Bug 32083 - [regression] Shaders not working in the game MARS
Summary: [regression] Shaders not working in the game MARS
Status: RESOLVED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: glsl-compiler (show other bugs)
Version: git
Hardware: All All
: medium normal
Assignee: Tapani Pälli
QA Contact:
URL: http://mars-game.sourceforge.net/
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-12-03 09:28 UTC by Sven Arvidsson
Modified: 2014-01-09 06:55 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments
Screenshot of bug (57.38 KB, image/jpeg)
2010-12-03 09:28 UTC, Sven Arvidsson
Details

Description Sven Arvidsson 2010-12-03 09:28:58 UTC
Created attachment 40784 [details]
Screenshot of bug

Turning on shaders in the game M.A.R.S. results in a mostly pink background image, which is an obvious misrendering.

This is probably a regression in glsl2 (swrast from Mesa 7.8 seems to work okay).

The game only uses two shaders, available here:
http://mars-game.svn.sourceforge.net/viewvc/mars-game/data/shaders/

I'm using git master, dbf996f30856c000e18e3700f741d4e92561b4ec
Comment 1 Sven Arvidsson 2013-09-18 11:11:43 UTC
Still a problem with latest version of the game 0.7.4 and Mesa 9.2. Turning on shaders locks the display, but the game seems to continue running.

On i965 I get:
 
 intel_do_flush_locked failed: No such file or directory

On r600g I get rejected CS with the following in dmesg:

 [160490.435642] radeon 0000:01:00.0: evergreen_cs_track_validate_cb:476 cb[0] bo too small (layer size 1376256, offset 0, max layer 1, bo size 4096, slice 5375)
 [160490.435646] radeon 0000:01:00.0: evergreen_cs_track_validate_cb:480 problematic surf: (768 448) (4 4 1 1 2 1024 2)
 [160490.435649] radeon 0000:01:00.0: evergreen_packet3_check:1958 invalid cmd stream 539
 [160490.435651] [drm:radeon_cs_ib_chunk] *ERROR* Invalid command stream !
 [160490.438695] radeon 0000:01:00.0: evergreen_cs_track_validate_texture:850 texture bo too small (layer size 16777216, offset 0, max layer 1, depth 1, bo size 65536) (2048 2048)
 [160490.438697] [drm:radeon_cs_ib_chunk] *ERROR* Invalid command stream !
 [160490.439243] [drm:radeon_cs_parser_relocs] *ERROR* gem object lookup failed 0x2b
 [160490.439245] [drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -2!
Comment 2 Tapani Pälli 2014-01-07 10:43:04 UTC
I've tried to run precompiled binaries of the game but the ones from rpm and ones from zip relase have symbols that are not present in the bundled libraries or libraries from my OS (fedora19). I'm having also problems compiling the game, cmake configuration passes ok but compilation fails due to SFML headers not found correctly. Is there a known working 'zip release' (like mars_linux_0.7.4.tar.gz) of the game with all the deps included?
Comment 3 Sven Arvidsson 2014-01-07 16:11:46 UTC
(In reply to comment #2)
> I've tried to run precompiled binaries of the game but the ones from rpm and
> ones from zip relase have symbols that are not present in the bundled
> libraries or libraries from my OS (fedora19). I'm having also problems
> compiling the game, cmake configuration passes ok but compilation fails due
> to SFML headers not found correctly. Is there a known working 'zip release'
> (like mars_linux_0.7.4.tar.gz) of the game with all the deps included?

I had a quick look at this. Debian ships version 0.7.5 of the game, which does not seem to have been released anywhere else, but in this version the shaders seems to work fine.

Next I tried getting an apitrace from version 0.7.4, but replaying the trace does not trigger the bug.

Not sure if it's any point in keeping this bug open any longer.
Comment 4 Tapani Pälli 2014-01-08 06:30:04 UTC
OK, I'll try to get the version Debian ships to verify. I can't spot anything from shaders that would not be supported, the issues are very likely gone.
Comment 5 Tapani Pälli 2014-01-09 06:55:07 UTC
tested Mars 0.7.5 on debian jessie with Mesa master and shader effects enabled, graphics work fine, resolving as fixed


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.