[Bug 10221] New: "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

Summary: "make test" fails on d3d9/tests/visual.c
           Product: Wine
           Version: CVS/GIT
          Platform: Other
        OS/Version: other
            Status: NEW
          Keywords: download, source
          Severity: normal
          Priority: P2
         Component: wine-directx-d3d
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

On my Compaq Presario R3000 running Ubuntu Feisty, "make test"
fails because of the following errors in d3d9/tests/visual.c:

visual.c:2271: Test failed: quad 1 has color 00000000, expected 0x00ffffff
visual.c:2273: Test failed: quad 2 has color 00000000, expected 0x00ffff00
visual.c:2275: Test failed: quad 3 has color 00000000, expected 0x000000ff
visual.c:2277: Test failed: quad 4 has color 00000000, expected 0x00ffffff
visual.c:2345: Test failed: quad 1 has color 00000000, expected 0x0000ff00
visual.c:2349: Test failed: quad 3 has color 00000000, expected 0x00ff0000

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #1 from Vitaliy Margolen   2007-10-28 21:47:54 ---
You probably missing real 3d hardware and drivers...

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #2 from Dan Kegel   2007-10-29 07:20:20 ---
lspci | grep -i nvidia says
01:00.0 VGA compatible controller: nVidia Corporation NV17 [GeForce4 420 Go
32M] (rev a3)
/etc/X11/xorg.conf says
Section "Device"
        Identifier      "nVidia Corporation NV17 [GeForce4 420 Go 32M]"
        Driver          "nvidia"
http://www.xbitlabs.com/news/other/display/news1435.html
says this chip was a "lite" version of GeForce3
from around 2002.
glxgears runs and gets a proud 20 fps (on a busy system).

This is my fast home laptop, about two or three years old.
It ought to be possible to run the wine test suite on
this machine, so if that graphics isn't good enough,
the test suite should sense that and skip the tests.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #3 from Austin English   2008-02-11 13:30:12 ---
(In reply to comment #2)
> lspci | grep -i nvidia says
> 01:00.0 VGA compatible controller: nVidia Corporation NV17 [GeForce4 420 Go
> 32M] (rev a3)
> /etc/X11/xorg.conf says
> Section "Device"
>         Identifier      "nVidia Corporation NV17 [GeForce4 420 Go 32M]"
>         Driver          "nvidia"
> http://www.xbitlabs.com/news/other/display/news1435.html
> says this chip was a "lite" version of GeForce3
> from around 2002.
> glxgears runs and gets a proud 20 fps (on a busy system).
>
> This is my fast home laptop, about two or three years old.
> It ought to be possible to run the wine test suite on
> this machine, so if that graphics isn't good enough,
> the test suite should sense that and skip the tests.

I get the same in a feisty vm. glxgears shows average of 1788 FPS. Real video
card on windows host is an nvidia geforce 7600 gt.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #4 from franck   2008-02-19 05:09:37 ---
Created an attachment (id=10839)
 --> (http://bugs.winehq.org/attachment.cgi?id=10839)
"make test" d3d9-visual log + config detail

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

franck  changed:

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

--- Comment #5 from franck   2008-02-19 05:11:12 ---
same here on mandriva2008 + intel  945GM + wine-0.9.55-163-g4c8e218

visual.c:2554: Test failed: quad 1 has color 00ffff00, expected 0x00ffffff
visual.c:2559: Test failed: quad 3 has color 00000000, expected 0x000000ff
visual.c:2561: Test failed: quad 4 has color 00ffff00, expected 0x00ffffff

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #6 from Austin English   2008-04-27 21:25:10 ---
Still see this in my feisty vm and my hardy laptop (ATI Radeon Mobility 7000).

