Re: [PATCH v2] l10n: reformat some localized strings for v2.23.0

[Junio C Hamano <gitster@pobox.com>]

Jeff King <peff@peff.net> writes:

>>  		OPT_STRING('s', "source", &opts.from_treeish, "<tree-ish>",
>> -			   N_("where the checkout from")),
>> +			   N_("where the checkout is from")),
>
> I think your original "where to checkout from" is better.

Would we even want to do s/where/which tree-ish/?

> translators, but note that the output will be different. The original
> would be something like:
>
>   warning: Fetch normally indicates...
>   warning: To re-enable...
>
> where now we'd get:
>
>   warning: Fetch normally indicates...
>   check has been disabled...
>   or run 'git config...
>
> which might be a bit harder to read because the wrapped lines lose the
> prefix. For advise() we nicely pick out the newlines and prefix each
> line individually, but warning(), error(), etc, don't do that. Maybe
> they should.

Yeah, I'd be surprised that nobody thought of doing that, so perhaps
somebody tried and failed with a possible fallout.  I do not offhand
see any downside of teaching them to do the prefixing.

For existing multi-line warnings that uses a single call to
warning(), I think the preparer of the message manually indents the
second and subsequent line by a run of SPs to match the screen width
of "warning:" prefix (and expect translators to do the same with
their language), and we need to get rid of that kind of "hack" when
we insert a middle layer between *_builtin() and vreportf() to do
the line chomping to produce output similar to advise() code.

> That's too big for this late in the -rc cycle, I think.

Agreed.
Thanks.