Message-Id: <199609231535.KAA09755@sierra.zyzzyva.com>
To: new-httpd@hyperreal.com
Subject: Re: bad effect of opening the error log late 
In-reply-to: brian's message of Sun, 22 Sep 1996 22:19:13 -0700.
         <Pine.SGI.3.93.960922221832.5136a-100000@fully.organic.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Mon, 23 Sep 1996 10:35:18 -0500
From: Randy Terbush <randy@zyzzyva.com>
Sender: owner-new-httpd@apache.org
Precedence: bulk
Reply-To: new-httpd@hyperreal.com

> On Thu, 12 Sep 1996, Mark Brown wrote:
> > On some unixes, fopen(path, "a") seeks to the end-of-file
> > rather than opening the file in O_APPEND mode.  SunOS
> > seems to have this affliction.
> > 
> > This doesn't work if another process opens the same file
> > for append, as opposed to sharing the file descriptor.
> > 
> > Because mod_fastcgi starts its process manager process from a
> > module init proc, and the error log isn't open at the time
> > that module init procs are called, mod_fastcgi is being bitten by
> > this problem.
> > 
> > Either the error log should be opened before the module
> > init procs are called, or pfopen should work around the problem by
> > doing the correct open (with O_APPEND) and then calling fdopen.
> 
> Didn't see any comment on this - the error log should be opened as early as
> possible, in my opinion, so I'd go with the first solution.
> 
> 	Brian

Last I knew, the error_log fd was being initialized to stderr, so
it should not fail. Do I misunderstand the problem?