[I'm not at all convinced that CVS should contain kludges for systems
as horribly broken as this.  I mean, we already have one such for AIX,
but I'm not sure the existence of one justifies spreading the infection.

-kingdon]

Date: Mon, 28 Apr 1997 13:07:49 +0200 (METDST)
From: Scott BERG <Scott.Berg@cern.ch>
To: bug-cvs@gnu.org
Subject: Re: cvs: rmdir call sets errno to bad value under afs

On Sun, 27 Apr 1997, Jim Kingdon wrote:

> > It may be better to find where AFS buries #define's for errno values
> > for the AFS-patched
> 
> I don't know what I think of this idea, but I will say this:
> 
> I do not like the idea of putting AFSCVS ifdefs back in.  We just got
> rid of the old ones a few months ago.

Well, the two operating systems I have afs for (hp/ux and aix) both
give an errno of 93 when attempting to do a rmdir on a non-empty
directory.  On another system running an afs-to-nfs converter, it
correctly gives EEXIST.

Since I don't have any kind of system level documentation for what AFS
does, I don't know if there's any kind of guarantee on that value of
93.  It may even be different on some other operating systems that I
haven't been able to check.  Doing a quick grep on all the afs include
files I could find has failed to yield any kind of straightforward
symbol definition.

The easiest fix seemed to be just to not bother checking errno, and
assume the rmdir failed because the directory was non-empty, but it
may not be the most robust.

Thanks

				-Scott Berg

--
J. Scott Berg                      email: Scott.Berg@cern.ch
Work Phone: (41-22-)767-94-67      Real mail: CERN; SL Division; Bldg 30-2002
                                              CH-1211 Geneva 23; Switzerland