From Bugzilla Helper: User-Agent: Mozilla/4.76 [en] (X11; U; Linux 2.4.2-0.1.28 i686) I've got an OpenGL application that reads tries to make a texture with the internal format and external format the same (GL_RGB5_A1 and GL_UNSIGNED_SHORT_5_5_5_1 respectively). The result is a garbled texture - it looks like random memory (not just mis-formatted). Using a non-packed format (like RGBA8) image (converting to RGB5_A1) seems to work fine. Reproducible: Always Steps to Reproduce: 1. make a solid color texture array (like 0x5959) 2. make a texture with above params 3. display it Actual Results: As a bonus it sometimes kills X - although I'm guessing this problem will go away once it is rendered correctly.
Which graphics card are you using? Are you using DRI or software rendering? Which application is causing this?
Unfortunately I no longer work for the company where I produced the problem, so that complicates things a bit. The corruption was with Mesa on a NeoMagic chipset - software rendering. The problem was reproducable with OpenGL calls, so the original application should not be needed. A simple test app that creates a 5551 texture (as describe in the original bug report) and draws a polygon using it should exhibit the problem (although it may be fixed in newer versions of Mesa/X).
I was unsure whether to classify this a new bug or not... perhaps someone could direct me whether or not to do so. I am an administrator for an educational research lab... we are having similar problems with X restarting when running our opengl apps. We have also been successful at almost always getting X to restart when closing down a java3d app. We develop numerous opengl applications, and while they are less consistent, one of our developers has an app that restarts X _almost_ every time he runs it. The same behavior does not occur on RedHat 6.2. It _does_ occur on any machines we have tried it on regardless of the hardware. We even have a couple of boxes with NVIDIA cards/drivers on them, and it is reproducible. I have also tried on our Universities open labs with RedHat 7.1 and reproduced the bug (they have different hardware and monitors settings, etc) One thing I tried was upgrading to X 4.1.0 (from rawhide, along with the updated Mesa and glibc 2.2.3). The problem persisted.
Without an attached small standalone test case, I cannot do very much about this. If you attach a small application source that is just enough code to trigger the bug, then it can be looked into more easily. If you do so, please REOPEN the bug report, otherwise, please report the bug upstream to the Mesa development team on http://mesa3d.sourceforge.net