONLY
+extern template FMT_API void vformat_to(buffer<char>&, string_view,
+                                        typename vformat_args<>::type,
+                                        locale_ref);
 extern template FMT_API auto thousands_sep_impl<char>(locale_ref)
     -> thousands_sep_result<char>;
 extern template FMT_API auto thousands_sep_impl<wchar_t>(locale_ref)
@@ -4145,7 +4431,7 @@ extern template FMT_API auto decimal_point_impl(locale_ref) -> char;
 extern template FMT_API auto decimal_point_impl(locale_ref) -> wchar_t;
 #endif  // FMT_HEADER_ONLY
 
-FMT_END_DETAIL_NAMESPACE
+}  // namespace detail
 
 #if FMT_USE_USER_DEFINED_LITERALS
 inline namespace literals {
@@ -4165,7 +4451,7 @@ template <detail_exported::fixed_string Str> constexpr auto operator""_a() {
   return detail::udl_arg<char_t, sizeof(Str.data) / sizeof(char_t), Str>();
 }
 #  else
-constexpr auto operator"" _a(const char* s, size_t) -> detail::udl_arg<char> {
+constexpr auto operator""_a(const char* s, size_t) -> detail::udl_arg<char> {
   return {s};
 }
 #  endif
@@ -4182,7 +4468,7 @@ template <typename Locale, typename... T,
           FMT_ENABLE_IF(detail::is_locale<Locale>::value)>
 inline auto format(const Locale& loc, format_string<T...> fmt, T&&... args)
     -> std::string {
-  return vformat(loc, string_view(fmt), fmt::make_format_args(args...));
+  return fmt::vformat(loc, string_view(fmt), fmt::make_format_args(args...));
 }
 
 template <typename OutputIt, typename Locale,
@@ -4193,7 +4479,7 @@ auto vformat_to(OutputIt out, const Locale& loc, string_view fmt,
   using detail::get_buffer;
   auto&& buf = get_buffer<char>(out);
   detail::vformat_to(buf, fmt, args, detail::locale_ref(loc));
-  return detail::get_iterator(buf);
+  return detail::get_iterator(buf, out);
 }
 
 template <typename OutputIt, typename Locale, typename... T,
@@ -4204,7 +4490,39 @@ FMT_INLINE auto format_to(OutputIt out, const Locale& loc,
   return vformat_to(out, loc, fmt, fmt::make_format_args(args...));
 }
 
-FMT_MODULE_EXPORT_END
+template <typename Locale, typename... T,
+          FMT_ENABLE_IF(detail::is_locale<Locale>::value)>
+FMT_NODISCARD FMT_INLINE auto formatted_size(const Locale& loc,
+                                             format_string<T...> fmt,
+                                             T&&... args) -> size_t {
+  auto buf = detail::counting_buffer<>();
+  detail::vformat_to<char>(buf, fmt, fmt::make_format_args(args...),
+                           detail::locale_ref(loc));
+  return buf.count();
+}
+
+FMT_END_EXPORT
+
+template <typename T, typename Char>
+template <typename FormatContext>
+FMT_CONSTEXPR FMT_INLINE auto
+formatter<T, Char,
+          enable_if_t<detail::type_constant<T, Char>::value !=
+                      detail::type::custom_type>>::format(const T& val,
+                                                          FormatContext& ctx)
+    const -> decltype(ctx.out()) {
+  if (specs_.width_ref.kind == detail::arg_id_kind::none &&
+      specs_.precision_ref.kind == detail::arg_id_kind::none) {
+    return detail::write<Char>(ctx.out(), val, specs_, ctx.locale());
+  }
+  auto specs = specs_;
+  detail::handle_dynamic_spec<detail::width_checker>(specs.width,
+                                                     specs.width_ref, ctx);
+  detail::handle_dynamic_spec<detail::precision_checker>(
+      specs.precision, specs.precision_ref, ctx);
+  return detail::write<Char>(ctx.out(), val, specs, ctx.locale());
+}
+
 FMT_END_NAMESPACE
 
 #ifdef FMT_HEADER_ONLY
diff --git a/dep/fmt/include/fmt/os.h b/dep/fmt/include/fmt/os.h
index d82be1125a4..3c7b3ccb481 100644
--- a/dep/fmt/include/fmt/os.h
+++ b/dep/fmt/include/fmt/os.h
@@ -13,12 +13,14 @@
 #include <cstdio>
 #include <system_error>  // std::system_error
 
+#include "format.h"
+
 #if defined __APPLE__ || defined(__FreeBSD__)
-#  include <xlocale.h>  // for LC_NUMERIC_MASK on OS X
+#  if FMT_HAS_INCLUDE(<xlocale.h>)
+#    include <xlocale.h>  // for LC_NUMERIC_MASK on OS X
+#  endif
 #endif
 
-#include "format.h"
-
 #ifndef FMT_USE_FCNTL
 // UWP doesn't provide _pipe.
 #  if FMT_HAS_INCLUDE("winapifamily.h")
@@ -46,6 +48,7 @@
 
 // Calls to system functions are wrapped in FMT_SYSTEM for testability.
 #ifdef FMT_SYSTEM
+#  define FMT_HAS_SYSTEM
 #  define FMT_POSIX_CALL(call) FMT_SYSTEM(call)
 #else
 #  define FMT_SYSTEM(call) ::call
@@ -71,7 +74,7 @@
 #define FMT_RETRY(result, expression) FMT_RETRY_VAL(result, expression, -1)
 
 FMT_BEGIN_NAMESPACE
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
 /**
   \rst
@@ -114,57 +117,19 @@ template <typename Char> class basic_cstring_view {
   basic_cstring_view(const std::basic_string<Char>& s) : data_(s.c_str()) {}
 
   /** Returns the pointer to a C string. */
-  const Char* c_str() const { return data_; }
+  auto c_str() const -> const Char* { return data_; }
 };
 
 using cstring_view = basic_cstring_view<char>;
 using wcstring_view = basic_cstring_view<wchar_t>;
 
-template <typename Char> struct formatter<std::error_code, Char> {
-  template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
-    return ctx.begin();
-  }
-
-  template <typename FormatContext>
-  FMT_CONSTEXPR auto format(const std::error_code& ec, FormatContext& ctx) const
-      -> decltype(ctx.out()) {
-    auto out = ctx.out();
-    out = detail::write_bytes(out, ec.category().name(),
-                              basic_format_specs<Char>());
-    out = detail::write<Char>(out, Char(':'));
-    out = detail::write<Char>(out, ec.value());
-    return out;
-  }
-};
-
 #ifdef _WIN32
 FMT_API const std::error_category& system_category() noexcept;
 
-FMT_BEGIN_DETAIL_NAMESPACE
-// A converter from UTF-16 to UTF-8.
-// It is only provided for Windows since other systems support UTF-8 natively.
-class utf16_to_utf8 {
- private:
-  memory_buffer buffer_;
-
- public:
-  utf16_to_utf8() {}
-  FMT_API explicit utf16_to_utf8(basic_string_view<wchar_t> s);
-  operator string_view() const { return string_view(&buffer_[0], size()); }
-  size_t size() const { return buffer_.size() - 1; }
-  const char* c_str() const { return &buffer_[0]; }
-  std::string str() const { return std::string(&buffer_[0], size()); }
-
-  // Performs conversion returning a system error code instead of
-  // throwing exception on conversion error. This method may still throw
-  // in case of memory allocation error.
-  FMT_API int convert(basic_string_view<wchar_t> s);
-};
-
+namespace detail {
 FMT_API void format_windows_error(buffer<char>& out, int error_code,
                                   const char* message) noexcept;
-FMT_END_DETAIL_NAMESPACE
+}
 
 FMT_API std::system_error vwindows_error(int error_code, string_view format_str,
                                          format_args args);
@@ -207,7 +172,7 @@ std::system_error windows_error(int error_code, string_view message,
 // Can be used to report errors from destructors.
 FMT_API void report_windows_error(int error_code, const char* message) noexcept;
 #else
-inline const std::error_category& system_category() noexcept {
+inline auto system_category() noexcept -> const std::error_category& {
   return std::system_category();
 }
 #endif  // _WIN32
@@ -244,7 +209,7 @@ class buffered_file {
     other.file_ = nullptr;
   }
 
-  buffered_file& operator=(buffered_file&& other) {
+  auto operator=(buffered_file&& other) -> buffered_file& {
     close();
     file_ = other.file_;
     other.file_ = nullptr;
@@ -258,9 +223,9 @@ class buffered_file {
   FMT_API void close();
 
   // Returns the pointer to a FILE object representing this file.
-  FILE* get() const noexcept { return file_; }
+  auto get() const noexcept -> FILE* { return file_; }
 
-  FMT_API int descriptor() const;
+  FMT_API auto descriptor() const -> int;
 
   void vprint(string_view format_str, format_args args) {
     fmt::vprint(file_, format_str, args);
@@ -310,7 +275,7 @@ class FMT_API file {
   file(file&& other) noexcept : fd_(other.fd_) { other.fd_ = -1; }
 
   // Move assignment is not noexcept because close may throw.
-  file& operator=(file&& other) {
+  auto operator=(file&& other) -> file& {
     close();
     fd_ = other.fd_;
     other.fd_ = -1;
@@ -321,24 +286,24 @@ class FMT_API file {
   ~file() noexcept;
 
   // Returns the file descriptor.
-  int descriptor() const noexcept { return fd_; }
+  auto descriptor() const noexcept -> int { return fd_; }
 
   // Closes the file.
   void close();
 
   // Returns the file size. The size has signed type for consistency with
   // stat::st_size.
-  long long size() const;
+  auto size() const -> long long;
 
   // Attempts to read count bytes from the file into the specified buffer.
-  size_t read(void* buffer, size_t count);
+  auto read(void* buffer, size_t count) -> size_t;
 
   // Attempts to write count bytes from the specified buffer to the file.
-  size_t write(const void* buffer, size_t count);
+  auto write(const void* buffer, size_t count) -> size_t;
 
   // Duplicates a file descriptor with the dup function and returns
   // the duplicate as a file object.
-  static file dup(int fd);
+  static auto dup(int fd) -> file;
 
   // Makes fd be the copy of this file descriptor, closing fd first if
   // necessary.
@@ -350,22 +315,29 @@ class FMT_API file {
 
   // Creates a pipe setting up read_end and write_end file objects for reading
   // and writing respectively.
+  // DEPRECATED! Taking files as out parameters is deprecated.
   static void pipe(file& read_end, file& write_end);
 
   // Creates a buffered_file object associated with this file and detaches
   // this file object from the file.
-  buffered_file fdopen(const char* mode);
+  auto fdopen(const char* mode) -> buffered_file;
+
+#  if defined(_WIN32) && !defined(__MINGW32__)
+  // Opens a file and constructs a file object representing this file by
+  // wcstring_view filename. Windows only.
+  static file open_windows_file(wcstring_view path, int oflag);
+#  endif
 };
 
 // Returns the memory page size.
-long getpagesize();
+auto getpagesize() -> long;
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 struct buffer_size {
   buffer_size() = default;
   size_t value = 0;
-  buffer_size operator=(size_t val) const {
+  auto operator=(size_t val) const -> buffer_size {
     auto bs = buffer_size();
     bs.value = val;
     return bs;
@@ -397,56 +369,61 @@ struct ostream_params {
 #  endif
 };
 
-FMT_END_DETAIL_NAMESPACE
+class file_buffer final : public buffer<char> {
+  file file_;
+
+  FMT_API void grow(size_t) override;
+
+ public:
+  FMT_API file_buffer(cstring_view path, const ostream_params& params);
+  FMT_API file_buffer(file_buffer&& other);
+  FMT_API ~file_buffer();
+
+  void flush() {
+    if (size() == 0) return;
+    file_.write(data(), size() * sizeof(data()[0]));
+    clear();
+  }
+
+  void close() {
+    flush();
+    file_.close();
+  }
+};
+
+}  // namespace detail
 
 // Added {} below to work around default constructor error known to
 // occur in Xcode versions 7.2.1 and 8.2.1.
 constexpr detail::buffer_size buffer_size{};
 
 /** A fast output stream which is not thread-safe. */
-class FMT_API ostream final : private detail::buffer<char> {
+class FMT_API ostream {
  private:
-  file file_;
-
-  void grow(size_t) override;
+  FMT_MSC_WARNING(suppress : 4251)
+  detail::file_buffer buffer_;
 
   ostream(cstring_view path, const detail::ostream_params& params)
-      : file_(path, params.oflag) {
-    set(new char[params.buffer_size], params.buffer_size);
-  }
+      : buffer_(path, params) {}
 
  public:
-  ostream(ostream&& other)
-      : detail::buffer<char>(other.data(), other.size(), other.capacity()),
-        file_(std::move(other.file_)) {
-    other.clear();
-    other.set(nullptr, 0);
-  }
-  ~ostream() {
-    flush();
-    delete[] data();
-  }
+  ostream(ostream&& other) : buffer_(std::move(other.buffer_)) {}
 
