http://bugs.winehq.org/show_bug.cgi?id=29008
Bug #: 29008 Summary: Age of Wonders fails to run because of a floating point error Product: Wine Version: 1.3.32 Platform: x86 OS/Version: Linux Status: NEW Severity: normal Priority: P2 Component: directx-ddraw AssignedTo: wine-bugs@winehq.org ReportedBy: stefan@codeweavers.com Classification: Unclassified
Age of Wonders does not start because of a floating point exception during wined3d initialization. This is a regression caused by this patch:
7b0ba5153f3e42f31b922e5eb997d5e1d0cb44d4 is the first bad commit commit 7b0ba5153f3e42f31b922e5eb997d5e1d0cb44d4 Author: Matteo Bruni mbruni@codeweavers.com Date: Thu Nov 3 15:26:36 2011 +0100
wined3d: Test more thoroughly for post-pixelshader blending support, try on more texture formats.
This problem happens is quite complicated: Before initializing DirectDraw, the game changes the FPU control word and removes the FPCW$INVALID flag(0x0001), among others. During wined3d initialization wined3d tests if WINED3DFMT_UYVY is FBO attachable, which is true because we emulate it with GL_LUMINANCE_ALPHA. After that the new code added by the patch tries to draw to it, which apparently generates a NaN somewhere, and since the game disallowed them an exception is raised. The full FPU control word that is set when DirectDrawCreate is called is 0x1372.
The game pops up an error dialog and tries to initialize DirectDraw again, and is stuck in a loop from this point on. It isn't even necessary to install the game to reproduce this issue, the aowsetup.exe program on the CD has the same crash.
The problem is not related to WINED3DFMT_UYVY per se, the other luminance-alpha formats like WINED3DFMT_YUY2, WINED3DFMT_L8A8 and WINED3DFMT_L4A4 affected as well.
My GPU driver is the Nvidia binary driver, version 290.06, on a GeForce 9600M GPU. A while ago this driver didn't support luminance-alpha fbo attachments, apparently this changed somewhen.
http://bugs.winehq.org/show_bug.cgi?id=29008
Stefan Dösinger stefan@codeweavers.com changed:
What |Removed |Added ---------------------------------------------------------------------------- URL| |http://www.triumphstudios.c | |om/ageofwonders/aow.html CC| |matteo.mystral@gmail.com Regression SHA1| |7b0ba5153f3e42f31b922e5eb99 | |7d5e1d0cb44d4
--- Comment #1 from Stefan Dösinger stefan@codeweavers.com 2011-11-06 12:32:44 CST --- I've set the app's website as URL. It has 3 demo links, but they seem to be dead. The appdb ID is 4589, but I don't know how to link the bug and appdb entries.
While looking for Matteo's bugzilla account I found bug 29000, which is caused by the same patch and may have the same mechanics. (A search for the SHA1 didn't find 29000, why?)
http://bugs.winehq.org/show_bug.cgi?id=29008
Henri Verbeet hverbeet@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Keywords| |regression
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #2 from Austin English austinenglish@gmail.com 2011-11-07 12:43:26 CST --- (In reply to comment #1)
I've set the app's website as URL. It has 3 demo links, but they seem to be dead. The appdb ID is 4589, but I don't know how to link the bug and appdb entries.
You have to do it from the AppDB side (which I just did for this bug).
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #3 from Matteo Bruni matteo.mystral@gmail.com 2011-11-07 19:44:20 CST --- I looked into bug 29000 (actually, I didn't notice this bug until I essentially replicated all the debugging Stefan already did...) and yes, the issue is the same.
I think this is a bug in Nvidia drivers, as the FPU exception is generated inside libGL and this works fine on fglrx. Anyway, it may also be I am doing something fishy, prompting such behavior.
In this case, probably there isn't much point in testing the problematic texture formats anyway (e.g. red_mask is 0 for those luminance formats, so the test is not checking that part), I could just skip those and be fine with that.
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #4 from Matteo Bruni matteo.mystral@gmail.com 2011-11-07 20:00:01 CST --- Created attachment 37375 --> http://bugs.winehq.org/attachment.cgi?id=37375 Workaround the issue + some improvements
This should help, it also includes a couple of fixes (aka that's how it should have been in the first place). I'll rework this a bit before submitting to wine-patches, maybe introducing a new WINED3DFMT_FLAG to mark those formats which should be tested on DX10+ GPUs.
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #5 from Matteo Bruni matteo.mystral@gmail.com 2011-11-16 06:02:57 CST --- Should be fixed in current git by 62a2ad3d0f8d8435619924a6ff4a19adf94c16cc (this bug can be marked duplicate of bug 29000, or the other way around).
http://bugs.winehq.org/show_bug.cgi?id=29008
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Resolution| |DUPLICATE
--- Comment #6 from Austin English austinenglish@gmail.com 2011-11-16 13:13:36 CST --- (In reply to comment #5)
Should be fixed in current git by 62a2ad3d0f8d8435619924a6ff4a19adf94c16cc (this bug can be marked duplicate of bug 29000, or the other way around).
Dupe.
*** This bug has been marked as a duplicate of bug 29000 ***
http://bugs.winehq.org/show_bug.cgi?id=29008
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #7 from Austin English austinenglish@gmail.com 2011-11-17 00:10:36 CST --- Closing.
http://bugs.winehq.org/show_bug.cgi?id=29008
Austin English austinenglish@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Component|directx-ddraw |directx-d3d
http://bugs.winehq.org/show_bug.cgi?id=29008
keks9n@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |keks9n@gmail.com
--- Comment #8 from keks9n@gmail.com 2013-08-28 09:15:53 CDT --- This bug has retured. At least on wine 1.6. AoW.exe says "Invalid floating point operation".
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #9 from Stefan Dösinger stefan@codeweavers.com 2013-08-28 09:25:01 CDT --- Can you run a regression test?
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #10 from keks9n@gmail.com 2013-08-28 09:26:54 CDT --- What kind of test? If you want me to bisect it will take some time.
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #11 from Bruno Jesus 00cpxxx@gmail.com 2013-08-28 09:42:23 CDT --- (In reply to comment #10)
What kind of test? If you want me to bisect it will take some time.
http://wiki.winehq.org/RegressionTesting
But IMO you should retry in wine git first.
http://bugs.winehq.org/show_bug.cgi?id=29008
--- Comment #12 from Stefan Dösinger stefan@codeweavers.com 2013-09-01 16:30:36 CDT --- I cannot reproduce this on Wine 1.7.1, Nvidia 325.15. AoW starts without issues.
-
wine-bugs＠winehq.org