Re: how portable is fchdir?

[Christopher Faylor]

On Mon, Mar 03, 2003 at 07:52:52PM -0800, Paul Eggert wrote:
>Christopher Faylor <address@hidden> writes:
>
>> >It's broken in Cygwin, at least for 1.3.6-6.  See:
>> 
>> Quoting a year-old message hardly qualifies as "broken on cygwin".
>
>A one-year-old bug report is recent by Autoconf standards.  We are
>still worried about porting to systems that are a decade or more old.
>
>> >None of these systems are exactly spring chickens, except for Cygwin.
>> 
>> and the version of cygwin that you are referring to isn't exactly a
>> spring chicken either.
>
>It is quite a bit younger than the other systems I mentioned.  The
>first mention that I see of it on deja.com is 2001-12-20.  By
>contrast, SunOS 4.1.4's first mention is 1993-02-26, AIX 3.2.5's first
>mention is 1993-02-17, and IRIX 5.3's first mention is 1994-01-16.
>By the standards of these other systems, Cygwin 1.3.6-6 is indeed a
>spring chicken.

As long as we're going by standards, then, by cygwin's 1.3.6-6 is
ancient.  That was 14+ revisions ago.  If you are going to base autoconf
rules on the fact that something was broken in a cygwin release at
some point then you might as well disqualify cygwin as a platform
entirely.  IMO, different rules should apply to cygwin.

I understand that you can't rely on fchdir anyway, but please don't
use cygwin as an excuse for why this is so unless you want to be
completely accommodating to all cygwin versions from B16.0 to 1.3.21.
I don't want that and I doubt that anyone wants that.

cgf