Bug 107747 - [Intel GFX CI] spec@arb_texture_multisample@sample-position@4 intermittently fail
Summary: [Intel GFX CI] spec@arb_texture_multisample@sample-position@4 intermittently ...
Status: CLOSED WORKSFORME
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: 18.0
Hardware: Other All
: medium normal
Assignee: Intel 3D Bugs Mailing List
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2018-08-30 10:51 UTC by Martin Peres
Modified: 2019-03-15 15:40 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments

Description Martin Peres 2018-08-30 10:51:34 UTC
On pig-snb-2600, we started seeing intermitent failures for the test spec@arb_texture_multisample@sample-position@4:

.../piglit/bin/arb_texture_multisample-sample-position 4 -auto -fbo

Stderr	
sample 0 position incorrect (y axis): observed 0.093750, expected 0.125000
sample 1 position incorrect (y axis): observed 0.093750, expected 0.375000
sample 2 position incorrect (y axis): observed 0.093750, expected 0.625000
sample 3 position incorrect (y axis): observed 0.093750, expected 0.875000

This never happened before, but it just happened twice yesterday, 3 hours appart, which suggests something has changed. Given that we only changed the kernel, it is possible that this would be a kernel change that led to this instability, but I fail to see how it could happen as this is supposed to be fixed functions...

This is why I wrote the bug for mesa, waiting for your input!
Comment 1 Denis 2018-08-30 13:12:32 UTC
hi. I have i5-2520M and latest 4.18.5 kernel. Mesa 18.3.0, built from source (git-8c0e3f3822). I can't reproduce the failure on this test.

upd - I have ubuntu 16.04 (xserver 1.19.6)

>intermitent failures
could you please provide approximate fail rate please?
Comment 2 Mark Janes 2018-08-30 15:27:01 UTC
Mesa CI has no intermittent results for this test:

http://mesa-ci-results.jf.intel.com/mesa_master/test/303a541c2e7cb9068b8c7a31de7236e6/history
Comment 3 Martin Peres 2018-09-11 15:22:29 UTC
(In reply to Mark Janes from comment #2)
> Mesa CI has no intermittent results for this test:
> 
> http://mesa-ci-results.jf.intel.com/mesa_master/test/
> 303a541c2e7cb9068b8c7a31de7236e6/history

Good to know, but I assume we have different SNBs: https://intel-gfx-ci.01.org/hardware/shard-snb1/

What kind of information are you looking for? We had 3 failures so far (CI_DRM_4749_full (1 week, 3 days / 143 runs ago), CI_DRM_4729_full (1 week, 5 days / 194 runs ago) and CI_DRM_4726_full (1 week, 6 days / 202 runs ago)).

What I would like to know is: Can the kernel actually influence this test.
Comment 4 Mark Janes 2018-09-11 15:39:14 UTC
I'll ask Mesa devs to chime in on the kernel implications.

We could possibly verify that the kernel is affecting this test by deploying your kernel in Mesa CI to see if we can repro the failures.  What branch should we build?
Comment 5 Martin Peres 2018-09-11 16:35:11 UTC
(In reply to Mark Janes from comment #4)
> I'll ask Mesa devs to chime in on the kernel implications.

Thanks :)

> 
> We could possibly verify that the kernel is affecting this test by deploying
> your kernel in Mesa CI to see if we can repro the failures.  What branch
> should we build?

Good idea! Here is all the information:
 - https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_4749/kernel.config.bz2
 - 4a46c18fad0de38a78b4b0c848892de494324a17 @ git://anongit.freedesktop.org/gfx-ci/linux (Should be tag CI_DRM_4726)
Comment 6 Martin Peres 2018-09-20 12:47:23 UTC
I also get failures with spec@arb_texture_multisample@texelfetch@2-gs-isampler2dms:

Command line	
/opt/igt/piglit/bin/texelFetch gs isampler2DMS 2 -auto -fbo

Stdout	
65x32x1
Probe at (35,27)
Expected: 0.546875 0.870968 0.000000 0.000000
Observed: 0.000000 0.000000 0.000000 0.000000

I did not want to create another bug to not clutter your bugtracker with CI failures that seem related anyway.

I hope it helps!
Comment 7 Mark Janes 2018-09-20 15:54:09 UTC
I do see intermittent failures with
spec@arb_texture_multisample@texelfetch@2-gs-isampler2dms
Comment 8 Martin Peres 2019-03-08 12:41:32 UTC
(In reply to Mark Janes from comment #7)
> I do see intermittent failures with
> spec@arb_texture_multisample@texelfetch@2-gs-isampler2dms

Do you still see this happening? We got nothing since CI_DRM_4842_full (5 months, 2 weeks / 3044 runs ago).
Comment 9 Mark Janes 2019-03-08 20:05:16 UTC
No recent failures in the logs of Mesa i965 CI.
Comment 10 Martin Peres 2019-03-15 15:40:23 UTC
(In reply to Mark Janes from comment #9)
> No recent failures in the logs of Mesa i965 CI.

Thanks, I guess we'll just close this for good then. Thanks!
Comment 11 CI Bug Log 2019-03-15 15:40:53 UTC
The CI Bug Log issue associated to this bug has been archived.

New failures matching the above filters will not be associated to this bug anymore.


Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.