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #7 from Dan Kegel   2008-05-13 10:48:30 ---
With today's git, and with a fresh metacity per 
http://bugs.winehq.org/show_bug.cgi?id=9916#c69, I think
the only "make test" failures on my Compaq r3000 laptop running Gutsy
will be these four from d3d9/visual:

fixme:d3d:tex_alphaop Attempt to enable unsupported stage!
visual.c:8694: Test failed: TSSARGTEMP test returned color 0x00ff0000, expected
0x00FFFF00
fixme:d3d:tex_colorop >>>>>>>>>>>>>>>>> GL_INVALID_OPERATION (0x502) from
glActiveTextureARB @ state.c / 1833
visual.c:9074: Test failed: stretchrect: Pixel 160,360 has color 0x00000000,
expected 0x00ff0000
visual.c:9076: Test failed: stretchrect: Pixel 480,360 has color 0x00000000,
expected 0x0000ff00
visual.c:9078: Test failed: stretchrect: Pixel 480,120 has color 0x00000000,
expected 0x000000ff

(but see also bug 12246 for a d3d9/stateblock problem under warn+heap).

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

Maarten Lankhorst  changed:

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

--- Comment #8 from Maarten Lankhorst   2008-05-13 15:35:00 ---
I still need the following snippet for the visual tests to pass:

--- a/dlls/d3d9/tests/visual.c
+++ b/dlls/d3d9/tests/visual.c
@@ -104,6 +104,7 @@ static DWORD getPixelColor(IDirect3DDevice9 *device, UIN
         trace("Can't create an offscreen plain surface to read the render t
         return 0xdeadbeef;
     }
+    Sleep(100);

hr = IDirect3DDevice9_GetFrontBufferData(device, 0, surf);
     if(FAILED(hr))

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

Stefan Dösinger  changed:

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

--- Comment #9 from Stefan Dösinger   2008-05-15 14:35:46 ---
Wrt the TSSARGTEMP failure - the Geforce 3 was the first card that supported
this I think, maybe we're doing something wrong and hit a hardware limitation.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #10 from Dan Kegel   2008-05-21 10:13:33 ---
Using an approved WM, on a GeForce 4 (440 Go 64M),
these are indeed the only test failures left on my machine now:

visual.c:8695: Test failed: TSSARGTEMP test returned color 0x00ff0000, expected
0x00FFFF00
fixme:d3d:tex_colorop >>>>>>>>>>>>>>>>> GL_INVALID_OPERATION (0x502) from
glActiveTextureARB @ state.c /
 1833
visual.c:9075: Test failed: stretchrect: Pixel 160,360 has color 0x00000000,
expected 0x00ff0000
visual.c:9077: Test failed: stretchrect: Pixel 480,360 has color 0x00000000,
expected 0x0000ff00
visual.c:9079: Test failed: stretchrect: Pixel 480,120 has color 0x00000000,
expected 0x000000ff

Maarten's Sleep didn't change anything for me.

I'm using the real nvidia driver, not nv.  This is Gutsy, and
it's using version 1.0-9639 (Apr 16 2007) of the nvidia driver.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #11 from Dan Kegel   2008-06-15 20:30:15 ---
On my new box (http://kegel.com/new-computer-2008.html ,
Hardy, nvidia driver, with an approved metacity),
the visual test passes fairly often, but if you run
it enough times, it fails.  e.g.

#!/bin/sh
set -x
set -e
cd dlls/d3d9/tests
for a in 1 2 3 4 5 6 7 8 9
do
   rm -f visual.ok
   make visual.ok 
done
echo ok

pretty much always fails with errors like

visual.c:8973: Test failed: has color 0x00cc0000, expected 0x00ff0000 (case 0)
visual.c:8975: Test failed: has color 0x00ea0000, expected 0x00ff0000 (case 0)
visual.c:8973: Test failed: has color 0x00cc0000, expected 0x00ff0000 (case 1)
visual.c:8975: Test failed: has color 0x00ea0000, expected 0x00ff0000 (case 1)
...

