| Summary: | [830M] rendering errors with UXA (intel 2.7) | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Product: | xorg | Reporter: | ahph1uxsb8rae | ||||||
| Component: | Driver/intel | Assignee: | Carl Worth <cworth> | ||||||
| Status: | RESOLVED FIXED | QA Contact: | Xorg Project Team <xorg-team> | ||||||
| Severity: | normal | ||||||||
| Priority: | medium | ||||||||
| Version: | 7.4 (2008.09) | ||||||||
| Hardware: | x86 (IA32) | ||||||||
| OS: | Linux (All) | ||||||||
| Whiteboard: | |||||||||
| i915 platform: | i915 features: | ||||||||
| Attachments: |
|
||||||||
|
Description
ahph1uxsb8rae
2009-06-14 13:25:49 UTC
Fortunately I had some time to do additional test runs... SUBJECT FONT RENDERING ERRORS WHEN USING UXA ON INTEL INTEGRATED GRAPHICS SYSTEM SUMMARY Chipset: i82830M [2001, Almador] CPU: Intel Celeron 1.13 Ghz Single Core (2.2k BOGOMIPS) RAM: 768MB, no swap Architecture: i686 Kernel: different Distribution: Ubuntu 9.04 (jaunty, Alternate-CD) Machine: Acer Travelmate 223 (Laptop) Display: Internal (DVI) 1024x768 #lspci -nn 00:00.0 Host bridge [0600]: Intel Corporation 82830 830 Chipset Host Bridge [8086:3575] (rev 04) 00:02.0 VGA compatible controller [0300]: Intel Corporation 82830 CGC [Chipset Graphics Controller] [8086:3577] (rev 04) 00:02.1 Display controller [0380]: Intel Corporation 82830 CGC [Chipset Graphics Controller] [8086:3577] 00:1d.0 USB Controller [0c03]: Intel Corporation 82801CA/CAM USB Controller #1 [8086:2482] (rev 02) 00:1d.1 USB Controller [0c03]: Intel Corporation 82801CA/CAM USB Controller #2 [8086:2484] (rev 02) 00:1e.0 PCI bridge [0604]: Intel Corporation 82801 Mobile PCI Bridge [8086:2448] (rev 42) 00:1f.0 ISA bridge [0601]: Intel Corporation 82801CAM ISA Bridge (LPC) [8086:248c] (rev 02) 00:1f.1 IDE interface [0101]: Intel Corporation 82801CAM IDE U100 Controller [8086:248a] (rev 02) 00:1f.3 SMBus [0c05]: Intel Corporation 82801CA/CAM SMBus Controller [8086:2483] (rev 02) 00:1f.5 Multimedia audio controller [0401]: Intel Corporation 82801CA/CAM AC'97 Audio Controller [8086:2485] (rev 02) 00:1f.6 Modem [0703]: Intel Corporation 82801CA/CAM AC'97 Modem Controller [8086:2486] (rev 02) 01:05.0 Ethernet controller [0200]: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ [10ec:8139] (rev 10) 01:09.0 CardBus bridge [0607]: O2 Micro, Inc. OZ601/6912/711E0 CardBus/SmartCardBus Controller [1217:6972] TESTING SCENARIO Firefox with Flashplayer: "Big Buck Bunny" on youtube [http://www.youtube.com/watch?v=PcBECFLXmuc] Totem Video Player: big_buck_bunny_480p_stereo.ogg [http://mirror.bigbuckbunny.de/peach/bigbuckbunny_movies/big_buck_bunny_480p_stereo.ogg] Compiz Test [SKIP_CHECKS=yes compiz --replace] Normal Surfing & Using for at least half an hour UXA forced by xorg.conf. File is immutable that Ubuntu can't play around with. Framebuffers (CONFIG_FB) are blacklisted by default in Ubuntu. Compositing/Compiz is disabled as it's also blacklisted by default. I don't know, if it matters, but xfs (X Font Server) is not installed. Is it deprecated today? Installed a "CLI" version from CD onto separate partition. Very basic debian/ubuntu system. Saved this partition (490MB) to an image with "partimage". This image is restored after any driver upgrade to have a fresh install to start from. ------------------------------------------------------------- - SYSTEM A [SCORE: +] Ubuntu 9.04 with X, gdm and icewm kernel: Linux nb1 2.6.28-11-generic #42-Ubuntu SMP Fri Apr 17 01:57:59 UTC 2009 i686 GNU/Linux xserver-xorg-video-intel: 2.6.3-0ubuntu9 xserver-xorg: 7.4~5ubuntu18 libdrm: 2.4.5-0ubuntu4 libgl1-mesa-dri: 7.4-0ubuntu3.1 BBB in Firefox and Totem are fluent, CPU load 50-90% Compiz skipped, switching to VT /dev/ttyX skipped. One freeze in Firefox on heise.de. I guess it was a flash commercial (normally eaten by adblock ;) Reboot only with SysRq possible. Not a drama at all. No rendering errors, interface fast and responsive. ------------------------------------------------------------- - SYSTEM B [SCORE: -] Ubuntu 9.04 with default GNOME versions as SYSTEM A BBB in Firefox and Totem are fluent, CPU load 50-90% Compiz starts, blanks anything but background image and freezes X before reloading the desktop. Switching to VT /dev/ttyX and back works. Font rendering errors occur fast. First some glyphs