WineHQ Bugzilla – Bug 35579 – d3d8:visual test_scalar_instructions() fails on Windows

Bugzilla

Last modified: 2014-02-20 12:37:11 UTC

d3d8:visual test_scalar_instructions() fails on Windows

Bug 35579 - d3d8:visual test_scalar_instructions() fails on Windows


Summary:	d3d8:visual test_scalar_instructions() fails on Windows

Status:	CLOSED FIXED

AppDB:	Show Apps affected by this bug
Product:	Wine
Classification:	Unclassified
Component:	d3d
Version:	1.7.12
Hardware:	x86 Windows

Importance:	P2 normal
Target Milestone:	---
Assigned To:	Mr. Bugs

URL:	http://test.winehq.org/data/tests/d3d...
Keywords:	regression, source

Depends on:
Blocks:
	Show dependency tree

Reported:	2014-02-11 10:51 UTC by François Gouget
Modified:	2014-02-20 12:37 UTC (History)
CC List:	0 users

See Also:
Regression SHA1:	27dc41d0dc31f02476a008cbe1a517bc9a425a30
Fixed by SHA1:	b22ee99fe948a30dc0c0aad5f3df711cffaa9596
Distribution:	---
Staged patchset:

Attachments
Add an attachment (back traces, logs, proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.

Description François Gouget 2014-02-11 10:51:49 UTC

The conformance test errors are:
visual.c:1803: Test failed: rcp_test: Got unexpected color 0x00ffffff, expected 0x00808080.
visual.c:1803: Test failed: rsq_test: Got unexpected color 0x00ffffff, expected 0x00b4b4b4.
visual.c:1803: Test failed: exp_test: Got unexpected color 0x00d6d6d6, expected 0x00404040.
visual.c:1803: Test failed: expp_test: Got unexpected color 0x00ffffff, expected 0x00404040.
visual.c:1803: Test failed: log_test: Got unexpected color 0x00000000, expected 0x00ffffff.
visual.c:1803: Test failed: logp_test: Got unexpected color 0x0000ff00, expected 0x00ffffff.

These errors occur on my Acer laptop: Windows 8.1 running on real hardware with an Intel HD4000 graphics card. It turns out they also happen on my fg-win2000-uplast and fg-win7u64-0sp VMware VMs. And interestingly also on the WineTestBot w864 VM which should help with tests.

http://test.winehq.org/data/tests/d3d8:visual.html

The commit that introduced these tests is:

commit 27dc41d0dc31f02476a008cbe1a517bc9a425a30
Author: Henri Verbeet <hverbeet@codeweavers.com>
Date:   Thu Dec 12 10:23:29 2013 +0100

    wined3d: Fix the source swizzle for WINED3DSIH_LOG and WINED3DSIH_LOGP.

Comment 1 Henri Verbeet 2014-02-11 11:23:43 UTC

At first sight it looks like it's using the .x component instead of .w. I'm afraid that pretty much qualifies as driver bug, or at least unwanted behaviour, since (several) applications depend on using .w here. I can add broken() cases for this if that's really what we want, although at least in the past I think we've always let the tests just fail for driver issues, with the testbot being the one big exception.

Comment 2 François Gouget 2014-02-11 17:11:01 UTC

I have retested this on my laptop with the latest Intel driver which I got from:

https://downloadcenter.intel.com/SearchResult.aspx?lang=eng&ProductFamily=Graphics&ProductLine=Laptop+graphics+drivers&ProductProduct=3rd+Generation+Intel%C2%AE+Core%E2%84%A2+Processors+with+Intel%C2%AE+HD+Graphics+4000

(the page labels it as 15.33.8.64.3345 presumably it's a bundle with the audio driver)

But the test still fail.


> I'm afraid that pretty much qualifies as driver bug, or at least
> unwanted behaviour

So there are two impacted drivers:
 * The VMware SVGA 3D 7.14.1.5025 (2013/12/7) driver.
   (as well as the older 7.14.1.1134 (2011/11/11) one).
 * The Intel HD4000 10.18.10.3345 (2013/10/31) driver.
   (the corresponding driver download is tagged as 15.33.8.64.3345 presumably it's a bundle with the audio driver)

If it's valid but unwanted behavior then it MUST be tagged as broken() rather than left to fail.

If it's a driver bug then it should really be reported so it's fixed and we don't have to deal with this failure forever. The report should be made by someone who will be able to answer technical questions, know which applications are impacted, etc. (so definitely not me).

That leaves the question of what to do until the drivers are fixed. This impacts more than just the WineTestBot.
I'm trying to get my laptop under the new 50 failing tests limit and while one test is not an issue, it would not do to leave the 40 DirectX/OpenGL tests to fail just because a very maintream driver is buggy.
Also even one failure means no green for Windows 8 on the test.winehq.org results, which in turns makes it harder to know when there are new failing tests.

Comment 3 Henri Verbeet 2014-02-12 06:14:50 UTC

(In reply to comment #2)
> So there are two impacted drivers:
>  * The VMware SVGA 3D 7.14.1.5025 (2013/12/7) driver.
>    (as well as the older 7.14.1.1134 (2011/11/11) one).
>  * The Intel HD4000 10.18.10.3345 (2013/10/31) driver.
>    (the corresponding driver download is tagged as 15.33.8.64.3345
> presumably it's a bundle with the audio driver)
> 
> If it's valid but unwanted behavior then it MUST be tagged as broken()
> rather than left to fail.
> 
Well, there isn't any public spec for this, or most parts of the Win32 API really, so in that regard it's always hard to say what's valid and what isn't. What we've always cared about is applications, and failing these tests means breaking them. For example, exp_test is for bug 27850, and IIRC failing the rcp and rsq ones would break Battlefield 1942.

> If it's a driver bug then it should really be reported so it's fixed and we
> don't have to deal with this failure forever. The report should be made by
> someone who will be able to answer technical questions, know which
> applications are impacted, etc. (so definitely not me).
> 
I wish anyones who wants to try that good luck, but my experience it pretty much that if you want something fixed you need to do it yourself, with very few exceptions.

> That leaves the question of what to do until the drivers are fixed. This
> impacts more than just the WineTestBot.
> I'm trying to get my laptop under the new 50 failing tests limit and while
> one test is not an issue, it would not do to leave the 40 DirectX/OpenGL
> tests to fail just because a very maintream driver is buggy.
> Also even one failure means no green for Windows 8 on the test.winehq.org
> results, which in turns makes it harder to know when there are new failing
> tests.
Every time I write a new test, it seems more and more tempting to either just skip the D3D tests completely on anything that isn't AMD or NVIDIA, or to just stop writing new tests.

Comment 4 Henri Verbeet 2014-02-15 07:29:29 UTC

Should no longer happen since commit b22ee99fe948a30dc0c0aad5f3df711cffaa9596.

Comment 5 Alexandre Julliard 2014-02-20 12:37:11 UTC

Closing bugs fixed in 1.7.13.