http://bugs.winehq.org/show_bug.cgi?id=13643
Summary: ddraw tests fail on XP SP2 with Radeon Mobility 9000 card Product: Wine Version: CVS/GIT Platform: PC OS/Version: Windows XP Status: NEW Severity: normal Priority: P2 Component: -unknown AssignedTo: wine-bugs@winehq.org ReportedBy: juan_lang@yahoo.com
Created an attachment (id=13624) --> (http://bugs.winehq.org/attachment.cgi?id=13624) d3d test failures
The ddraw d3d tests produce 22 failures, and the dsurface test produces 1 failure, on my laptop running Windows XP SP2. The graphics card is a Radeon Mobility 9000. The d3d failures are attached.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #1 from Juan Lang juan_lang@yahoo.com 2008-06-03 10:33:10 --- Created an attachment (id=13625) --> (http://bugs.winehq.org/attachment.cgi?id=13625) dsurface test failure
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #2 from James Hawkins truiken@gmail.com 2008-06-03 11:28:23 --- When I look at these tests, my gut feeling is that some member(s) of the vp_data struct is being messed with in the first call to SetViewport (I'm particularly thinking about vp_data.dwSize) and that messes up the remaining calls to SetViewport. When I tried to run ddraw_test.exe built by me, and even winetest.exe built by me, I don't get the failures (thus I can't debug the members of vp_data). These failures happen on real machines and VMs, so I don't think it's a driver bug.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #3 from James Hawkins truiken@gmail.com 2008-06-03 11:43:38 --- Note that I *do* get the failures when run with Paul Millar's compiled winetest.exe, just not my winetest.exe. So the variable here is the compiler, thus why I think it has to do with dwSize or similar.
http://bugs.winehq.org/show_bug.cgi?id=13643
Paul Vriens Paul.Vriens.Wine@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |Paul.Vriens.Wine@gmail.com
--- Comment #4 from Paul Vriens Paul.Vriens.Wine@gmail.com 2008-06-03 13:38:30 --- Same thing here on my vmware box. It produces errors with the winetest from Paul Millar but not when I compile the test myself.
So what can we do about this. Add some tracing to the test, get it into GIT and wait for Paul Millar's new executable? Seems like a long process.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #5 from James Hawkins truiken@gmail.com 2008-06-03 13:52:06 --- We could send Paul Millar a patch to apply to his tree and ask him to create a special build for us (sent to the list or whatever, not for general use).
http://bugs.winehq.org/show_bug.cgi?id=13643
Paul Millar paulm@astro.gla.ac.uk changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |paulm@astro.gla.ac.uk
--- Comment #6 from Paul Millar paulm@astro.gla.ac.uk 2008-06-04 16:23:27 --- Hi guys,
I've manually built the ddraw_test.exe. It's available from:
http://quisquiliae.physics.gla.ac.uk/cross/special/
I couldn't say how likely it is that the build environment is triggering this problem. I seems unlikely, but quisquiliae does appear to be the common factor.
One possible way to proceed is to replicate my build environment. The MinGW-building script and patch-set are both available from the bottom of this page:
http://www.astro.gla.ac.uk/~paulm/Cross/
The native- and cross- compiling information is embedded within the executable (see top of the "[file]" link). These values are generated dynamically, so they should always be correct. For the sake of completeness, I've copied across the salient points:
Native gnu toolchain: GNU ar 2.17 Debian GNU/Linux gcc version 4.0.4 20060507 (prerelease) (Debian 4.0.3-3) ccache version 2.4 distcc 2.18.3 i486-pc-linux-gnu
Cross-compiling gnu toolchain: i686-mingw32msvc-gcc (GCC) 3.4.6 GNU ar 2.16.91 20060119 ccache version 2.4 distcc 2.18.3 i486-pc-linux-gnu
I'm happy to create a special patched version of ddraw_test.exe if that would help. Just send me the patch and I'll make the test available at same location /cross/special/
Cheers,
Paul.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #7 from Juan Lang juan_lang@yahoo.com 2008-06-04 22:58:16 --- Hi Paul, thanks for the new build. This still has the same failures in the d3d test. The dsurface test doesn't fail, but it didn't before, either :-/ I was just smoking something when I said it had a failure, I guess.
In any case, I get many failures on my machine irrespective of the build environment. I can't submit the whole batch of them because the tests hang my machine.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #8 from Paul Vriens Paul.Vriens.Wine@gmail.com 2008-06-05 00:57:57 --- With my own compiled version:
d3d: 969 tests executed (0 marked as todo, 0 failures), 1 skipped.
With the one from Paul Millar:
d3d: 969 tests executed (0 marked as todo, 19 failures), 1 skipped.
