Example:
==8753== Possible data race during read of size 8 at 0x2e803e60 by thread #8
==8753== at 0xDD0FFC: PreparedResultSet::operator[](unsigned int) const (QueryResult.h:83)
==8753== by 0x131D82D: WorldSession::HandleCharEnum(Trinity::AutoPtr<PreparedResultSet, ACE_Null_Mutex>) (CharacterHandler.cpp:217)
==8753== by 0x118F8C3: WorldSession::ProcessQueryCallbacks() (WorldSession.cpp:1087)
{..}
==8753== Address 0x2e803e60 is 0 bytes inside a block of size 16 alloc'd
==8753== at 0x4C28B75: operator new(unsigned long) (vg_replace_malloc.c:261)
==8753== by 0x137D515: __gnu_cxx::new_allocator<Field*>::allocate(unsigned long, void const*) (new_allocator.h:92)
==8753== by 0x137D378: std::_Vector_base<Field*, std::allocator<Field*> >::_M_allocate(unsigned long) (in /home/aokromes/bin/bin/worldserver)
==8753== by 0x137CEC6: std::vector<Field*, std::allocator<Field*> >::_M_fill_insert(__gnu_cxx::__normal_iterator<Field**, std::vector<Field*, std::allocator<Field*> > >, unsigned long, Field* const&) (vector.tcc:416)
==8753== by 0x137CB27: std::vector<Field*, std::allocator<Field*> >::insert(__gnu_cxx::__normal_iterator<Field**, std::vector<Field*, std::allocator<Field*> > >, unsigned long, Field* const&) (stl_vector.h:944)
==8753== by 0x137C964: std::vector<Field*, std::allocator<Field*> >::resize(unsigned long, Field*) (stl_vector.h:632)
==8753== by 0x137BFCA: PreparedResultSet::PreparedResultSet(st_mysql_stmt*, st_mysql_res*, unsigned long, unsigned int) (QueryResult.cpp:98)
==8753== by 0x137A12B: MySQLConnection::Query(PreparedStatement*) (MySQLConnection.cpp:493)
==8753== by 0x137F663: PreparedStatementTask::Execute() (PreparedStatement.cpp:412)
Have a feeling this is a false positive, though. Nevertheless good to get rid of.
- Remove redundant locking in PingOperation. Since these are delayed to async threads with their own MySQL connection, no concurrent access here either.
- Codestyle cleanup
- Documentation refining
Most of the cores queries have been converted to prepared statements now. If your code contains queries, please either use the exiting or implement new statements.
The only exclusions are one-time queries (e.g. for core startup) and queries that can't be prepared due to limitations.
Note: The default config file action for clients failing the checks can be changed for each check via the characters.warden_action table
Credits to TOM_RUS
Fix priority of the player when its added to the lfg group
Better implementation of the Dungeon Deserter debuff
Rewrite the NeedBeforeGreed loot
Fixed players being shown as Unknown Entity when entering the lfg group
Some incremental optimizations after original patch
Thanks to Retriman and Paecman for base implementation
Signed-off-by: Machiavelli <machiavelli.trinity@gmail.com>
Spline movement controls movements of server-side controlled units (monster movement, taxi movement, etc).
Proper implementation of effects such as charge, jump, cyclic movement will rely on it.
However, need improve our states system before.
Technical changes:
* Added linear, catmullrom and bezier3 splines which based on client's algorthims. They can be reused for proper transport position interpolation.
* Precission increased. There are no more position desync issues since client's position calculation formulas used.
* Now possible to move by paths with multiple points, send whole path to client.
--
Original author of research and implementation: SilverIce. Massive kudos.
Original port for Trinity (ref #4629) Chaplain and Venugh
With the following incremental fixes during my review:
- Restore flightmaster end grid pre-loading
- Fix uninitialized Creature::m_path_id
- Add missing trinity_string entries for .movegens command
- Fix a bug in WaypointMovementGenerator that would trigger unexpected pausing at waypoints for various amounts of time
Known issues:
- Errors like WaypointMovementGenerator::LoadPath creature XXX (Entry: YYYYY GUID: ZZZZZZ) doesn't have waypoint path id: 0.
This is caused by bad DB data. This commit didn't "break" it.
Do not forget to re-run CMake before compiling.
Machiavelli