Comment 31 for bug 91849

Hi Sergio, thanks for the backtrace; looks like it also involves compPaintWindowBackground. It would be helpful if you installed -dbg versions of xserver and your video driver, to try and get additional details on the crash.

Yes, I dropped -intel as an involved package because reports have indicated that the same problem affects other drivers beyond -intel (S3 for instance). Also, from the backtraces we aren't yet seeing a failure in driver code. It may very well end up being a driver issue, but we don't yet have solid enough evidence of that yet.

The following is at the end of the xfce-terminal strace:

connect(4, {sa_family=AF_FILE, path="/tmp/.X11-unix/X0"}, 19) = 0
...
write(4, "\24\0\6\0\1\0@\2\354\0\0\0\4\0\0\0\0\0\0\0@B\17\0", 24) = 24
read(4, "\34P\236\1 \0@\2\364\0\0\0]\274K\326\0\0\0\0\2\0\0\0\0"..., 32) = 32
read(4, "\34P\236\1 \0@\2_\1\0\0]\274K\326\0\1\0\0\2\0\0\0\0\0\0"..., 32) = 32
read(4, "\34P\236\1 \0@\2a\1\0\0]\274K\326\0\1\0\0\2\0\0\0\0\0\0"..., 32) = 32
read(4, "\34P\236\1 \0@\2\356\0\0\0]\274K\326\0\0\0\0\2\0\0\0\0"..., 32) = 32
read(4, "\25\0\236\1 \0@\2 \0@\2\24\r\340\0\6\0\33\0\0?\t\10\200"..., 32) = 32
read(4, "\34P\236\1E\0\0\0\37\1\0\0_\274K\326\0\1\0\0\2\0\0\0\0"..., 32) = 32
read(4, "\34P\236\1E\0\0\0Z\1\0\0_\274K\326\0\1\0\0\2\0\0\0\0\0"..., 32) = 32
read(4, "\34P\236\1 \0@\2F\1\0\0_\274K\326\0\1\0\0\2\0\0\0\0\0\0"..., 32) = 32
read(4, "\26\0\236\1 \0@\2 \0@\2\"\r\340\0\6\0\33\0\341\2\312\1"..., 32) = 32
read(4, "\226\216\236\1 \0@\2 \0@\2\24\r\340\0\26\1+\1\341\2\312"..., 32) = 32
read(4, 0xbfcadbfc, 32) = -1 EAGAIN (Resource temporarily unavailable)
poll([{fd=4, events=POLLIN, revents=POLLIN|POLLERR|POLLHUP}], 1, -1) = 1
read(4, 0xbfcadbfc, 32) = -1 ECONNRESET (Connection reset by peer)
write(2, "xfce4-terminal: Fatal IO error 1"..., 96) = 96
close(7) = 0
--- SIGCHLD (Child exited) @ 0 (0) ---
sigreturn() = ? (mask now [])
kill(4959, SIGTERM) = 0
exit_group(1) = ?

I notice that first write() statement is writing 24 bits... I wonder if that relates at all to the 16/24 DefaultDepth distinction being noticed with this bug.

But anyway, the connection to X is dying after doing some IO with it. Unfortunately the backtraces so far have not pinpointed where in the server this IO is being received, so it's hard to identify why it is causing a crash. So we'll definitely need to have someone install -dbg versions and run gdb against this to get a backtrace (and maybe step through code to identify what lines are involved.)