Re: Using NLS in the regression checker

[Philipp Thomas <pthomas at suse dot de>]

* Geoff Keating (geoffk@cygnus.com) [20001024 01:06]:

> I've added GNU gettext to the regression checker's build environment.
> It'd be good if someone could glance over the output, at
> <http://www.cygnus.com/~geoffk/gcc-regression/lastbuild.txt>, and
> check that it really is testing whatever should be tested; all I get
> is:
> 
> make[2]: Entering directory `/sloth/delay/tbox/build-gcc/gcc/po'
> /sloth/delay/tbox/boot/bin/msgfmt -o en_GB.mo /sloth/delay/tbox/cvs-gcc/egcs/gcc/po/en_GB.po
> make[2]: Leaving directory `/sloth/delay/tbox/build-gcc/gcc/po'
> 
> which may or may not be right.


Damn it! I just realized that this isn't enough :( With gettext installed,
only the translations will be compiled. Currently the *only* way to get the
checking would be to use the CVS version of gettext and then configure with
--enable-maintainer-mode as then the master catalog gcc.pot will be rebuilt
and *this* depends on a correct POTFILES.in.

This is starting to be a mess :( Just to reiterate: what is needed is a
target for checking POTFILES.in. As this list contains source files from all
languages, the check can only be done if a complete source tree is
present.

So what I'd need is way to check if indeed a complete source tree is
present. Hmm, I just got the idea that checking for the existance of
<language>/Makefile should suffice as a test, wouldn't it?

Philipp

-- 
Philipp Thomas <pthomas@suse.de>
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