* Core/Metric: Log detailed metrics about each opcode handler
* Add new panel to Performance profiling dashboard and use fill(0) instead of fill(none)
* Add new settings Metric.Threshold.* to be able to specify the minimum threshold for the specified metrics
* Update dashboard
* Change thresholds to be required to send the metrics. A TC_METRIC_DETAILED_TIMER metric with an expected threshold not configured will be ignored
* Use typedef Milliseconds
* Refresh realms on load
* Core/Metric: Add new cmake option WITH_DETAILED_METRICS to enable more metrics
Add more detailed metrics in World::UpdateSessions().
* Attempt using C++17 features
* Fix cmake typo
Co-authored-by: NoName <322016+Faq@users.noreply.github.com>
* Fix build with WITHOUT_METRICS
* Update Performance profiling dashboard with Update sessions panel
* Add panel to show sessions with update time above 100 ms
* Move legends on the right and add max update time diff in the legend of sessions
Co-authored-by: NoName <322016+Faq@users.noreply.github.com>
uint64 CalculateTime(uint64 t_offset) has been replaced with Milliseconds CalculateTime(Milliseconds t_offset).
Also add the std::chrono-ified overload void AddEvent(BasicEvent*, Milliseconds, bool)
DelayEvents(Milliseconds delay) had different semantics than
DelayEvents(Milliseconds delay, uint32 group).
The first method delayed the events only in the case the internal timer
already ticked at least for the amount of delay. In contrast the latter method
delayed events regardless of the internal timer value.
Use the latter semantics for DelayEvents(Milliseconds delay) as well which makes
the outcome more predictable. Adapt tests accordingly.
With the switch to std::chrono return type of GetTimeUntilEvent
we don't run into an overflow condition which happend with the previous
uint32 return value if the events scheduled execution time is in the past.
Test for this case.
- Did you know BigNumber quietly assumes every byte array it gets is little-endian, even though openssl bignums use big-endian? Now you do!
- In entirely unrelated news, make the above behavior explicit through a default-true boolean, same as existing GetBytes derivatives.
- Also, if you are in the enlightened openssl 1.1 crowd, there's no more endian wrangling involved, because openssl now does all of that for us. Progress!
- Fix a handful of 1/256 bugs with most significant byte zero in BigNumber
- Get rid of (most of) the C-style arrays in authserver
- CryptoRandom as a unified source for cryptographic randomness
- Bring our other crypto APIs into 2020
- BigNumber usability improvements
- Authserver is now actually readable as a result of all of the above
To enable the test suite, make sure to configure CMake with -DBUILD_TESTING=1 , since it is disabled by default. The catch2 dependency will be downloaded during configure time.
Also add a new target "tests-common", which includes unit tests for the "common" project. To finally run the tests use the "test" target.
CircleCI: Run unit tests
* Core/Metric: Add more metrics about World::Update() loop
* Add new Performance profiling grafana dashboard
* Add new cmake setting WITHOUT_METRICS that disables all metrics
As new metrics are added, someone might want to disable them if unused.
* Add more metrics about World::Update() loop
* Remove old performance profiling features
Remove "server set difftime" command, "-RecordUpdateTimeDiffInterval" and "MinRecordUpdateTimeDiff" worldserver config settings and their related code.
The whole UpdateTime class could be removed too.
* Update and rename 2020_07_99_99_world.sql to 2020_07_24_00_world.sql
* Core/Combat: Disable triggers entering combat
* Move check to database flags_extra flag in creature_template with value 0x00002000
* Fix Rotface puddle stalker too
* Rename 9999_99_99_99_world.sql to 2020_07_22_00_world.sql
The intention here is to add additional 30 seconds of delay to the current delay of EVENT_FLY_TO_RANDOM_PILLAR.
Instead of EventMap::GetNextEventTime, EventMap::GetTimeUntilEvent has to be used which returns the currently
scheduled delay until the event occurs.
EventMap::GetNextEventTime instead exposes the timer value internal to EventMap at which the event should occur.
