Message-Id: <199604010324.VAA12431@sierra.zyzzyva.com>
To: new-httpd@hyperreal.com
Subject: Re: SIGPIPE and timeout on Solaris 
In-reply-to: rst's message of Sun, 31 Mar 1996 13:01:37 -0500.
         <199603311801.NAA05686@volterra.ai.mit.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Sun, 31 Mar 1996 21:24:49 -0600
From: Randy Terbush <randy@zyzzyva.com>
Sender: owner-new-httpd@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

> One thing I discovered when playing with my threaded server is that 
> there is a potential for an infinite loop in the timeout() routine in
> http_main.c --- it does not block SIGPIPE, so when it tries to flush
> the connection before closing it, I found at least my threaded version
> getting (the threaded equivalent of) another SIGPIPE, leading to infinite
> recursion.  In the threaded case, with the stack in the data segment,
> this led to a quick core dump, but on other systems it might last quite
> a while.  
> 
> The fix (I think) was to put the moral equivalent of signal (SIGPIPE, SIG_IGN)
> inside the timeout() routine.  At any rate, it may be worth trying.
> 
> rst

Adding 'signal (SIGPIPE, SIG_IGN)' to the top of timeout() seems to
have fixed this. Thanks Rob!

Some others ought to review this before we commit it, but it seems
that it works.