Bug 36014 - [bisected SNB] compiz is broken
Summary: [bisected SNB] compiz is broken
Status: VERIFIED FIXED
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: git
Hardware: All Linux (All)
: high critical
Assignee: Chris Wilson
QA Contact: Xorg Project Team
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2011-04-05 20:25 UTC by fangxun
Modified: 2011-04-12 21:26 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
fbo-alphatest-formats output (1.59 KB, text/plain)
2011-04-05 20:25 UTC, fangxun
no flags Details

Description fangxun 2011-04-05 20:25:10 UTC
Created attachment 45333 [details]
fbo-alphatest-formats output

System Environment:
--------------------------
Arch:           x86_64
Platform:       SNB
Libdrm:         (master)2.4.24-10-gf395b0a25c9f99688e7b2799ac02f1f930f24e09
Mesa:           (7.10)868ca30235548bd09c2a6c3e88fe3585113ef55d
Xserver:        (master)xorg-server-1.10.0
Xf86_video_intel: (master)2.14.902-3-g537a836dd6db384d53b52eb457a7d257c440217f
Kernel: (drm-intel-backport)c94249d2a6911daf74f329e05c42e076af2cd024


Bug detailed description:
-------------------------
Many piglit cases failed with latest code: fbo/fbo-alphatest-formats, fbo/fbo-clear-formats, fbo/fbo-fragcoord2, bugs/r300-readcache, bugs/fdo22540... 
It happens on all our SNB machines. Bisect shows d2106384be6f9df498392127c3ff64d0a2b17457 is the first bad commit. 
commit d2106384be6f9df498392127c3ff64d0a2b17457
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Apr 4 12:33:04 2011 +0100

    i965: Convert to relative relocations for state
    
    References: https://bugs.freedesktop.org/show_bug.cgi?id=35733
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>


Bug detailed description:
-------------------------
1. start X
2. ./fbo-alphatest-formats  -auto
Comment 1 fangxun 2011-04-05 23:02:47 UTC
Commit d210638 also causes messing up on gnome desktop with compiz. So change priority to critical.
Comment 2 Chris Wilson 2011-04-06 00:38:34 UTC
fbo/fbo-alphatest-formats -auto fails for me with d2106384be6f9df498392127c3ff64d0a2b17457^ or 2.14.901 on HuronRiver.
Comment 3 Chris Wilson 2011-04-06 01:26:31 UTC
Having just updated to mesa.git master (forgot I had some wacky experimental version on that machine) and using xf86-video-intel.git master:

=0 ickle:/usr/src/piglit (master)$ DISPLAY=:0 bin/fbo-alphatest-formats -auto
Using test set: Core formats
Testing 3
Testing 4
Testing GL_RGB
Testing GL_RGBA
Testing GL_ALPHA
Testing GL_LUMINANCE - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE_ALPHA - fbo incomplete (status = 0x8cdd)
Testing GL_INTENSITY - fbo incomplete (status = 0x8cdd)
Testing GL_ALPHA4
Testing GL_ALPHA8
Testing GL_ALPHA12
Testing GL_ALPHA16
Testing GL_LUMINANCE4 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE8 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE12 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE16 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE4_ALPHA4 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE8_ALPHA8 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE12_ALPHA12 - fbo incomplete (status = 0x8cdd)
Testing GL_LUMINANCE16_ALPHA16 - fbo incomplete (status = 0x8cdd)
Testing GL_INTENSITY4 - fbo incomplete (status = 0x8cdd)
Testing GL_INTENSITY8 - fbo incomplete (status = 0x8cdd)
Testing GL_INTENSITY12 - fbo incomplete (status = 0x8cdd)
Testing GL_INTENSITY16 - fbo incomplete (status = 0x8cdd)
Testing GL_R3_G3_B2
Testing GL_RGB4
Testing GL_RGB5
Testing GL_RGB8
Testing GL_RGB10
Testing GL_RGB12
Testing GL_RGB16
Testing GL_RGBA2
Testing GL_RGBA4
Testing GL_RGB5_A1
Testing GL_RGBA8
Testing GL_RGB10_A2
Testing GL_RGBA12
Testing GL_RGBA16
PIGLIT: {'result': 'pass' }
Comment 4 Chris Wilson 2011-04-06 01:53:29 UTC
Ho hum. I can confirm that:

mesa.git + xf86-video-intel.git: pass
mesa-7.10 + xf86-video-intel.git: fail
mesa-7.10 + xf86-video-intel-2.14.901: pass
Comment 5 Chris Wilson 2011-04-06 05:42:56 UTC
To confuse matters further, fbo-alphatest-formats (and co) fail on the first run on a fresh boot be it with xf86-video-intel.git or xf86-video-intel-2.14.902.

Which explains the oddity of my previous test. So after being run once, I do not see the failure of fbo-alphatest-formats with either xf86-video-intel.git or xf86-video-intel-2.14.902
Comment 6 Chris Wilson 2011-04-07 08:21:02 UTC
After re-installing f15 on HuronRiver to get a working compiz, I did confirm that breakage. I'm still not convinced that the fault lies within the relative relocations of the ddx, but I could not find any other fix despite banging my head against it for too long. So reverted for the time being.

commit 59ed6b05db99d7e5144f4f8499c67cee9b833673
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Apr 7 15:28:46 2011 +0100

    Revert "i965: Convert to relative relocations for state"
    
    This reverts commit d2106384be6f9df498392127c3ff64d0a2b17457.
    
    Breaks compiz (but not mutter/gnome-shell) on gen6. Not sure if this is
    not seem deep interaction issue with multiple clients sharing the GPU or
    just with compiz, but for now we have to revert and suffer the inane
    performance hit. It looks suspiciously like another deferred damage
    issue...
    
    Bugzilla: 51a27e88b073cff229fff4362cb6ac22835c4044
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Comment 7 fangxun 2011-04-12 21:26:24 UTC
Verified with commit: 4fa35dd5e13a58070220f787fc0678f1c679808d


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.