then more and more over a few minutes. The "d" is always affected early, "X", "O" too. If you have a serif and a sans-serif font in Firefox, only one "family" is garbled. I played with font antialiasing ("smoothing"). Ubuntu has "None", "Grayscale" and "Subpixel". None: Rendering errors. Glyphs are garbled. Grayscale: The same. But glyphs become black boxes. Subpixel: Like GS. But it takes longer until errors appear. Switching between them does not really help. It seems to recognize its errors on each AA algorithm (pixmap cache?). Changing the resolution itself to i.e. 75dpi helps as smaller font size is needed. Until these are "tainted", too. Overall performace is fine, but slower than SYSTEM A. This time no freezes. ------------------------------------------------------------- - SYSTEM C [SCORE: o] "Safe Configuration" from http://ubuntuforums.org/showthread.php?t=1130582 including changes to xorg.conf, but tiling eq "false" Ubuntu 9.04 with default GNOME kernel: Linux nb1 2.6.28-11-generic #42-Ubuntu SMP Fri Apr 17 01:57:59 UTC 2009 i686 GNU/Linux xserver-xorg-video-intel: 2.7.1-0ubuntu1~xup~1 xserver-xorg: 7.4~5ubuntu18 libdrm: 2.4.9-1ubuntu1~xup~1 libgl1-mesa-dri: 7.4-0ubuntu3.1 BBB in Firefox and Totem are fluent, CPU load a bit higher. The WM gets slow. Compiz starts, blanks anything but background image and freezes X before reloading the desktop. Switching to VT /dev/ttyX and back works. Again, font errors. But not as bad as before. It now begins with normal glyphs having a gray background, instead of transparent. I think it's a wrong alpha value, because links in Firefox have a lightblue background, if the link-color is blue. Later I had different alpha values on glyphs. I have a screenshot of it. Overall performace is better, close to SYSTEM A. This time no freezes. ------------------------------------------------------------- - SYSTEM D [SCORE: o] "Optimal Configuration" from http://ubuntuforums.org/showthread.php?t=1130582 including changes to xorg.conf, but tiling eq "true" and kernel 2.6.30 Ubuntu 9.04 with default GNOME kernel: Linux nb1 2.6.30-020630-generic #020630 SMP Wed Jun 10 09:45:40 UTC 2009 i686 GNU/Linux the rest like SYSTEM C Well, it's like SYSTEM C. With one interesting difference on the font errors. Now the wrong alpha channel is only half the width of a glyph on the left side. Or topside and solid blue. When I wanted to switch to Firefox by Alt+Tab, X froze. A restart by Alt+Print+K resulted in a nice colorful pattern and a black screen afterwards. GDM seems not to be there as I was not able to login. Unfortunately X has no log of that. ------------------------------------------------------------- - SYSTEM D2 [SCORE: o] "Optimal Configuration" from http://ubuntuforums.org/showthread.php?t=1130582 including changes to xorg.conf, but tiling eq "true" and kernel 2.6.30 Ubuntu 9.04 with default KDE4 kernel: Linux nb1 2.6.30-020630-generic #020630 SMP Wed Jun 10 09:45:40 UTC 2009 i686 GNU/Linux the rest like SYSTEM C KDE is affected, too. Here the artifacts are boxes placed inside the glyph. Their color matches the text itself. Title bars tend to have a wrong background image repeating. There is one pixel space between the tiles. Desktop effects work, but rather slow on this.... ancient machine. No freezing. ------------------------------------------------------------- - SYSTEM D3 [SCORE: o] "Optimal Configuration" from http://ubuntuforums.org/showthread.php?t=1130582 including changes to xorg.conf, but tiling eq "true" and kernel 2.6.30 Ubuntu 9.04 with default icewm kernel: Linux nb1 2.6.30-020630-generic #020630 SMP Wed Jun 10 09:45:40 UTC 2009 i686 GNU/Linux the rest like SYSTEM C Well, Firefox and Konqueror (as member of Qt-Party) produced errors again. But on icewm it seems to be less than other WMs. As third browser I installed Opera. It's statically linked against Qt3. Surprise: All fonts were clean during testing. It only became "infected", if the errors were triggered by another process. Opera never did that himself. Coincidence? Maybe. ------------------------------------------------------------- - SYSTEM E [SCORE: -] "Bleeding Edge" from http://ubuntuforums.org/showthread.php?t=1130582 including changes to xorg.conf, but tiling eq "true" and kernel 2.6.30 Ubuntu 9.04 with default GNOME kernel: