Fix some static analysis issues about uninitialized values. Most of them are false positives, always initialized before being accessed, while some of them are real issues spotted by valgrind too.
Fix some static analysis issues, mostly false positive about fields not initialized in the constructor. It's good practice anyway to always initialize them.
Properly save transport related data to database when saving a Player .
On Player login update the current position as relative to transport, in case the transport moved.
Fix a database race condition between authserver saving session key to database asynchronously and worldserver reading it; session key is now saved synchronously.
- Changed default loggers and appenders
- '.' determines the relation between loggers ("type.subtype" inherits "type" logger setting if logger "type.subtype" is not defined)
- When core logs a message it search for the correct logger (root is the default one)
ie: a message logged with "type.subtype"
* Core will try to find a logger with name "type.subtype", if its not found then will search for "type", again if its not found it will return the default one "root"
Add additional assert to assure that helgrind reports related to SQLQueryHolderTask::Execute() and WorldSession::ProcessQueryCallbacks() are false positives.
The only chance for WorldSession::ProcessQueryCallbacks() to access the SQL queries before they are completed is if m_result.ready() is true before the end of SQLQueryHolderTask::Execute() .
Helgrind false positive log:
Possible data race during read of size 8 at 0x2BD641E0 by thread #7
Locks held: none
at : SQLQueryHolder::GetPreparedResult(unsigned long) (QueryHolder.cpp:107)
by : Player::LoadFromDB(unsigned int, SQLQueryHolder*) (Player.cpp:16917)
by : WorldSession::HandlePlayerLogin(LoginQueryHolder*) (CharacterHandler.cpp:807)
by : WorldSession::ProcessQueryCallbacks() (WorldSession.cpp:1112)
by : WorldSession::Update(unsigned int, PacketFilter&) (WorldSession.cpp:401)
by : World::UpdateSessions(unsigned int) (World.cpp:2646)
by : World::Update(unsigned int) (World.cpp:2003)
by : WorldRunnable::run() (WorldRunnable.cpp:60)
by : ACE_Based::Thread::ThreadTask(void*) (Threading.cpp:186)
by : ACE_OS_Thread_Adapter::invoke() (in /usr/lib/libACE-6.0.3.so)
by : mythread_wrapper (hg_intercepts.c:233)
by : start_thread (pthread_create.c:311)
This conflicts with a previous write of size 8 by thread #4
Locks held: none
at : SQLQueryHolder::SetPreparedResult(unsigned long, PreparedResultSet*) (QueryHolder.cpp:140)
by : SQLQueryHolderTask::Execute() (QueryHolder.cpp:196)
by : SQLOperation::call() (SQLOperation.h:65)
by : DatabaseWorker::svc() (DatabaseWorker.cpp:45)
by : ACE_Task_Base::svc_run(void*) (in /usr/lib/libACE-6.0.3.so)
by : ACE_Thread_Adapter::invoke_i() (in /usr/lib/libACE-6.0.3.so)
by : ACE_Thread_Adapter::invoke() (in /usr/lib/libACE-6.0.3.so)
by : mythread_wrapper (hg_intercepts.c:233)
Address 0x2BD641E0 is 16 bytes inside a block of size 792 alloc'd
at : operator new(unsigned long) (vg_replace_malloc.c:319)
by : __gnu_cxx::new_allocator<std::pair<SQLElementData, SQLResultSetUnion> >::allocate(unsigned long, void const*) (new_allocator.h:104)
by : std::_Vector_base<std::pair<SQLElementData, SQLResultSetUnion>, std::allocator<std::pair<SQLElementData, SQLResultSetUnion> > >::_M_allocate(unsigned long) (in /home/jackpoz/trinity/bin/worldserver)
by : std::vector<std::pair<SQLElementData, SQLResultSetUnion>, std::allocator<std::pair<SQLElementData, SQLResultSetUnion> > >::_M_fill_insert(__gnu_cxx::__normal_iterator<std::pair<SQLElementData, SQLResultSetUnion>*, std::vector<std::pair<SQLElementData, SQLResultSetUnion>, std::allocator<std::pair<SQLElementData, SQLResultSetUnion> > > >, unsigned long, std::pair<SQLElementData, SQLResultSetUnion> const&) (vector.tcc:483)
