You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@httpd.apache.org by Doug MacEachern <do...@opengroup.org> on 1997/07/25 22:01:55 UTC

Re: child_exit?

Dean Gaudet <dg...@arctic.org> wrote:

> 
> 
> On Fri, 25 Jul 1997, Doug MacEachern wrote:
> 
> > Now that we have the new child_init hook, I think a child_exit hook
> > is in order to give modules a chance to cleanup, etc., instead of just
> > exit'ing in child_main() once we hit max_request_per_child, etc.  I
> > can't recall if this has been discussed before, a simple patch is
> > below for you to consider.  If there are +1s I'll add the NULL slot to
> > the rest of the modules. 
> 
> 
> You should mask USR1 and HUP during child_exit_modules.  

Right, will this do?

#ifdef SIGHUP
    signal (SIGHUP, SIG_IGN);
#endif 
#ifdef SIGUSR1
    signal (SIGUSR1, SIG_IGN);
#endif 

> Documentation
> should include a note that there's absolutely no guarantee that the exit
> function will be called.

I'll add something.

> 
> Did you miss just_die() ? 

yes, let me know if you see others.

> I hope the pconf pool is sane at that point.  I'm worried that a USR1/HUP
> could come in while it's in the middle of a block_alarm() critical
> section...

hmm, not sure about that either.

-Doug