Re: RFA: start of i18n

[Tom Tromey <tromey at redhat dot com>]

Andrew> I suspect GDB is going to have some fun with ui-out.  Some
Andrew> messages are not contigious strings but rather a concatenation
Andrew> of several strings.

Yeah, that could hurt.

Andrew> Should the primary catalogue even be kept in CVS (at least on
Andrew> the trunk)?

It depends.  It is in bfd (etc) because, back in the day, I put
gettext into maintainer-tools (a Cygnus internal thing for those who
don't know), and the policy was that you didn't need maintainer-tools
to do an ordinary build.  That decision lives on in the rest of the
src tree.  Doing it that way is definitely easiest.  It doesn't seem
notably difficult.

Andrew> Hmm, how do things handle a situtation where the primary
Andrew> catalogue gets changed but not the translated catalogs?

You just ignore it.  When you upload a new primary catalog to the
translation site, the translation teams get email.  Then they use
their tools (which are also part of gettext) to merge the new
translations.  I think these tools also help you find out which
translations changed just a little (e.g. if a typo is fixed in English
it requires a minor change in the translated version) or were removed.

gettext always falls back to whatever is in the source (English) if it
can't find a given translation.  So shipping an incomplete .po file is
not fatal; this happens all the time in fact.

This is why Gnome has a string freeze before release.  It lets the
active translation teams ensure 100% coverage, by having some time for
the upload/translate/download cycle of translation.  Whether gdb will
need this is hard to say in advance.

Andrew> Given that GDB is interactive it is likely going to see more
Andrew> string changes then say BINUTILS or GCC.

gcc error messages seem to churn constantly :-).
But, yeah, I'd expect even more in gdb land.


Anyway, back to the patch: it is pretty short and I think
uncontroversial.  It seems unlikely to break the build, given that
very similar code is already littered throughout the src tree.
Could someone approve it in its entirety?

Tom