Re: out-of-mem handler

[Havoc Pennington <hp redhat com>]

Tim Janik <timj@gtk.org> writes:
> that's just a random game, you try to put some questionable handling
> for *some* case where a rare error could take effect. the point is,
> you either need to do this for libc as well or not do anything at
> all.

That isn't true at all. If you can avoid 50% of crashes that's much
better than avoiding 0% of crashes.

> being held during your malloc sleep etc., there are probably a lot
> of else situations where you'd rather bail than sleep.
> 

gnome-session bailing is catastrophic; sleep() can't possibly cause
worse results. If gnome-session bails your whole session implodes.

It is totally bogus to say that because you can't recover 100% of the
time you shouldn't try to recover. It's just not true. If users avoid
50% of session crashes then users are happier.

Yes if you can think of a way to recover 100% of the time that might
be nice, but I can't think of a way.

Havoc