[Bug 26763] New: Wolfenstein: crashes on start

http://bugs.winehq.org/show_bug.cgi?id=26763

Summary: Wolfenstein: crashes on start
           Product: Wine
           Version: 1.3.17
          Platform: x86
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=34086)
 --> (http://bugs.winehq.org/attachment.cgi?id=34086)
Log of crash

Hi,

I'm trying to get the game Wolfenstein running with Wine, a title which
according to the appdb entry should be well supported.

Unfortunately I'm getting nowhere, the game crashes directly after the splash
screen appears, and I don't even get a backtrace...

wine: Unhandled page fault on read access to 0x00000000 at address 0x10185da4
(thread 0009), starting debugger...
 fixme:dbghelp_dwarf:dwarf2_lookup_type Unable to load forward reference for
tag 26
 fixme:dbghelp_dwarf:dwarf2_lookup_type Unable to load forward reference for
tag 26
 fixme:dbghelp_dwarf:dwarf2_lookup_type Unable to load forward reference for
tag 26
 Unhandled exception: page fault on read access to 0x00000000 in 32-bit code
(0x10185da4).

Full log is attached.

If I use winedbg I also get these messages:

wine client error:26: write: Bad file descriptor
 Inconsistency detected by ld.so: dl-open.c: 221: dl_open_worker: Assertion
`_dl_debug_initialize (0, args->nsid)->r_state == RT_CONSISTENT' failed!

If I try the multiplayer binary instead, I get an error dialog "Error creating
game rendering context", and a console with these warnings:

WARNING: idRenderContextWGL::Create : wglChoosePixelFormatARB failed to find a
suitable format.
 WARNING: idRenderContextWGL::Create : wglChoosePixelFormatARB failed to find a
suitable format.

After some googling, I found bug 9186 which mentions the same error in game
using the same engine, so I made another log with WINEDEBUG=+wgl.

For graphics I use Mesa with the r300g driver (tried both 7.10 and using git)
which I guess could be the cause of the problem. If so it would be great to
know what the problem is, so I can file a bug with the Mesa devs.

I'm using Wine 1.3.17 (tried both the Ubuntu packages and compiling myself) in
a clean wineprefix (the only addition is native d3dx9_39.dll for
GetShaderConstantTableEx support) on Debian unstable 32-bit.

The game is updated to version 1.11 and Securom worked around with a patched
binary.

[Bug 26763] Wolfenstein: crashes on start

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #1 from Sven Arvidsson  2011-04-13 12:10:13 CDT ---
Created an attachment (id=34087)
 --> (http://bugs.winehq.org/attachment.cgi?id=34087)
wgl trace

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Sven Arvidsson  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|-unknown                    |opengl
            Summary|Wolfenstein: crashes on     |Wolfenstein / ETQW: "Error
                   |start                       |creating game rendering
                   |                            |context" with Mesa

--- Comment #2 from Sven Arvidsson  2011-04-14 12:12:03 CDT ---
I just gave the ETQW demo a try, and I'm getting the same error, "Error
creating
game rendering context"

It would be interesting to know if this is a regression, or if it still works
with drivers from Nvidia or ATI?

I'm reassigning the bug to the opengl component, I hope that's correct.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Roderick Colenbrander  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #3 from Roderick Colenbrander  2011-04-16 21:37:09 CDT ---
The single and multiplayer binaries should be considered two different games.
The single player version is a Direct3D game and for some reason there is a
crash inside game code (perhaps driver related, but hard to say).

The multiplayer version is in fact an OpenGL game and for that one a +wgl log
could help. The unable to find a pixel format error could be a bug in the wine
WGL layer, but it may also reveal a bug in the Mesa GLX layer.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Sven Arvidsson  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #34087|0                           |1
        is obsolete|                            |

--- Comment #4 from Sven Arvidsson  2011-04-18 14:08:51 CDT ---
Created an attachment (id=34172)
 --> (http://bugs.winehq.org/attachment.cgi?id=34172)
wgl log from wolf2mp

(In reply to comment #3)
> The single and multiplayer binaries should be considered two different games.
> The single player version is a Direct3D game and for some reason there is a
> crash inside game code (perhaps driver related, but hard to say).
>
> The multiplayer version is in fact an OpenGL game and for that one a +wgl log
> could help. The unable to find a pixel format error could be a bug in the wine
> WGL layer, but it may also reveal a bug in the Mesa GLX layer.

