Re: [Po4a-dev] Re: [Po4a-cvs] po4a/po/pod fr.po po4a-pod.pot

[Denis Barbier]

On Sat, Nov 29, 2003 at 04:56:51PM +0100, Martin Quinson wrote:
> On Thu, Nov 27, 2003 at 08:05:25AM +0100, Denis Barbier wrote:
> > On Wed, Nov 26, 2003 at 06:54:35PM -0500, Barbier wrote:
> > > Modified files:
> > >   po/pod         : fr.po po4a-pod.pot 
> > > 
> > > Log message:
> > >   Update, and add some translated strings.
> > 
> > But some blank lines are missing in the translated POD files.
> 
> This is because the pod parser used as ground for the corresponding po4a
> module is a damned bastard, eating stuff without repporting to the using
> program. 
> 
> So, po4a has no way to see when empty lines were silencely eaten, and I have
> to rely  on tricks to reintroduce those lines afterward. I see 3 solutions:
>  - fix Pod::Parser to let it stop eating blank lines.
>  - do not rely on this module in order to be able to fix it without having
>    to wait 2 years for the fix to go into the debian stable archive
>  - add too much lines. But it will break the tests trying to prove that the
>    module is idempotent, ie po4a-normalize will never output what it did read.
>    
> The third solution may be the better way to go. We could always cheat on the
> tests by giving a file on which there is no pb as input to -normalize, ie a
> file with already too much lines which mimicks the bugs of that kind.
> And of course, test break is better than functionnality break ;)

Savannah ate some commits I performed yesterday.  Here is the changelog
entry I added:
  * Some newlines were missing with the POD format, because trailing
    newlines were removed in Locale::Po4a::Pod, and some (but not all)
    were added in Locale::Po4a::Po.  These changes are removed,
    and po4a-normalize does no more lose newlines on POD files.

You can run 'cvs diff' on these 2 files to see my changes.  The pod format
seems to work pretty well now, but I cannot tell for others.

> Could you please provide me an example of missing blank lines? blank lines
> before a command are recommanded and produce a warning if missing, but
> shouldn't induce error, IIRC.

This request could have been easily dealt with if translated files were
stored in CVS ;)
IIRC erroneous output was shown when there was no blank lines after
=head1 commands, in particular when followed by =over as in
  =head1 OPTIONS
  =over 4

Denis