This is the mail archive of the
mailing list for the GCC project.
Re: Using NLS in the regression checker
* Geoff Keating (firstname.lastname@example.org) [20001021 23:10]:
> [Aargh! Was that me? Sorry!]
Nope, you didn't do something wrong ;-) I possibly should have made that
> Sure. Just change configure.in so that NLS is switched on by default,
> and the regression tester will test it automatically. I'm happy to
> add any tools to the regression tester's environment that it might need.
NLS *is* on by default as long as build == host, i.e. no canadian cross. The
gettext tools need to be installed though, otherwise no catalogs will be
built. If they're present, a build will fail if POTFILES.in wasn't kept up
to date. So no special check target is necessary.
> > Note that you don't need the the gettext tools installed in order to check.
> > Just do a 'make stamp-check-po' in gcc/po and it'll fail if POTFILES.in
> > hasn't been kept up-to-date.
> ... or add 'make stamp-check-po' to 'make check'.
> I don't want to have the regression tester differ greatly from the
> normal 'configure && make && make check' build process, because
> otherwise it'll be testing something different to what most people
That's a very valid point to make. But I can't include that in make check
because POTFILES.in currently lists *all* files and thus an NLS enabled (the
default) build will only succeed if a complete source tree is present.
For changing this I would have to think about a way to split POTFILES.in
into language specific parts (matching the tarballs gcc is available as) and
then combine them at configure time.
But as I wrote, only an installed gettext package is needed for the
regression checker and you said that you had no problems with installing
Philipp Thomas <email@example.com>
Development, SuSE GmbH, Schanzaecker Str. 10, D-90443 Nuremberg, Germany
#define NINODE 50 /* number of in core inodes */
#define NPROC 30 /* max number of processes */
-- Version 7 UNIX for PDP 11, /usr/include/sys/param.h