> >>>>> INADA Naoki <email address hidden> writes:
> 
>     >> 46      + # NOTE: If cmd_gettext translates ':Usage:\n', the section
> will
>     >> 47      + # be shown after "Description" section.
>     >> 48      + doc = cmd_gettext(doc)
>     >>
>     >> /me blinks
>     >>
>     >> I'm sure this will be eaiser to understand if you added a
>     >> corresponding test.
> 
>     > Is the test for garbage input and garbage output really needed?
> 
> later you have:
> 
> 221     + def filter(p):
> 222     + # ':Usage:' has special meaning in help topics.
> 223     + # This is usage example of command and should not be translated.
> 224     + if p.splitlines()[0] == ':Usage:':
> 225     + return True
> 
> Reading that, I think I understand. The later is needed to prepare the
> file for translators while the former is making sure we don't try to
> translate when we display the command help. Is that right ?

Yes, I want to avoid translating ':Usage:' because it is used as a keyword
in Command.get_help_text().

> 
>     > I've modified export-pot commands to don't exports paragraph
>     > starting with ":Usage:". I think this is not beautiful but enough.
> 
> Indeed, now it makes sense, sorry for being a bit obtuse here :)
> 
> So I think what I wanted is:
> 
>     >> 46      + # NOTE: If cmd_gettext translates ':Usage:\n', the section
> will
>     >> 47      + # be shown after "Description" section.
> 
> # be shown after "Description" section and we don't want to translate the
> usage string.

OK, I'll fix the comment.

> 
> But we still want to translate 'Usage' itself right ? Like we want to
> translate 'Description', 'Tips', 'See Also' and so on. Do we reuse these
> specifics strings from somewhere or are they part of the strings our
> translators will process ?
> 
> (Also see below for other related points)

Yes, get_help_text() removes ":Usage:" in the docstring and write separately.

>            result += gettext(':Usage:\n%s') % (usage,) + '\n'


>     >> 135     - sys.stdout.write(self.get_help_text())
>     >> 136     + self._setup_outf()
>     >> 137     + self.outf.write(self.get_help_text())
>     >> 138     return 0
>     >> 139     if 'usage' in opts: # e.g. bzr add --usage
>     >> 140     sys.stdout.write(self.get_help_text(verbose=False))
>     >>
>     >> /me blinks
>     >>
>     >> Nice to address this bug, but why on earth is it needed ? This
>     >> should really be addressed far earlier no ? Can you find where or
>     >> do you need help here ?
> 
>     > It was not problem before because get_help_text() returns bytestring
>     > containing only ascii characters.
>     > Since i18n is enabled, get_help_text() returns unicode though `outf`
>     > is needed.
> 
> Sorry, badly phrased. I understand why we need to use outf, I don't
> understand why we need to call set_outf (or rather why it has not been
> already called).
> 
>     >>
>     >> Also, it looks like the exact same bug affect 'usage'. I realize
>     >> this may be out of scope for this proposal, so may be we want a
>     >> different proposal that, just let me know if you want to do it or
>     >> leave it to me.
> 
>     > Ah, I've forgot to address --usage. I'll fix it.
> 
> Right, so same fix twice, hence my worries about why we need to fix that
> as it seems to reveal a deeper issue...
> 

There are no path that doesn't call self._setup_outf(). I'll call it earlier.