[Bug 16412] New: 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Summary: 3dmark2000 Environment Bump Mapping test broken in
                    1.1.10
           Product: Wine
           Version: 1.1.10
          Platform: PC
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: -unknown
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

In wine-1.1.10 (slackware package, wine-1.1.10-i486-1kjz.tgz) i saw new bug in
3dmark2000's last test, Environment Bump Mapping. Test itself rendered very
dark, and console spammed a bit with messages:

fixme:d3d_shader:register_for_arg Unknown source argument 255
{many lines}
fixme:d3d_shader:gen_ati_shader Swizzles not implemented
fixme:d3d_shader:register_for_arg Unknown source argument 255
{repeated again}

My OpenGL info:

OpenGL vendor string: Tungsten Graphics, Inc.
OpenGL renderer string: Mesa DRI R200 20060602 AGP 4x x86/MMX+/3DNow!+ TCL
OpenGL version string: 1.3 Mesa 7.3-devel

wine 1.1.8 works OK.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #1 from Andrew   2008-12-08 01:46:46 ---
Created an attachment (id=17742)
 --> (http://bugs.winehq.org/attachment.cgi?id=17742)
copy/paste from terminal

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #2 from Andrew   2008-12-08 01:48:45 ---
Created an attachment (id=17743)
 --> (http://bugs.winehq.org/attachment.cgi?id=17743)
glxinfo output

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |regression

--- Comment #3 from Austin English   2008-12-08 03:16:10 ---
Please run a regression test:
http://wiki.winehq.org/RegressionTesting

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #4 from Rico   2008-12-08 15:34:09 ---
WORKSFORME.

wine-1.1.10-74-gd3f7f5e
nvidia proprietary driver 180.08 (8800GTS)

I tried with and without GLSL.

I couldn't see any visual difference between these two versions (1.1.8 and
1.1.10) and I couldn't get the FIXMEs to show up in the console. Probably
depends on the renderpath which is used. Could you add two screen shots, which
show the problem if there is any? Or are there only the FIXEMEs which appear?

Note: FIXMEs aren't bugs.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #5 from Andrew   2008-12-08 17:06:47 ---
I think this bug is only visible on r200 class hardware.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #6 from Andrew   2008-12-08 17:08:14 ---
Created an attachment (id=17761)
 --> (http://bugs.winehq.org/attachment.cgi?id=17761)
incorrect rendering

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #7 from Andrew   2008-12-08 17:18:06 ---
Created an attachment (id=17762)
 --> (http://bugs.winehq.org/attachment.cgi?id=17762)
Fo me, this version looks correct

wine-1.1.8

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Andrew  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #17762|Fo me, this version looks   |For me, this version looks
        description|correct                     |correct

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Pauli Nieminen  changed:

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

--- Comment #8 from Pauli Nieminen   2008-12-08 22:36:04 ---
For me all programs using shaders with wine-1.1.10 (That includes civ-IV Beyond
The Sword and 3DMark2001SE any test using vertex shaders)

Crash is always including messange about unknown target for fragment shader
operation and then sometimes it includes cut messange about assertion failure
in r200_maos_arrays.c:403: r200EmitArrays: Assertion `0' failed. This error
messange about assertion failure loses X number of chars from begin so it isn't
vissible unless path is very long like with civ IV under steam.

That assertion is about wrong size parameter to driver.. which might be related
to wine error that selects GL_NONE as target register for fragment shader
operation.

I can't run regresion test because my self compiled wine is unstable and always
crashes when steam tryes to create process for game (with memory access
violation to random address number)

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #9 from H. Verbeet   2008-12-09 09:23:27 ---
Created an attachment (id=17771)
 --> (http://bugs.winehq.org/attachment.cgi?id=17771)
patch

Does the attached patch help?

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #10 from Pauli Nieminen   2008-12-09 10:51:29 ---
Created an attachment (id=17772)
 --> (http://bugs.winehq.org/attachment.cgi?id=17772)
d3d8 visual test error log

89139b75898fb5ac7094a641b0d91a06936a6c51 is first bad commit
commit 89139b75898fb5ac7094a641b0d91a06936a6c51
Author: Henri Verbeet 
Date:   Wed Dec 3 14:53:43 2008 +0100

wined3d: Make shader texture format fixups more generic.

Based on a patch by Stefan Dösinger. This is more flexible, and allows
    the shader backend implementation to be simpler, since it doesn't have
    to know about specific formats. The next patch makes use of this.

:040000 040000 206d90d013257ba1fda8517302d0b91530aba96d
9b0bf9d78dc4786c9fc888e784e2b33dd05986d0 M      dlls

Ok. I found that this bug is present when running d3d8 tests so I was able to
run bisect and found this commit broke ati fragment shader implementation.

I also noticed that my r200 assertion failure isn't related to this. I will
soon open another bug report after I have completed regression test for crash.

Attached patch doesn't at least fix the fixme about Unknow source argument.

Attachment includes my d3d8 test log with commit d3f7f5ebb6ff6 and attached
patch.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Austin English  changed:

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

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Stefan Dösinger  changed:

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

--- Comment #11 from Stefan Dösinger   2008-12-09 12:27:33 ---
The patch works for me

The "Unknown source argument 255" fixme's are most likely unrelated. I'll take
a look at them

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #12 from Stefan Dösinger   2008-12-09 13:08:50 ---
The "Unknown source argument 255" fixme's are related to the patch as well,
although they are not the fault of the patch. The patch uncovered a bug in the
rest of the code, where 0xffffffff is used instead of ARG_UNUSED. Since the
size of the color/alpha op members in the ffp desc structure was changed this
difference surfaced.

The problem is harmless, the FIXME is all the harm it does. I'll send a patch
that fixes it.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #13 from Andrew   2008-12-09 18:35:37 ---
Patch from comment #9 + wine-1.1.10-130-gfba5c4f works for me, too ...
Sorry for delay, even with ccache my computer can compile wine in ~2 hours

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

James Huk  changed:

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

--- Comment #14 from James Huk   2009-05-07 07:25:15 ---
(In reply to comment #13)
> Patch from comment #9 + wine-1.1.10-130-gfba5c4f works for me, too ...
> Sorry for delay, even with ccache my computer can compile wine in ~2 hours

Seems to work just fine with wine-1.1.20 and Nvidia GeForce9800GT (driver
180.51), no glitches at all. Andrew, could You re-test it with your hardware?

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

--- Comment #15 from Stefan Dösinger   2009-05-07 07:42:12 ---
Testing this bug on a geforce card is moot, this is an radeon 8500-9200
specific codepath. The patch is committed and it works for me, so I think we
can close this bug.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Austin English  changed:

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

--- Comment #16 from Austin English   2009-05-07 11:36:42 ---
Marking fixed, per Stefan.

[Bug 16412] 3dmark2000 Environment Bump Mapping test broken in 1.1.10

http://bugs.winehq.org/show_bug.cgi?id=16412

Alexandre Julliard  changed:

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

--- Comment #17 from Alexandre Julliard   2009-05-08 12:49:33 ---
Closing bugs fixed in 1.1.21.