Greg Ames wrote:
>
>
> In the mean time, I'll trim down this patch to the most essential piece
> - writing to the pipe of death - and commit it. Please speak up ASAP if
> you have a problem with this.
>
Well, I have a problem with this.
Just writing to the pipe of death isn't enough apparently. I did that
in the shutdown leg, then noticed that "apachectl restart" didn't work
right. That looked easy - no pipe of death in that section of code - so
I did pipe of death stuff in the restart leg as well, but it was still
flaky.
If you do an "apachectl graceful" first, that seems to work fine (patch
or no steenkin' patch). But then if you follow it up with "apachectl
restart" or "apachectl stop", this one strange process or thread won't
die. "ps ax -O wchan" says it is in "unix_a", whatever that is. It is
the first thing listed under the parent after you start apache. And the
parent bails out on a restart sometimes :-(
My trimmed down patch might be a slight improvement over what is in CVS
at the moment but it is nowhere near the solution our users need. So
I'm not committing it as it stands now.
Ryan or Jeff - what's up with a missing "workers_may_exit=1" on line
732? That's in make_child with the current CVS; no comprende.
Greg
