GNOME Bugzilla – Bug 352520
blue screen & tfp errors on aiglx
Last modified: 2008-01-20 19:03:20 UTC
Please describe the problem: if I enable the compositor I'm getting a blue screen similar to https://bugs.freedesktop.org/show_bug.cgi?id=5999 and following error output: Log level 16: No texture created - maybe Texture From Pixmap extension is not present? however tfp must be present in some way, since compiz works on this configuration. (compiz --indirect-rendering) metacity 2.15.34 xorg 7.1 xf86-video-ati git20060812 radeon9800 Steps to reproduce: 1. 2. 3. Actual results: Expected results: Does this happen every time? Other information:
I can confirm it. my setup: xserver 1.1.1 (with aiglx fixes from fedora), r300 dri Mesa HEAD, xorg-driver-video-ati 6.6.2. metacity is built against libcm-0.0.22. On this setup compiz (on aiglx!) runs without big problems. Enabling compositor gconf key causes exacly same problems as Pavel described.
small update: I tried todays cvs of metacity and libcm. the error message above was created by libcm as it is gone now. The problem also looks different now: after enabeling compositor I can still see the windows however neither them nor the desktop gets repainted, meaning if I drag a window I can only see a trail of its shadows. The window content is not being moved (only overpainted by the shadows) so Id say theres something wrong with the actual compositing
Same issue with xorg-server 1.1.1, nvidia beta drivers, metacity 2.16.1 and CVS libcm from this day. I got several similar reports.
Me too! Compiz with aiglx/tfp/cow works fine but metacity gives blue windows. --- xorg-xserver 7.1.0 nvidia 9625 beta drivers metacity CVS today libcm CVS today
here the same, with i810 OSS drivers
Same problem here: xorg-xserver 7.1.0 nvidia 9625 beta drivers metacity 2.16.2 libcm CVS today
I have the same problem (blue windows). Configuration: xorg 7.1.0 NVIDIA 9625 Beta metacity CVS (2006-10-02) libcm CVS (2006-10-02)
Same here (Gentoo ~x86 on Athlon 64 X2): XOrg 7.1 Nvidia 9652 Beta Metacity 2.16.2 libcm 0.0.22 from 2006-07-14
I'm seeing this, too: xorg-server-1.1.1 nVidia 9625 beta Metacity 2.16.3 libcm from 2006-09-29
Look's like problems in config script
Got exactly the same problem on Gentoo ~amd64 with * X.org 7.1 with nVidia 1.0-9625 beta drivers * metacity-2.16.2 * libcm CVS HEAD
Same here, when I start beryl with beryl manager, I got this problem. Nvidia 9625 driver xorg 1.1.1 and if I start beryl with bery-start, seems good, but beryl-start doesn't work on KDE.
a response of the gnome-developers would be very nice
There is a thread on the desktop mailing list about compiz-beryl-metacity and the 3D compositor. Some would like to shoot metacity and use compiz and/or beryl instead. Others would like to tidy metacity in order to support cleanly the different possible combinaisons with a plugin architecture: bare X11/compositor/compositor+3D. Till their is no decision on that, I think this bug will persist, unless somebody else than an "official" gnome dev fixes it.
Comment 12 makes this sound like an AIGLX or GLX problem. At any rate, despite being a metacity developer, I know virtually nothing about the metacity/compiz/beryl situation; I really have no idea what the plans of others are and I haven't personally worked on the compositing stuff. What I know was basically what I saw on mailing lists (in some cases, having someone else point me to them). Wish I could help, but I can't, and I don't know how closely Soeren is following the metacity-compositor bugs. Anyway, I just thought I'd chime in since so many are having this problem. *shrug*
hitting the same issue. all windows appear blue ( reminds me of an xv video window ) using fc6.i386 NVIDIA-Linux-x86-1.0-9625-pkg1.run libcm head metacity 2.6.17
After recompiling libcm *after* installing nvidia 9625 and thus the GLX TFP extension, the »maybe TFP extension not present« debug message disappears. Windows still blue. Compiz is fine. I'd volunteer to provide some more information as I know C and gdb well, but I don't know where to start here.
Pull down CVS metacity, CVS libcm and you are good to go. There is a document in metacity doc to guide you in metacity dev.
Created attachment 74680 [details] log of a GDB session with libcm's ./scenetest I found nothing interesting in metacity's verbose output, but the scenetest included with libcm fails to work for me. I'm on a 64-bit system if that might matter.
I am not sure this is even a metacity bug opposed to an libcm bug. I cant run all the test on a amd64 machine with Xorg 7.1 and 9xxx series nvidia drivers.
According to Sören, this is fixed in CVS HEAD for libcm and metacity, can anyone verify this?
No, it still doesn't work for me. Building libcm with the Mesa headers still shows the blue windows. Building it with the nvidia headers didn't work out of the box, only after defining GLX_FONT_LEFT_EXT to 0x20DE in the beginning of wspixmap.c. But this also shows blue windows. What I didn't try yet is using Mesa from CVS. I'm on Gentoo though, so I'm by no means a reference. ;)
Built it freshly from anonymous CVS, blue windows again. However, it compiled out of the box with nvidia headers. libcm/src $ ./scenetest depth: 24 ** ERROR **: Unexpected X error: BadValue (integer parameter out of range for operation) serial 46 error_code 2 request_code 156 minor_code 3) aborting... Is this something that isn't implemented by the nvidia drivers yet?
This is a bug I am also experiencing, and judging by the number of people, perhaps this bug should be marked unconfermed --> new.
BSOD...
Can someone update this to reflect devel, and not 2.15.x series . and just curious, why is this unconfirmed ?
(In reply to comment #26) > Can someone update this to reflect devel, and not 2.15.x series . I've given perms to your account to do such things in the future (please take a look at the bugsquad triage guide at http://live.gnome.org/Bugsquad/TriageGuide before making changes to bugs in other products). > and just curious, why is this unconfirmed ? Because unconfirmed vs. confirmed is completely ignored for metacity bugs. I never query on it or sort according to it, and I doubt anyone else does. (I tend to get annoyed if bugsquadders mark metacity bugs confirmed and do nothing else, because all it does for me is generate some bugzilla spam. Not that I have much room to complain, having been a bugsquadder that generated thousands of emails for other maintainers in the past...)
I have the same problem. Screenshot from another user shows the same screen I am getting (blue, instead of white or black): http://img355.imageshack.us/my.php?image=compizbugmz2.png I am running nvidia's binary driver 1.0.9639; same behavior with latest 9755 driver. metacity 2.19.5 libcm 0.1.1 xorg.conf: ServerLayout section enables AIGLX nvidia Device section enables all of the following: AddARGBGLXVisuals, RenderAccel, AllowGLXWithComposite, backingstore, TripleBuffer, DamageEvents The latter three were enabled attempting to fix the problem. Kernel version 2.6.20 (amd64 arch), with gentoo-sources patches. This bug has been open for the better part of a year. Will this be fixed any time soon?
I have the same issue. Able to help with debugging.
*** Bug 363845 has been marked as a duplicate of this bug. ***
Bluescreen issue was a problem with X extensions that Søren's compositor used, but Iain's compositor does not. -> OBSOLETE If anyone sees it with current Metacity, please do reopen.