From: "Cliff Woolley" <[EMAIL PROTECTED]>
Sent: Thursday, August 02, 2001 1:51 PM
> > wrowe 01/08/01 12:15:22
> >
> > Modified: server core.c
> > Log:
> > This patch fixes more untold breakage than you can shake a stick at.
>
> APR_BRIGADE_FOREACH(e,b) provides a perfectly usable e as long as you
> don't MODIFY e within the loop, because the foreach gets to the next
> bucket in the brigade using APR_BUCKET_NEXT(e). I haven't looked yet
> at that loop (but I will)... was e getting modified? If so, then
> yeah, that would cause loads of funky problems.
>
> Are there other planned changes to the threaded MPM or to the MPM API (as
> in the current ap_graceful_stop_signalled() discussion)?
No, we were treating 'e' as legit AFTER we had dropped out of the APR_BRIGADE_FOREACH,
this was very broken.