-  void flush() {
-    if (size() == 0) return;
-    file_.write(data(), size());
-    clear();
-  }
+  ~ostream();
+
+  void flush() { buffer_.flush(); }
 
   template <typename... T>
-  friend ostream output_file(cstring_view path, T... params);
+  friend auto output_file(cstring_view path, T... params) -> ostream;
 
-  void close() {
-    flush();
-    file_.close();
-  }
+  void close() { buffer_.close(); }
 
   /**
     Formats ``args`` according to specifications in ``fmt`` and writes the
     output to the file.
    */
   template <typename... T> void print(format_string<T...> fmt, T&&... args) {
-    vformat_to(detail::buffer_appender<char>(*this), fmt,
+    vformat_to(std::back_inserter(buffer_), fmt,
                fmt::make_format_args(args...));
   }
 };
@@ -467,12 +444,12 @@ class FMT_API ostream final : private detail::buffer<char> {
   \endrst
  */
 template <typename... T>
-inline ostream output_file(cstring_view path, T... params) {
+inline auto output_file(cstring_view path, T... params) -> ostream {
   return {path, detail::ostream_params(params...)};
 }
 #endif  // FMT_USE_FCNTL
 
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_OS_H_
diff --git a/dep/fmt/include/fmt/ostream.h b/dep/fmt/include/fmt/ostream.h
index c3cdd4a61b2..26fb3b5ac07 100644
--- a/dep/fmt/include/fmt/ostream.h
+++ b/dep/fmt/include/fmt/ostream.h
@@ -8,61 +8,58 @@
 #ifndef FMT_OSTREAM_H_
 #define FMT_OSTREAM_H_
 
-#include <fstream>
-#include <ostream>
-#if defined(_WIN32) && defined(__GLIBCXX__)
-#  include <ext/stdio_filebuf.h>
-#  include <ext/stdio_sync_filebuf.h>
-#elif defined(_WIN32) && defined(_LIBCPP_VERSION)
-#  include <__std_stream>
+#include <fstream>  // std::filebuf
+
+#ifdef _WIN32
+#  ifdef __GLIBCXX__
+#    include <ext/stdio_filebuf.h>
+#    include <ext/stdio_sync_filebuf.h>
+#  endif
+#  include <io.h>
 #endif
 
 #include "format.h"
 
 FMT_BEGIN_NAMESPACE
-
-template <typename OutputIt, typename Char> class basic_printf_context;
-
 namespace detail {
 
-// Checks if T has a user-defined operator<<.
-template <typename T, typename Char, typename Enable = void>
-class is_streamable {
+template <typename Streambuf> class formatbuf : public Streambuf {
  private:
-  template <typename U>
-  static auto test(int)
-      -> bool_constant<sizeof(std::declval<std::basic_ostream<Char>&>()
-                              << std::declval<U>()) != 0>;
-
-  template <typename> static auto test(...) -> std::false_type;
+  using char_type = typename Streambuf::char_type;
+  using streamsize = decltype(std::declval<Streambuf>().sputn(nullptr, 0));
+  using int_type = typename Streambuf::int_type;
+  using traits_type = typename Streambuf::traits_type;
 
-  using result = decltype(test<T>(0));
+  buffer<char_type>& buffer_;
 
  public:
-  is_streamable() = default;
+  explicit formatbuf(buffer<char_type>& buf) : buffer_(buf) {}
+
+ protected:
+  // The put area is always empty. This makes the implementation simpler and has
+  // the advantage that the streambuf and the buffer are always in sync and
+  // sputc never writes into uninitialized memory. A disadvantage is that each
+  // call to sputc always results in a (virtual) call to overflow. There is no
+  // disadvantage here for sputn since this always results in a call to xsputn.
+
+  auto overflow(int_type ch) -> int_type override {
+    if (!traits_type::eq_int_type(ch, traits_type::eof()))
+      buffer_.push_back(static_cast<char_type>(ch));
+    return ch;
+  }
 
-  static const bool value = result::value;
+  auto xsputn(const char_type* s, streamsize count) -> streamsize override {
+    buffer_.append(s, s + count);
+    return count;
+  }
 };
 
-// Formatting of built-in types and arrays is intentionally disabled because
-// it's handled by standard (non-ostream) formatters.
-template <typename T, typename Char>
-struct is_streamable<
-    T, Char,
-    enable_if_t<
-        std::is_arithmetic<T>::value || std::is_array<T>::value ||
-        std::is_pointer<T>::value || std::is_same<T, char8_type>::value ||
-        std::is_convertible<T, fmt::basic_string_view<Char>>::value ||
-        std::is_same<T, std_string_view<Char>>::value ||
-        (std::is_convertible<T, int>::value && !std::is_enum<T>::value)>>
-    : std::false_type {};
-
 // Generate a unique explicit instantion in every translation unit using a tag
 // type in an anonymous namespace.
 namespace {
 struct file_access_tag {};
 }  // namespace
-template <class Tag, class BufType, FILE* BufType::*FileMemberPtr>
+template <typename Tag, typename BufType, FILE* BufType::*FileMemberPtr>
 class file_access {
   friend auto get_file(BufType& obj) -> FILE* { return obj.*FileMemberPtr; }
 };
@@ -71,36 +68,40 @@ class file_access {
 template class file_access<file_access_tag, std::filebuf,
                            &std::filebuf::_Myfile>;
 auto get_file(std::filebuf&) -> FILE*;
-#elif defined(_WIN32) && defined(_LIBCPP_VERSION)
-template class file_access<file_access_tag, std::__stdoutbuf<char>,
-                           &std::__stdoutbuf<char>::__file_>;
-auto get_file(std::__stdoutbuf<char>&) -> FILE*;
 #endif
 
-inline bool write_ostream_unicode(std::ostream& os, fmt::string_view data) {
+inline auto write_ostream_unicode(std::ostream& os, fmt::string_view data)
+    -> bool {
+  FILE* f = nullptr;
 #if FMT_MSC_VERSION
   if (auto* buf = dynamic_cast<std::filebuf*>(os.rdbuf()))
-    if (FILE* f = get_file(*buf)) return write_console(f, data);
+    f = get_file(*buf);
+  else
+    return false;
 #elif defined(_WIN32) && defined(__GLIBCXX__)
   auto* rdbuf = os.rdbuf();
-  FILE* c_file;
-  if (auto* fbuf = dynamic_cast<__gnu_cxx::stdio_sync_filebuf<char>*>(rdbuf))
-    c_file = fbuf->file();
+  if (auto* sfbuf = dynamic_cast<__gnu_cxx::stdio_sync_filebuf<char>*>(rdbuf))
+    f = sfbuf->file();
   else if (auto* fbuf = dynamic_cast<__gnu_cxx::stdio_filebuf<char>*>(rdbuf))
-    c_file = fbuf->file();
+    f = fbuf->file();
   else
     return false;
-  if (c_file) return write_console(c_file, data);
-#elif defined(_WIN32) && defined(_LIBCPP_VERSION)
-  if (auto* buf = dynamic_cast<std::__stdoutbuf<char>*>(os.rdbuf()))
-    if (FILE* f = get_file(*buf)) return write_console(f, data);
 #else
-  ignore_unused(os, data);
+  ignore_unused(os, data, f);
+#endif
+#ifdef _WIN32
+  if (f) {
+    int fd = _fileno(f);
+    if (_isatty(fd)) {
+      os.flush();
+      return write_console(fd, data);
+    }
+  }
 #endif
   return false;
 }
-inline bool write_ostream_unicode(std::wostream&,
-                                  fmt::basic_string_view<wchar_t>) {
+inline auto write_ostream_unicode(std::wostream&,
+                                  fmt::basic_string_view<wchar_t>) -> bool {
   return false;
 }
 
@@ -121,18 +122,19 @@ void write_buffer(std::basic_ostream<Char>& os, buffer<Char>& buf) {
 }
 
 template <typename Char, typename T>
-void format_value(buffer<Char>& buf, const T& value,
-                  locale_ref loc = locale_ref()) {
+void format_value(buffer<Char>& buf, const T& value) {
   auto&& format_buf = formatbuf<std::basic_streambuf<Char>>(buf);
   auto&& output = std::basic_ostream<Char>(&format_buf);
 #if !defined(FMT_STATIC_THOUSANDS_SEPARATOR)
-  if (loc) output.imbue(loc.get<std::locale>());
+  output.imbue(std::locale::classic());  // The default is always unlocalized.
 #endif
   output << value;
   output.exceptions(std::ios_base::failbit | std::ios_base::badbit);
 }
 
-template <typename T> struct streamed_view { const T& value; };
+template <typename T> struct streamed_view {
+  const T& value;
+};
 
 }  // namespace detail
 
@@ -145,7 +147,7 @@ struct basic_ostream_formatter : formatter<basic_string_view<Char>, Char> {
   auto format(const T& value, basic_format_context<OutputIt, Char>& ctx) const
       -> OutputIt {
     auto buffer = basic_memory_buffer<Char>();
-    format_value(buffer, value, ctx.locale());
+    detail::format_value(buffer, value);
     return formatter<basic_string_view<Char>, Char>::format(
         {buffer.data(), buffer.size()}, ctx);
   }
@@ -174,19 +176,12 @@ struct formatter<detail::streamed_view<T>, Char>
   \endrst
  */
 template <typename T>
-auto streamed(const T& value) -> detail::streamed_view<T> {
+constexpr auto streamed(const T& value) -> detail::streamed_view<T> {
   return {value};
 }
 
 namespace detail {
 
-// Formats an object of type T that has an overloaded ostream operator<<.
-template <typename T, typename Char>
-struct fallback_formatter<T, Char, enable_if_t<is_streamable<T, Char>::value>>
-    : basic_ostream_formatter<Char> {
-  using basic_ostream_formatter<Char>::format;
-};
-
 inline void vprint_directly(std::ostream& os, string_view format_str,
                             format_args args) {
   auto buffer = memory_buffer();
@@ -196,7 +191,7 @@ inline void vprint_directly(std::ostream& os, string_view format_str,
 
 }  // namespace detail
 
-FMT_MODULE_EXPORT template <typename Char>
+FMT_EXPORT template <typename Char>
 void vprint(std::basic_ostream<Char>& os,
             basic_string_view<type_identity_t<Char>> format_str,
             basic_format_args<buffer_context<type_identity_t<Char>>> args) {
@@ -215,7 +210,7 @@ void vprint(std::basic_ostream<Char>& os,
     fmt::print(cerr, "Don't {}!", "panic");
   \endrst
  */
-FMT_MODULE_EXPORT template <typename... T>
+FMT_EXPORT template <typename... T>
 void print(std::ostream& os, format_string<T...> fmt, T&&... args) {
   const auto& vargs = fmt::make_format_args(args...);
   if (detail::is_utf8())
@@ -224,7 +219,7 @@ void print(std::ostream& os, format_string<T...> fmt, T&&... args) {
     detail::vprint_directly(os, fmt, vargs);
 }
 
-FMT_MODULE_EXPORT
+FMT_EXPORT
 template <typename... Args>
 void print(std::wostream& os,
            basic_format_string<wchar_t, type_identity_t<Args>...> fmt,
@@ -232,6 +227,19 @@ void print(std::wostream& os,
   vprint(os, fmt, fmt::make_format_args<buffer_context<wchar_t>>(args...));
 }
 
+FMT_EXPORT template <typename... T>
+void println(std::ostream& os, format_string<T...> fmt, T&&... args) {
+  fmt::print(os, "{}\n", fmt::format(fmt, std::forward<T>(args)...));
+}
+
+FMT_EXPORT
+template <typename... Args>
+void println(std::wostream& os,
+             basic_format_string<wchar_t, type_identity_t<Args>...> fmt,
+             Args&&... args) {
+  print(os, L"{}\n", fmt::format(fmt, std::forward<Args>(args)...));
+}
+
 FMT_END_NAMESPACE
 
 #endif  // FMT_OSTREAM_H_
diff --git a/dep/fmt/include/fmt/printf.h b/dep/fmt/include/fmt/printf.h
index 70a592dc26d..07e81577cf3 100644
--- a/dep/fmt/include/fmt/printf.h
+++ b/dep/fmt/include/fmt/printf.h
@@ -14,24 +14,24 @@
 #include "format.h"
 
 FMT_BEGIN_NAMESPACE
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
-template <typename T> struct printf_formatter { printf_formatter() = delete; };
-
-template <typename Char>
-class basic_printf_parse_context : public basic_format_parse_context<Char> {
-  using basic_format_parse_context<Char>::basic_format_parse_context;
+template <typename T> struct printf_formatter {
+  printf_formatter() = delete;
 };
 
-template <typename OutputIt, typename Char> class basic_printf_context {
+template <typename Char> class basic_printf_context {
  private:
-  OutputIt out_;
+  detail::buffer_appender<Char> out_;
   basic_format_args<basic_printf_context> args_;
 
+  static_assert(std::is_same<Char, char>::value ||
+                    std::is_same<Char, wchar_t>::value,
+                "Unsupported code unit type.");
+
  public:
   using char_type = Char;
-  using format_arg = basic_format_arg<basic_printf_context>;
-  using parse_context_type = basic_printf_parse_context<Char>;
+  using parse_context_type = basic_format_parse_context<Char>;
   template <typename T> using formatter_type = printf_formatter<T>;
 
