Bug 233690

I have the same problem on my i915 and i945-based machines. Other systems appear
to be unaffected.

All gdb tells me is that Xorg gets a SIGPIPE and then decides to die.

There's no obvious errors in Xorg.0.log - the problem happens shortly after
AIGLX gives its 'Loaded and initialized /usr/lib/dri/i915_dri.so' message. No
SELinux audit messages are found. I'm fairly well stumped on how to debug this
further.

SIGPIPE is just the signal we get when clients disconnect.  it's completely normal.

The way to determine if this is X's fault is to boot to runlevel 3 and run 'X
:0' from the console by itself.  if X comes up, then it's probably not a bug in
X, but rather, some client being started as part of the X session is failing,
and therefore the startup script just aborts.

Attaching /var/log/Xorg.0.log would help, but it's likely to just tell us that
the server starts correctly and it's actually the session that fails.

"X :0" in runlevel 3 works OK, so I'm also thinking it's the session. I'm still
puzzled as to why this only happens on intel/i810 chipset machines, though.

It doesn't seem to be directly tied to the new version of the i810 driver. I
found a mirro that had yet to be updated and havne now installed
xorg-x11-drv-i810-1.6.5-13.fc7 and the error still occurs.

I also tried booting with the oldest kernel I have (2.6.20-1.2987) to see if
that made a difference, but it did not.

I installed some update yesterday and today I booted, logged in and installed
more updates. So logically I guess it has to be an update from today that did it
- however I can't find a likely suspect among the following (output from rpm -qa
--last):

xorg-x11-drv-i810-1.6.5-13.fc7                Fri 23 Mar 2007 09:08:34 PM CET
gocr-0.44-2.fc7                               Fri 23 Mar 2007 05:52:18 PM CET
NetworkManager-vpnc-0.6.4-3.fc7               Fri 23 Mar 2007 05:52:03 PM CET
elinks-0.11.1-5.1                             Fri 23 Mar 2007 03:55:30 PM CET
anaconda-11.2.0.40-1                          Fri 23 Mar 2007 02:43:27 PM CET
redhat-artwork-5.0.12-1.fc7                   Fri 23 Mar 2007 02:43:12 PM CET
fedora-logos-6.0.97-1.fc7                     Fri 23 Mar 2007 02:42:12 PM CET
kernel-2.6.20-1.3016.fc7                      Fri 23 Mar 2007 02:41:50 PM CET
kernel-headers-2.6.20-1.3016.fc7              Fri 23 Mar 2007 02:41:38 PM CET
fedora-release-notes-6.92-3                   Fri 23 Mar 2007 02:41:25 PM CET
yum-updatesd-3.1.5-1.fc7                      Fri 23 Mar 2007 02:41:23 PM CET
selinux-policy-targeted-2.5.9-5.fc7           Fri 23 Mar 2007 02:41:14 PM CET
selinux-policy-2.5.9-5.fc7                    Fri 23 Mar 2007 02:41:09 PM CET
yum-3.1.5-1.fc7                               Fri 23 Mar 2007 02:41:04 PM CET
inkscape-0.45.1-1.fc7                         Thu 22 Mar 2007 06:31:50 PM CET
firstboot-1.4.34-1.fc7                        Thu 22 Mar 2007 05:04:30 PM CET
fedora-release-6.92-1                         Thu 22 Mar 2007 05:04:29 PM CET
kernel-2.6.20-1.3003.fc7                      Thu 22 Mar 2007 05:03:59 PM CET
firstboot-tui-1.4.34-1.fc7                    Thu 22 Mar 2007 05:03:51 PM CET

I have also tried adding a new session to gdm which only runs xterm - but the
login process dies just as fast with the same error.

Ah! This is apparently actually a problem with anaconda interfering with the X
server session startup. It should be fixed in tomorrow's rawhide. See the
following mail:

https://www.redhat.com/archives/fedora-test-list/2007-March/msg00476.html