Oh, I see. In that case, let's focus on the OpenGL issues. I'm attaching +wgl
logs from both Wolf2MP.exe and the ETQW demo.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #5 from Sven Arvidsson  2011-04-18 14:09:14 CDT ---
Created an attachment (id=34173)
 --> (http://bugs.winehq.org/attachment.cgi?id=34173)
wgl log from etqw demo

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #6 from Roderick Colenbrander  2011-04-18 15:52:06 CDT ---
I had quick glance but don't see anything suspicious on the opengl side in
those logs. The last calls in the log are to dxdiag / devenum, so perhaps it is
something related to that.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #7 from Henri Verbeet  2011-04-18 16:54:00 CDT ---
For the ETQW issue, does the native Linux version work? IIRC that one is
supposed to work with r300g, provided you have the necessary s3tc bits setup.
If the native linux version is broken as well it may make more sense to try to
get that fixed first, but if it isn't I suppose it makes sense to look at this
from the Wine side.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #8 from Sven Arvidsson  2011-04-18 17:10:25 CDT ---
The native version is working fine.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #9 from Sven Arvidsson  2011-04-22 09:14:23 CDT ---
Not sure how relevant this is but:

Forcing software rendering with LIBGL_ALWAYS_SOFTWARE=1 makes no difference.

According to the game log, only 64 MB of video memory is detected, manually
setting the correct value in VideoMemorySize makes no difference.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #10 from Sven Arvidsson  2011-04-23 11:37:42 CDT ---
Just FYI Wolfenstein singleplayer started working. I did a quick bisect and it
seems this is the commit fixing things:
http://cgit.freedesktop.org/mesa/mesa/commit/?id=8e28d842d192e69ba8cae4f9754...

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #11 from Sven Arvidsson  2011-04-24 16:03:53 CDT ---
Turns out this is a regression, both ETQW and Wolfenstein MP is working with
1.1.32. I'm going to try bisecting to find the regression when I have more
time.

Guess I got confused when I first looked up Wolfenstein in AppDB and saw it was
rated platinum. But on closer examination most submitters haven't tried running
the multiplayer version.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #12 from Sven Arvidsson  2011-04-26 14:37:39 CDT ---
The winner is:

e7590fcfb51bb149a6cf8272f5576861f2193183 is the first bad commit
commit e7590fcfb51bb149a6cf8272f5576861f2193183
Author: Roderick Colenbrander 
Date:   Fri Nov 13 12:28:48 2009 +0100

wgl: Make sure we set a valid value for GLX_DRAWABLE_TYPE. Right now we
default to 0 which is illegal.

:040000 040000 18c42aa22adede3c11c377d311dd9e1ed90a38ad
6988f46168fb6b19f72688f7e0984f9d29e3eb06 M    dlls

I confirmed this by commenting out this change from git master, and without it
both ETQW and Wolfenstein multiplayer runs.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #13 from Roderick Colenbrander  2011-04-26 18:30:40 CDT ---
This part of the log clearly shows the 'failure':
trace:wgl:ConvertAttribWGLtoGLX pAttr[0] = 2010
trace:wgl:ConvertAttribWGLtoGLX pAttr[1] = WGL_SUPPORT_OPENGL_ARB: 1
trace:wgl:ConvertAttribWGLtoGLX pAttr[2] = 2015
trace:wgl:ConvertAttribWGLtoGLX pAttr[3] = GLX_RED_SIZE: 8
trace:wgl:ConvertAttribWGLtoGLX pAttr[4] = 2017
trace:wgl:ConvertAttribWGLtoGLX pAttr[5] = GLX_GREEN_SIZE: 8
trace:wgl:ConvertAttribWGLtoGLX pAttr[6] = 2019
trace:wgl:ConvertAttribWGLtoGLX pAttr[7] = GLX_BLUE_SIZE: 8
trace:wgl:ConvertAttribWGLtoGLX pAttr[8] = 201b
trace:wgl:ConvertAttribWGLtoGLX pAttr[9] = GLX_ALPHA_SIZE: 8
trace:wgl:ConvertAttribWGLtoGLX pAttr[10] = 2022
trace:wgl:ConvertAttribWGLtoGLX pAttr[11] = GLX_DEPTH_SIZE: 24
trace:wgl:ConvertAttribWGLtoGLX pAttr[12] = 2023
trace:wgl:ConvertAttribWGLtoGLX pAttr[13] = GLX_STENCIL_SIZE: 8
trace:wgl:ConvertAttribWGLtoGLX pAttr[14] = 2011
trace:wgl:ConvertAttribWGLtoGLX pAttr[15] = GLX_DOUBLEBUFFER: 1
trace:wgl:ConvertAttribWGLtoGLX pAttr[?] = GLX_DRAWABLE_TYPE: 0xffffffff
trace:wgl:ConvertAttribWGLtoGLX pAttr[?] = GLX_RENDER_TYPE: 0xffffffff
warn:wgl:X11DRV_wglChoosePixelFormatARB Compatible Pixel Format not found

I have an idea on what's wrong though and after another read of the GLX specs,
I think it is a Mesa bug. Just in case can you attach the output of 'glxinfo
-c'?

