> On Mon, Dec 15, 2003 at 09:04:35PM +0100, Erik Huelsmann wrote:
> > I'm afraid I don't understand what you try to say with the issue.
> Talking to
> > Greg Hudson did not help, since he doesn't either. The problem is that
> the
> > word exit does not appear in serve.c at all and the abort function is
> there
> > once, but on a dead code path under normal operation. main.c uses the
> exit
> > function, but does so when it has thread creation problems or socket
> problems;
> > nothing related to protocol errors. Could you elaborate a bit, so that I
> could
> > have a look and try to fix this?
> >
>
> accept() has certain non-fatal errors; EMFILE and ENFILE in particular.
> It would be a shame if a threaded svnserve tore down all the threads just
> because it couldn't allocate another fd. Even a non-threaded daemon
> mode svnserve shouldn't exit; it should just log the error and continue;
> perhaps more fds will be available in the future.
>
> Brane may have been talking about different cases, though.
I didn't know there were non-critical errors there, but this is what I was
talking about: the threaded, forked and non-threaded versions of svnserve use
exit in the main listener loop. This may be a problem, but is not related to
protocol errors. That's why I wasn't understanding the issue. Brane
explicitly was talking about users hitting ^C during some operation thus causing a
protocol error and svnserve being torn down completely.
I agree that the use of exit() in the listener-loop might be something to
look at, but won't tear down svnserve at users hitting ^C.
Unfortunately I still don't understand what the issue is about. Thanks
though.
bye,
Erik.
--
+++ GMX - die erste Adresse für Mail, Message, More +++
Neu: Preissenkung für MMS und FreeMMS! http://www.gmx.net
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Dec 15 23:57:18 2003