   /**
@@ -40,75 +40,77 @@ template <typename OutputIt, typename Char> class basic_printf_context {
     stored in the context object so make sure they have appropriate lifetimes.
     \endrst
    */
-  basic_printf_context(OutputIt out,
+  basic_printf_context(detail::buffer_appender<Char> out,
                        basic_format_args<basic_printf_context> args)
       : out_(out), args_(args) {}
 
-  OutputIt out() { return out_; }
-  void advance_to(OutputIt it) { out_ = it; }
+  auto out() -> detail::buffer_appender<Char> { return out_; }
+  void advance_to(detail::buffer_appender<Char>) {}
 
-  detail::locale_ref locale() { return {}; }
+  auto locale() -> detail::locale_ref { return {}; }
 
-  format_arg arg(int id) const { return args_.get(id); }
+  auto arg(int id) const -> basic_format_arg<basic_printf_context> {
+    return args_.get(id);
+  }
 
   FMT_CONSTEXPR void on_error(const char* message) {
     detail::error_handler().on_error(message);
   }
 };
 
-FMT_BEGIN_DETAIL_NAMESPACE
+namespace detail {
 
 // Checks if a value fits in int - used to avoid warnings about comparing
 // signed and unsigned integers.
 template <bool IsSigned> struct int_checker {
-  template <typename T> static bool fits_in_int(T value) {
+  template <typename T> static auto fits_in_int(T value) -> bool {
     unsigned max = max_value<int>();
     return value <= max;
   }
-  static bool fits_in_int(bool) { return true; }
+  static auto fits_in_int(bool) -> bool { return true; }
 };
 
 template <> struct int_checker<true> {
-  template <typename T> static bool fits_in_int(T value) {
+  template <typename T> static auto fits_in_int(T value) -> bool {
     return value >= (std::numeric_limits<int>::min)() &&
            value <= max_value<int>();
   }
-  static bool fits_in_int(int) { return true; }
+  static auto fits_in_int(int) -> bool { return true; }
 };
 
-class printf_precision_handler {
- public:
+struct printf_precision_handler {
   template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
-  int operator()(T value) {
+  auto operator()(T value) -> int {
     if (!int_checker<std::numeric_limits<T>::is_signed>::fits_in_int(value))
-      FMT_THROW(format_error("number is too big"));
+      throw_format_error("number is too big");
     return (std::max)(static_cast<int>(value), 0);
   }
 
   template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value)>
-  int operator()(T) {
-    FMT_THROW(format_error("precision is not integer"));
+  auto operator()(T) -> int {
+    throw_format_error("precision is not integer");
     return 0;
   }
 };
 
 // An argument visitor that returns true iff arg is a zero integer.
-class is_zero_int {
- public:
+struct is_zero_int {
   template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
-  bool operator()(T value) {
+  auto operator()(T value) -> bool {
     return value == 0;
   }
 
   template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value)>
-  bool operator()(T) {
+  auto operator()(T) -> bool {
     return false;
   }
 };
 
 template <typename T> struct make_unsigned_or_bool : std::make_unsigned<T> {};
 
-template <> struct make_unsigned_or_bool<bool> { using type = bool; };
+template <> struct make_unsigned_or_bool<bool> {
+  using type = bool;
+};
 
 template <typename T, typename Context> class arg_converter {
  private:
@@ -132,22 +134,23 @@ template <typename T, typename Context> class arg_converter {
     if (const_check(sizeof(target_type) <= sizeof(int))) {
       // Extra casts are used to silence warnings.
       if (is_signed) {
-        arg_ = detail::make_arg<Context>(
-            static_cast<int>(static_cast<target_type>(value)));
+        auto n = static_cast<int>(static_cast<target_type>(value));
+        arg_ = detail::make_arg<Context>(n);
       } else {
         using unsigned_type = typename make_unsigned_or_bool<target_type>::type;
-        arg_ = detail::make_arg<Context>(
-            static_cast<unsigned>(static_cast<unsigned_type>(value)));
+        auto n = static_cast<unsigned>(static_cast<unsigned_type>(value));
+        arg_ = detail::make_arg<Context>(n);
       }
     } else {
       if (is_signed) {
         // glibc's printf doesn't sign extend arguments of smaller types:
         //   std::printf("%lld", -42);  // prints "4294967254"
         // but we don't have to do the same because it's a UB.
-        arg_ = detail::make_arg<Context>(static_cast<long long>(value));
+        auto n = static_cast<long long>(value);
+        arg_ = detail::make_arg<Context>(n);
       } else {
-        arg_ = detail::make_arg<Context>(
-            static_cast<typename make_unsigned_or_bool<U>::type>(value));
+        auto n = static_cast<typename make_unsigned_or_bool<U>::type>(value);
+        arg_ = detail::make_arg<Context>(n);
       }
     }
   }
@@ -175,8 +178,8 @@ template <typename Context> class char_converter {
 
   template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
   void operator()(T value) {
-    arg_ = detail::make_arg<Context>(
-        static_cast<typename Context::char_type>(value));
+    auto c = static_cast<typename Context::char_type>(value);
+    arg_ = detail::make_arg<Context>(c);
   }
 
   template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value)>
@@ -186,122 +189,131 @@ template <typename Context> class char_converter {
 // An argument visitor that return a pointer to a C string if argument is a
 // string or null otherwise.
 template <typename Char> struct get_cstring {
-  template <typename T> const Char* operator()(T) { return nullptr; }
-  const Char* operator()(const Char* s) { return s; }
+  template <typename T> auto operator()(T) -> const Char* { return nullptr; }
+  auto operator()(const Char* s) -> const Char* { return s; }
 };
 
 // Checks if an argument is a valid printf width specifier and sets
 // left alignment if it is negative.
 template <typename Char> class printf_width_handler {
  private:
-  using format_specs = basic_format_specs<Char>;
-
-  format_specs& specs_;
+  format_specs<Char>& specs_;
 
  public:
-  explicit printf_width_handler(format_specs& specs) : specs_(specs) {}
+  explicit printf_width_handler(format_specs<Char>& specs) : specs_(specs) {}
 
   template <typename T, FMT_ENABLE_IF(std::is_integral<T>::value)>
-  unsigned operator()(T value) {
+  auto operator()(T value) -> unsigned {
     auto width = static_cast<uint32_or_64_or_128_t<T>>(value);
     if (detail::is_negative(value)) {
       specs_.align = align::left;
       width = 0 - width;
     }
     unsigned int_max = max_value<int>();
-    if (width > int_max) FMT_THROW(format_error("number is too big"));
+    if (width > int_max) throw_format_error("number is too big");
     return static_cast<unsigned>(width);
   }
 
   template <typename T, FMT_ENABLE_IF(!std::is_integral<T>::value)>
-  unsigned operator()(T) {
-    FMT_THROW(format_error("width is not integer"));
+  auto operator()(T) -> unsigned {
+    throw_format_error("width is not integer");
     return 0;
   }
 };
 
+// Workaround for a bug with the XL compiler when initializing
+// printf_arg_formatter's base class.
+template <typename Char>
+auto make_arg_formatter(buffer_appender<Char> iter, format_specs<Char>& s)
+    -> arg_formatter<Char> {
+  return {iter, s, locale_ref()};
+}
+
 // The ``printf`` argument formatter.
-template <typename OutputIt, typename Char>
+template <typename Char>
 class printf_arg_formatter : public arg_formatter<Char> {
  private:
   using base = arg_formatter<Char>;
-  using context_type = basic_printf_context<OutputIt, Char>;
-  using format_specs = basic_format_specs<Char>;
+  using context_type = basic_printf_context<Char>;
 
   context_type& context_;
 
-  OutputIt write_null_pointer(bool is_string = false) {
+  void write_null_pointer(bool is_string = false) {
     auto s = this->specs;
     s.type = presentation_type::none;
-    return write_bytes(this->out, is_string ? "(null)" : "(nil)", s);
+    write_bytes(this->out, is_string ? "(null)" : "(nil)", s);
   }
 
  public:
-  printf_arg_formatter(OutputIt iter, format_specs& s, context_type& ctx)
-      : base{iter, s, locale_ref()}, context_(ctx) {}
+  printf_arg_formatter(buffer_appender<Char> iter, format_specs<Char>& s,
+                       context_type& ctx)
+      : base(make_arg_formatter(iter, s)), context_(ctx) {}
 
-  OutputIt operator()(monostate value) { return base::operator()(value); }
+  void operator()(monostate value) { base::operator()(value); }
 
   template <typename T, FMT_ENABLE_IF(detail::is_integral<T>::value)>
-  OutputIt operator()(T value) {
+  void operator()(T value) {
     // MSVC2013 fails to compile separate overloads for bool and Char so use
     // std::is_same instead.
-    if (std::is_same<T, Char>::value) {
-      format_specs fmt_specs = this->specs;
-      if (fmt_specs.type != presentation_type::none &&
-          fmt_specs.type != presentation_type::chr) {
-        return (*this)(static_cast<int>(value));
-      }
-      fmt_specs.sign = sign::none;
-      fmt_specs.alt = false;
-      fmt_specs.fill[0] = ' ';  // Ignore '0' flag for char types.
-      // align::numeric needs to be overwritten here since the '0' flag is
-      // ignored for non-numeric types
-      if (fmt_specs.align == align::none || fmt_specs.align == align::numeric)
-        fmt_specs.align = align::right;
-      return write<Char>(this->out, static_cast<Char>(value), fmt_specs);
+    if (!std::is_same<T, Char>::value) {
+      base::operator()(value);
+      return;
     }
-    return base::operator()(value);
+    format_specs<Char> fmt_specs = this->specs;
+    if (fmt_specs.type != presentation_type::none &&
+        fmt_specs.type != presentation_type::chr) {
+      return (*this)(static_cast<int>(value));
+    }
+    fmt_specs.sign = sign::none;
+    fmt_specs.alt = false;
+    fmt_specs.fill[0] = ' ';  // Ignore '0' flag for char types.
+    // align::numeric needs to be overwritten here since the '0' flag is
+    // ignored for non-numeric types
+    if (fmt_specs.align == align::none || fmt_specs.align == align::numeric)
+      fmt_specs.align = align::right;
+    write<Char>(this->out, static_cast<Char>(value), fmt_specs);
   }
 
   template <typename T, FMT_ENABLE_IF(std::is_floating_point<T>::value)>
-  OutputIt operator()(T value) {
-    return base::operator()(value);
+  void operator()(T value) {
+    base::operator()(value);
   }
 
   /** Formats a null-terminated C string. */
-  OutputIt operator()(const char* value) {
-    if (value) return base::operator()(value);
-    return write_null_pointer(this->specs.type != presentation_type::pointer);
+  void operator()(const char* value) {
+    if (value)
+      base::operator()(value);
+    else
+      write_null_pointer(this->specs.type != presentation_type::pointer);
   }
 
   /** Formats a null-terminated wide C string. */
-  OutputIt operator()(const wchar_t* value) {
-    if (value) return base::operator()(value);
-    return write_null_pointer(this->specs.type != presentation_type::pointer);
+  void operator()(const wchar_t* value) {
+    if (value)
+      base::operator()(value);
+    else
+      write_null_pointer(this->specs.type != presentation_type::pointer);
   }
 
-  OutputIt operator()(basic_string_view<Char> value) {
-    return base::operator()(value);
-  }
+  void operator()(basic_string_view<Char> value) { base::operator()(value); }
 
   /** Formats a pointer. */
-  OutputIt operator()(const void* value) {
-    return value ? base::operator()(value) : write_null_pointer();
+  void operator()(const void* value) {
+    if (value)
+      base::operator()(value);
+    else
+      write_null_pointer();
   }
 
