Bug 91110

Summary: Textures missing iff character in specific areas in dolphin running Xenoblades Chronicles
Product: Mesa Reporter: Christian Weinz <christian>
Component: Drivers/Gallium/radeonsiAssignee: Default DRI bug account <dri-devel>
Status: RESOLVED INVALID QA Contact: Default DRI bug account <dri-devel>
Severity: normal    
Priority: medium CC: EoD
Version: git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
URL: https://bugs.dolphin-emu.org/issues/8688
Whiteboard:
i915 platform: i915 features:
Attachments: frame 1 of the apitrace dump with R600_DEBUG=switch_on_eop
frame 2 of the apitrace dump with R600_DEBUG=switch_on_eop

Description Christian Weinz 2015-06-25 19:49:59 UTC 
Iff a character is in specific areas some textures on him are not rendered. In the entry of colony 9 is such an area.

The textures are rendered properly on the Intel GPU.

There are screenshots, a fifolog and an apitrace at the end of the linked bugreport.

The problem appears in Debian Jessie, Sid and mesa-git the last time checked, which was 9 days ago.

I compiled mesa-git with

CFLAGS = -g -O2 -Wall -std=c99 -Werror=implicit-function-declaration -Werror=missing-prototypes -fno-strict-aliasing -fno-builtin-memcmp -DDEBUG

and set

$ export LIBGL_DEBUG=verbose
$ export MESA_DEBUG=1
$ export EGL_LOG_LEVEL=debug

before running dolphin. Running dolphin with the game until the issue appears gives the following output

$ dolphin-emu
Gtk-Message: Failed to load module "canberra-gtk-module"
libGL: screen 0 does not appear to be DRI3 capable
libGL: pci id for fd 22: 1002:679a, driver radeonsi
libGL: OpenDriver: trying lib/tls/radeonsi_dri.so
libGL: OpenDriver: trying lib/radeonsi_dri.so
libGL: Can't open configuration file /home/user/.drirc: Arquivo ou diretório não encontrado.
libGL: Can't open configuration file /home/user/.drirc: Arquivo ou diretório não encontrado.
libGL: Using DRI2 for screen 0

I'm sorry, I don't know how to write a simple GLUT-based test program.
Comment 1 Michel Dänzer 2015-06-26 08:19:17 UTC 
I can't reproduce this problem on Kaveri, the apitrace plays back correctly. Maybe the problem is Tahiti specific.
Comment 2 Michel Dänzer 2015-06-29 00:29:08 UTC 
Does setting the environment variable R600_DEBUG=switch_on_eop work around the problem when playing back the apitrace or running the game?
Comment 3 Christian Weinz 2015-06-29 01:50:50 UTC 
Created attachment 116776 [details]
frame 1 of the apitrace dump with R600_DEBUG=switch_on_eop
Comment 4 Christian Weinz 2015-06-29 01:51:27 UTC 
Created attachment 116777 [details]
frame 2 of the apitrace dump with R600_DEBUG=switch_on_eop
Comment 5 Christian Weinz 2015-06-29 01:57:43 UTC 
No.

I tested the replay of the trace and the game both with mesa from Sid and mesa-git, each time with R600_DEBUG=switch_on_eop set. The issue remained unaffected.

Are you sure you meant R600_DEBUG=switch_on_eop? I'm asking because I'm using radeonsi.
Comment 6 EoD 2016-02-14 05:09:43 UTC 
(In reply to Christian Weinz from comment #5)
> No.
> 
> Are you sure you meant R600_DEBUG=switch_on_eop? I'm asking because I'm
> using radeonsi.

r600 and radeonsi share the same R600_DEBUG.

Apparently the apitrace file has been deleted. Can you upload it again? Some people use google drive for apitraces for instance.
Comment 7 Timothy Arceri 2019-02-17 23:33:07 UTC 
apitrace never reuploaded. Given the age of the bug and no follow-up bug reports I'm just going to close this bug as invalid.

Feel free to reopen if it is still a problem, adding a fresh apitrace and updated information.

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.