[Aptitude-devel] I18n questions in aptitude development tree

[Daniel Burrows]

On Wed, Feb 27, 2008 at 04:19:51PM +0100, Jens Seidel <jensseidel at users.sf.net> was heard to say:
> On Wed, Feb 27, 2008 at 06:52:43AM -0800, Daniel Burrows wrote:
> > On Wed, Feb 27, 2008 at 10:02:59AM +0100, Sven Joachim <svenjoac at gmx.de> was heard to say:
> > > On 2008-02-27 07:10 +0100, Daniel Burrows wrote:
> > > > On Sun, Feb 24, 2008 at 02:34:29PM +0100, Sven Joachim <svenjoac at gmx.de> was heard to say:
> > > >> 1) There is no Makefile in the po/ directory, so `make' failed there.
> > > >>    How comes that?
> > > 
> > > There is no po/POTFILES.in from which po/POTFILES could be created, why
> > > is it missing?
> > 
> >   I see what happened.  I changed POTFILES.in to an autogenerated file
> > recently so that it would stay in sync with the source code.  But
> > apparently it needs to be present in order for autoconf to generate
> > po/Makefile.in. (not sure why) Just creating an empty file with
> > "touch po/POTFILES.in" seems to be sufficient, and I've modified
> > autogen.sh to do so.
> 
> Hhm, I'm confused. I think I had problems with this file as well but
> don't remember exactly. Probably only the automake update after the last
> "hg update" which affected your POTFILES changes failed so I did
> invoking automake and some Makefile editing manually.

  What probably happened is that you already had a Makefile generated
before I removed POTFILES.in.  So if it did get removed in your local
copy when you fetched changes (I assume Mercurial does this, but I'm
not 100% sure), the Makefile was still present.  When it noticed that
Makefile.in.in had changed, the next attempt to run "make" in that
directory re-ran the autotools, and then when POTFILES.in was needed by
another target, it regenerated it using the Makefile rule.  The only bad
effect here is that "configure" itself won't create po/Makefile if
POTFILES.in is missing (apparently in the belief that it's not a real
po/ directory).

  At least, that seems to be the reason I didn't notice this problem.

> Neverthless it should work out of the box if one checks out a fresh
> source tree. po/POTFILES.in exists in the repository and is not empty!
> At least it has proper content for me and hg diff shows no changes.
> So what happened? I'm confused ...

  I believe that if you do "hg clone . /tmp/test-aptitude-checkout",
you'll find that the new copy doesn't contain a po/POTFILES.in.  It
looks to me like "hg diff" just exits silently if the file you give it
isn't in the repository, as does "hg status".  "hg log <file>" shows
changes to the file but not its removal.

  POTFILES.in was removed in revision 3c3411a6440e.

  Daniel