   /** Formats an argument of a custom (user-defined) type. */
-  OutputIt operator()(typename basic_format_arg<context_type>::handle handle) {
-    auto parse_ctx =
-        basic_printf_parse_context<Char>(basic_string_view<Char>());
+  void operator()(typename basic_format_arg<context_type>::handle handle) {
+    auto parse_ctx = basic_format_parse_context<Char>({});
     handle.format(parse_ctx, context_);
-    return this->out;
   }
 };
 
 template <typename Char>
-void parse_flags(basic_format_specs<Char>& specs, const Char*& it,
-                 const Char* end) {
+void parse_flags(format_specs<Char>& specs, const Char*& it, const Char* end) {
   for (; it != end; ++it) {
     switch (*it) {
     case '-':
@@ -314,9 +326,7 @@ void parse_flags(basic_format_specs<Char>& specs, const Char*& it,
       specs.fill[0] = '0';
       break;
     case ' ':
-      if (specs.sign != sign::plus) {
-        specs.sign = sign::space;
-      }
+      if (specs.sign != sign::plus) specs.sign = sign::space;
       break;
     case '#':
       specs.alt = true;
@@ -328,8 +338,8 @@ void parse_flags(basic_format_specs<Char>& specs, const Char*& it,
 }
 
 template <typename Char, typename GetArg>
-int parse_header(const Char*& it, const Char* end,
-                 basic_format_specs<Char>& specs, GetArg get_arg) {
+auto parse_header(const Char*& it, const Char* end, format_specs<Char>& specs,
+                  GetArg get_arg) -> int {
   int arg_index = -1;
   Char c = *it;
   if (c >= '0' && c <= '9') {
@@ -344,7 +354,7 @@ int parse_header(const Char*& it, const Char* end,
       if (value != 0) {
         // Nonzero value means that we parsed width and don't need to
         // parse it or flags again, so return now.
-        if (value == -1) FMT_THROW(format_error("number is too big"));
+        if (value == -1) throw_format_error("number is too big");
         specs.width = value;
         return arg_index;
       }
@@ -355,7 +365,7 @@ int parse_header(const Char*& it, const Char* end,
   if (it != end) {
     if (*it >= '0' && *it <= '9') {
       specs.width = parse_nonnegative_int(it, end, -1);
-      if (specs.width == -1) FMT_THROW(format_error("number is too big"));
+      if (specs.width == -1) throw_format_error("number is too big");
     } else if (*it == '*') {
       ++it;
       specs.width = static_cast<int>(visit_format_arg(
@@ -365,13 +375,53 @@ int parse_header(const Char*& it, const Char* end,
   return arg_index;
 }
 
+inline auto parse_printf_presentation_type(char c, type t)
+    -> presentation_type {
+  using pt = presentation_type;
+  constexpr auto integral_set = sint_set | uint_set | bool_set | char_set;
+  switch (c) {
+  case 'd':
+    return in(t, integral_set) ? pt::dec : pt::none;
+  case 'o':
+    return in(t, integral_set) ? pt::oct : pt::none;
+  case 'x':
+    return in(t, integral_set) ? pt::hex_lower : pt::none;
+  case 'X':
+    return in(t, integral_set) ? pt::hex_upper : pt::none;
+  case 'a':
+    return in(t, float_set) ? pt::hexfloat_lower : pt::none;
+  case 'A':
+    return in(t, float_set) ? pt::hexfloat_upper : pt::none;
+  case 'e':
+    return in(t, float_set) ? pt::exp_lower : pt::none;
+  case 'E':
+    return in(t, float_set) ? pt::exp_upper : pt::none;
+  case 'f':
+    return in(t, float_set) ? pt::fixed_lower : pt::none;
+  case 'F':
+    return in(t, float_set) ? pt::fixed_upper : pt::none;
+  case 'g':
+    return in(t, float_set) ? pt::general_lower : pt::none;
+  case 'G':
+    return in(t, float_set) ? pt::general_upper : pt::none;
+  case 'c':
+    return in(t, integral_set) ? pt::chr : pt::none;
+  case 's':
+    return in(t, string_set | cstring_set) ? pt::string : pt::none;
+  case 'p':
+    return in(t, pointer_set | cstring_set) ? pt::pointer : pt::none;
+  default:
+    return pt::none;
+  }
+}
+
 template <typename Char, typename Context>
 void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
              basic_format_args<Context> args) {
-  using OutputIt = buffer_appender<Char>;
-  auto out = OutputIt(buf);
-  auto context = basic_printf_context<OutputIt, Char>(out, args);
-  auto parse_ctx = basic_printf_parse_context<Char>(format);
+  using iterator = buffer_appender<Char>;
+  auto out = iterator(buf);
+  auto context = basic_printf_context<Char>(out, args);
+  auto parse_ctx = basic_format_parse_context<Char>(format);
 
   // Returns the argument with specified index or, if arg_index is -1, the next
   // argument.
@@ -387,26 +437,24 @@ void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
   const Char* end = parse_ctx.end();
   auto it = start;
   while (it != end) {
-    if (!detail::find<false, Char>(it, end, '%', it)) {
-      it = end;  // detail::find leaves it == nullptr if it doesn't find '%'
+    if (!find<false, Char>(it, end, '%', it)) {
+      it = end;  // find leaves it == nullptr if it doesn't find '%'.
       break;
     }
     Char c = *it++;
     if (it != end && *it == c) {
-      out = detail::write(
-          out, basic_string_view<Char>(start, detail::to_unsigned(it - start)));
+      write(out, basic_string_view<Char>(start, to_unsigned(it - start)));
       start = ++it;
       continue;
     }
-    out = detail::write(out, basic_string_view<Char>(
-                                 start, detail::to_unsigned(it - 1 - start)));
+    write(out, basic_string_view<Char>(start, to_unsigned(it - 1 - start)));
 
-    basic_format_specs<Char> specs;
+    auto specs = format_specs<Char>();
     specs.align = align::right;
 
     // Parse argument index, flags and width.
     int arg_index = parse_header(it, end, specs, get_arg);
-    if (arg_index == 0) parse_ctx.on_error("argument not found");
+    if (arg_index == 0) throw_format_error("argument not found");
 
     // Parse precision.
     if (it != end && *it == '.') {
@@ -417,7 +465,7 @@ void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
       } else if (c == '*') {
         ++it;
         specs.precision = static_cast<int>(
-            visit_format_arg(detail::printf_precision_handler(), get_arg(-1)));
+            visit_format_arg(printf_precision_handler(), get_arg(-1)));
       } else {
         specs.precision = 0;
       }
@@ -426,20 +474,19 @@ void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
     auto arg = get_arg(arg_index);
     // For d, i, o, u, x, and X conversion specifiers, if a precision is
     // specified, the '0' flag is ignored
-    if (specs.precision >= 0 && arg.is_integral())
-      specs.fill[0] =
-          ' ';  // Ignore '0' flag for non-numeric types or if '-' present.
-    if (specs.precision >= 0 && arg.type() == detail::type::cstring_type) {
-      auto str = visit_format_arg(detail::get_cstring<Char>(), arg);
+    if (specs.precision >= 0 && arg.is_integral()) {
+      // Ignore '0' for non-numeric types or if '-' present.
+      specs.fill[0] = ' ';
+    }
+    if (specs.precision >= 0 && arg.type() == type::cstring_type) {
+      auto str = visit_format_arg(get_cstring<Char>(), arg);
       auto str_end = str + specs.precision;
       auto nul = std::find(str, str_end, Char());
-      arg = detail::make_arg<basic_printf_context<OutputIt, Char>>(
-          basic_string_view<Char>(
-              str, detail::to_unsigned(nul != str_end ? nul - str
-                                                      : specs.precision)));
+      auto sv = basic_string_view<Char>(
+          str, to_unsigned(nul != str_end ? nul - str : specs.precision));
+      arg = make_arg<basic_printf_context<Char>>(sv);
     }
-    if (specs.alt && visit_format_arg(detail::is_zero_int(), arg))
-      specs.alt = false;
+    if (specs.alt && visit_format_arg(is_zero_int(), arg)) specs.alt = false;
     if (specs.fill[0] == '0') {
       if (arg.is_arithmetic() && specs.align != align::left)
         specs.align = align::numeric;
@@ -451,7 +498,6 @@ void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
     // Parse length and convert the argument to the required type.
     c = it != end ? *it++ : 0;
     Char t = it != end ? *it : 0;
-    using detail::convert_arg;
     switch (c) {
     case 'h':
       if (t == 'h') {
@@ -490,7 +536,7 @@ void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
     }
 
     // Parse type.
-    if (it == end) FMT_THROW(format_error("invalid format string"));
+    if (it == end) throw_format_error("invalid format string");
     char type = static_cast<char>(*it++);
     if (arg.is_integral()) {
       // Normalize type.
@@ -500,32 +546,25 @@ void vprintf(buffer<Char>& buf, basic_string_view<Char> format,
         type = 'd';
         break;
       case 'c':
-        visit_format_arg(
-            detail::char_converter<basic_printf_context<OutputIt, Char>>(arg),
-            arg);
+        visit_format_arg(char_converter<basic_printf_context<Char>>(arg), arg);
         break;
       }
     }
-    specs.type = parse_presentation_type(type);
+    specs.type = parse_printf_presentation_type(type, arg.type());
     if (specs.type == presentation_type::none)
-      parse_ctx.on_error("invalid type specifier");
+      throw_format_error("invalid format specifier");
 
     start = it;
 
     // Format argument.
-    out = visit_format_arg(
-        detail::printf_arg_formatter<OutputIt, Char>(out, specs, context), arg);
+    visit_format_arg(printf_arg_formatter<Char>(out, specs, context), arg);
   }
-  detail::write(out, basic_string_view<Char>(start, to_unsigned(it - start)));
+  write(out, basic_string_view<Char>(start, to_unsigned(it - start)));
 }
-FMT_END_DETAIL_NAMESPACE
-
-template <typename Char>
-using basic_printf_context_t =
-    basic_printf_context<detail::buffer_appender<Char>, Char>;
+}  // namespace detail
 
-using printf_context = basic_printf_context_t<char>;
-using wprintf_context = basic_printf_context_t<wchar_t>;
+using printf_context = basic_printf_context<char>;
+using wprintf_context = basic_printf_context<wchar_t>;
 
 using printf_args = basic_format_args<printf_context>;
 using wprintf_args = basic_format_args<wprintf_context>;
@@ -542,26 +581,21 @@ inline auto make_printf_args(const T&... args)
   return {args...};
 }
 
-/**
-  \rst
-  Constructs an `~fmt::format_arg_store` object that contains references to
-  arguments and can be implicitly converted to `~fmt::wprintf_args`.
-  \endrst
- */
+// DEPRECATED!
 template <typename... T>
 inline auto make_wprintf_args(const T&... args)
     -> format_arg_store<wprintf_context, T...> {
   return {args...};
 }
 
-template <typename S, typename Char = char_t<S>>
+template <typename Char>
 inline auto vsprintf(
-    const S& fmt,
-    basic_format_args<basic_printf_context_t<type_identity_t<Char>>> args)
+    basic_string_view<Char> fmt,
+    basic_format_args<basic_printf_context<type_identity_t<Char>>> args)
     -> std::basic_string<Char> {
-  basic_memory_buffer<Char> buffer;
-  vprintf(buffer, detail::to_string_view(fmt), args);
-  return to_string(buffer);
+  auto buf = basic_memory_buffer<Char>();
+  detail::vprintf(buf, fmt, args);
+  return to_string(buf);
 }
 
 /**
@@ -576,20 +610,19 @@ inline auto vsprintf(
 template <typename S, typename... T,
           typename Char = enable_if_t<detail::is_string<S>::value, char_t<S>>>
 inline auto sprintf(const S& fmt, const T&... args) -> std::basic_string<Char> {
-  using context = basic_printf_context_t<Char>;
   return vsprintf(detail::to_string_view(fmt),
-                  fmt::make_format_args<context>(args...));
+                  fmt::make_format_args<basic_printf_context<Char>>(args...));
 }
 
-template <typename S, typename Char = char_t<S>>
+template <typename Char>
 inline auto vfprintf(
-    std::FILE* f, const S& fmt,
-    basic_format_args<basic_printf_context_t<type_identity_t<Char>>> args)
+    std::FILE* f, basic_string_view<Char> fmt,
+    basic_format_args<basic_printf_context<type_identity_t<Char>>> args)
     -> int {
-  basic_memory_buffer<Char> buffer;
-  vprintf(buffer, detail::to_string_view(fmt), args);
-  size_t size = buffer.size();
-  return std::fwrite(buffer.data(), sizeof(Char), size, f) < size
+  auto buf = basic_memory_buffer<Char>();
+  detail::vprintf(buf, fmt, args);
+  size_t size = buf.size();
+  return std::fwrite(buf.data(), sizeof(Char), size, f) < size
              ? -1
              : static_cast<int>(size);
 }
@@ -605,17 +638,16 @@ inline auto vfprintf(
  */
 template <typename S, typename... T, typename Char = char_t<S>>
 inline auto fprintf(std::FILE* f, const S& fmt, const T&... args) -> int {
-  using context = basic_printf_context_t<Char>;
   return vfprintf(f, detail::to_string_view(fmt),
-                  fmt::make_format_args<context>(args...));
+                  fmt::make_format_args<basic_printf_context<Char>>(args...));
 }
 
-template <typename S, typename Char = char_t<S>>
-inline auto vprintf(
-    const S& fmt,
-    basic_format_args<basic_printf_context_t<type_identity_t<Char>>> args)
+template <typename Char>
+FMT_DEPRECATED inline auto vprintf(
+    basic_string_view<Char> fmt,
+    basic_format_args<basic_printf_context<type_identity_t<Char>>> args)
     -> int {
-  return vfprintf(stdout, detail::to_string_view(fmt), args);
+  return vfprintf(stdout, fmt, args);
 }
 
 /**
@@ -627,14 +659,17 @@ inline auto vprintf(
     fmt::printf("Elapsed time: %.2f seconds", 1.23);
   \endrst
  */
-template <typename S, typename... T, FMT_ENABLE_IF(detail::is_string<S>::value)>
-inline auto printf(const S& fmt, const T&... args) -> int {
-  return vprintf(
-      detail::to_string_view(fmt),
-      fmt::make_format_args<basic_printf_context_t<char_t<S>>>(args...));
+template <typename... T>
+inline auto printf(string_view fmt, const T&... args) -> int {
+  return vfprintf(stdout, fmt, make_printf_args(args...));
+}
+template <typename... T>
+FMT_DEPRECATED inline auto printf(basic_string_view<wchar_t> fmt,
+                                  const T&... args) -> int {
+  return vfprintf(stdout, fmt, make_wprintf_args(args...));
 }
 
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_PRINTF_H_
diff --git a/dep/fmt/include/fmt/ranges.h b/dep/fmt/include/fmt/ranges.h
index dea7d60dd6b..3638fffb83b 100644
--- a/dep/fmt/include/fmt/ranges.h
+++ b/dep/fmt/include/fmt/ranges.h
@@ -1,13 +1,9 @@
-// Formatting library for C++ - experimental range support
+// Formatting library for C++ - range and tuple support
 //
-// Copyright (c) 2012 - present, Victor Zverovich
+// Copyright (c) 2012 - present, Victor Zverovich and {fmt} contributors
 // All rights reserved.
 //
 // For the license information refer to format.h.
-//
-// Copyright (c) 2018 - present, Remotion (Igor Schulz)
-// All Rights Reserved
-// {fmt} support for ranges, containers and types tuple interface.
 
