Assertion crash in Unit destructor

[silviu20092]

Description:

Assertion crash in Unit destructor.
https://gist.github.com/silviu20092/8d98e52f2a8be380f5c83aa99790c96f

Steps to reproduce the problem:

Seems to be random and pretty rare, I don't know the exact steps to reproduce it.

Branch(es):

3.3.5

TC rev. hash/commit:

80f30d5

Operating system:

Debian 10

[jackpoz]

do you have any custom change ? does Server.log show any error before crashing ?

[silviu20092]

I do have a couple of extra spell scripts and some custom player scripts (pvp/duels hooks related) and the Autobalance patch from here: https://github.com/trickerer/Autobalance_for_NPCBots

[acidmanifesto]

TrinityCore/src/server/game/Entities/Unit/Unit.cpp

Line 414 in 80f30d5

ASSERT(m_Controlled.empty());

Im just pasting this here for quicker reference on the issue to the line that is present in that gist. I have not encountered it personally. just posting the referenced file and line link.

[jackpoz]

@silviu20092 when reporting an issue please mention if you have custom changes and if possible include a link to the public github fork with the version you are running.

I assume you have npc bots too as you have autobalance ?

[silviu20092]

@jackpoz sorry, nope there is no NPC bots, that patch is meant to work with the npc bots aswell but I don't use them. For this test server there is just the autobalance and a couple of spellscripts (just written for fun really, not meant to mean anything). Besides the autobalance changes there are 0 changes compared to 3.3.5 TC branch.

[silviu20092]

I see that in Map.cpp at line 3620 we have:
RemoveFromMap(obj->ToCreature(), true);
Inside this function is the following code:
obj->RemoveFromWorld();
This will remove everything in m_Controlled (if the object is in world at the time of call). Then later in the same RemoveFromMap function DeleteFromWorld(obj); is called which essentially does a delete obj and Unit destructor gets called.

So how is m_Controlled not empty at the time of object deletion if it was cleared just earlier? Maybe at the time of obj->RemoveFromWorld(); the creature was no longer in world (but had m_Controlled not empty)?

[jackpoz]

did Server.log show any message by the way before the crash ?

[silviu20092]

Unfortunately I had it on overwrite mode and I didn't get to see anything before the crash.

[jackpoz]

Server.log should contain more information, either change to append mode or add the date to the log name from the settings.

I always create a new log file with date & time included in the file name.
Related lines in my worldserver.conf file:

#  Appender config values: Given an appender "name"
#    Appender.name
#        Description: Defines 'where to log'.

Appender.Console=1,3,0
Appender.Server=2,2,15,World/Server.log
Appender.GM=2,2,15,gm/gm_%s.log
Appender.DBErrors=2,2,15,DBErrors/DBErrors.log

#  Logger config values: Given a logger "name"
#    Logger.name
#        Description: Defines 'What to log'

Logger.root=5,Console Server
Logger.server=3,Console Server
Logger.commands.gm=3,GM
Logger.scripts.hotswap=3,Console Server
Logger.sql.sql=5,Console DBErrors
Logger.sql.updates=3,Console Server
Logger.mmaps=3,Server

[silviu20092]

Yeah I have set it up on append mode now, however I did not encounter this crash anymore, seems to be really isolated.

[ckegg]

Here are 4 crash logs on different time but always on the same creature 37134, Ymirjar Huntress
crash-2021-04-21-23-52-19.log
crash-2021-04-22-00-22-20.log
crash-2021-05-04-22-07-30.log
crash-2021-05-04-22-36-01.log