[Bug 20591] New: FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

Summary: FIFA 08: crashes when starting a match
           Product: Wine
           Version: 1.1.31
          Platform: PC
               URL: http://largedownloads.ea.com/pub/demos/Fifa/08/FIFA08_
                    PC_Demo.exe
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: directx-d3d
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
                CC: [e-mail]

FIFA 08 (also demo version, link in URL) crashes when starting a match (you see
some frames, and when the game is going to show players, it crashes) in Wine
1.1.31 and later (also in current git). This was introduced in this commit:

wined3d: Don't reserve clipping constants needlessly.
author    Stefan Dösinger 
     Tue, 6 Oct 2009 18:59:44 +0000 (20:59 +0200)
commit    8a68f418dbea0967003818e5bbac10e5ffc22512

I added the following code right before the new if..else clause in
dlls/wined3d/arb_program_shader.c in current git:

FIXME("clip_limit: gl_info->limits.clipplanes = %u\n",
gl_info->limits.clipplanes);
FIXME("clip_limit: ctx->target_version = %u (ARB = %d)\n", ctx->target_version,
ARB);
FIXME("clip_limit: use_nv_clip(gl_info) = %d, ctx->target_version >= NV2 =
%d\n", use_nv_clip(gl_info), ctx->target_version >= NV2);
unsigned int __mask = ctx->cur_vs_args->boolclip.clipplane_mask;
FIXME("clip_limit: mask = %u\n", __mask);
FIXME("clip_limit: count_bits(mask) = %u\n", count_bits(__mask));

And got those messages right before the crash:

fixme:d3d_shader:shader_generate_arb_declarations clip_limit:
gl_info->limits.clipplanes = 6
fixme:d3d_shader:shader_generate_arb_declarations clip_limit:
ctx->target_version = 0 (ARB = 0)
fixme:d3d_shader:shader_generate_arb_declarations clip_limit:
use_nv_clip(gl_info) = 0, ctx->target_version >= NV2 = 0
fixme:d3d_shader:shader_generate_arb_declarations clip_limit: mask = 0
fixme:d3d_shader:shader_generate_arb_declarations clip_limit: count_bits(mask)
= 0

And by the way I've also tried adding the following code in place of this
if..else clause in the current git:
clip_limit = gl_info->limits.clipplanes;
if(ctx->target_version == ARB) clip_limit = min(clip_limit, 4);
And FIFA doesn't crash then, but I guess it's not the proper way to fix the
problem.

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

Bartosz Brachaczek  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download, regression

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #1 from Bartosz Brachaczek   2009-11-06 07:22:37 ---
And by the way I have Radeon 9550 (RV350) running on open source radeon driver
if it changes anything.

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #2 from Stefan Dösinger   2009-11-06 15:53:17 ---
Can you post a backtrace of the crash? I guess we're not checking a NULL
pointer somewhere, or wrongly assume that certain extensions or GPU features
are available, and thus run into a NULL pointer dereference because some object
is not created.

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #3 from Bartosz Brachaczek   2009-11-06 18:51:14 ---
Hm, I realized that actually it doesn't crash. It just exits with error code
255 (or -1). I'm attaching a log (probably useless).

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #4 from Bartosz Brachaczek   2009-11-06 18:52:50 ---
Created an attachment (id=24572)
 --> (http://bugs.winehq.org/attachment.cgi?id=24572)
simple log from wine

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #5 from Bartosz Brachaczek   2009-11-06 19:27:44 ---
I think that it might be a bug in Mesa, not in Wine. I thought that this
"Params exhasted" message is interesting so I looked at source code of
r300VertexProgUpdateParams() function in Mesa DRI r300 driver and found this
code:

if (vp->code.constants.Count * 4 > VSF_MAX_FRAGMENT_LENGTH) {
        /* Should have checked this earlier... */
        fprintf(stderr, "%s:Params exhausted\n", __FUNCTION__);
        _mesa_exit(-1);
    }

If someone with different GPU is able to run FIFA 08 successfully (I cannot
test), then it's definitely a bug in Mesa.

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #6 from Bartosz Brachaczek   2009-11-06 20:09:30 ---
I compiled libdrm and mesa from git and now FIFA 08 doesn't exit. Mesa bug was
fixed in
http://cgit.freedesktop.org/mesa/mesa/commit/?id=3594b53c0173ac810106f667604...
. But now players aren't displayed (it look like there were no players at all;
all other textures are ok) and thus the game is still unplayable. Can someone
with another GPU check this?

