Runesoft games based on ogre engine broken with fglrx (?)

[ivanovic]

Hiho!
Maybe someone of you can help me, though I fear this is a general error with fglrx these days. My problem: it is not possible for me to start games from Runesoft that are based on the Ogre engine. I now own Ankh, Ankh2 as well as Jack Keane and neither of those games wants to start, all exist with a lovely segfault. About one year ago I was still able to at least play Ankh2 on my amd64 linux box, so I fear some regression got into, but maybe I am just doing something wrong.

Here is what I do:
Start any of the runesoft games with a system with fglrx (latest, some older ones make no difference either). The result is a segfault. When starting the game (in this example Jack Keane) in gdb, the result is as follows:

$ gdb ./jk
GNU gdb 6.8
Copyright (C) 2008 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law. Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-pc-linux-gnu"...
(gdb) run
Starting program: /home/nils/JackKeane/bin/release/jk
[Thread debugging using libthread_db enabled]
Starting JK 1.0.3...(release)Starting intro...
[New Thread 0xf74256c0 (LWP 20835)]
[New Thread 0xf6cb2b70 (LWP 20844)]
[Thread 0xf6cb2b70 (LWP 20844) exited]
Starting application...class created.
Desktop settings: 1920x1200
Log manager created.
Setting log level to 1
System: Linux 2.6.30 #3 SMP Sat Jun 13 16:33:28 CEST 2009
Available app memory (KB, maybe virtual): 4046740
Using RenderSystem: OpenGL Rendering Subsystem
Ogre initialized.
Inizializzo CEGUI...Ok.
CEGUI intialized.
LUA scripting initialized.
InputSystem initialized.
AudioManager initialized.
Sound effects/speech/music intialized.
GUIWindowManager construction...Ok.
Initializing Render Task...Ok
Initializing mouse interface...Ok
Initializing audio subsystem...OpenAL intialized, device:
[New Thread 0xf6cb2b70 (LWP 20909)]
Allocated 64 sound sources.
Distance factor set to: 0.1
Rolloff factor set to: 0.1
Distance factor set to: 0.01
Rolloff factor set to: 1
Ok.
Initializing compositor...Ok.
Initializing listener...Ok.
GetTotalMemory (MB) returned: 3951
Displaying intro scene...
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0xf74256c0 (LWP 20835)]
0xf54355af in ?? () from /usr/lib32/dri/fglrx_dri.so
(gdb) kill
Kill the program being debugged? (y or n) y
(gdb) quit

It makes no difference if I have composite active or not. You can test if it works for you using the Ankh2 demo (which crashes as well!). It is available via torrent and the torrent is eg linked at linux-gamers.net.

My system:
* Gentoo amd64 (unstable) with multilib installed (required for the runesoft games since they are 32bit only)
* Core2Quad Q9300
* HD3850 512MB with latest fglrx (9.5 crashes, too IIRC, downgrading further is not really possible without downgrading xorg-server...)
* xorg-server 1.6.2-rc1 (running kde 4.2.4)
* kernel 2.6.30 (yes, some patches for fglrx required, but with an older kernel it crashes, too)
* glibc 2.10.1
If composite is active or not seems not to affect the situation.

Can anybody of you confirm this problem? Does someone maybe have a solution (beside some "get an nvidia card!") for the problem? Does it also happen with fglrx on 32bit, or is it fine there? Or do you even use fglrx on 64bit and don't observe this problem?

I tried starting ankh2 with the open source drivers and, oh wonder, it starts, but since there is no 3D support for r6xx so far everything is done via the mesa software rasterizer and thus is unbearable slow (~1fps). So the game itself can at least start in a (very) recent 64bit environment with multilib...

[Kano]

You can downgrade down to 9-4.

[ivanovic]

You can downgrade down to 9-4.

And with this one I get the same error.

EDIT:
Just tried using radeon (yeah, 3D via pure software...) and then it is possible to start the game without segfault (unplayable with ~1fps, but it does start and show the intros and title screen). So I really assume that it is basically fglrx playing tricks on me.

[rogerdv]

Im developing a game with ogre too, but my problem is that with 9.x drivers Ogre doesnt works properly. I dont get a segfault, but just a black screen with patches of pixels. I got the same with 3 different cards, when I use the opengl renderer (yep, it is broken too under windows). Im sorry, but Im thinking to recommend Nvidia for playing. So, dont feel bad, you are not alone.
If your card is not too recent, try with some 8.x driver.

[bridgman]

I did some searching on the Ogre site looking for more info but the only issue I saw there was related to hardware mipmap generation, with a conclusion that Ogre seemed to be one of the few apps actually using it and maybe it wasn't worth the bother. I didn't see any reports related to Ogre on the ati.cchtml.com tracker either.

I doubt that the reported mipmap problem would explain what rogerdv is seeing, and I'm sure it wouldn't explain the segfault though. Can someone confirm this happens on a directly supported OS (released versions of OpenSuSE or Ubuntu, I guess) and then get cross-linked tickets into both Ogre and fglrx trackers ?

[rogerdv]

Yep, I read about the mipmapping problem. So, do you think it is not the cause of my problem? Too bad, I had hope it would be solved in Ogre 1.6.3 with some dirty hack that disables hardware mipmapping.

[bridgman]

I don't really know; all I have to go on is that the forum posts seemed to describe the problem as moderate corruption while your problem sounded like essentially nothing being drawn. The only way to be sure I guess is to patch out the use of mipmap generation.