Thanks dinosaur_. It would be helpful if you could provide a core file from the crash along with nvidia-bug-report.log.gz, or a backtrace from gdb along with the output of the gdb commands “info registers” and "info sharedlibrary.
Here is the gdb session log for a crash I just got with a third party viewer using the very same rendering engine as Linden Lab’s official viewer (i.e. I would have gotten the same crash with Linden Lab’s viewer, but I used this TPV because because that’s the only one I compile myself and could therefore have the (262Mb large) debug symbols file for: Linden Lab doesn’t provide the debug symbols file for their release binaries which can’t therefore be debugged efficiently with gdb).
I alas cannot provide you with a core dump for this crash, because I got core dumps disabled in my Linux kernel (will enable them for the next kernel release so to provide you with one).
In the gdb log, note that the values for mResX and mResY are correct (they indeed correspond to the actual size of the viewer window), so the glCopyTexSubImage2D() function that crashed during the desktop switching was apparently called with valid parameters.
gdb-session.txt (9.68 KB)
nvidia-bug-report.log.gz (62.1 KB)
Any news on this front ?
I almost thought that v331.13 (beta) got it fixed (no crash in over one week running this driver), but I just got one crash (see the attachments)…
nvidia-bug-report.log.gz (75.9 KB)
stack_trace.log (709 Bytes)
Hi dinosaur_. The bug has been reproduced here and is being investigated.
That’s really great news ! Thanks a bunch !
I just wanted to let you know that since it has been out, I have been using v331.20, and I didn’t yet encounter a single crash with it ! :-) It could still be pure luck, but I’m starting to hope that a proper fix finally made its way in this driver version…