Heap corruption happens

[ohlidalp]

To laymen: heap corruption is a problem with dynamic RAM allocation and practically means: there's a bug somewhere in RoR which trashes our data in RAM, it's extra hard to find and it can theoretically crash RoR anytime for anybody 😟

Links:

• http://www.ogre3d.org/forums/viewtopic.php?f=4&t=93326 - OGRE forum thread about this problem.
• Fixed crash when loading "miniredmars" map. #1365 (comment) --- user tritonas00, linux, crash spawning 'bombinette.load' from "miniredmars" map.
• Crash on loading miniredmars map #1256 (comment) --- user Bob-Z, linux, spawning 'bombinette.load' from "miniredmars" map.
• (update) Crash on loading miniredmars map #1256 (comment)

I'm trying to use Microsoft Application Verifier to track it down, but I'm getting stuck on renderer init - both renderers exhibit some memory issues which halt the AppVerifier.

[tritonas00]

So you can reproduce the crash with opengl renderer?

[ohlidalp]

@tritonas00 Nope, miniredmars spawns OK for me with OpenGL. But I'm getting other weird crashes.

UPDATE: I just ran my OGRE testbed application () with the same debug tool --> Same output!

[tritonas00]

Same output!

Nice, now you don't need to search the whole RoR 😄

[ohlidalp]

It turned out a lot of the heap corruption issues with my "unlimited-nb" were my fault - I simply allocated less memory than was needed. There are still mysterious crashes going on, but it's not as freaky anymore.

[tritonas00]

The proposed solution here #1256 (comment) fixes the miniredmars issue here also.

[ohlidalp]

@tritonas00 The proposed solution here #1256 (comment) fixes the miniredmars issue here also.

Looks like all this breakage started with my recent refactor: 8eeb060, but there are 2 possible scenarios:

• Either the bad addressing was there from the start, but it only became a problem after the refactor - I changed data layout and removed several unused slabs of dynamically allocated memory.
• Or, I simply made a mistake during the refactor and introduced the adressing bug.

I need to take a close look, and I'm that won't happen anytime soon (busy).

Anyway, @Bob-Z @tritonas00 Thanks a lot for testing!

[ohlidalp]

Closing as resolved:

• There was a bug in 'wheels[2]' - if it had odd number of rays, memory corruption happened. This was causing Miniredmars map to crash: Crash on loading miniredmars map #1256 (comment) - Fixed.
• When removing the N/B limits, I made several bugs in memory size calculation, causing more corruption. Fixed.