+ Add HandleOutgoingData, which tests for m_Protocol before calling DataPrepared.
* Change std::function to bool + if/else to handle incoming data; it's almost certainly faster.
* Fixes#5260
* Update player list gamemode on world change
* Fix invisibility for spectators, use entity metadata
* Populate m_World for cPlayers on load
- Remove SendPlayerMaxSpeed, a duplicate of SendEntityProperties
* Flush out all pending, buffered changes at the end of each tick, after every chunk is ticked. This makes every block update client-side in unison, instead of unlucky ones only being sent 1 tick later.
* Re-add buffer for outgoing network data; IOCP async WSASend has higher overhead than expected... Fixes regression introduced in 054a89dd9
First one: add missing exception handler in ProcessProtocolIn
Second: remove faulty logic dealing with incomplete packets.
`a_Data = a_Data.substr(m_Buffer.GetUsedSpace() - m_Buffer.GetReadableSpace());`
was incorrect; it attempted to apply a length derived from m_Buffer to an unrelated a_Data. Its purpose was to give cProtocol the data the client sent, minus initial handshake bytes. However, we can use the knowledge that during initial handshake, there is no encryption and every byte can be written unchanged into m_Buffer, to just call cProtocol with a data length of zero. This will cause it to parse from m_Buffer - wherein we have already written everything the client sent - with no a_Data manipulation needed.
Additionally, removed UnsupportedButPingableProtocolException (use of exception as control flow) and encode this state as m_Protocol == nullptr, id est "no protocol for this unsupported version", which is then handled by cMultiVersionProtocol itself.
+ A cPlayer, once created, has a strong pointer to the cClientHandle. The player ticks the clienthandle. If he finds the handle destroyed, he destroys himself in turn. Nothing else can kill the player.
* The client handle has a pointer to the player. Once a player is created, the client handle never outlasts the player, nor does it manage the player's lifetime. The pointer is always safe to use after FinishAuthenticate, which is also the point where cProtocol is put into the Game state that allows player manipulation.
+ Entities are once again never lost by constructing a chunk when they try to move into one that doesn't exist.
* Fixed a forgotten Super invocation in cPlayer::OnRemoveFromWorld.
* Fix SaveToDisk usage in destructor by only saving things cPlayer owns, instead of accessing cWorld.
- Removed RemoveFromAllChunks. On destruction cWorld::RemovePlayer calls RemoveClientFromChunks already, and there's no need to manually clear the chunk lists.
* Turn off global-constructors warning. These are needed to implement cRoot signal handler functionality
* Add Clang flags based on version lookup instead of a compile test. The CMake config process is single threaded and slow enough already
* Reduced GetStackValue verbosity
+ Clarify EnchantmentLevel, StayCount, AlwaysTicked, ViewDistance signedness
+ Give SettingsRepositoryInterface a move constructor to simplify main.cpp code
- Remove do {} while (false) construction in redstone handler
* Tried to fix a small issue...
Ended up rewriting a bunch of god awful, opaque code with no source
and no sense. Who names a function GetPlayerRelativeBlockHardness???
It's gone now. We're safe again.
* Testing anti-cheat.
* Tidy up debug logging.
* Remove empty member declaration.
* Rewrite GetDigSpeed slightly for better readability.
* GetMiningProgressPerTick now returns 1 when instantly mined. Fixed hasily written typo.
* Comment style and typo fixes.
* Use lapis for enchanting, subtract correct number of levels, ClientHandle now selects from pregenerated list.
Co-authored-by: Tiger Wang <ziwei.tiger@outlook.com>
* Use lapis for enchanting, subtract correct number of levels.
* Enforced distancing regulation better than our government.
Co-authored-by: Tiger Wang <ziwei.tiger@outlook.com>
Rorkh
Ethan Jones
Tiger Wang
Mattes D
KingCol13
nshah25
Damián Imrich
12xx12
mathiascode
peterbell10