Bug 58033

Try: RADEON_DEBUG=nozmask

Also try reverting the commit e866bd1adea2c3b4971ad68e69c6.

(In reply to comment #2)
> Also try reverting the commit e866bd1adea2c3b4971ad68e69c6.

I've tried this, and have also tried RADEON_DEBUG=nohiz,nozmask,nocbzb. However, I'm still seeing the problem. Interestingly, the problem does not happen with Fedora's "stock" Mesa drivers mesa-dri-drivers-8.0.4-1.fc17.i686.

Can you bisect?

(In reply to comment #4)
> Can you bisect?

Apparently, this is the bad commit:

39737e17e7a61535a35669756161005a7a5c887b is the first bad commit
commit 39737e17e7a61535a35669756161005a7a5c887b
Author: Marek Olšák <maraeo@gmail.com>
Date:   Mon Dec 3 01:26:22 2012 +0100

    st/dri: always allocate private depth-stencil buffers
    
    This disables DRI2 sharing of zbuffers. The window zbuffer is allocated just
    like any other texture - through resource_create.
    
    The idea of allocating a zbuffer through DRI2 isn't very useful with MSAA,
    where a single-sample zbuffer is useless.
    
    IIRC, the Intel driver does the same thing.
    
    Reviewed-by: Brian Paul <brianp@vmware.com>

:040000 040000 e8045e2d2c7e9a1ebd6eb1f22c93b97b9c3e8ce7 4deb5755e011d592366e6140bf385bfefe157f3f M	src

This bug is still present after Marek's latest r300g commits: HEAD is

commit e3e1ffb2520498584ef402213d0c8aa4303a46a3
Author: Marek Olšák <maraeo@gmail.com>
Date:   Mon Jan 14 05:51:05 2013 +0100

    r300g: set a dummy vertex buffer in context_create
    
    so that the driver doesn't crash if an app doesn't set any vertex buffers.

Marek Olšák post some patches on mailing list that implementing separate depth-stencil clear. Maybe they will help.

Still broken as of git HEAD:

commit ca39c0f94a4e3cc25b6cc9507fb729b85140733a
Author: Ian Romanick <ian.d.romanick@intel.com>
Date:   Wed Jan 16 15:34:49 2013 -0800

    mesa/es3: Don't check dimensions in _mesa_es3_error_check_format_and_type

Could you please attach a screenshot showing the issue?

(In reply to comment #9)
> Could you please attach a screenshot showing the issue?

Sorry, the corruption is not showing up which I press the "Print Screen" button. I shall attempt to describe instead:

Imagine moving towards a forest, with the trees in the centre of your screen. As you approach the forest, you expect the trees to get larger and appear to slide towards the left and right edges of your screen. (That's perfectly ordinary depth-perception). The corruption seems to be that nothing is immediately rendered where the trees used to be, leaving large black gaps. When you stop moving, the black gaps are quickly filled again.

Created attachment 73347 [details]
Screenshot of corruption in WoW

(In reply to comment #9)
> Could you please attach a screenshot showing the issue?

I've managed to capture a screenshot using GIMP instead of the Print Screen button.

Created attachment 73348 [details]
Screenshot of corruption in WoW

Created attachment 73349 [details]
More corruption in WoW

Another screenshot, to give you a better idea.

Created attachment 73350 [details]
Yet more corruption in WoW

All of these screenshots were taken while moving. The corruption disappears in a stationary shot.

I've just upgraded one of my 32 bit boxes to Fedora 18, which has allowed me to compile Mesa git here. And I can now report that the corruption happens with an RV730 chip too.

My CAICOS (HD6450) is not affected by this bug (64 bit).

My RV790 (HD4890) is also unaffected by this bug. And this box is 64 bit too. Could it just be a coincidence that both of my affected boxes are 32 bit?

This bug is still happening on my 32 bit T60p after #66921 has been fixed.

(In reply to comment #15)
> I've just upgraded one of my 32 bit boxes to Fedora 18, which has allowed me
> to compile Mesa git here. And I can now report that the corruption happens
> with an RV730 chip too.

This corruption appears with all shader backends (sb,llvm,sb+llvm) ?

(In reply to comment #19)
> This corruption appears with all shader backends (sb,llvm,sb+llvm) ?

Actually, the RV730 corruption has changed: instead of the "shadow-like" artifacts, the window beneath the UI briefly turns completely black instead. (Or sometime completely blue - I guess it depends on my location in the game).

This corruptions happens both with and without RADEON_DEBUG=nosb.

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/427.