 #ifndef FMT_RANGES_H_
 #define FMT_RANGES_H_
@@ -22,27 +18,25 @@ FMT_BEGIN_NAMESPACE
 
 namespace detail {
 
-template <typename RangeT, typename OutputIterator>
-OutputIterator copy(const RangeT& range, OutputIterator out) {
+template <typename Range, typename OutputIt>
+auto copy(const Range& range, OutputIt out) -> OutputIt {
   for (auto it = range.begin(), end = range.end(); it != end; ++it)
     *out++ = *it;
   return out;
 }
 
-template <typename OutputIterator>
-OutputIterator copy(const char* str, OutputIterator out) {
+template <typename OutputIt>
+auto copy(const char* str, OutputIt out) -> OutputIt {
   while (*str) *out++ = *str++;
   return out;
 }
 
-template <typename OutputIterator>
-OutputIterator copy(char ch, OutputIterator out) {
+template <typename OutputIt> auto copy(char ch, OutputIt out) -> OutputIt {
   *out++ = ch;
   return out;
 }
 
-template <typename OutputIterator>
-OutputIterator copy(wchar_t ch, OutputIterator out) {
+template <typename OutputIt> auto copy(wchar_t ch, OutputIt out) -> OutputIt {
   *out++ = ch;
   return out;
 }
@@ -69,7 +63,7 @@ template <typename T> class is_map {
   template <typename> static void check(...);
 
  public:
-#ifdef FMT_FORMAT_MAP_AS_LIST
+#ifdef FMT_FORMAT_MAP_AS_LIST  // DEPRECATED!
   static constexpr const bool value = false;
 #else
   static constexpr const bool value =
@@ -82,7 +76,7 @@ template <typename T> class is_set {
   template <typename> static void check(...);
 
  public:
-#ifdef FMT_FORMAT_SET_AS_LIST
+#ifdef FMT_FORMAT_SET_AS_LIST  // DEPRECATED!
   static constexpr const bool value = false;
 #else
   static constexpr const bool value =
@@ -157,8 +151,9 @@ template <typename T>
 struct has_mutable_begin_end<
     T, void_t<decltype(detail::range_begin(std::declval<T>())),
               decltype(detail::range_end(std::declval<T>())),
-              enable_if_t<std::is_copy_constructible<T>::value>>>
-    : std::true_type {};
+              // the extra int here is because older versions of MSVC don't
+              // SFINAE properly unless there are distinct types
+              int>> : std::true_type {};
 
 template <typename T>
 struct is_range_<T, void>
@@ -188,7 +183,7 @@ template <size_t N> using make_index_sequence = std::make_index_sequence<N>;
 template <typename T, T... N> struct integer_sequence {
   using value_type = T;
 
-  static FMT_CONSTEXPR size_t size() { return sizeof...(N); }
+  static FMT_CONSTEXPR auto size() -> size_t { return sizeof...(N); }
 };
 
 template <size_t... N> using index_sequence = integer_sequence<size_t, N...>;
@@ -211,41 +206,61 @@ class is_tuple_formattable_ {
   static constexpr const bool value = false;
 };
 template <typename T, typename C> class is_tuple_formattable_<T, C, true> {
-  template <std::size_t... I>
-  static std::true_type check2(index_sequence<I...>,
-                               integer_sequence<bool, (I == I)...>);
-  static std::false_type check2(...);
-  template <std::size_t... I>
-  static decltype(check2(
-      index_sequence<I...>{},
-      integer_sequence<
-          bool, (is_formattable<typename std::tuple_element<I, T>::type,
-                                C>::value)...>{})) check(index_sequence<I...>);
+  template <std::size_t... Is>
+  static auto check2(index_sequence<Is...>,
+                     integer_sequence<bool, (Is == Is)...>) -> std::true_type;
+  static auto check2(...) -> std::false_type;
+  template <std::size_t... Is>
+  static auto check(index_sequence<Is...>) -> decltype(check2(
+      index_sequence<Is...>{},
+      integer_sequence<bool,
+                       (is_formattable<typename std::tuple_element<Is, T>::type,
+                                       C>::value)...>{}));
 
  public:
   static constexpr const bool value =
       decltype(check(tuple_index_sequence<T>{}))::value;
 };
 
-template <class Tuple, class F, size_t... Is>
-void for_each(index_sequence<Is...>, Tuple&& tup, F&& f) noexcept {
+template <typename Tuple, typename F, size_t... Is>
+FMT_CONSTEXPR void for_each(index_sequence<Is...>, Tuple&& t, F&& f) {
   using std::get;
-  // using free function get<I>(T) now.
-  const int _[] = {0, ((void)f(get<Is>(tup)), 0)...};
-  (void)_;  // blocks warnings
+  // Using a free function get<Is>(Tuple) now.
+  const int unused[] = {0, ((void)f(get<Is>(t)), 0)...};
+  ignore_unused(unused);
 }
 
-template <class T>
-FMT_CONSTEXPR make_index_sequence<std::tuple_size<T>::value> get_indexes(
-    T const&) {
-  return {};
+template <typename Tuple, typename F>
+FMT_CONSTEXPR void for_each(Tuple&& t, F&& f) {
+  for_each(tuple_index_sequence<remove_cvref_t<Tuple>>(),
+           std::forward<Tuple>(t), std::forward<F>(f));
+}
+
+template <typename Tuple1, typename Tuple2, typename F, size_t... Is>
+void for_each2(index_sequence<Is...>, Tuple1&& t1, Tuple2&& t2, F&& f) {
+  using std::get;
+  const int unused[] = {0, ((void)f(get<Is>(t1), get<Is>(t2)), 0)...};
+  ignore_unused(unused);
 }
 
-template <class Tuple, class F> void for_each(Tuple&& tup, F&& f) {
-  const auto indexes = get_indexes(tup);
-  for_each(indexes, std::forward<Tuple>(tup), std::forward<F>(f));
+template <typename Tuple1, typename Tuple2, typename F>
+void for_each2(Tuple1&& t1, Tuple2&& t2, F&& f) {
+  for_each2(tuple_index_sequence<remove_cvref_t<Tuple1>>(),
+            std::forward<Tuple1>(t1), std::forward<Tuple2>(t2),
+            std::forward<F>(f));
 }
 
+namespace tuple {
+// Workaround a bug in MSVC 2019 (v140).
+template <typename Char, typename... T>
+using result_t = std::tuple<formatter<remove_cvref_t<T>, Char>...>;
+
+using std::get;
+template <typename Tuple, typename Char, std::size_t... Is>
+auto get_formatters(index_sequence<Is...>)
+    -> result_t<Char, decltype(get<Is>(std::declval<Tuple>()))...>;
+}  // namespace tuple
+
 #if FMT_MSC_VERSION && FMT_MSC_VERSION < 1920
 // Older MSVC doesn't get the reference type correctly for arrays.
 template <typename R> struct range_reference_type_impl {
@@ -269,45 +284,37 @@ using range_reference_type =
 template <typename Range>
 using uncvref_type = remove_cvref_t<range_reference_type<Range>>;
 
-template <typename Range>
-using uncvref_first_type =
-    remove_cvref_t<decltype(std::declval<range_reference_type<Range>>().first)>;
-
-template <typename Range>
-using uncvref_second_type = remove_cvref_t<
-    decltype(std::declval<range_reference_type<Range>>().second)>;
-
-template <typename OutputIt> OutputIt write_delimiter(OutputIt out) {
-  *out++ = ',';
-  *out++ = ' ';
-  return out;
-}
-
-template <typename Char, typename OutputIt>
-auto write_range_entry(OutputIt out, basic_string_view<Char> str) -> OutputIt {
-  return write_escaped_string(out, str);
-}
-
-template <typename Char, typename OutputIt, typename T,
-          FMT_ENABLE_IF(std::is_convertible<T, std_string_view<char>>::value)>
-inline auto write_range_entry(OutputIt out, const T& str) -> OutputIt {
-  auto sv = std_string_view<Char>(str);
-  return write_range_entry<Char>(out, basic_string_view<Char>(sv));
-}
-
-template <typename Char, typename OutputIt, typename Arg,
-          FMT_ENABLE_IF(std::is_same<Arg, Char>::value)>
-OutputIt write_range_entry(OutputIt out, const Arg v) {
-  return write_escaped_char(out, v);
+template <typename Formatter>
+FMT_CONSTEXPR auto maybe_set_debug_format(Formatter& f, bool set)
+    -> decltype(f.set_debug_format(set)) {
+  f.set_debug_format(set);
 }
+template <typename Formatter>
+FMT_CONSTEXPR void maybe_set_debug_format(Formatter&, ...) {}
+
+// These are not generic lambdas for compatibility with C++11.
+template <typename ParseContext> struct parse_empty_specs {
+  template <typename Formatter> FMT_CONSTEXPR void operator()(Formatter& f) {
+    f.parse(ctx);
+    detail::maybe_set_debug_format(f, true);
+  }
+  ParseContext& ctx;
+};
+template <typename FormatContext> struct format_tuple_element {
+  using char_type = typename FormatContext::char_type;
+
+  template <typename T>
+  void operator()(const formatter<T, char_type>& f, const T& v) {
+    if (i > 0)
+      ctx.advance_to(detail::copy_str<char_type>(separator, ctx.out()));
+    ctx.advance_to(f.format(v, ctx));
+    ++i;
+  }
 
-template <
-    typename Char, typename OutputIt, typename Arg,
-    FMT_ENABLE_IF(!is_std_string_like<typename std::decay<Arg>::type>::value &&
-                  !std::is_same<Arg, Char>::value)>
-OutputIt write_range_entry(OutputIt out, const Arg& v) {
-  return write<Char>(out, v);
-}
+  int i;
+  FormatContext& ctx;
+  basic_string_view<char_type> separator;
+};
 
 }  // namespace detail
 
@@ -321,29 +328,20 @@ template <typename T, typename C> struct is_tuple_formattable {
       detail::is_tuple_formattable_<T, C>::value;
 };
 
-template <typename TupleT, typename Char>
-struct formatter<TupleT, Char,
-                 enable_if_t<fmt::is_tuple_like<TupleT>::value &&
-                             fmt::is_tuple_formattable<TupleT, Char>::value>> {
+template <typename Tuple, typename Char>
+struct formatter<Tuple, Char,
+                 enable_if_t<fmt::is_tuple_like<Tuple>::value &&
+                             fmt::is_tuple_formattable<Tuple, Char>::value>> {
  private:
+  decltype(detail::tuple::get_formatters<Tuple, Char>(
+      detail::tuple_index_sequence<Tuple>())) formatters_;
+
   basic_string_view<Char> separator_ = detail::string_literal<Char, ',', ' '>{};
   basic_string_view<Char> opening_bracket_ =
       detail::string_literal<Char, '('>{};
   basic_string_view<Char> closing_bracket_ =
       detail::string_literal<Char, ')'>{};
 
-  // C++11 generic lambda for format().
-  template <typename FormatContext> struct format_each {
-    template <typename T> void operator()(const T& v) {
-      if (i > 0) out = detail::copy_str<Char>(separator, out);
-      out = detail::write_range_entry<Char>(out, v);
-      ++i;
-    }
-    int i;
-    typename FormatContext::iterator& out;
-    basic_string_view<Char> separator;
-  };
-
  public:
   FMT_CONSTEXPR formatter() {}
 
@@ -359,17 +357,21 @@ struct formatter<TupleT, Char,
 
   template <typename ParseContext>
   FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
-    return ctx.begin();
+    auto it = ctx.begin();
+    if (it != ctx.end() && *it != '}')
+      FMT_THROW(format_error("invalid format specifier"));
+    detail::for_each(formatters_, detail::parse_empty_specs<ParseContext>{ctx});
+    return it;
   }
 