by : std::vector<std::pair<SQLElementData, SQLResultSetUnion>, std::allocator<std::pair<SQLElementData, SQLResultSetUnion> > >::insert(__gnu_cxx::__normal_iterator<std::pair<SQLElementData, SQLResultSetUnion>*, std::vector<std::pair<SQLElementData, SQLResultSetUnion>, std::allocator<std::pair<SQLElementData, SQLResultSetUnion> > > >, unsigned long, std::pair<SQLElementData, SQLResultSetUnion> const&) (stl_vector.h:1024)
by : std::vector<std::pair<SQLElementData, SQLResultSetUnion>, std::allocator<std::pair<SQLElementData, SQLResultSetUnion> > >::resize(unsigned long, std::pair<SQLElementData, SQLResultSetUnion>) (stl_vector.h:707)
by : SQLQueryHolder::SetSize(unsigned long) (QueryHolder.cpp:167)
by : LoginQueryHolder::Initialize() (CharacterHandler.cpp:66)
by : WorldSession::HandlePlayerLoginOpcode(WorldPacket&) (CharacterHandler.cpp:788)
by : WorldSession::Update(unsigned int, PacketFilter&) (WorldSession.cpp:363)
by : World::UpdateSessions(unsigned int) (World.cpp:2646)
by : World::Update(unsigned int) (World.cpp:2003)
- Drop groups (roles than can have inherited roles) and roles (set of
permissions)
- Permissions can now have inherited permissions (those act as roles)
RBAC DB structure is now limited to four tables
- rbac_permissions: Contains permissions and roles
- rbac_linked_permissions: Contains the relation between permissions and
linked permissions (those permissions that have linked permissions are
called roles)
- rbac_default_permissions: Contains the list of permissions to be granted
to each security level [Added to maintain compatibility in an easy way]
- rbac_account_permissions: Contains the list of permissions granted or
denied for a particular account.
NOTE: IF YOU ARE USING CUSTOM PERMISSIONS, ROLES OR GROUPS CHECK THE SQL
BEFORE APPLYING...
Change DatabaseWorkerPool<T>::DirectExecute() to handle PreparedStatement disposal in the same way of DatabaseWorkerPool<T>::Query() , thanks Joschiwald for patch.
Restore 2 disabled warnings, 1 already disabled in CMake with the WITH_WARNINGS flag set to False and the other disabled by default as stated in VS documentation http://msdn.microsoft.com/en-us/library/aa984150.aspx
Fix race condition in Log by using atomic operators.
Helgrind log:
Possible data race during read of size 8 at 0x7379D98 by thread #1
Locks held: none
at 0x15AE9C7: AppenderFile::_write(LogMessage const&) (AppenderFile.cpp:59)
by 0x15ADFF8: Appender::write(LogMessage&) (Appender.cpp:106)
by 0x159F14E: Logger::write(LogMessage&) (Logger.cpp:83)
by 0x15A215B: Log::write(LogMessage*) (Log.cpp:290)
by 0x15A200F: Log::vlog(LogFilterType, LogLevel, char const*, __va_list_tag*) (Log.cpp:272)
by 0x15A2682: Log::outInfo(LogFilterType, char const*, ...) (Log.cpp:364)
by 0xF7DA28: Master::Run() (Master.cpp:296)
by 0xF835E8: main (Main.cpp:142)
This conflicts with a previous write of size 8 by thread #10
Locks held: none
at 0x15AE9D7: AppenderFile::_write(LogMessage const&) (AppenderFile.cpp:59)
by 0x15ADFF8: Appender::write(LogMessage&) (Appender.cpp:106)
by 0x159F14E: Logger::write(LogMessage&) (Logger.cpp:83)
by 0x15A215B: Log::write(LogMessage*) (Log.cpp:290)
by 0x15A200F: Log::vlog(LogFilterType, LogLevel, char const*, __va_list_tag*) (Log.cpp:272)
by 0x15A2682: Log::outInfo(LogFilterType, char const*, ...) (Log.cpp:364)
by 0xF7EC1F: FreezeDetectorRunnable::run() (Master.cpp:98)
by 0x15A5B3E: ACE_Based::Thread::ThreadTask(void*) (Threading.cpp:186)
Address 0x7379D98 is 88 bytes inside a block of size 96 alloc'd
at 0x4C2C857: operator new(unsigned long) (in /usr/lib/valgrind/vgpreload_helgrind-amd64-linux.so)
by 0x15A0FE3: Log::CreateAppenderFromConfig(char const*) (Log.cpp:150)
by 0x15A1AED: Log::ReadAppendersFromConfig() (Log.cpp:244)
by 0x15A31B4: Log::LoadFromConfig() (Log.cpp:469)
by 0x15A053B: Log::Log() (Log.cpp:35)
by 0xF75CD1: ACE_Singleton<Log, ACE_Thread_Mutex>::ACE_Singleton() (Singleton.inl:13)
by 0xF754A4: ACE_Singleton<Log, ACE_Thread_Mutex>::instance() (Singleton.cpp:91)
by 0xF8351A: main (Main.cpp:135)
Change IntervalTimer::Reset() behavior to handle system clock changes forward and backward.