[Bug 20591] FIFA 08: crashes when starting a match

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #7 from Stefan Dösinger   2009-11-07 03:24:27 ---
Its pretty likely that this patch triggers an out-of-constants situation,
especially if the driver is managing them inefficiently or using some private
constants. After all this patch was intended to free up 6 constants  that
aren't actually used, so the app has more of them available.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Bartosz Brachaczek  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|FIFA 08: crashes when       |FIFA 08: players are not
                   |starting a match            |displayed

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #8 from Jörg Höhle   2009-12-09 06:18:32 ---
Created an attachment (id=25138)
 --> (http://bugs.winehq.org/attachment.cgi?id=25138)
GL errors messages from the console triggered by the regression
> But now players aren't displayed

Maybe it's the same issue with Lego Star Wars II on Intel.  The regression is
the same:
commit 8a68f418dbea0967003818e5bbac10e5ffc22512
Author: Stefan Dösinger 
Date:   Tue Oct 6 20:59:44 2009 +0200
    wined3d: Don't reserve clipping constants needlessly.

Since then in LSWII, the minifigs display without body, legs, arms, mouth or
eyes on Intel i915 on Linux.  Previously the rendering was perfect.  Now only 3
pieces are rendered: the head (without eyes), the hair and the hips.  Looks
like ghosts moving.  The parts that are now missing are the animated ones (legs
move, eyes roll, body twists).

The attached log snippet shows:
fixme:d3d_shader:shader_arb_generate_vshader HW VertexShader Error at position
218: "Too many parameter variables"

No regression on MacOS with nVidia graphics.

Using Ubuntu Intrepid.  Beside that, I saw no other visual glitch in the app.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Michael Stefaniuc  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
                 CC|                            |[e-mail]
     Ever Confirmed|0                           |1

--- Comment #9 from Michael Stefaniuc  2010-11-15 17:31:20 CST ---
Confirming as per comment #8

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #10 from Stefan Dösinger  2011-07-26 16:54:47 CDT ---
After reading the posts in this bug again I think we should close this bug as
invalid. The complaint from mesa that program parameters are exhausted strongly
suggests that mesa can't provide the number of parameters it claims it
supports.

If this was a binary driver I'd add a quirk in our driver quirk table that
subtracts a few parameters from the number the driver claims it supports, like
we do on OSX or fglrx. But Since this is Mesa there is a decent chance that the
driver gets fixed.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #11 from Austin English  2011-07-26 17:28:32 CDT ---
(In reply to comment #10)
> After reading the posts in this bug again I think we should close this bug as
> invalid. The complaint from mesa that program parameters are exhausted strongly
> suggests that mesa can't provide the number of parameters it claims it
> supports.

A mesa bug should be filed and linked here, before closing this as invalid.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Henri Verbeet  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
    Regression SHA1|                            |8a68f418dbea0967003818e5bba
                   |                            |c10e5ffc22512

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Stefan Dösinger  changed:

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

--- Comment #12 from Stefan Dösinger  2012-01-21 14:54:28 CST ---
Does this bug still occur with wine git and a modern Mesa release? A few weeks
ago I removed a fglrx hack that lowered the amount of uniforms, although I'd
expect this change to make this specific problem worse, not better. Please test
with the glsl shader backend, although I'm interested in the ARB shader result
as well.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Jörg Höhle  changed:

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

--- Comment #13 from Jörg Höhle  2012-02-22 16:55:31 CST ---
No modern mesa here, sorry. Lego Star Wars 2 still shows ghosts only, with:
fixme:d3d_shader:shader_arb_generate_vshader HW VertexShader Error at position
234: "Too many parameter variables"
and thousands of
err:d3d_draw:drawStridedFast >>>>>>>>>>>>>>>>> GL_INVALID_OPERATION (0x502)
from glDrawElements @ drawprim.c / 52

Graphics have regressed: In wine-1.4rc4 I observe the inverted "picture in
picture" known from Mac OS bug #22029, comment #4 now on Intel too, e.g. when
hitting Esc or during some animations or transitions.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

lizhenbo  changed:

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

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

--- Comment #14 from lizhenbo  2013-01-31 20:09:19 CST ---
Works fine in wine-1.5.22-254-gc14bdaf
Fedora 16 32-bit, AMD A6-3670 (Radeon HD 6530D)
Catalyst 12.10
Maybe you should try to update Wine.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Henri Verbeet  changed:

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

--- Comment #15 from Henri Verbeet  2013-06-12 05:41:12 CDT ---
I'm going to close this as UPSTREAM, since the original crash has been fixed
for a long time in Mesa. The constant issue is essentially a hardware limit,
although with some GL extensions it could perhaps be mitigated enough. That
part is pretty much a duplicate of bug 7991.

[Bug 20591] FIFA 08: players are not displayed

http://bugs.winehq.org/show_bug.cgi?id=20591

Henri Verbeet  changed:

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

--- Comment #16 from Henri Verbeet  2013-06-12 05:41:31 CDT ---
Closing.