-  template <typename FormatContext = format_context>
-  auto format(const TupleT& values, FormatContext& ctx) const
+  template <typename FormatContext>
+  auto format(const Tuple& value, FormatContext& ctx) const
       -> decltype(ctx.out()) {
-    auto out = ctx.out();
-    out = detail::copy_str<Char>(opening_bracket_, out);
-    detail::for_each(values, format_each<FormatContext>{0, out, separator_});
-    out = detail::copy_str<Char>(closing_bracket_, out);
-    return out;
+    ctx.advance_to(detail::copy_str<Char>(opening_bracket_, ctx.out()));
+    detail::for_each2(
+        formatters_, value,
+        detail::format_tuple_element<FormatContext>{0, ctx, separator_});
+    return detail::copy_str<Char>(closing_bracket_, ctx.out());
   }
 };
 
@@ -398,12 +400,10 @@ template <typename Context> struct range_mapper {
 };
 
 template <typename Char, typename Element>
-using range_formatter_type = conditional_t<
-    is_formattable<Element, Char>::value,
+using range_formatter_type =
     formatter<remove_cvref_t<decltype(range_mapper<buffer_context<Char>>{}.map(
                   std::declval<Element>()))>,
-              Char>,
-    fallback_formatter<Element, Char>>;
+              Char>;
 
 template <typename R>
 using maybe_const_range =
@@ -413,45 +413,32 @@ using maybe_const_range =
 #if !FMT_MSC_VERSION || FMT_MSC_VERSION >= 1910
 template <typename R, typename Char>
 struct is_formattable_delayed
-    : disjunction<
-          is_formattable<uncvref_type<maybe_const_range<R>>, Char>,
-          has_fallback_formatter<uncvref_type<maybe_const_range<R>>, Char>> {};
+    : is_formattable<uncvref_type<maybe_const_range<R>>, Char> {};
 #endif
-
 }  // namespace detail
 
+template <typename...> struct conjunction : std::true_type {};
+template <typename P> struct conjunction<P> : P {};
+template <typename P1, typename... Pn>
+struct conjunction<P1, Pn...>
+    : conditional_t<bool(P1::value), conjunction<Pn...>, P1> {};
+
 template <typename T, typename Char, typename Enable = void>
 struct range_formatter;
 
 template <typename T, typename Char>
 struct range_formatter<
     T, Char,
-    enable_if_t<conjunction<
-        std::is_same<T, remove_cvref_t<T>>,
-        disjunction<is_formattable<T, Char>,
-                    detail::has_fallback_formatter<T, Char>>>::value>> {
+    enable_if_t<conjunction<std::is_same<T, remove_cvref_t<T>>,
+                            is_formattable<T, Char>>::value>> {
  private:
   detail::range_formatter_type<Char, T> underlying_;
-  bool custom_specs_ = false;
   basic_string_view<Char> separator_ = detail::string_literal<Char, ',', ' '>{};
   basic_string_view<Char> opening_bracket_ =
       detail::string_literal<Char, '['>{};
   basic_string_view<Char> closing_bracket_ =
       detail::string_literal<Char, ']'>{};
 
-  template <class U>
-  FMT_CONSTEXPR static auto maybe_set_debug_format(U& u, int)
-      -> decltype(u.set_debug_format()) {
-    u.set_debug_format();
-  }
-
-  template <class U>
-  FMT_CONSTEXPR static void maybe_set_debug_format(U&, ...) {}
-
-  FMT_CONSTEXPR void maybe_set_debug_format() {
-    maybe_set_debug_format(underlying_, 0);
-  }
-
  public:
   FMT_CONSTEXPR range_formatter() {}
 
@@ -473,31 +460,24 @@ struct range_formatter<
   FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
     auto it = ctx.begin();
     auto end = ctx.end();
-    if (it == end || *it == '}') {
-      maybe_set_debug_format();
-      return it;
-    }
 
-    if (*it == 'n') {
+    if (it != end && *it == 'n') {
       set_brackets({}, {});
       ++it;
     }
 
-    if (*it == '}') {
-      maybe_set_debug_format();
-      return it;
+    if (it != end && *it != '}') {
+      if (*it != ':') FMT_THROW(format_error("invalid format specifier"));
+      ++it;
+    } else {
+      detail::maybe_set_debug_format(underlying_, true);
     }
 
-    if (*it != ':')
-      FMT_THROW(format_error("no other top-level range formatters supported"));
-
-    custom_specs_ = true;
-    ++it;
     ctx.advance_to(it);
     return underlying_.parse(ctx);
   }
 
-  template <typename R, class FormatContext>
+  template <typename R, typename FormatContext>
   auto format(R&& range, FormatContext& ctx) const -> decltype(ctx.out()) {
     detail::range_mapper<buffer_context<Char>> mapper;
     auto out = ctx.out();
@@ -507,9 +487,9 @@ struct range_formatter<
     auto end = detail::range_end(range);
     for (; it != end; ++it) {
       if (i > 0) out = detail::copy_str<Char>(separator_, out);
-      ;
       ctx.advance_to(out);
-      out = underlying_.format(mapper.map(*it), ctx);
+      auto&& item = *it;
+      out = underlying_.format(mapper.map(item), ctx);
       ++i;
     }
     out = detail::copy_str<Char>(closing_bracket_, out);
@@ -520,13 +500,14 @@ struct range_formatter<
 enum class range_format { disabled, map, set, sequence, string, debug_string };
 
 namespace detail {
-template <typename T> struct range_format_kind_ {
-  static constexpr auto value = std::is_same<range_reference_type<T>, T>::value
-                                    ? range_format::disabled
-                                : is_map<T>::value ? range_format::map
-                                : is_set<T>::value ? range_format::set
-                                                   : range_format::sequence;
-};
+template <typename T>
+struct range_format_kind_
+    : std::integral_constant<range_format,
+                             std::is_same<uncvref_type<T>, T>::value
+                                 ? range_format::disabled
+                             : is_map<T>::value ? range_format::map
+                             : is_set<T>::value ? range_format::set
+                                                : range_format::sequence> {};
 
 template <range_format K, typename R, typename Char, typename Enable = void>
 struct range_default_formatter;
@@ -601,9 +582,6 @@ template <typename Char, typename... T> struct tuple_join_view : detail::view {
       : tuple(t), sep{s} {}
 };
 
-template <typename Char, typename... T>
-using tuple_arg_join = tuple_join_view<Char, T...>;
-
 // Define FMT_TUPLE_JOIN_SPECIFIERS to enable experimental format specifiers
 // support in tuple_join. It is disabled by default because of issues with
 // the dynamic width and precision.
@@ -673,7 +651,45 @@ struct formatter<tuple_join_view<Char, T...>, Char> {
   }
 };
 
-FMT_MODULE_EXPORT_BEGIN
+namespace detail {
+// Check if T has an interface like a container adaptor (e.g. std::stack,
+// std::queue, std::priority_queue).
+template <typename T> class is_container_adaptor_like {
+  template <typename U> static auto check(U* p) -> typename U::container_type;
+  template <typename> static void check(...);
+
+ public:
+  static constexpr const bool value =
+      !std::is_void<decltype(check<T>(nullptr))>::value;
+};
+
+template <typename Container> struct all {
+  const Container& c;
+  auto begin() const -> typename Container::const_iterator { return c.begin(); }
+  auto end() const -> typename Container::const_iterator { return c.end(); }
+};
+}  // namespace detail
+
+template <typename T, typename Char>
+struct formatter<
+    T, Char,
+    enable_if_t<conjunction<detail::is_container_adaptor_like<T>,
+                            bool_constant<range_format_kind<T, Char>::value ==
+                                          range_format::disabled>>::value>>
+    : formatter<detail::all<typename T::container_type>, Char> {
+  using all = detail::all<typename T::container_type>;
+  template <typename FormatContext>
+  auto format(const T& t, FormatContext& ctx) const -> decltype(ctx.out()) {
+    struct getter : T {
+      static auto get(const T& t) -> all {
+        return {t.*(&getter::c)};  // Access c through the derived class.
+      }
+    };
+    return formatter<all>::format(getter::get(t), ctx);
+  }
+};
+
+FMT_BEGIN_EXPORT
 
 /**
   \rst
@@ -716,7 +732,7 @@ auto join(std::initializer_list<T> list, string_view sep)
   return join(std::begin(list), std::end(list), sep);
 }
 
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_RANGES_H_
diff --git a/dep/fmt/include/fmt/std.h b/dep/fmt/include/fmt/std.h
index 41d2b2838b6..7cff1159201 100644
--- a/dep/fmt/include/fmt/std.h
+++ b/dep/fmt/include/fmt/std.h
@@ -8,10 +8,18 @@
 #ifndef FMT_STD_H_
 #define FMT_STD_H_
 
+#include <atomic>
+#include <bitset>
+#include <cstdlib>
+#include <exception>
+#include <memory>
 #include <thread>
 #include <type_traits>
+#include <typeinfo>
 #include <utility>
+#include <vector>
 
+#include "format.h"
 #include "ostream.h"
 
 #if FMT_HAS_INCLUDE(<version>)
@@ -25,94 +33,258 @@
 #  if FMT_HAS_INCLUDE(<variant>)
 #    include <variant>
 #  endif
+#  if FMT_HAS_INCLUDE(<optional>)
+#    include <optional>
+#  endif
+#endif
+
+#if FMT_CPLUSPLUS > 201703L && FMT_HAS_INCLUDE(<source_location>)
+#  include <source_location>
+#endif
+
+// GCC 4 does not support FMT_HAS_INCLUDE.
+#if FMT_HAS_INCLUDE(<cxxabi.h>) || defined(__GLIBCXX__)
+#  include <cxxabi.h>
+// Android NDK with gabi++ library on some architectures does not implement
+// abi::__cxa_demangle().
+#  ifndef __GABIXX_CXXABI_H__
+#    define FMT_HAS_ABI_CXA_DEMANGLE
+#  endif
+#endif
+
+// Check if typeid is available.
+#ifndef FMT_USE_TYPEID
+// __RTTI is for EDG compilers. In MSVC typeid is available without RTTI.
+#  if defined(__GXX_RTTI) || FMT_HAS_FEATURE(cxx_rtti) || FMT_MSC_VERSION || \
+      defined(__INTEL_RTTI__) || defined(__RTTI)
+#    define FMT_USE_TYPEID 1
+#  else
+#    define FMT_USE_TYPEID 0
+#  endif
 #endif
 
-#ifdef __cpp_lib_filesystem
+// For older Xcode versions, __cpp_lib_xxx flags are inaccurately defined.
+#ifndef FMT_CPP_LIB_FILESYSTEM
+#  ifdef __cpp_lib_filesystem
+#    define FMT_CPP_LIB_FILESYSTEM __cpp_lib_filesystem
+#  else
+#    define FMT_CPP_LIB_FILESYSTEM 0
+#  endif
+#endif
+
+#ifndef FMT_CPP_LIB_VARIANT
+#  ifdef __cpp_lib_variant
+#    define FMT_CPP_LIB_VARIANT __cpp_lib_variant
+#  else
+#    define FMT_CPP_LIB_VARIANT 0
+#  endif
+#endif
+
+#if FMT_CPP_LIB_FILESYSTEM
 FMT_BEGIN_NAMESPACE
 
 namespace detail {
 
-template <typename Char>
-void write_escaped_path(basic_memory_buffer<Char>& quoted,
-                        const std::filesystem::path& p) {
-  write_escaped_string<Char>(std::back_inserter(quoted), p.string<Char>());
-}
-#  ifdef _WIN32
-template <>
-inline void write_escaped_path<char>(basic_memory_buffer<char>& quoted,
-                                     const std::filesystem::path& p) {
-  auto s = p.u8string();
-  write_escaped_string<char>(
-      std::back_inserter(quoted),
-      string_view(reinterpret_cast<const char*>(s.c_str()), s.size()));
+template <typename Char, typename PathChar>
+auto get_path_string(const std::filesystem::path& p,
+                     const std::basic_string<PathChar>& native) {
+  if constexpr (std::is_same_v<Char, char> && std::is_same_v<PathChar, wchar_t>)
+    return to_utf8<wchar_t>(native, to_utf8_error_policy::replace);
+  else
+    return p.string<Char>();
 }
-#  endif
-template <>
-inline void write_escaped_path<std::filesystem::path::value_type>(
-    basic_memory_buffer<std::filesystem::path::value_type>& quoted,
-    const std::filesystem::path& p) {
-  write_escaped_string<std::filesystem::path::value_type>(
-      std::back_inserter(quoted), p.native());
+
+template <typename Char, typename PathChar>
+void write_escaped_path(basic_memory_buffer<Char>& quoted,
+                        const std::filesystem::path& p,
+                        const std::basic_string<PathChar>& native) {
+  if constexpr (std::is_same_v<Char, char> &&
+                std::is_same_v<PathChar, wchar_t>) {
+    auto buf = basic_memory_buffer<wchar_t>();
+    write_escaped_string<wchar_t>(std::back_inserter(buf), native);
+    bool valid = to_utf8<wchar_t>::convert(quoted, {buf.data(), buf.size()});
+    FMT_ASSERT(valid, "invalid utf16");
+  } else if constexpr (std::is_same_v<Char, PathChar>) {
+    write_escaped_string<std::filesystem::path::value_type>(
+        std::back_inserter(quoted), native);
+  } else {
+    write_escaped_string<Char>(std::back_inserter(quoted), p.string<Char>());
+  }
 }
 