This fixes IntervalTimer:.Passed() returning true till it catches up to the new time, triggering the event up to "std::numeric_limits<time_t>::max() / _interval" times.
Fixes https://github.com/TrinityCore/TrinityCore/issues/5816
Replace thread-unsafe localtime() http://www.cplusplus.com/reference/ctime/localtime/ with thread-safe portable ACE_OS::localtime_r() .
Helgrind log:
Possible data race during read of size 4 at 0x6F183C0 by thread #1
Locks held: none
at 0x14E72E3: World::InitDailyQuestResetTime() (World.cpp:2772)
by 0x14E3A01: World::SetInitialWorldSettings() (World.cpp:1790)
by 0x101122A: Master::Run() (Master.cpp:164)
by 0x101740C: main (Main.cpp:142)
This conflicts with a previous write of size 4 by thread #2
Locks held: none
at 0x6C2D3BA: __tzfile_compute (tzfile.c:797)
by 0x6C2D036: __tz_convert (tzset.c:627)
by 0x164146C: LogMessage::getTimeStr(long) (Appender.cpp:23)
by 0x1641550: LogMessage::getTimeStr() (Appender.cpp:31)
by 0x1641722: Appender::write(LogMessage&) (Appender.cpp:80)
by 0x1633FCE: Logger::write(LogMessage&) (Logger.cpp:83)
by 0x16433D8: LogOperation::call() (LogOperation.cpp:29)
by 0x16428A4: LogWorker::svc() (LogWorker.cpp:45)
Settings within worldserver.conf:
Three settings for secruity level:
0 - None - No change to current system
1 - Email - Always requires the email entered on registration for confirming.
2 - RBAC - Groups applied with the RBAC role always require the email entered on registration for confirming.
RBAC default to every group. Changed some logs to make it more clear what is going on at all.
Emails may now no longer exceed 64 chars. Current email is used as regmail.
On account creation, two emails are saved. Registration email and normal email. Normal email is relevant afterwards. Registration email can be changed by console ONLY.
Includes new commands and changes to existing ones:
.account fulfills several new functions:
* Still prints GM Level.
* If account has permission, it displays the current email. This is not defaulted to any group.
* Security level is displayed. Also displays if user has RBAC perm if RBAC security mode is selected
.account email allows user to change email with sufficient confirmation
.account set sec email allows higher sec with higher sec than account to change the normal email. Registrationemail remains untouched here.
.account set sec regmail allows console to change registration email.
.pinfo now displays the registration and normal mail.
Also fixes .learn all crafts.
Closes#10558
New column in account table is a base32 of token key bytes,
coincidentally it is the same format Google's Authenticator Android app uses.
If you want that to work, set system time on server correctly and use ntpd.
Closes#10527
Signed-off-by: Nay <dnpd.dd@gmail.com>
Increment the reference count of m_task in Thread::start() before spawning the actual Thread that will execute the task, otherwise the thread might finish, decRef the task and delete it.
Valgrind log of the issue:
Invalid read of size 8
at 0x1314CAD: ACE_Atomic_Op_GCC<long>::operator++() (Atomic_Op_GCC_T.inl:34)
by 0x15933FB: ACE_Based::Runnable::incReference() (Threading.h:36)
by 0x1592D2D: ACE_Based::Thread::start() (Threading.cpp:136)
by 0x1592C37: ACE_Based::Thread::Thread(ACE_Based::Runnable*) (Threading.cpp:111)
by 0xF6C463: Master::Run() (Master.cpp:195)
by 0xF725D0: main (Main.cpp:142)
Address 0x26137278 is 8 bytes inside a block of size 24 free'd
at 0x4C2B59C: operator delete(void*) (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0xF67FDB: RARunnable::~RARunnable() (RARunnable.cpp:55)
by 0x1593441: ACE_Based::Runnable::decReference() (Threading.h:40)
by 0x1592E92: ACE_Based::Thread::ThreadTask(void*) (Threading.cpp:186)
by 0x515EA35: ACE_OS_Thread_Adapter::invoke() (in /usr/lib/libACE-6.0.3.so)
by 0x5F19F8D: start_thread (pthread_create.c:311)
by 0x6A46E1C: clone (clone.S:113)
Closes#10619
Remove mutex from BigNumber class - it didn't do what it was advertised to do - consider using the "locked" array outside of the function in which it was "locked".
