Date: Fri, 23 Apr 1999 07:27:19 -0400 (EDT)
From: Ryan Bloom <rbb@raleigh.ibm.com>
To: "new-httpd@apache.org" <new-httpd@apache.org>
Subject: Re: cvs commit: apache-apr/apr/file_io/unix filedup.c
In-Reply-To: <199904230327.NAA06015@silk.apana.org.au>
Message-ID: <Pine.A41.4.05.9904230722130.23958-100000@webdev7.raleigh.ibm.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: new-httpd-owner@apache.org
Precedence: bulk
Reply-To: new-httpd@apache.org

> >  Log:
> >  A new file function to allow me to specify a file descriptor to be put into
> >  an apr_file.  I am still trying to figure out if this needs to be abstracted
> >  more, and how to do it.  But, the function is needed for the process stuff,
> >  so I am committing it now.
>    
> >  +apr_status_t apr_pushfile(apr_file_t *, int)
> >  +       Push a platform specific file descriptor into the apr file type.
> >  +     Arguments:
> >  +       arg 1)  the file descriptor to modify.
> >  +       arg 2)  The file descriptor to push.
> >  +       return) APR_SUCCESS or APR_FAILURE
> 
> The problem with that is that a file descriptor's type varies by platform. In
> unix it may be an int but in OS/2 it's an unsigned long and in Win32 it's a
> HANDLE (which is actually a void *). I don't know about BeOS or Mac.
> 
> Maybe if you describe the problem you're trying to solve we can come up with
> a suitable method.

I knew this one wasn't going to fly.  It's okay.  I figured out how to
solve the problem last night, so I will be removing this patch in a little
while.  I was trying to keep the process stuff in apr from knowing about
the file_io stuff in apr, but I am reasonable convinced that isn't going
to work very well.

The basic problem is that I am creating a new process, and I need to be
able to setup pipes to the new process, for stdin, stdout, and stderr.
Those pipes are apr_file_t types, and I needed to dup2 the child's std*
fileno to the apr_file_t, and this patch let me do this with an apr
function.  I don't think I need or want an apr function to do this though.
What I want, is to have the apr_create_process function have access to the
internals of apr_file_t for the platform it was written for.

Does that make sense?

Ryan

_______________________________________________________________________
Ryan Bloom		rbb@raleigh.ibm.com
4205 S Miami Blvd	
RTP, NC 27709		It's a beautiful sight to see good dancers 
			doing simple steps.  It's a painful sight to
			see beginners doing complicated patterns.