Linux nb1 2.6.30-020630-generic #020630 SMP Wed Jun 10 09:45:40 UTC 2009 i686 GNU/Linux xserver-xorg-video-intel: 2.7.99.901+git20090615.3da549f5-0ubuntu0sarvatt~jaunty xserver-xorg: 7.4~5ubuntu18 libdrm: 2.4.11+git20090519.f355ad89-0ubuntu0sarvatt~jaunty libgl1-mesa-dri: 7.5.0~git20090615+mesa-7-5-branch.d027e8fe-0ubuntu0sarvatt Pretty much like SYSTEM C, again. Fortunately font errors are very low or at least didn't occur during testing. Sometimes the text in the menus turn gray, but get back to black after some time. But some icons are now "infected". Because it only appears in Firefox, I think it's probably not caused by X alone. A restart of FF helps. Basically it's the "semi-alpha-on-one-side" variant. While I tried to prepare the screenshot using GIMP, X became very unstable and froze continuosly. After four reboots I gave up. I could not isolate a specific reason. It happened on window moving/resizing, choosing a color or tool. XVideo Output in VLC of BBB in a short test caused freezing again. First it worked, then the overlay got blue and then crashed. After rebooting, it worked. After some testing this driver release comes close to performance of SYSTEM A. The "redrawing" speed of windows is somewhat slower. My current "rescue" system with EXA is faster here (although slower in total). Freezes are frequent. Alt+Print+K useless. ------------------------------------------------------------- CONCLUSION UXA obviously has problems on i830M. But new software always have. GTK and Qt are afflicted by font corruptions and/or wrong background repeating. On "native" icewm less errors happened (unless using gtk/qt-apps), although became very unstable and prone to X freezes when using the experimental 2.7.99-git intel driver. 2.7.1 is more stable but unfortunately errors are possible and more or less likely. Best results: SYSTEM A and SYSTEM D[123]. Unfortunately, I cannot find a specific reason for these glitches. Sometimes they appear real fast, next time it takes ten or more minutes. Here it began after window switching, there just during writing this report. Or after scrolling the text in gedit/kate. If a glyph gets "tainted", it remains in this condition. Dirty font/pixmap cache? On 2.7.99-git they sometimes "healed" on their own. A miracle? Only one place was clear all the time: Anything in xterm. Most of the time I wrote in nano... ANNEX I clicked a bit through bugs.launchpad.net and found some with similar experience. Font Smearing: https://bugs.launchpad.net/xserver-xorg-video-intel/+bug/304514 Gradient Banding: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/305909 Fonts Corrupts after a while: https://bugs.launchpad.net/ubuntu/+source/xfonts-base/+bug/326487 Fonts unreadable by the time: https://bugs.launchpad.net/ubuntu/+source/gnome-system-monitor/+bug/378119 Font Drawing Corruption: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/293059 Interesting - Font Corruption *INSIDE* terminal (but ATI graphic): https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-ati/+bug/291040 Suspend-To-RAM (STR) does not work. but as far as I remember it never has on linux here. It woke up, but video card is not reactivated. WinXP did it when I bought the machine. STD untested. I spared EXA testing as it's obsolete now. Thanks for reading :) #EOD Created attachment 26916 [details]
More examples of font glitches on intel-2.7.1
I have several logs and debugging info (dmesg/glxinfo). Should I upload all 13 files separately as attachment (as guide lines say) or pack them into a big text file, zip, tar whatever? Just wanted to ask before in this case... I updated the test partition to Karmic two weeks ago. First it had 2.7.99-git, but now 2.8 of intel drivers as it reached alpha-3 stage. All glitches, rendering errors I mentioned are gone! Hmm, is it okay for a non-developer to set the bug status to "fixed"? (In reply to comment #4) > I updated the test partition to Karmic two weeks ago. First it had 2.7.99-git, > but now 2.8 of intel drivers as it reached alpha-3 stage. > > All glitches, rendering errors I mentioned are gone! Hmm, is it okay for a > non-developer to set the bug status to "fixed"? Thanks for the update. I'm delighted to know that the bug is fixed! -Carl |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.