 }  // namespace detail
 
-template <typename Char>
-struct formatter<std::filesystem::path, Char>
-    : formatter<basic_string_view<Char>> {
+FMT_EXPORT
+template <typename Char> struct formatter<std::filesystem::path, Char> {
+ private:
+  format_specs<Char> specs_;
+  detail::arg_ref<Char> width_ref_;
+  bool debug_ = false;
+  char path_type_ = 0;
+
+ public:
+  FMT_CONSTEXPR void set_debug_format(bool set = true) { debug_ = set; }
+
+  template <typename ParseContext> FMT_CONSTEXPR auto parse(ParseContext& ctx) {
+    auto it = ctx.begin(), end = ctx.end();
+    if (it == end) return it;
+
+    it = detail::parse_align(it, end, specs_);
+    if (it == end) return it;
+
+    it = detail::parse_dynamic_spec(it, end, specs_.width, width_ref_, ctx);
+    if (it != end && *it == '?') {
+      debug_ = true;
+      ++it;
+    }
+    if (it != end && (*it == 'g')) path_type_ = *it++;
+    return it;
+  }
+
   template <typename FormatContext>
-  auto format(const std::filesystem::path& p, FormatContext& ctx) const ->
-      typename FormatContext::iterator {
-    basic_memory_buffer<Char> quoted;
-    detail::write_escaped_path(quoted, p);
-    return formatter<basic_string_view<Char>>::format(
-        basic_string_view<Char>(quoted.data(), quoted.size()), ctx);
+  auto format(const std::filesystem::path& p, FormatContext& ctx) const {
+    auto specs = specs_;
+#  ifdef _WIN32
+    auto path_string = !path_type_ ? p.native() : p.generic_wstring();
+#  else
+    auto path_string = !path_type_ ? p.native() : p.generic_string();
+#  endif
+
+    detail::handle_dynamic_spec<detail::width_checker>(specs.width, width_ref_,
+                                                       ctx);
+    if (!debug_) {
+      auto s = detail::get_path_string<Char>(p, path_string);
+      return detail::write(ctx.out(), basic_string_view<Char>(s), specs);
+    }
+    auto quoted = basic_memory_buffer<Char>();
+    detail::write_escaped_path(quoted, p, path_string);
+    return detail::write(ctx.out(),
+                         basic_string_view<Char>(quoted.data(), quoted.size()),
+                         specs);
   }
 };
 FMT_END_NAMESPACE
-#endif
+#endif  // FMT_CPP_LIB_FILESYSTEM
 
 FMT_BEGIN_NAMESPACE
+FMT_EXPORT
+template <std::size_t N, typename Char>
+struct formatter<std::bitset<N>, Char> : nested_formatter<string_view> {
+ private:
+  // Functor because C++11 doesn't support generic lambdas.
+  struct writer {
+    const std::bitset<N>& bs;
+
+    template <typename OutputIt>
+    FMT_CONSTEXPR auto operator()(OutputIt out) -> OutputIt {
+      for (auto pos = N; pos > 0; --pos) {
+        out = detail::write<Char>(out, bs[pos - 1] ? Char('1') : Char('0'));
+      }
+
+      return out;
+    }
+  };
+
+ public:
+  template <typename FormatContext>
+  auto format(const std::bitset<N>& bs, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    return write_padded(ctx, writer{bs});
+  }
+};
+
+FMT_EXPORT
 template <typename Char>
 struct formatter<std::thread::id, Char> : basic_ostream_formatter<Char> {};
 FMT_END_NAMESPACE
 
-#ifdef __cpp_lib_variant
+#ifdef __cpp_lib_optional
 FMT_BEGIN_NAMESPACE
-template <typename Char> struct formatter<std::monostate, Char> {
-  template <typename ParseContext>
-  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
+FMT_EXPORT
+template <typename T, typename Char>
+struct formatter<std::optional<T>, Char,
+                 std::enable_if_t<is_formattable<T, Char>::value>> {
+ private:
+  formatter<T, Char> underlying_;
+  static constexpr basic_string_view<Char> optional =
+      detail::string_literal<Char, 'o', 'p', 't', 'i', 'o', 'n', 'a', 'l',
+                             '('>{};
+  static constexpr basic_string_view<Char> none =
+      detail::string_literal<Char, 'n', 'o', 'n', 'e'>{};
+
+  template <class U>
+  FMT_CONSTEXPR static auto maybe_set_debug_format(U& u, bool set)
+      -> decltype(u.set_debug_format(set)) {
+    u.set_debug_format(set);
+  }
+
+  template <class U>
+  FMT_CONSTEXPR static void maybe_set_debug_format(U&, ...) {}
+
+ public:
+  template <typename ParseContext> FMT_CONSTEXPR auto parse(ParseContext& ctx) {
+    maybe_set_debug_format(underlying_, true);
+    return underlying_.parse(ctx);
+  }
+
+  template <typename FormatContext>
+  auto format(const std::optional<T>& opt, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    if (!opt) return detail::write<Char>(ctx.out(), none);
+
+    auto out = ctx.out();
+    out = detail::write<Char>(out, optional);
+    ctx.advance_to(out);
+    out = underlying_.format(*opt, ctx);
+    return detail::write(out, ')');
+  }
+};
+FMT_END_NAMESPACE
+#endif  // __cpp_lib_optional
+
+#ifdef __cpp_lib_source_location
+FMT_BEGIN_NAMESPACE
+FMT_EXPORT
+template <> struct formatter<std::source_location> {
+  template <typename ParseContext> FMT_CONSTEXPR auto parse(ParseContext& ctx) {
     return ctx.begin();
   }
 
   template <typename FormatContext>
-  auto format(const std::monostate&, FormatContext& ctx) const
+  auto format(const std::source_location& loc, FormatContext& ctx) const
       -> decltype(ctx.out()) {
     auto out = ctx.out();
-    out = detail::write<Char>(out, "monostate");
+    out = detail::write(out, loc.file_name());
+    out = detail::write(out, ':');
+    out = detail::write<char>(out, loc.line());
+    out = detail::write(out, ':');
+    out = detail::write<char>(out, loc.column());
+    out = detail::write(out, ": ");
+    out = detail::write(out, loc.function_name());
     return out;
   }
 };
+FMT_END_NAMESPACE
+#endif
 
+#if FMT_CPP_LIB_VARIANT
+FMT_BEGIN_NAMESPACE
 namespace detail {
 
 template <typename T>
 using variant_index_sequence =
     std::make_index_sequence<std::variant_size<T>::value>;
 
-// variant_size and variant_alternative check.
-template <typename T, typename U = void>
-struct is_variant_like_ : std::false_type {};
-template <typename T>
-struct is_variant_like_<T, std::void_t<decltype(std::variant_size<T>::value)>>
-    : std::true_type {};
+template <typename> struct is_variant_like_ : std::false_type {};
+template <typename... Types>
+struct is_variant_like_<std::variant<Types...>> : std::true_type {};
 
-// formattable element check
+// formattable element check.
 template <typename T, typename C> class is_variant_formattable_ {
-  template <std::size_t... I>
+  template <std::size_t... Is>
   static std::conjunction<
-      is_formattable<std::variant_alternative_t<I, T>, C>...>
-      check(std::index_sequence<I...>);
+      is_formattable<std::variant_alternative_t<Is, T>, C>...>
+      check(std::index_sequence<Is...>);
 
  public:
   static constexpr const bool value =
@@ -140,6 +312,21 @@ template <typename T, typename C> struct is_variant_formattable {
       detail::is_variant_formattable_<T, C>::value;
 };
 
+FMT_EXPORT
+template <typename Char> struct formatter<std::monostate, Char> {
+  template <typename ParseContext>
+  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
+    return ctx.begin();
+  }
+
+  template <typename FormatContext>
+  auto format(const std::monostate&, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    return detail::write<Char>(ctx.out(), "monostate");
+  }
+};
+
+FMT_EXPORT
 template <typename Variant, typename Char>
 struct formatter<
     Variant, Char,
@@ -156,16 +343,195 @@ struct formatter<
     auto out = ctx.out();
 
     out = detail::write<Char>(out, "variant(");
-    std::visit(
-        [&](const auto& v) {
-          out = detail::write_variant_alternative<Char>(out, v);
-        },
-        value);
+    FMT_TRY {
+      std::visit(
+          [&](const auto& v) {
+            out = detail::write_variant_alternative<Char>(out, v);
+          },
+          value);
+    }
+    FMT_CATCH(const std::bad_variant_access&) {
+      detail::write<Char>(out, "valueless by exception");
+    }
     *out++ = ')';
     return out;
   }
 };
 FMT_END_NAMESPACE
+#endif  // FMT_CPP_LIB_VARIANT
+
+FMT_BEGIN_NAMESPACE
+FMT_EXPORT
+template <typename Char> struct formatter<std::error_code, Char> {
+  template <typename ParseContext>
+  FMT_CONSTEXPR auto parse(ParseContext& ctx) -> decltype(ctx.begin()) {
+    return ctx.begin();
+  }
+
+  template <typename FormatContext>
+  FMT_CONSTEXPR auto format(const std::error_code& ec, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    auto out = ctx.out();
+    out = detail::write_bytes(out, ec.category().name(), format_specs<Char>());
+    out = detail::write<Char>(out, Char(':'));
+    out = detail::write<Char>(out, ec.value());
+    return out;
+  }
+};
+
+FMT_EXPORT
+template <typename T, typename Char>
+struct formatter<
+    T, Char,  // DEPRECATED! Mixing code unit types.
+    typename std::enable_if<std::is_base_of<std::exception, T>::value>::type> {
+ private:
+  bool with_typename_ = false;
+
+ public:
+  FMT_CONSTEXPR auto parse(basic_format_parse_context<Char>& ctx)
+      -> decltype(ctx.begin()) {
+    auto it = ctx.begin();
+    auto end = ctx.end();
+    if (it == end || *it == '}') return it;
+    if (*it == 't') {
+      ++it;
+      with_typename_ = FMT_USE_TYPEID != 0;
+    }
+    return it;
+  }
+
+  template <typename OutputIt>
+  auto format(const std::exception& ex,
+              basic_format_context<OutputIt, Char>& ctx) const -> OutputIt {
+    format_specs<Char> spec;
+    auto out = ctx.out();
+    if (!with_typename_)
+      return detail::write_bytes(out, string_view(ex.what()), spec);
+
+#if FMT_USE_TYPEID
+    const std::type_info& ti = typeid(ex);
+#  ifdef FMT_HAS_ABI_CXA_DEMANGLE
+    int status = 0;
+    std::size_t size = 0;
+    std::unique_ptr<char, void (*)(void*)> demangled_name_ptr(
+        abi::__cxa_demangle(ti.name(), nullptr, &size, &status), &std::free);
+
+    string_view demangled_name_view;
+    if (demangled_name_ptr) {
+      demangled_name_view = demangled_name_ptr.get();
+
+      // Normalization of stdlib inline namespace names.
+      // libc++ inline namespaces.
+      //  std::__1::*       -> std::*
+      //  std::__1::__fs::* -> std::*
+      // libstdc++ inline namespaces.
+      //  std::__cxx11::*             -> std::*
+      //  std::filesystem::__cxx11::* -> std::filesystem::*
+      if (demangled_name_view.starts_with("std::")) {
+        char* begin = demangled_name_ptr.get();
+        char* to = begin + 5;  // std::
+        for (char *from = to, *end = begin + demangled_name_view.size();
+             from < end;) {
+          // This is safe, because demangled_name is NUL-terminated.
+          if (from[0] == '_' && from[1] == '_') {
+            char* next = from + 1;
+            while (next < end && *next != ':') next++;
+            if (next[0] == ':' && next[1] == ':') {
+              from = next + 2;
+              continue;
+            }
+          }
+          *to++ = *from++;
+        }
+        demangled_name_view = {begin, detail::to_unsigned(to - begin)};
+      }
+    } else {
+      demangled_name_view = string_view(ti.name());
+    }
+    out = detail::write_bytes(out, demangled_name_view, spec);
+#  elif FMT_MSC_VERSION
+    string_view demangled_name_view(ti.name());
+    if (demangled_name_view.starts_with("class "))
+      demangled_name_view.remove_prefix(6);
+    else if (demangled_name_view.starts_with("struct "))
+      demangled_name_view.remove_prefix(7);
+    out = detail::write_bytes(out, demangled_name_view, spec);
+#  else
+    out = detail::write_bytes(out, string_view(ti.name()), spec);
+#  endif
+    *out++ = ':';
+    *out++ = ' ';
+    return detail::write_bytes(out, string_view(ex.what()), spec);
 #endif
+  }
+};
+
+namespace detail {
+
+template <typename T, typename Enable = void>
+struct has_flip : std::false_type {};
+
+template <typename T>
+struct has_flip<T, void_t<decltype(std::declval<T>().flip())>>
+    : std::true_type {};
+
+template <typename T> struct is_bit_reference_like {
+  static constexpr const bool value =
+      std::is_convertible<T, bool>::value &&
+      std::is_nothrow_assignable<T, bool>::value && has_flip<T>::value;
+};
+
+#ifdef _LIBCPP_VERSION
+
+// Workaround for libc++ incompatibility with C++ standard.
+// According to the Standard, `bitset::operator[] const` returns bool.
+template <typename C>
+struct is_bit_reference_like<std::__bit_const_reference<C>> {
+  static constexpr const bool value = true;
+};
+
+#endif
+
+}  // namespace detail
 
