Re: [PATCH v2 09/14] i18n: notes: mark error messages for translation

["Jean-Noël AVILA" <jn.avila@free.fr>]

On mardi 13 septembre 2016 16:35:05 CEST Vasco Almeida wrote:
> A Seg, 12-09-2016 às 14:23 +0200, Jean-Noël Avila escreveu:
> > Not sure this one will be easy to localize. The verb is passed as a
> > parameter : see line 366 "list", line 426 "add", line 517 "copy",
> > line
> > 658 "show", line 816 "merge", line 908 "remove" or line 595 with
> > argv[0].
> > 
> > If all the verbs are real subcommands, then the translators should be
> > warned that this is some english twisting, but that they need to
> > refer
> > to the subcommand on the command line.
> 
> Yes, these verbs are git notes subcommands. I will add a Translators
> comment to it explaining so. Or we can unfold that error messages like
> 
> if (!strcmp(subcommand, "add")
> 	die(_("Refusing to add notes in %s (outside of refs/notes/)"),
> ref);
> elseif ...
> 
> else
> 	die(_("Refusing to %s notes in %s (outside of refs/notes/)"),
> subcommand, ref);

This would be counter productive to use the inject strings as keys just to 
test them just after.

> 
> This is more verbose but translations would benefit from it being more
> natural. What do we prefer: (1) concise source and a little unnatural
> translations or (2) verbose code and natural translations?
> 
> Compare, imaging that English is a target translation language, the
> user would read:
> "Refusing to do add of notes in /path [...]" (1)
> "Refusing do add notes in /path [...]" (2)

Having one sentence per action is cumbersome, but avoiding sentence lego is 
mandatory for proper i18n. How about  just adding quotes around the subcommand 
and warn translators ?