0.23.0 Bugs

Support for running, installing or compiling OpenMW

Before you submit a bug report for the first time, please read: Bug reporting guidelines
User avatar
psi29a
Posts: 4807
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: 0.23.0 Bugs

Post by psi29a » 02 May 2013, 11:12

OK... managed to duplicate the memory usage abuse, 70 % usage of 4GiB of ram, then a segfault. I ran gdb on it and got this backtrace:
Spoiler: Show

Looks like the bug in the other thread.

My machine, 64-bit Ubuntu raring using 32-bit tar.gz build. Using opensource radeon graphics.

The Ogre.log ballooned to 20MiB but the last lines (that repeated) was this:
12:06:29: Error prior to using GLSL Program Object : invalid operation
12:06:29: GLSL compile log: openmw_objects_vertex_20287399230:218(68): error: `materialEmissive' undeclared
0:218(72): error: type mismatch
0:218(72): error: Operands to arithmetic operators must be numeric
0:218(72): error: Operands to arithmetic operators must be numeric
0:218(72): error: type mismatch
0:221(54): error: `materialAmbient' undeclared
0:221(59): error: type mismatch
0:221(59): error: Operands to arithmetic operators must be numeric
0:221(71): error: Operands to arithmetic operators must be numeric
0:221(71): error: Operands to arithmetic operators must be numeric
0:221(71): error: type mismatch

12:06:29: Error prior to using GLSL Program Object : invalid operation
12:06:29: GLSL compile log: openmw_objects_vertex_20287399230:218(68): error: `materialEmissive' undeclared
0:218(72): error: type mismatch
0:218(72): error: Operands to arithmetic operators must be numeric
0:218(72): error: Operands to arithmetic operators must be numeric
0:218(72): error: type mismatch
0:221(54): error: `materialAmbient' undeclared
0:221(59): error: type mismatch
0:221(59): error: Operands to arithmetic operators must be numeric
0:221(71): error: Operands to arithmetic operators must be numeric
0:221(71): error: Operands to arithmetic operators must be numeric
0:221(71): error: type mismatch

12:06:29: Error prior to using GLSL Program Object : invalid operation
Ogre.log.bz2: https://dl.dropboxusercontent.com/u/396161/Ogre.log.bz2

suspicious file: https://dl.dropboxusercontent.com/u/396 ... 2028739923

Update: tested the 64-bit tar.gz release, and it holds stable at 8,5% mem usage and high FPS rate in comparison to 32-bit build.

User avatar
scrawl
Posts: 2152
Joined: 18 Feb 2012, 11:51

Re: 0.23.0 Bugs

Post by scrawl » 02 May 2013, 11:43

That gets us somewhere. There's 2 problems here:

- There's a bug in Ogre with GLSL programs. If one fails to compile, it tries recompiling it every frame and apparently causes some leakage while doing so, hence the low fps, huge Ogre.log and memory usage. This bug has been around for ages and I have no idea how to fix it.
- The shader preprocessor seems to behave non-deterministically in the 32 bit build. It seems to accept some #if statements regardless of whether they're true or not. Either there's something wrong with the build, or boost 1.53 (which the .tar.gz build uses) maybe broke something with our included copy of boost::wave (unlikely, since the 64 bit build works)

User avatar
psi29a
Posts: 4807
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: 0.23.0 Bugs

Post by psi29a » 02 May 2013, 11:55

How comes this wasn't a problem in 0.22.0 release?

User avatar
scrawl
Posts: 2152
Joined: 18 Feb 2012, 11:51

Re: 0.23.0 Bugs

Post by scrawl » 02 May 2013, 12:11

0.22 has the problem too for me. 0.21 works fine. It might be boost's fault, K1ll seems to have switched his package from boost 1.40 to 1.53 starting with the 0.22 release.

User avatar
psi29a
Posts: 4807
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: 0.23.0 Bugs

Post by psi29a » 02 May 2013, 12:37

Because Sir_Herrbatka won't post it, I will:
(01:34:15 PM) Sir_Herrbatka: i don't see the problem with memory is 0.22.0
(01:34:28 PM) Sir_Herrbatka: we may have more problems than it seems

User avatar
sirherrbatka
Posts: 2157
Joined: 07 Aug 2011, 17:21

Re: 0.23.0 Bugs

Post by sirherrbatka » 02 May 2013, 12:48

Because i have disable object shaders. This is a different issue.

User avatar
psi29a
Posts: 4807
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: 0.23.0 Bugs

Post by psi29a » 02 May 2013, 12:56

sirherrbatka wrote:Because i have disable object shaders. This is a different issue.
Still an issue... :)

User avatar
sirherrbatka
Posts: 2157
Joined: 07 Aug 2011, 17:21

Re: 0.23.0 Bugs

Post by sirherrbatka » 02 May 2013, 16:18

I can't reproduce this every time i run 0.23.0 :|

gsedej
Posts: 6
Joined: 12 May 2012, 23:33

Re: 0.23.0 Bugs

Post by gsedej » 04 May 2013, 10:45

I see black instead of a loading screen. (the game works OK)

Ubuntu 12.04, 64b, PPA, Intel 4500MHD

update: I tested on similar system with i5 cpu and intel HD3000 graphics and menu showed just for a moment before gameplay (most of time was still black)

User avatar
psi29a
Posts: 4807
Joined: 29 Sep 2011, 10:13
Location: Belgium
Gitlab profile: https://gitlab.com/psi29a/
Contact:

Re: 0.23.0 Bugs

Post by psi29a » 08 May 2013, 16:01

Any movement on the 32-bit Linux builds? Is this a problem with boost 1.53 or our custom 'wave' ?

I sent an email to k1ll to ask him about building new releases with an older boost version to test.

Locked