Re: pg_rewind and log messages

Fujii Masao wrote:
> On Mon, Apr 6, 2015 at 5:33 PM, Michael Paquier
> <michael(dot)paquier(at)gmail(dot)com> wrote:
> > On Mon, Apr 6, 2015 at 1:41 PM, Michael Paquier wrote:
> >> I guess that you are working on a patch? If not, you are looking for one?
> >
> > Code-speaking, this gives the patch attached.
>
> Thanks! Here are the review comments:
>
> I'm not familiar with native language support (sorry), but don't we need to
> add the shortcut of gettext into every calls of pg_log and pg_fatal, e.g.,
> change pg_fatal("xxx") to pg_fatal(_("xxx"))? I know that fprintf() in
> pg_Log_v() has such shortcut, but I'm not sure if that's enough or not.

It's not necessary for pg_fatal and the like, because those functions
are marked to have their first argument automatically translated in
nls.mk. This means that the string literal is automatically extracted
into pg_rewind.pot for translators. Of course, the function itself must
call _() (or some variant thereof) to actually fetch the translated
string at run time.

I'm not sure about translation of generic strings such as "%s: %s". My
first impression is that they shouldn't be translated, but maybe it is
important that they are for languages I don't know nothing about such as
Japanese.

Another thing is compound messages like "foo has happened\nSee --help
for usage.\n" and "bar didn't happen\.See --help for usage". In those
cases, the "see --help" part would need to be translated over and over,
so it's best to separate them in phrases to avoid repetitive work for
translators. Not sure how to do this -- maybe something like
_("foo has happened\n") _("See --help")
but I'm not sure how to appease the compiler. Having them in two
separate pg_log() calls (or whatever) was handy for this reason.

--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services