Fwd: Re: Remove annoying fatal_error() after wineserver crash

[Malte Starostik]

Seems I'm too used to mailing lists which set Reply-to to the list...

Am Wednesday 10 October 2001 01:15 schrieben Sie:
> Malte Starostik  writes:
> > not sure about the reason for that "the socket exists, but I can't
> > connect" message, but IMHO it's pretty annoying. I hit this fatal_exit()
> > very often while debugging reaktivate. And in such cases - or when
> > starting an app from a desktop / startmenu icon - when wine's STDERR goes
> > to ~/.xsession-errors instead of a shell, the user has no chance to
> > figure why an app won't start. Well, long story, short patch which Works
> > for Me (TM) but I don't know if there's anything wrong with doing this...
>
> The problem with this is that it creates a race condition and we may
> end up with two servers running in the same directory, which is bad.
>
> The error message should only happen when a previous server has
> crashed, which is a critical error (the server should never crash no
> matter what the app does). In that case you should also have a core
> file in the same directory as the socket, which you can use to debug
> and fix the crash.

Thanks for the hint, I'll try to reproduce it with an appropriate ulimit set.
But right, an actual wineserver crash didn't happen in a while. Unfortunately
I have some system instability currently, so many times the message was
caused by a stale socket lying around after a solid lockup of my box.
What tends to happen rather often though is that wineserver keeps running
after I quit all windows/winelib apps and consumes 100% CPU. In some of such
cases, it'll only die on SIGKILL, which also leaves a socket around.
I agree that neither a system lockup nor a hanging wineserver should happen
though :)
Maybe there could be a wineserver.pid file in the same dir and only if the
process identified therein doesn't exist anymore, the socket will be
autoremoved?
But after all it's not too big a problem, guess there are more important
issues.
--
Malte Starostik
PGP: 1024D/D2F3C787 [C138 2121 FAF3 410A 1C2A  27CD 5431 7745 D2F3 C787]

[Andreas Mohr]

On Wed, Oct 10, 2001 at 03:40:56AM +0200, Malte Starostik wrote:
> Maybe there could be a wineserver.pid file in the same dir and only if the
> process identified therein doesn't exist anymore, the socket will be
> autoremoved?
> But after all it's not too big a problem, guess there are more important
> issues.

Guess what ?
This is exactly what I implemented (wineserver.pid).
It got rejected, though.

Just to make sure you know about it...