On Jun 17, 2001, Martin Dickopp <[EMAIL PROTECTED]> wrote:
> I made this assumption without checking it, because the current
> (1.4-p4) missing makes the same assumption. It contains the line
> files=`sed -n 's/^[ ]*A[CM]_CONFIG_HEADER(\([^)]*\)).*/\1/p' configure.in`
> in the autoheader case.
Autoheader is indeed always run with PWD = top_builddir.
> Looking closer at the script, I notice that it also seems to assume
> $PWD = $top_srcdir = $top_builddir.
I doubt it. I've certainly used it with srcdir != builddir.
It's likely that assumptions change depending on which program is
`missing'.
Anyway, I've just realized that your change affects only aclocal,
autoconf, autoheader and automake emulations, and all of these are
always run off top_srcdir, so the change is safe, even though it sets
a precedent that might lead others to conclude missing is always run
off top_srcdir (since the test for configure.ac is always done). I'd
appreciate a comment before the test explaining why it is safe to do
the test there even though missing isn't always guaranteed to run in
the directory containing configure.{in,ac}.
--
Alexandre Oliva Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer aoliva@{cygnus.com, redhat.com}
CS PhD student at IC-Unicamp oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist *Please* write to mailing lists, not to me
