[ the gnu list server seems down ATM; expect delays ] Hi Stefano,
* Stefano Lattarini wrote on Wed, Jul 28, 2010 at 11:07:20AM CEST: > At Wednesday 28 July 2010, Ralf Wildenhues wrote: > > But even then, when developing Automake you will need to be extra > > careful to never let autoconf be invoked from the rebuild rules; > > because then, it will run plain "autoconf" and not the $AUTOCONF > > that you set. > Why? AUTOCONF is AC_SUBST by configure, so if I do: > AUTOCONF=autoconf-2.67 ./boostrap > AUTOCONF=autoconf-2.67 ./configure (please pass variables as arguments to configure, so config.status --recheck remembers them.) > there should be no problem. Am I missing something? You're right. Only, that AUTOM4TE should be set, too, so that aclocal.tmp uses the right one. Maybe it's prudent to set that here, too. > (BTW, if my understanding is correct, this would make a nice addition > to HACKING, too). Yep. > > Changing that > > from an absolute file name to a non-absolute one is a bad idea, it > > breaks several test setups I have, > How? The $PERL in bootstrap should be only substituted in the shebang > lines of automake.tmp and aclocal.tmp, and both of these scripts are > removed just after being run once by boostrap itself. Overlooked by me. > The amended patches are attached, with a new follow-up patch doing > some fixlets for bootstrap. Thanks, OK for maint with $AUTOM4TE handling folded in as above. Cheers, Ralf > From bf4e7ecea7feae34056b2e4382a2bef94642cb07 Mon Sep 17 00:00:00 2001 > From: Stefano Lattarini <stefano.lattar...@gmail.com> > Date: Tue, 27 Jul 2010 22:44:54 +0200 > Subject: [PATCH 1/3] Bootstrap: let the user choose which autoconf to use. > > * bootstrap ($AUTOCONF): New variable, from the environment. > Use it instead of calling "autoconf" directly. [...]
