I get black stripes on the Moon and Mars with render path : OpenGL vertex program (this does not happen with the 2 other render path : basic and multitexture or with other planets)
See images:
http://godard.b.free.fr/echange/celestia/mars.jpg
http://godard.b.free.fr/echange/celestia/moon.jpg
System :
Mandriva 2007, linux kernel 2.6.17 SMP
AMD Athlon 64x2
Nvidia GeForce 7600GT + driver 1.0-9631
Xorg 7.1.1 with AIGLX
Celestia 1.4.1 (from autopackage)
Moon + Mars texture bug with OpenGL vertex program [SOLVED]
Moon + Mars texture bug with OpenGL vertex program [SOLVED]
Last edited by astro on 16.02.2007, 22:39, edited 1 time in total.
Thanks for your reply Selden.
I just upgraded to 1.0.9746 and it didn't change a thing.
If it is a driver issue, I can post bug reports on the NVIDIA forum. However, how can you be so certain the bug is with the display driver? I don't have similar bugs with other games/applications. The fact that the bugs don't appear on cards other than the 7600 GT doesn't necessarily mean the bug is with the display driver.
I just upgraded to 1.0.9746 and it didn't change a thing.
If it is a driver issue, I can post bug reports on the NVIDIA forum. However, how can you be so certain the bug is with the display driver? I don't have similar bugs with other games/applications. The fact that the bugs don't appear on cards other than the 7600 GT doesn't necessarily mean the bug is with the display driver.
- t00fri
- Developer
- Posts: 8772
- Joined: 29.03.2002
- Age: 22
- With us: 22 years 7 months
- Location: Hamburg, Germany
astro wrote:Thanks for your reply Selden.
I just upgraded to 1.0.9746 and it didn't change a thing.
If it is a driver issue, I can post bug reports on the NVIDIA forum. However, how can you be so certain the bug is with the display driver? I don't have similar bugs with other games/applications. The fact that the bugs don't appear on cards other than the 7600 GT doesn't necessarily mean the bug is with the display driver.
Apart from the fact that we are talking here mainly about Celestia 1.5.0 which has MANY improvements & fixes, indeed, I have never observed any of your strange bugs in 1.4.1 , despite using similar NVIDIA cards and always the latest drivers. I use BOTH Linux and Windows and compile the code almost daily...
It then rather looks like a bug that perhaps has been "patched" into one of your graphics libraries by your Linux distributor. I have encountered such annoying "features" quite a few times during the many years of working with Linux.
You may provide cel://urls of your various bug observations, so other users/devs may check the configs easily.
Bye Fridger
Celestia's v2 Render Path is designed to make use of many of the latest features defined in OpenGL, so it stresses many of the most recent code additions to the drivers. The primary author of Celestia is essentially the only person developing its OpenGL code, so he has to depend on the drivers doing the right things. Game producers have many people working on the code and can take the time to find and write workarounds for underlying bugs.
Also, you need to make sure your system actually is using the Nvidia driver. Too often people have reported slowness and other problems, later to discover that their system was not using the driver they thought it was using.
Unfortunately, I'm not a Linux expert, so I can't help in detail. However, you might try the non-proprietary software-only drivers that don't use hardware acceleration to see what difference they make.
Also, you need to make sure your system actually is using the Nvidia driver. Too often people have reported slowness and other problems, later to discover that their system was not using the driver they thought it was using.
Unfortunately, I'm not a Linux expert, so I can't help in detail. However, you might try the non-proprietary software-only drivers that don't use hardware acceleration to see what difference they make.
Selden
t00fri, you are probably right. I just tested celestia 1.4.1 (both the autopackage from the celestia website and the gentoo ebuild) on Sabayon Linux x86 3.25 and I don't get these bugs on same hardware.
kernel 2.6.19-gentoo-r4 #1 SMP
Xorg 7.1.99.903 + nvidia 1.0-9631
I will probably report the bug on the Mandriva forums.
Thanks
kernel 2.6.19-gentoo-r4 #1 SMP
Xorg 7.1.99.903 + nvidia 1.0-9631
I will probably report the bug on the Mandriva forums.
Thanks
- t00fri
- Developer
- Posts: 8772
- Joined: 29.03.2002
- Age: 22
- With us: 22 years 7 months
- Location: Hamburg, Germany
astro wrote:t00fri, you are probably right. I just tested celestia 1.4.1 (both the autopackage from the celestia website and the gentoo ebuild) on Sabayon Linux x86 3.25 and I don't get these bugs on same hardware.
kernel 2.6.19-gentoo-r4 #1 SMP
Xorg 7.1.99.903 + nvidia 1.0-9631
I will probably report the bug on the Mandriva forums.
Thanks
I am using Linux from SuSE since many years. Right now I run the latest 10.1 OpenSuse version and don't see any spectacular bugs in Celestia.
Bye Fridger
After discussing this on mandriva's plf-discuss mailing list, the problem is that celestia uses a hardcoded library path.
chrpath /usr/bin/celestia
/usr/bin/celestia: RPATH=${ORIGIN}/../lib:${ORIGIN}/../lib/autopackage
Mandriva's nvidia package from the PLF source install nvidia's libGL in /usr/lib/nvidia and the mesa libGL is still present in /usr/lib but the loader ld is configured so that /usr/lib/nvidia is searched before /usr/lib.
However with its hardcoded library path, celestia uses /usr/lib/libGL.so (the MESA lib)
Thus the problem is probably in Celestia's Makefiles.
Contrary to what I said before, the problem doesn't appear with Mandriva Celestia 1.4.1 RPM package (they have removed the RPATH for that package)
https://www.zarb.org/pipermail/plf-disc ... 04356.html
chrpath /usr/bin/celestia
/usr/bin/celestia: RPATH=${ORIGIN}/../lib:${ORIGIN}/../lib/autopackage
Mandriva's nvidia package from the PLF source install nvidia's libGL in /usr/lib/nvidia and the mesa libGL is still present in /usr/lib but the loader ld is configured so that /usr/lib/nvidia is searched before /usr/lib.
However with its hardcoded library path, celestia uses /usr/lib/libGL.so (the MESA lib)
Thus the problem is probably in Celestia's Makefiles.
Contrary to what I said before, the problem doesn't appear with Mandriva Celestia 1.4.1 RPM package (they have removed the RPATH for that package)
https://www.zarb.org/pipermail/plf-disc ... 04356.html
Thank you, astro. I had this problem pending for a long time. Last time a handcrafted installation of Mesa solved it in an older system but I didn't feel like diving in the Mess again. After reading your message I've compiled today's CVS two times and "./configure --with-kde --disable-rpath" made the difference: the stripes disappeared and all five render paths are back again without touching anything else. Neat.