bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bug-gnulib] utime.c patch for the !HAVE_UTIMES_NULL case


From: Paul Eggert
Subject: Re: [Bug-gnulib] utime.c patch for the !HAVE_UTIMES_NULL case
Date: 09 Sep 2003 10:39:52 -0700
User-agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.3

Bruno Haible <address@hidden> writes:

> Paul Eggert wrote:
> > While testing some other changes I noticed that utime.c won't build at
> > all in the !HAVE_UTIMES_NULL case.
> 
> Indeed. Which platforms are/were those that have the !HAVE_UTIMES_NULL flaw?

Traditionally, BSD had the problem; I'm fairly sure the problem
persisted through 4.3BSD.

I did some software archaeology by looking at Usenet.  It appears that
that the problem was still present in SunOS 3, and that Sun mostly
fixed it by SunOS 4 but it still had some problems with remote
filesystems.

Here are the details.  Guy Harris (then of Sun) reported in
<http://groups.google.com/groups?selm=28784%40sun.uucp> (1987-09-22)
that SunOS 3.2 did not support utimes (x, NULL), and that they planned
to add support (presumably to SunOS 4) but there were some limitations
in using it over NFS.  Dan Heller (then of Sun) wrote in 1990 that
utime(x, NULL) still wasn't supported over RFS (see
<http://groups.google.com/groups?selm=87539%40uunet.UU.NET>), but I
don't think we need to worry about RFS any more.

So this one is a borderline issue.  I suspect that it was mostly fixed
in SunOS 4, but jm_FUNC_UTIMES_NULL will still report failure on a
SunOS 4 host if the source files are accessed via an NFS server that
does not support the "special hack" mentioned by Guy Harris.

Perhaps Jim Meyering can cite other examples of the problem.


> A look into coreutils/old/fileutils/ChangeLog indicates that it might be a
> SunOS 4 issue? But we still support SunOS 4 when used with gcc, right?

Yes, that's the intention.  For example, the 'error' module patch
<http://mail.gnu.org/archive/html/bug-gnulib/2003-09/msg00041.html>
assumes vprintf, which wouldn't be a valid assumption if we merely
assumed freestanding C89, but Usenet archaeology says SunOS 4 had
vprintf so it should be OK.

I'm flying blind to some extent here, as I no longer have a SunOS 4.x
host to test with.  Twin Sun decommissioned its last SunOS 4.x host in
August 2000, and nobody I know at UCLA has one.




reply via email to

[Prev in Thread] Current Thread [Next in Thread]