Hi everybody,

Like I said earlier, I've personally been using Precise with Unity on one of my main test machines since the alpha stage, and haven't seen any X crashes that weren't due to bugs that have since been fixed, and I certainly haven't seen random, frequent X crashes like what seems to be described here. That machine has its drivers and GPUs swapped frequently, and for as long as I have been using Precise, I have always used 295 or later drivers on it. (Compiz does seem to crash frequently, but it always recovers, without disturbing any of the applications or windows I have open.)

In addition to my own personal experience, which is obviously a tiny sample, the driver goes through extensive automated and manual testing on many different hardware configurations, and on a variety of distributions, including Ubuntu 12.04. We haven't seen any issues that cause X to crash sporadically, as seems to be described here.

With that in mind, we're obviously missing something important that's required to reproduce this bug (if it is even just a single bug: it's hard to tell, since the symptom is fairly generic). If we can collect as much detail as possible from everybody who is affected, that would help us understand what is happening better. The more information you provide, the better our chances for reproducing the issue at NVIDIA. Please provide as many of the following as possible:

1) An nvidia-bug-report.log file, generated immediately after a crash. This can be generated by running the script `nvidia-bug-report.log.sh`, which is included as part of the driver install. The script requires superuser privileges to collect system information, so you should run it with sudo. The script will save a file called nvidia-bug-report.log.gz to your current working directory.
2) A detailed description of what you were doing at the time of the crash.
3) If you are familiar with using gdb, stack traces from the crashing server, with X debug symbols installed, would be very helpful. If you're not familiar with gdb, but want to try to provide a stack trace anyway, you can try following these steps:
    a) Install X debug symbols (sudo apt-get install xserver-xorg-core-dbg)
    b) Start your desktop session, as normal
    c) For good measure, capture the DSO load base of the NVIDIA driver, in case it shows up in the stack trace of the crashing server. (grep nvidia_drv.so /proc/`pidof X`/maps) - you may need to do this as root. Stash this information away, and if nvidia_drv.so shows up in your backtrace when you get one, submit it along with everything else.
    d) Log into your system *FROM ANOTHER MACHINE* and attach a debugger to the already running X server (gdb Xorg `pidof X`): it's important that you do this from another machine over ssh, because after the debugger is attached, X will be frozen until you continue. If you start the debugger from a VT on the same machine, when X crashes later, it will be frozen, and you probably won't be able to switch to your VT running the debugger
    e) After the debugger is attached, type "continue" (or just "c") at the gdb prompt and hit enter
    f) Use your system, leaving the debugger attached the whole time. If X crashes, the debugger will stop. Go to your machine running the debugger, and type "backtrace" (or just "bt") and hit enter to get a backtrace at the time of your crash. If the backtrace is long, you may have to hit enter a few more times to see the rest of it. Copy and paste this backtrace, and submit it along with the nvidia-bug-report.log.

It does sound like some boards are affected, and some are not, though the boards reported so far don't have a whole lot in common with each other. We will do our best to match the configurations reported here, in hopes of reproducing this crash. Thanks in advance for your help. In the meantime, for those who are affected, it does sound like unity-2d is more stable, so perhaps you can switch to unity-2d as a temporary workaround while waiting for this issue to be resolved.

Alberto, has anybody at Canonical been able to reproduce this?