On Thu, Jul 07, 2005 at 12:24:51PM +0200, Stepan Kasal wrote: > Hello Paul, > I cannot agree with this change: > > On Wed, Jul 06, 2005 at 10:02:56AM -0700, Paul Eggert wrote: > > * tests/local.at (AT_CONFIG_CMP): Ignore lines like "LIBS=''" too. > > They are generated by the Tru64 v5.1B shell. > ... > > --- tests/local.at 24 Jun 2005 00:47:20 -0000 1.18 > > +++ tests/local.at 6 Jul 2005 16:51:30 -0000 > > @@ -331,6 +331,7 @@ m4_define([AT_CONFIG_CMP], > > do > > $SED '/^ac_cv_/ b > > /^m4_defn([m4_re_word])=./ !d > > + /^[[^=]]*='\''\''$/ d > > /^a[[cs]]_/ d > > /^PPID=/ d > > /^RANDOM=/ d > > First, you made a typo, so you are in fact deleting variables with values > starting with a dollar. > Second, the intention to delete all lines matching /^[^=]*=''$/ should be > documented in the comment above the macro. > But I think that we cannot ignore these lines. They mean that the variable > is set to an empty string, right? > > For now, I have reverted that change in the CVS. > > So the most important question is why LIBS was empty after the first run > of configure -C, while it was unset after the second run?
For the first run, without cache, LIBS is empty because macros try to save its value to be restored later. For the second run, the same macros use the already cached values without checking again, and without setting LIBS at the end ... > Nicolas, could you investigate this? The problem is that Tru64 ksh `set' subcommand seems to report LIBS in a diffrent way other shells do : OSF1 medusa.sis.pasteur.fr V5.1 2650 alpha : [EMAIL PROTECTED] [~]> /bin/ksh $ set | grep LIBS $ LIBS= $ set | grep LIBS LIBS='' NetBSD lanfeust.sis.pasteur.fr 3.99.7 amd64 : [EMAIL PROTECTED] [~]> /bin/ksh $ set | grep LIBS $ LIBS= $ set | grep LIBS LIBS= -- Nicolas Joly Biological Software and Databanks. Institut Pasteur, Paris.