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #12 from H. Verbeet   2008-06-16 01:57:29 ---
That's stream_test() failing... what specific version of the nvidia drivers are
you using? A while ago I had some similar looking failures in stream_test() as
well, although consistently. The geometry appeared to be getting corrupted, and
disabling the VBO extension could be used to work around this, as could writing
less data to the vertex buffer in one of the earlier tests. As far as I could
see we were not doing anything wrong there though, and newer drivers fixed the
issue for me.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #13 from Dan Kegel   2008-06-16 08:43:02 ---
Created an attachment (id=14098)
 --> (http://bugs.winehq.org/attachment.cgi?id=14098)
Output of nvidia-bug-report.sh on my new machine

NVRM version: NVIDIA UNIX x86 Kernel Module  169.12  Thu Feb 14 17:53:07 PST
2008

Thanks for the tip.  I see an update for Hardy is proposed,
http://albertomilone.com/wordpress/?p=210

I'll try that and see if it helps!

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #14 from Austin English   2008-12-17 13:03:59 ---
I get:

austin@work:~/wine-git/dlls/d3d9/tests$ rm -rf visual.ok && make visual.ok
../../../tools/runtest -q -P wine -M d3d9.dll -T ../../.. -p d3d9_test.exe.so
visual.c && touch visual.ok
fixme:win:EnumDisplayDevicesW ((null),0,0x32f128,0x00000000), stub!
err:x11settings:X11DRV_ChangeDisplaySettingsEx No matching mode found
640x480x32 @0! (XRandR)
err:d3d:WineD3D_ChoosePixelFormat Can't find a suitable iPixelFormat
visual.c:9906: Tests skipped: Sanity check returned an incorrect
color(0000dfef), can't assure the correctness of the tests, skipping

in git. Not ideal, but not a failure. How is it for you Dan?

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #15 from Dan Kegel   2008-12-17 22:17:33 ---
It's fine for me on newer nvidia with Ubuntu Intrepid, I think.

http://test.winehq.org/data/tests/d3d9:visual.html shows
occasional failures, e.g.
http://test.winehq.org/data/5646fec9b14b2d039994cdeeef141eca889e649f/#group_...
on mstefani's Intel 945GME graphics card,
opengl.c:581: OpenGL renderer: Mesa DRI Intel(R) 945GME GEM 20080716
x86/MMX/SSE2

http://test.winehq.org/data/704f9ea5d18cc005b576e1da302cac05463b50f2/#group_...
on ALH's vm,
opengl.c:581: OpenGL renderer: Mesa GLX Indirect
(i.e. not hardware accelerated),

http://test.winehq.org/data/9960ab2418b18245dcf249da3a0ea097cdf62553/#group_...
on a Dell 630,
opengl.c:576: OpenGL renderer: Quadro NVS 135M/PCI/SSE2
opengl.c:577: OpenGL driver version: 2.1.2 NVIDIA 177.80
opengl.c:578: OpenGL vendor: NVIDIA Corporation

The failure on that last one was

visual.c:8834: Test failed: Offscreen failed for D3DFMT_G16R16: Got color
0x181800, expected 0x1818ff.

and seems related to this recent change:

http://source.winehq.org/git/wine.git/?a=commitdiff;h=e04556bb166804edc5c6cb...

We can probably discount the failure on the Intel card and the
virtual card as due to driver problems.
The Nvidia failure seems more serious.
Stefan, what's up there?

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #16 from Stefan Dösinger   2008-12-18 09:39:52 ---
That patch had a bug; I sent a patch to fix it a few days ago, can you retry?

Dan, can you paste the full glxinfo output? Probably there's some extension
oddity on your card which wined3d misses and doesn't enable the fragment
pipeline replacement, but does use G16R16 textures.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

Dan Kegel  changed:

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

--- Comment #17 from Dan Kegel   2008-12-18 10:04:28 ---
Not my system, that was just somebody's system on test.winehq.org.

Let's call this bug fixed for now!

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

Alexandre Julliard  changed:

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

--- Comment #18 from Alexandre Julliard   2008-12-20 09:03:13 ---
Closing bugs fixed in 1.1.11.

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|CVS/GIT                     |unspecified

[Bug 10221] "make test" fails on d3d9/tests/visual.c

http://bugs.winehq.org/show_bug.cgi?id=10221

--- Comment #19 from Dan Kegel   2009-07-04 10:39:10 ---
See also bug 19183.