+// We can't use std::vector<bool, Allocator>::reference and
+// std::bitset<N>::reference because the compiler can't deduce Allocator and N
+// in partial specialization.
+FMT_EXPORT
+template <typename BitRef, typename Char>
+struct formatter<BitRef, Char,
+                 enable_if_t<detail::is_bit_reference_like<BitRef>::value>>
+    : formatter<bool, Char> {
+  template <typename FormatContext>
+  FMT_CONSTEXPR auto format(const BitRef& v, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    return formatter<bool, Char>::format(v, ctx);
+  }
+};
+
+FMT_EXPORT
+template <typename T, typename Char>
+struct formatter<std::atomic<T>, Char,
+                 enable_if_t<is_formattable<T, Char>::value>>
+    : formatter<T, Char> {
+  template <typename FormatContext>
+  auto format(const std::atomic<T>& v, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    return formatter<T, Char>::format(v.load(), ctx);
+  }
+};
+
+#ifdef __cpp_lib_atomic_flag_test
+FMT_EXPORT
+template <typename Char>
+struct formatter<std::atomic_flag, Char> : formatter<bool, Char> {
+  template <typename FormatContext>
+  auto format(const std::atomic_flag& v, FormatContext& ctx) const
+      -> decltype(ctx.out()) {
+    return formatter<bool, Char>::format(v.test(), ctx);
+  }
+};
+#endif  // __cpp_lib_atomic_flag_test
+
+FMT_END_NAMESPACE
 #endif  // FMT_STD_H_
diff --git a/dep/fmt/include/fmt/xchar.h b/dep/fmt/include/fmt/xchar.h
index 3b5bc15ca0a..f609c5c41a2 100644
--- a/dep/fmt/include/fmt/xchar.h
+++ b/dep/fmt/include/fmt/xchar.h
@@ -12,13 +12,32 @@
 
 #include "format.h"
 
+#ifndef FMT_STATIC_THOUSANDS_SEPARATOR
+#  include <locale>
+#endif
+
 FMT_BEGIN_NAMESPACE
 namespace detail {
+
 template <typename T>
 using is_exotic_char = bool_constant<!std::is_same<T, char>::value>;
+
+inline auto write_loc(std::back_insert_iterator<detail::buffer<wchar_t>> out,
+                      loc_value value, const format_specs<wchar_t>& specs,
+                      locale_ref loc) -> bool {
+#ifndef FMT_STATIC_THOUSANDS_SEPARATOR
+  auto& numpunct =
+      std::use_facet<std::numpunct<wchar_t>>(loc.get<std::locale>());
+  auto separator = std::wstring();
+  auto grouping = numpunct.grouping();
+  if (!grouping.empty()) separator = std::wstring(1, numpunct.thousands_sep());
+  return value.visit(loc_writer<wchar_t>{out, specs, separator, grouping, {}});
+#endif
+  return false;
 }
+}  // namespace detail
 
-FMT_MODULE_EXPORT_BEGIN
+FMT_BEGIN_EXPORT
 
 using wstring_view = basic_string_view<wchar_t>;
 using wformat_parse_context = basic_format_parse_context<wchar_t>;
@@ -33,7 +52,9 @@ inline auto runtime(wstring_view s) -> wstring_view { return s; }
 #else
 template <typename... Args>
 using wformat_string = basic_format_string<wchar_t, type_identity_t<Args>...>;
-inline auto runtime(wstring_view s) -> basic_runtime<wchar_t> { return {{s}}; }
+inline auto runtime(wstring_view s) -> runtime_format_string<wchar_t> {
+  return {{s}};
+}
 #endif
 
 template <> struct is_char<wchar_t> : std::true_type {};
@@ -41,15 +62,16 @@ template <> struct is_char<detail::char8_type> : std::true_type {};
 template <> struct is_char<char16_t> : std::true_type {};
 template <> struct is_char<char32_t> : std::true_type {};
 
-template <typename... Args>
-constexpr format_arg_store<wformat_context, Args...> make_wformat_args(
-    const Args&... args) {
+template <typename... T>
+constexpr auto make_wformat_args(const T&... args)
+    -> format_arg_store<wformat_context, T...> {
   return {args...};
 }
 
 inline namespace literals {
 #if FMT_USE_USER_DEFINED_LITERALS && !FMT_USE_NONTYPE_TEMPLATE_ARGS
-constexpr detail::udl_arg<wchar_t> operator"" _a(const wchar_t* s, size_t) {
+constexpr auto operator""_a(const wchar_t* s, size_t)
+    -> detail::udl_arg<wchar_t> {
   return {s};
 }
 #endif
@@ -78,9 +100,9 @@ template <typename Char, FMT_ENABLE_IF(!std::is_same<Char, char>::value)>
 auto vformat(basic_string_view<Char> format_str,
              basic_format_args<buffer_context<type_identity_t<Char>>> args)
     -> std::basic_string<Char> {
-  basic_memory_buffer<Char> buffer;
-  detail::vformat_to(buffer, format_str, args);
-  return to_string(buffer);
+  auto buf = basic_memory_buffer<Char>();
+  detail::vformat_to(buf, format_str, args);
+  return to_string(buf);
 }
 
 template <typename... T>
@@ -90,10 +112,10 @@ auto format(wformat_string<T...> fmt, T&&... args) -> std::wstring {
 
 // Pass char_t as a default template parameter instead of using
 // std::basic_string<char_t<S>> to reduce the symbol size.
-template <typename S, typename... Args, typename Char = char_t<S>,
+template <typename S, typename... T, typename Char = char_t<S>,
           FMT_ENABLE_IF(!std::is_same<Char, char>::value &&
                         !std::is_same<Char, wchar_t>::value)>
-auto format(const S& format_str, Args&&... args) -> std::basic_string<Char> {
+auto format(const S& format_str, T&&... args) -> std::basic_string<Char> {
   return vformat(detail::to_string_view(format_str),
                  fmt::make_format_args<buffer_context<Char>>(args...));
 }
@@ -108,11 +130,10 @@ inline auto vformat(
   return detail::vformat(loc, detail::to_string_view(format_str), args);
 }
 
-template <typename Locale, typename S, typename... Args,
-          typename Char = char_t<S>,
+template <typename Locale, typename S, typename... T, typename Char = char_t<S>,
           FMT_ENABLE_IF(detail::is_locale<Locale>::value&&
                             detail::is_exotic_char<Char>::value)>
-inline auto format(const Locale& loc, const S& format_str, Args&&... args)
+inline auto format(const Locale& loc, const S& format_str, T&&... args)
     -> std::basic_string<Char> {
   return detail::vformat(loc, detail::to_string_view(format_str),
                          fmt::make_format_args<buffer_context<Char>>(args...));
@@ -126,14 +147,14 @@ auto vformat_to(OutputIt out, const S& format_str,
     -> OutputIt {
   auto&& buf = detail::get_buffer<Char>(out);
   detail::vformat_to(buf, detail::to_string_view(format_str), args);
-  return detail::get_iterator(buf);
+  return detail::get_iterator(buf, out);
 }
 
-template <typename OutputIt, typename S, typename... Args,
+template <typename OutputIt, typename S, typename... T,
           typename Char = char_t<S>,
           FMT_ENABLE_IF(detail::is_output_iterator<OutputIt, Char>::value&&
                             detail::is_exotic_char<Char>::value)>
-inline auto format_to(OutputIt out, const S& fmt, Args&&... args) -> OutputIt {
+inline auto format_to(OutputIt out, const S& fmt, T&&... args) -> OutputIt {
   return vformat_to(out, detail::to_string_view(fmt),
                     fmt::make_format_args<buffer_context<Char>>(args...));
 }
@@ -149,18 +170,18 @@ inline auto vformat_to(
   auto&& buf = detail::get_buffer<Char>(out);
   vformat_to(buf, detail::to_string_view(format_str), args,
              detail::locale_ref(loc));
-  return detail::get_iterator(buf);
+  return detail::get_iterator(buf, out);
 }
 
-template <
-    typename OutputIt, typename Locale, typename S, typename... Args,
-    typename Char = char_t<S>,
-    bool enable = detail::is_output_iterator<OutputIt, Char>::value&&
-        detail::is_locale<Locale>::value&& detail::is_exotic_char<Char>::value>
+template <typename OutputIt, typename Locale, typename S, typename... T,
+          typename Char = char_t<S>,
+          bool enable = detail::is_output_iterator<OutputIt, Char>::value &&
+                        detail::is_locale<Locale>::value &&
+                        detail::is_exotic_char<Char>::value>
 inline auto format_to(OutputIt out, const Locale& loc, const S& format_str,
-                      Args&&... args) ->
+                      T&&... args) ->
     typename std::enable_if<enable, OutputIt>::type {
-  return vformat_to(out, loc, to_string_view(format_str),
+  return vformat_to(out, loc, detail::to_string_view(format_str),
                     fmt::make_format_args<buffer_context<Char>>(args...));
 }
 
@@ -171,36 +192,36 @@ inline auto vformat_to_n(
     OutputIt out, size_t n, basic_string_view<Char> format_str,
     basic_format_args<buffer_context<type_identity_t<Char>>> args)
     -> format_to_n_result<OutputIt> {
-  detail::iterator_buffer<OutputIt, Char, detail::fixed_buffer_traits> buf(out,
-                                                                           n);
+  using traits = detail::fixed_buffer_traits;
+  auto buf = detail::iterator_buffer<OutputIt, Char, traits>(out, n);
   detail::vformat_to(buf, format_str, args);
   return {buf.out(), buf.count()};
 }
 
-template <typename OutputIt, typename S, typename... Args,
+template <typename OutputIt, typename S, typename... T,
           typename Char = char_t<S>,
           FMT_ENABLE_IF(detail::is_output_iterator<OutputIt, Char>::value&&
                             detail::is_exotic_char<Char>::value)>
-inline auto format_to_n(OutputIt out, size_t n, const S& fmt,
-                        const Args&... args) -> format_to_n_result<OutputIt> {
+inline auto format_to_n(OutputIt out, size_t n, const S& fmt, T&&... args)
+    -> format_to_n_result<OutputIt> {
   return vformat_to_n(out, n, detail::to_string_view(fmt),
                       fmt::make_format_args<buffer_context<Char>>(args...));
 }
 
-template <typename S, typename... Args, typename Char = char_t<S>,
+template <typename S, typename... T, typename Char = char_t<S>,
           FMT_ENABLE_IF(detail::is_exotic_char<Char>::value)>
-inline auto formatted_size(const S& fmt, Args&&... args) -> size_t {
-  detail::counting_buffer<Char> buf;
+inline auto formatted_size(const S& fmt, T&&... args) -> size_t {
+  auto buf = detail::counting_buffer<Char>();
   detail::vformat_to(buf, detail::to_string_view(fmt),
                      fmt::make_format_args<buffer_context<Char>>(args...));
   return buf.count();
 }
 
 inline void vprint(std::FILE* f, wstring_view fmt, wformat_args args) {
-  wmemory_buffer buffer;
-  detail::vformat_to(buffer, fmt, args);
-  buffer.push_back(L'\0');
-  if (std::fputws(buffer.data(), f) == -1)
+  auto buf = wmemory_buffer();
+  detail::vformat_to(buf, fmt, args);
+  buf.push_back(L'\0');
+  if (std::fputws(buf.data(), f) == -1)
     FMT_THROW(system_error(errno, FMT_STRING("cannot write to file")));
 }
 
@@ -217,13 +238,22 @@ template <typename... T> void print(wformat_string<T...> fmt, T&&... args) {
   return vprint(wstring_view(fmt), fmt::make_wformat_args(args...));
 }
 
+template <typename... T>
+void println(std::FILE* f, wformat_string<T...> fmt, T&&... args) {
+  return print(f, L"{}\n", fmt::format(fmt, std::forward<T>(args)...));
+}
+
+template <typename... T> void println(wformat_string<T...> fmt, T&&... args) {
+  return print(L"{}\n", fmt::format(fmt, std::forward<T>(args)...));
+}
+
 /**
   Converts *value* to ``std::wstring`` using the default format for type *T*.
  */
 template <typename T> inline auto to_wstring(const T& value) -> std::wstring {
   return format(FMT_STRING(L"{}"), value);
 }
-FMT_MODULE_EXPORT_END
+FMT_END_EXPORT
 FMT_END_NAMESPACE
 
 #endif  // FMT_XCHAR_H_