0.23.0 Bugs

[Chris]

Do we still have any developers who work on a 32 bit Linux system? If yes, does OpenMW work there? And if it doesn't, any chance we could get it git bisect'ed?

OpenMW works on 32-bit Linux here. I'm not using the packaged builds, of course, but aside from a possible memory leak, I don't have any unexpected issues.

[Zini]

Okay. Then it is a packaging issue. Nothing more I can do here from my end.

[psi29a]

Someone claimed that win32 builds too had an issue (in another thread).

We need more testers on the 32-bit packages to track this down because I cannot reproduce it.

sirherrbatka says it is failing for him, I suggest we get a full and detailed dump of what is causing this.

Does it crash? Do you get a seg fault? Can you run gdb and get a back trace?

[sirherrbatka]

It does not crash or segfault. It simply rans out of memory very fast.

[psi29a]

Can you run valgrind on it? What is your process, just wondering around or do you just start it, not moving and watch mem usage go through the roof? Does the OOM killer come along and eventually kill it? Nvidia or AMD drivers or one of the open source drivers?

[sirherrbatka]

Binary packages are not builded with debug symbols?

[psi29a]

Binary packages are not builded with debug symbols?

No, they are not.. due to bug in Ogre 1.8.x from a bad assert that crashes it to desktop.

Attempts were made to fix it from scrawl to zini, but it was considered a bug in Ogre and we carried on.

[sirherrbatka]

So no valgrind.

I use nvidia binary blob.

PS
Staying in one place seems to prohibits this issue. But I once ran out of memory without leaving the starting cell.

[psi29a]

You cannot duplicate this with 64-bit deb/tar.gz ?

[sirherrbatka]

I don't have 64 bits OS on my computers (yet).