Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!auspex!guy
From: guy@auspex.auspex.com (Guy Harris)
Newsgroups: comp.sys.apollo
Subject: Re: Please mail and post specific changes for GCC on Apollo
Message-ID: <1733@auspex.auspex.com>
Date: 2 Jun 89 21:00:09 GMT
References: <1318@novavax.UUCP>
Reply-To: guy@auspex.auspex.com (Guy Harris)
Organization: Auspex Systems, Santa Clara
Lines: 23

>What needs to be done with includes of <sys/file.h>?

By and large, what needs to be done with includes like that in programs
that don't poke around and read the AT&T-derived "struct file" data
structures in the kernel (I would not be surprised to find that no such
data structures exist on Apollo systems) is to completely nuke them.

The problem is that 4.2BSD picked up "fcntl", the three-argument "open",
and various "open"/"fcntl" flags from System III (O_RDONLY, O_WRONLY,
O_RDWR, O_APPEND, O_TRUNC, O_CREAT, etc.); however, they documented it
differently, for no good reason.  In S3/S5, you include <fcntl.h> or
possibly <sys/fcntl.h> to get those flags; this works perfectly well
under BSD, but they decided to tell you to include <sys/file.h> and use
some other internal names for those flags.

POSIX specifies that you get them from <fcntl.h>, so Berkeley'll have to
clean up their documentation act at some point.

I suspect the <sys/file.h> stuff is just that (I have difficulty
imagining why GCC would have to look at internal kernel data structures
- especially since it's supposed to work under VMS!), so you should
apply the appropriate changes and then fold those fixes back into the
*mainline* GCC, so nobody has to do them all over again.