Created attachment 39071 [details]
Far view of Titan - atmosphere transparent
The atmosphere of some planets (not all) gets transparent when approaching the planets. As intended, the atmosphere of Titan should be opaque at all distances (except when standing on the surface). However, currently it gets more transparent as the observer is moving AWAY from the planet. From the distance, atmosphere is invisible, near the planet it is opaque.
Created attachment 39072 [details]
Near view of Titan - atmosphere almost opaque
This did not happen in Mesa 7.8.2.
It happens with Mesa 7.9-rc2, kernel 18.104.22.168, radeon driver 6.13.2, X.org 1.7.7.
OpenGL renderer string: Mesa DRI R600 (RV710 954F) 20090101 x86/MMX+/3DNow!+/SSE2 TCL DRI2
OpenGL version string: 2.1 Mesa 7.9-rc2
OpenGL shading language version string: 1.20
(In reply to comment #2)
> This did not happen in Mesa 7.8.2.
Any chance you could bisect it?
It also does not happen on 7.9 with software rendering.
I could try to bisect by date, but I don't know where to download the nightly snapshots.
It would be much easier to checkout the mesa tree with git and use git bisect.
From here: git://anongit.freedesktop.org/mesa/mesa ?
yes. see the mesa3d section of this guide for more info:
I have tried it, and managed several iterations. But now I have hit a range where celestia crashes on start with this error:
celestia: radeon_texture.c:703: radeon_store_teximage: Predpoklad `dstRowStride' nesplnený. (Prerequisite `dstRowStride' not met)
I have tried to get out of this range using "git bisect skip" on hundreds of commits, but I was not successful.
If you can help me get out of this hole I could continue bisecting.
Created attachment 39138 [details]
Directory of git bisect data which I created in the process
(In reply to comment #8)
> I have tried it, and managed several iterations. But now I have hit a range
> where celestia crashes on start with this error:
> celestia: radeon_texture.c:703: radeon_store_teximage: Predpoklad
> `dstRowStride' nesplnený. (Prerequisite `dstRowStride' not met)
> I have tried to get out of this range using "git bisect skip" on hundreds of
> commits, but I was not successful.
> If you can help me get out of this hole I could continue bisecting.
Driver broken for hundreds of commits? Sounds weird... Could you try "make distclean" and "make realclean" and then compile previously-broken-commit?
Didn't help. No change.
After trying bisecting mesa 7.9 and finally returning to the official snapshot, the problem vanished at the planet. It is still visible at some nebulas, but I am not sure it is the same problem. Both seem like a problem with some depth (distance from screen) determination.
The broken driver problem seems to be Bug 28284 according to the error message. I probably returned the tree into that time range.
Please try r600g, classic driver is no longer actively supported
I have not seen it since the report so let's close it for now.
And I am on R600g as it gets usable and the default.