This last one is the same for all XP/W2K3 systems seen at test.winehq.org.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #9 from James Hawkins truiken@gmail.com 2008-06-05 09:59:39 --- Thanks Paul M., I'll whip up a debugging patch for you to apply and we can see what values we're getting with your compile.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #10 from James Hawkins truiken@gmail.com 2008-06-10 17:52:49 --- Created an attachment (id=13893) --> (http://bugs.winehq.org/attachment.cgi?id=13893) debug tracing
Paul M., can you apply this patch to your tree and attach (or upload to your server) a build of a) the entire winetest.exe and b) just ddraw_test.exe. Thanks for your help.
http://bugs.winehq.org/show_bug.cgi?id=13643
James Hawkins truiken@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Attachment #13893|0 |1 is obsolete| |
--- Comment #11 from James Hawkins truiken@gmail.com 2008-06-10 17:57:41 --- Created an attachment (id=13894) --> (http://bugs.winehq.org/attachment.cgi?id=13894) fixed printf types
Better yet, try this patch. Forgot to check all the types first.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #12 from Paul Millar paul@astro.gla.ac.uk 2008-06-10 18:34:59 --- OK, please try:
http://quisquiliae.physics.gla.ac.uk/cross/special/winetest-ddraw-d3d-debug....
and
http://quisquiliae.physics.gla.ac.uk/cross/special/ddraw_test-d3d-debug.exe
HTH,
Paul.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #13 from James Hawkins truiken@gmail.com 2008-06-10 18:42:41 --- Thanks a lot Paul! This confirms my suspicions:
d3d.c:1187: vp_data.dwSize: 0 d3d.c:1188: vp_data.dwX: 0 d3d.c:1189: vp_data.dwY: 0 d3d.c:1190: vp_data.dwWidth: 0 d3d.c:1191: vp_data.dwHeight: 256 d3d.c:1192: vp_data.dvMaxX: 256.000000 d3d.c:1193: vp_data.dvMaxY: 256.000000 d3d.c:1194: vp_data.dvScaleX: 5.000000 d3d.c:1195: vp_data.dvScaleY: 5.000000 d3d.c:1196: vp_data.dvMinZ: -25.000000 d3d.c:1197: vp_data.dvMaxZ: 60.000000
I'll write up a patch real quick to fix it and ask if you can make one last build for me.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #14 from James Hawkins truiken@gmail.com 2008-06-10 18:46:58 --- Created an attachment (id=13895) --> (http://bugs.winehq.org/attachment.cgi?id=13895) fix
Paul, can you use this to make just a build of the ddraw test?
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #15 from Paul Vriens Paul.Vriens.Wine@gmail.com 2008-06-11 01:12:29 --- James, can you explain why we have to do this and how a compiler has effect on clearing vp_data? Isn't it also strange that only the first four DWORDS are cleared and not the rest?
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #16 from Paul Vriens Paul.Vriens.Wine@gmail.com 2008-06-11 05:11:44 --- Paul M, I think the real winetest build is affected by the patch.
http://quisquiliae.physics.gla.ac.uk/cross/logs/20080611-11:05.txt shows:
Ran command "git rebase origin", but it failed with following output: dlls/ddraw/tests/d3d.c: needs update
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #17 from Paul Millar paul@astro.gla.ac.uk 2008-06-14 08:31:19 --- I've created special builds of winetest.exe and ddraw_test.exe with the proposed "fix" patch (attachment #13895)
http://bugs.winehq.org/attachment.cgi?id=13895
The executables are available from:
http://quisquiliae.physics.gla.ac.uk/cross/special/winetest-fixed.exe
and
http://quisquiliae.physics.gla.ac.uk/cross/special/ddraw_test-fixed.exe
Cheers,
Paul.
PS. If I may comment, the proposed solution looks ugly as you're repeating the variable initialisation explicitly. A nicer solution would be to have a simple function that resets vp_data to "known, good values" (e.g., dwX to 10, dvMaxX to 256), which the individual tests can further modify.
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #18 from James Hawkins truiken@gmail.com 2008-06-14 12:49:36 --- (In reply to comment #17)
PS. If I may comment, the proposed solution looks ugly as you're repeating the variable initialisation explicitly. A nicer solution would be to have a simple function that resets vp_data to "known, good values" (e.g., dwX to 10, dvMaxX to 256), which the individual tests can further modify.
It's just a quick test to see if it works :-)
http://bugs.winehq.org/show_bug.cgi?id=13643
--- Comment #19 from Paul Vriens Paul.Vriens.Wine@gmail.com 2008-09-12 04:59:20 --- Isn't this one fixed? (after commit 5dcec4de497de9a9c5ffbd3ef4a5f289b95f52ba)
http://bugs.winehq.org/show_bug.cgi?id=13643
Juan Lang juan_lang@yahoo.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution| |FIXED
--- Comment #20 from Juan Lang juan_lang@yahoo.com 2008-09-12 10:14:58 --- Sorry, my Windows installation died, so I can't retest. If you think it's fixed, I'll go ahead and mark it fixed ;-)
http://bugs.winehq.org/show_bug.cgi?id=13643
Alexandre Julliard julliard@winehq.org changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #21 from Alexandre Julliard julliard@winehq.org 2008-09-19 11:17:26 --- Closing bugs fixed in 1.1.5.
http://bugs.winehq.org/show_bug.cgi?id=13643
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Version|CVS/GIT |unspecified
http://bugs.winehq.org/show_bug.cgi?id=13643
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- OS/Version|Windows XP |Windows