Message-ID: <369CB1CA.7B2C40A5@oranda.demon.co.uk> Date: Wed, 13 Jan 1999 14:46:34 +0000 From: "Gary V. Vaughan" Organization: Aethos Communication Systems ltd. X-Mailer: Mozilla 4.5 [en] (WinNT; I) X-Accept-Language: en MIME-Version: 1.0 To: Eli Zaretskii CC: Alexandre Oliva , Andris Pavenis , "Mark E." , autoconf AT gnu DOT org, djgpp-workers AT delorie DOT com, Libtool List Subject: Re: patch for autoheader.sh References: Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Reply-To: djgpp-workers AT delorie DOT com Eli Zaretskii wrote: > > On Wed, 13 Jan 1999, Gary V. Vaughan wrote: > > > > If /dev/null is verboten, maybe what Andris suggested with 1>&2 > > > is better? > > > > If it works in that particular case, then that is good, and should > > probably be done. However, as Alexandre pointed out, use of > > /dev/null in autoconf/automake/libtool/*.m4 is all-pervasive and if > > we try to change it, it will kill us all (and our families). > > I didn't suggest to change the existing uses of /dev/null (DJGPP > supports it). I suggested to use 1>&2 instead for the patch we are > discussing, and for it alone. Sorry... yes I agree with you (1>&- is not supported properly by ash, which is used as /bin/sh on older cygwins and early stampede linux). BUT, the fact that this patch is needed at all, probably means that the auto* tools should be audited to see whether the same problem is likely to occur at any of the other (many, many) points that it is used. Or else we might get other similar hard to find problems in the future. Alexandre, is this worth a note in the PORTing section of libtool? Or maybe we can come up with a test that searches for cases which might have the same problem? I might be persuaded to audit libtool (particularly if it prevents me from being killed =)O|), though there would be little point if the other tools are not checked. Cheers, Gary.