Memory leak WIP #7 & #8

[botmeister]

Looks like there's a memory leak with these two versions. Ever since upgrading to WIP #7 my server runs out of memory over a period of a couple days or so. I upgraded to WIP #8 and the same thing happens.

Anyone else running a bot server where the bots get dropped and added frequently?

[Onno Kreuzinger]

positive, my bot's don't get "older" than 16-30 hours, then the server restarts, the old realbot made it atleast for 48 hrs, sometimes 72+.
for the BU server this seems not to apply, perhaps because it is less often used ?

cheers

*edit*
oh it just has plenty of ram *g*
but it gets used ;-)

[T!m]

yup, same problem here, at leased with wip8, lots of crashes and bot drops, but I had no problems with wip7.

the server won't even dtay up for 36 hours, hell i'm glad if it stays up for 12 hours.
i thought it was the exp. files getting to big again.

[botmeister]

What I'm look for is if the memory usage keeps going up and up as the server runs, then if it drops back to normal as soon as you quit the hlds server. That's what I see happening. I'm not sure yet where the memory leak is, it could be seen on map changes, or during normal game play.

[T!m]

Originally Posted by botmeister

What I'm look for is if the memory usage keeps going up and up as the server runs, then if it drops back to normal as soon as you quit the hlds server.

1

i'l look for that tomorrow.

[Onno Kreuzinger]

hmm i could make a regression test with gdb running, setting memory limit to 64MB should crash it quite fast...

[Terran]

My server is running for 94h so far with 2-6 bots.
Memory usage of the engine: 636MB !!!

[DreamLord34]

My server has never crashed with just RB as the only bot. It would before when I had HPB in the mix. I do however get bots kicked for no reason.

[stefanhendriks]

*ATTENTION*
I have found code that would allocate memory using 'malloc' on map change. This is done on INITIALIZATION of the game dll but also on map change. On map change however the variable does not get freed. I suspect that due this accumilation of malloc's i will eventually eat up all your ram!

I have added a 'free memory' function before the initialization. I will release a WIP 8.5 in a few hours. I hope this leak will be solved then, perhaps this also solves the map change bug... (?)

[botmeister]

Originally Posted by stefanhendriks

*ATTENTION*
I have found code that would allocate memory using 'malloc' on map change. This is done on INITIALIZATION of the game dll but also on map change. On map change however the variable does not get freed. I suspect that due this accumilation of malloc's i will eventually eat up all your ram!

I have added a 'free memory' function before the initialization. I will release a WIP 8.5 in a few hours. I hope this leak will be solved then, perhaps this also solves the map change bug... (?)

Thanks for the quick fix stefan!