httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Doug MacEachern <>
Subject Re: mod_perl + threaded Perl
Date Fri, 20 Mar 1998 02:36:14 GMT
Dean Gaudet wrote:

> Until linuxthreads stops using SIGUSR1, no.  i.e. not in the linux 2.0/2.1
> time frames.  I think even the glibc threads are broken this way too...
> and I consider it to be a library/kernel bug rather than apache bug
> because SIGUSR1 is supposed to be there for the application.
> Incidentally given the general signal brokenness w.r.t. linux clone() and
> threads I'm not sure how well Apache, especially 1.3 will behave.  Apache
> assumes that signals work as they do in single threaded posix processes.
> As soon as you can end up with signal delivery going to another tid... I
> dunno what'll happen/go wrong.

bummer, I was afraid of this.  In the meantime, for testing/exploration, can we
consider a -DNO_SIGUSR1 to disable Apache's catching of SIGUSR1?  Is it needed for
anything other than graceful restarts?

> So be careful.  But do tell how things go!  'cause we're going to be there
> in the thick of it soon enough as we try to write a multithreaded apache.

 Will do!


View raw message