ConvertAttribWGLtoGLX is used as part of wglChoosePixelFormatARB which the game
uses to select a 'pixel format' to use for its OpenGL window. In Wine we use
glXChooseFBConfig to implement this call. In case of WGL if you don't specify a
'drawable type' it returns formats for all types of drawables it can support
(window, bitmap, pbuffer), but GLX only returns info for windows. If I remember
correctly this patch was for a part needed to make fglrx happy and to make
things more correctly.

The problem is that Mesa doesn't like me passing in '0xffffffff' as the
GLX_DRAWABLE_TYPE (actually it dislikes RENDER_TYPE of 0xffffffff as well). In
the end Mesa does about the following in its fbconfig checking:

for(i=0; i<config_list_size; i++)
    fbconfigs_compatible(requested_config, config_list[i]);

static int fbconfigs_compatible(config *a, config *b)
{
...
   MATCH_EXACT(level);

MATCH_MASK(drawableType);
   MATCH_MASK(renderType);
...
}

where MATCH_MASK:
#define MATCH_MASK(param)
do {
if ((a->param & ~b->param) != 0)                                                
      return False;
} while (0);

The problem is that the configurations in config_list have at max a few bits
set (e.g. GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT) while I'm passing
in ~0 (which is equal to GLX_DONT_CARE). Because a different number of bits is
set in a->param and b->param, MATCH_MASK fails and no configuration gets
selected.

There are two things in the GLX specs. First of all it mentions that when you
pass a bitmask to glXChooseFBConfig ALL bits set in the mask have to match. The
second thing it mentions is that if GLX_DONT_CARE is specified as a value for
one of the attributes (which is the case here), the option in question should
be ignored.

Some of the other MATCH_ macros in Mesa check for GLX_DONT_CARE, but not the
MATCH_MASK one, so I would guess the bug is there but I don't have a Mesa
system to test on. I could write a quick test app. (I could also just not set
GLX_RENDER_TYPE / GLX_DRAWABLE_TYPE when either of the two is 0, but not sure
what sort of issues that may cause ..)

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #14 from Sven Arvidsson  2011-04-27 06:22:16 CDT ---
Created an attachment (id=34372)
 --> (http://bugs.winehq.org/attachment.cgi?id=34372)
glxinfo -v output

Hmm, glxinfo doesn't have a "-c" option, at least not on my system, did you
mean "-v"?

As for the rest, I'm afraid I'm in over my head. Could you or Henri bring it up
with the Mesa devs directly?

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Wylda  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression
                 CC|                            |[e-mail]

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #15 from Sven Arvidsson  2011-06-28 10:41:24 CDT ---
By the way, instead of a test app, it might be possible to use apitrace to
trace the start of either of the games with a working driver and get a
reproducible testcase for the Mesa devs that way:
https://github.com/apitrace/apitrace

I might try this myself if I ever try out fglrx.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Henri Verbeet  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
    Regression SHA1|                            |e7590fcfb51bb149a6cf8272f55
                   |                            |76861f2193183

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

MD.IMAM HOSSAIN  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #16 from MD.IMAM HOSSAIN  2011-08-14 00:44:52 CDT ---
Same here for Intel GMA 4500 Mesa 7.12-devel with WINE 1.3.26

if I remove commit

e7590fcfb51bb149a6cf8272f5576861f2193183

wgl: Make sure we set a valid value for GLX_DRAWABLE_TYPE. Right now we
default to 0 which is illegal.

then the game works.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

--- Comment #17 from Sven Arvidsson  2012-03-18 12:57:23 CDT ---
(In reply to comment #15)
> By the way, instead of a test app, it might be possible to use apitrace to
> trace the start of either of the games with a working driver and get a
> reproducible testcase for the Mesa devs that way:
> https://github.com/apitrace/apitrace
>
> I might try this myself if I ever try out fglrx.

Doing this with apitrace didn't work. The few frames I managed to trace did
play back with Mesa so it was useless as a way of reproducing the bug.

I decided to file a bug with Mesa anyway:
https://bugs.freedesktop.org/show_bug.cgi?id=47478

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Jerome Leclanche  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]
           See Also|                            |https://bugs.freedesktop.or
                   |                            |g/show_bug.cgi?id=47478

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |UPSTREAM

--- Comment #18 from Alexandre Julliard  2013-07-02 04:08:30 CDT ---
Upstream bug then.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Jerome Leclanche  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #19 from Jerome Leclanche  2013-07-02 05:33:28 CDT ---
Fixed upstream by 9cda3560048e8595d3ffa315b76487f4479bff2c according to the
bug.

[Bug 26763] Wolfenstein / ETQW: "Error creating game rendering context" with Mesa

http://bugs.winehq.org/show_bug.cgi?id=26763

Bruno Jesus  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|regression                  |
    Regression SHA1|e7590fcfb51bb149a6cf8272f55 |
                   |76861f2193183               |