MegaGlest Forum

Archives (read only) => Glest Advanced Engine => Bug reports => Topic started by: silnarm on 2 June 2011, 22:14:54

Title: [Fixed] Heap corruption.
Post by: silnarm on 2 June 2011, 22:14:54

Just a heads up to anyone running from git, I discovered this morning we have another heap corruption problem. It seems that our leak dumper was hindering the mscrt checks somehow, and I have had leak dumping turned on for a long time now (until this morning...), so it may not necessarily be something very new.

This could be causing weirdness or crashing depending on the build, I don't have the time to investigate properly just now, but hopefully will get a chance to track it down tonight.

Title: Re: Heap corruption.
Post by: will on 3 June 2011, 06:35:48

(Running under valgrind on linux is also cool for tbis stuff (although my intel drivers cause me too much noise in the results, those witb better drivers might have more luck)

Title: Re: Heap corruption.
Post by: silnarm on 3 June 2011, 09:48:45

Yeah, valgrind is a great tool... Fortunately it turned out it was something new, so it didn't take long to find just using strategically placed _CrtCheckMemory() calls.

I am somewhat curious what valgrind might have had to say about it now though, as the problem was actually caused by some very dodgy duplicated class defs (which I did to avoid changing headers...), one of which had been modified in one file but not the other.

Lesson for the day: hacky shit that was done to avoid changing headers should always be cleaned up before comitting!