JDK-5033205 : PIT: OGL with Java2Demo and Radeon not displaying application demos properly
  • Type: Bug
  • Component: client-libs
  • Sub-Component: 2d
  • Affected Version: 5.0,6
  • Priority: P4
  • Status: Closed
  • Resolution: Cannot Reproduce
  • OS: linux_sun,windows_xp
  • CPU: x86
  • Submitted: 2004-04-16
  • Updated: 2007-01-12
  • Resolved: 2007-01-12
Related Reports
Duplicate :  
JDK-6256584 - PIT: OGL: Red, Blue, Yellow colors are swapped when drawing to offscreen, linux
Relates :  
JDK-5066318 - OGL: Enabling OGL pipeline on Windows with ATI board crashes VM at exit/shutdown
Relates :  
JDK-6251468 - OGL: GradientPaint slow on Nvidia GF2-series hardware
Description
Running Java2Demo with OGL pipeline enabled on WinXP and Radeon 9800XT Graphics card is not displaying application demos properly.

- The individual demo panels are drawn in very stretched and unusual sizes.  The demos then become very distorted but seem to still be running as you can see animation going on (no exceptions are thrown).

- This problem occurs right at launch even with the credits screen.

- If the main application window is reduced in size it the problem goes away and the demos are redrawn at proper size and scale.  Increasing the size of the main application window makes no difference though (the bug still persists).

- Bug was found in 2D Tiger Nightly PIT Build:
"2d-jcg-win-01-2004-04-14.tiger"

- Previously this bug was not observed due to blocking bug 5026756
- These errors are not seen when running with NVIDIA drivers (or when running without OGL pipline).
Comments
EVALUATION The image scaling issue seems to be specific to an older series of boards (e.g. Radeon 9500), as it is not reproducible on a Radeon 9800XT with Catalyst 6.4 or 7.1 drivers. And now that -Dsun.java2d.opengl.fbobject=true is the default in JDK 6 (see 6439320), the problem with scrollbars and other gradients is no longer visible on modern ATI boards (Radeon 9500 and later), although it is still intermittently visible on older boards where FBO is not available. This is clearly an ATI driver/hardware problem in their pbuffer support, but unfortunately ATI no longer supports hardware older than Radeon 9500, so pursuing a fix with them is futile (as it has been for the past three or so years). The issues mentioned later in this bug report have differed greatly from the ones in the original description; some have been worked around, and others would require fixes from ATI. At this point, I think it would be best to close this bug report (as "not reproducible" for lack of a better option) since there's not much else we can do from the JDK side. If necessary, more specific bugs can be filed later.
12-01-2007
WORK AROUND The issue with artifacts in vertical scrollbars can be worked around by enabling our FBO codepath: -Dsun.java2d.opengl=[T,t]rue -Dsun.java2d.opengl.fbobject=true Once 6439320 is fixed, this workaround should no longer be necessary, as we will automatically choose the FBO codepath by default at startup.
19-06-2006
EVALUATION The issue with artifacts in vertical scrollbars is due to a bug in the pbuffer support in their drivers. This issue can be worked around by enabling our FBO codepath instead, see workaround section for more details. The remaining issue with images not being scaled properly has no workaround at this time. We will continue to push the issue with ATI.
19-06-2006
EVALUATION The black GradientPaint issue was resolved on our side in 6251468, so that is no longer an issue. The red/blue swap issue was fixed in the ATI Catalyst 5.12 timeframe, so that is also no longer an issue. There are still a couple outstanding ATI driver bugs that are noticeable on WinXP with the JFC demo apps (tested most recently with Radeon 9500 Pro and Catalyst 6.2): - images are not scaled properly (see Colors tab in Java2Demo) - scrollbars sometimes contain garbage (see SwingSet2); this bug is not reproducible with -Dsun.java2d.opengl.fbobject=true I have filed bug reports with ATI for the above issues and they are being investigated. I will continue to keep this bug report open until all those rendering glitches are worked out.
27-02-2006
EVALUATION As mentioned in the evaluation of 5026756, there are still a number of outstanding problems in ATI's OGL drivers for Windows. The problem appears to be related to pbuffers and clipping. There is no obvious workaround. We will have to create smaller Java/native testcases that reproduce the problem and then will contact ATI for a solution. Hopefully by the time Tiger ships, they will have some fixes available in their Catalyst drivers. ###@###.### 2004-04-16 The OGL-based pipeline should be more stable on ATI boards these days due to the recent integration of the single-threaded rendering changes (see 6219284). There are just a couple issues remaining in ATI's drivers that cause rendering problems in Swing applications. (A related issue is the crash-on-exit bug, see 5066318.) The visual problems can be boiled down into: - GradientPaint is rendered incorrectly (only black is rendered, instead of the gradient colors); a bug report has been filed with ATI (see Ticket# 737-767994) - Sometimes images are rendered such that their red and blue components appear to be swapped; a bug report has been filed with ATI (see Ticket# 737-768050) Both of these issues can also be seen on ATI's proprietary Linux drivers. ###@###.### 2005-03-10 00:59:07 GMT There has been no update from ATI on the above issues. However, it appears that we can workaround the GradientPaint issue with a slight tweak in our code (see 6251468). I'll leave this bug open until all the open issues are resolved. ###@###.### 2005-04-06 18:19:10 GMT
06-04-2005
WORK AROUND
11-06-2004
SUGGESTED FIX
11-06-2004
PUBLIC COMMENTS
10-06-2004