Re: (no subject)

[Owen Taylor <otaylor redhat com>]

On Thu, 2003-05-22 at 08:44, Malcolm Tredinnick wrote:
> On Thu, May 22, 2003 at 08:28:59AM -0400, Owen Taylor wrote:
> > On Thu, 2003-05-22 at 08:07, Malcolm Tredinnick wrote
> > 
> > > [Note for those who may be concerned: not having completely up-to-date
> > > po files in tarballs is not a problem. The generated MO files used the
> > > original strings to look up the translations, not the line numbers in
> > > the comments of the PO file or anything like that.]
> > 
> > 
> > Does 'make update-po' use the line numbers?
> 
> Only as might happen in the call to msgmerge.
> 
> > It seems that at least potentially that the results of:
> > 
> >  Change code  ; msgmerge 
> >  Change code  ; msgmerge
> >  Change code  ; msgmerge
> > 
> > Might be different than:
> > 
> >  Change code ; change code ; change code ; msgmerge
> 
> I am not sure I understand what you are getting at here.  You may have
> raised a perfectly good insurmountable hurdle, but I think I am too
> stupid to understand it. :-(
> 
> My understanding is that possibly the results could be different, but
> does it matter? The point I was trying to make above (and I may have
> overstated the applicability of it) is that when the MO files are
> generated, none of the line number or filename comments matter.  The MO
> format stores each distinct string in a sorted table and looks them up
> via a binary search (at least in the GNU implementation).
> 
> What problem are you trying to avoid here, Owen?

Say, there was a GTK+ .po file for some language xy.po that hasn't
been updated since 1998. My question is whether the xy.po that
has had update-po run on it on every 'make dist' that I've done
since 1998 is going to be more "in sync" with current GTK+, then
what you'd get by running msgmerge on the 1998-vintage .po file
and the current gtk2.pot.

Regards,
                                          Owen