Bug#458858: few more hangs

[Brice Goglin <Brice.Goglin@ens-lyon.org>]

Ondrej Certik wrote:
> When X server hangs (for whatever reason), the log can always be
> accessed in the Xorg.0.log.old and it can also possibly
> leave the hardware in a bad state, so that when I restart, the X
> server hangs because of this bad state, right?
>   

Yes for instance some registers in the graphic board could be left in a
inconsistent state after the crash, and launching a new session will try
to understand these registers and fail.

That said, I tend to think that the driver should be able to cleanup
this mess instead of suffering from it.

> So you suggestion is to always to hard reboot (shutdown, wait,
> powerup), right? And then look into Xorg.0.log.old
>   

Well, you can look at the log before rebooting. But, you should avoid
trying to debug using a log from a session that you launched after the
crash of another session without a hard reboot.

If you do:
1) hard reboot
2) start X, crash
3) start X, crash
4) start X, crash
5) hard reboot
6) start X, crash
7) hard reboot
8) start X, crash
9)  start X, crash
...
The log from (3), (4) and (9) may contains errors caused by the crash of
the immediately-previous run, so you should not debug using this log.
But the log from (2), (6) and (8) should be fine.

Brice