GNAT: internationalization of error diagnosis messages

[Joseph S. Myers]

On Wed, 23 Apr 2003, Vadim Godunko wrote:

> Because it can't be extracted with "standard" utilities. Because 
> russian-speeking ada community want have localized messages only for 
> GNAT not for all GCC.

But can you speak for all the communities using different human languages
and using GCC as a compiler collection for multiple programming languages,
including Ada?  This is an i18n patch that needs to allow for more general
needs of GCC users and maintainers, not a l10n patch for one user group
only.

> > Does (e)xgettext know how to cope with .ad[bs] files? 
> > 
> No.
> 
> I extract messages with ASIS based application.

It is hard enough anyway to get the procedures for handling translations 
<http://gcc.gnu.org/translation.html> regularly followed, and for people 
submitting patches to make GCC's messages friendlier to i18n to get those 
patches reviewed and installed.  Adding additional mechanisms (new domain, 
new system for extracting messages) will just make matters worse.  While 
you could update translation.html to indicate the new procedures 
maintainers must follow to handle GNAT translations, in general we want to 
simplify the procedures that must be followed manually (e.g. automating 
more in the release script) rather than complicating them.  So I recommend 
designing i18n for GNAT to use the existing procedures:

* Teach gettext to understand Ada (it already does many languages), and
have a new release of gettext made which includes this support (then
update the documentation and configure to require that version of gettext,
and do anything necessary with xgettext).

* Use the existing textdomain and Translation Project machinery.  (If 
there should be different translators working on Ada, that's an internal 
matter for the language teams.)

-- 
Joseph S. Myers
jsm28@cam.ac.uk