Re: [PATCH] rebase: make warning less passive aggressive

Junio C Hamano [off-list ref] writes:

Jean-Noël AVILA [off-list ref] writes:

quoted

As a translator, I'm less bothered by editing a sentence to remove a question 
mark (maybe enforcing a language style and reformulating the sentence by the 
way), than by translating again and again similar sentences.

Sure, but if the original in C locale used to be "FOO BAR?" and you
translated it to "foo bar?" in your language, and then a patch
updates the string in the source to "FOO BAR", doesn't msgmerge
notice that the original as a "fuzzy" matching and offer you
something like

    #, fuzzy
    msgid "FOO BAR"
    msgstr "foo bar?"

so that all you have to do is to remove '?' anyway?  So I do not
think you'd need to translate the "FOO BAR" part again and again.

But the above assumes that for your language, the ONLY thing to turn
such a rhetorical "passive aggressive" question into grammatically
correct statement of a fact is to remove the question mark.  It may
not be universally true for all languages, and for some language,
even after msgmerge did its job correctly, you may need to do more
than just removing the question mark to adjust the remaining "foo
bar" part.

Hopefully the last message on this topic from me.  I just tried this:

 * apply v2 of "No rebase in progress?" -> "no rebase in progress"
   patch from Harmen.  This only changed _("") string, without
   touching any po/* files.

 * follow po/README.md and pretend as if I were a po/es.po team
   member, which involved running the command:

	$ make po-update PO_FILE=po/es.po

 * view the resulting po/es.po, which will be the starting point of
   a new translation.

        $ git diff po/es.po

The last one showed me this snippet:

    -msgid "No rebase in progress?"
    +#: builtin/rebase.c:1257
    +#, fuzzy
    +msgid "no rebase in progress"
     msgstr "¿No hay rebase en progreso?"

The "#, fuzzy" thing is a signal that an old msgid has been updated
and corresponding msgstr may need to be updated.

The primary problem I had with Harmen's v1 was that it touched po/
file in the first place (which was corrected in v2), but the updates
to po files it did looked like so:

    -msgid "No rebase in progress?"
    +msgid "No rebase in progress"
     msgstr "¿No hay rebase en progreso?"

Notice the lack of "#, fuzzy" marking?  If v1 were accepted with
these changes, then the "po-update" step I showed earlier would have
produced something like this instead:

    +#: builtin/rebase.c:1257
     msgid "no rebase in progress"
     msgstr "¿No hay rebase en progreso?"

As msgid alone was updated already, the "po-update" step has no idea
that the translated msgstr is out of date.  We do not get the
helpful "#, fuzzy" comment to tell the translator which msgstr is
out of date and needs attention.

So it is another reason why we want the developers to concentrate
only on writing good C-locale messages inside _("") and N_(""),
without touching po/ directory at all.  It would avoid making
changes that can confuse toolchain l10n people rely on (like the
presence of "fuzzy" comments).