OpenGL 3.0, GLSL 1.3

[agd5f]

I thought you are days from release so you can't fix it in 8.0.
Then tell me which version I should download. Thanks.

Depends on what the problem is and how complex the fixes are. Please test the 8.0 and master branches. If the issue is already fixed in master, we can cherrypick the fixes to the 8.0 branch. If it's broken in master, but working in the 8.0 branch, it would be helpful to identify the problematic commits. If it's working in 7.11 and broken in master and 8.0, it would be really helpful if you could bisect what commit broke things for the app in question.

[curaga]

BTW, who do I pester to get the postprocess fixes to 8.0?

In the current state of 8.0, depth filters crash. Master has Michel's fix for that, but that leaves the depth broken. My fixes for that've been waiting on the ML for a week now, no response on irc either.

Ideally I'd like to get all three pp patches to 8.0, Michel's + my two ones. Is it too late?

[aceman]

Depends on what the problem is and how complex the fixes are. Please test the 8.0 and master branches. If the issue is already fixed in master, we can cherrypick the fixes to the 8.0 branch. If it's broken in master, but working in the 8.0 branch, it would be helpful to identify the problematic commits. If it's working in 7.11 and broken in master and 8.0, it would be really helpful if you could bisect what commit broke things for the app in question.

On RV710, kernel 3.2.1, libdrm 2.4.30, X.org 1.9.5, radeon driver 6.14.3, R600g, here is the problem matrix:

<------><------>Celestia<------>scorched3d<---->google-e<------>glxgears
7.10.2(R600c)<->bug 33038<----->bug 28905<----->crash<-><------>OK
7.11.2<><------>OK<----><------>OK<----><------>crash<-><------>OK
8.0rc2<><------>bug 36782<----->OK<---><------->crash<-><------>OK
master<><------>no window<----->blank<-><------>crash<-><------>blank

So, as I said earlier I may have been trying some master version from after 8.0 was branched. I always grabbed from links in one of the commits having the hash in filename. master currently seems really broken.
However, 8.0rc2 does not have the glxgears and scorched3d problems I mentioned. But the celestia bug is real.
I'd discount Google Earth from this analysis as it seems completely broken. It may be some system lib interfering as I mentioned. I just remember it working on 7.10 in the past.

Master = mesa-d36f89c8cfb3776368f8acef46514421f454205c.tar.gz
Bug numbers as filed on https://bugs.freedesktop.org .
agd5f, is this of any use to you?

[agd5f]

Thanks. It would be great if you could file bugs for these so they can be tracked properly.

[aceman]

The bugs are filed, I have put their numbers in the table. I have already closed the obsolete ones.
I can't bisect the celestia regression between 7.11 and 8.0 currently. But when I get to it, I'll report it in its bug.
I just get many messages like this in the kernel log:
[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -12!
vmap allocation for size 178982912 failed: use vmalloc=<size> to increase size.

I am not sure if I should increase the vmalloc kernel parameter. Why would it be needed?