Hello Jean-Philippe,
On Tue, Apr 19, 2022 at 12:41:27AM +0200, Jean-Philippe MENGUAL wrote:
> Le 18/04/2022 à 13:17, Helge Kreutzmann a écrit :
> > Hello fellow manpage translators,
> > I made some observations when watching the French updates which I
> > would like to share with you. Please kindly disregard them if they are
> > obvious, done on purpose or do not apply; I'm not speaking French and
> > I'm not involved in any discussion regarding French translation, hence
> > these observations might be totally not applicable.
>
> Thanks for your message, it is useful to see how we can do better
>
> >
> > Without further ado:
> >
> > 1. After completing a translation, it is quite helpful to reformat it,
> > either before committing or in a second commit. This has two
> > benefits:
> > a) If there are any syntax errors, you immediately spot this.
> > b) Further scripts do not touch the file simply for reformatting,
> > making the "git diff" better to read.
>
> I apply hooks/pre-commit in my git command. Is this enough?
This should work, however, I'm not a git expert so I don't know all
the pros and the cons.
> > 2. If a translation has been proofread, it is usually quite helpful to
> > immediately (after reformatting) add it to the compendium
> > (../use-for-compendium.sh manx/foo.x.po)
> > and then update all translations
> > ../update-translations.sh
>
> To be honest I am not completely confortable with compendium stuff, so ok, I
> add those 2 commands to my automated committing scripts and we will see the
> result. I trust you.
If you encounter strings where different translations are necessary,
then you can disable the compendium for this string(s), see the
explanation for exclude.pot in CONTRIBUTING.md. You can find examples
for these in templates/exclude.pot.
If you encounter further problems, please let us know. The aim is
really to reduce translations. I have seen some (small) man pages
which are translated already > 80% simply from the compendium. Or if I
would start the German translation of "ld.1.po", it would aready have
513 strings (almost 40%) translated, just from the compendium. And
finally, I think consistency helps users, some strings/concepts with
same translations. And you nicely spot this if suddenly strings change
(at least in German we unified quite a bit using the compendium). And
where this does not work → exclude.pot.
> > Before committing, "git diff" is quite helpful.
> >
> > This has two benefits:
> > a) In many cases, other man page translations (fuzzy strings, missing
> > strings) are updated as well, easing the work and bringing more
> > pages over 80% (and thus the translations to users).
> > b) This is an extra check; in my recent updates I noticed errors
> > when performing these steps, especially during the final "git diff",
> > which I corrected, even though I do not speak French.
>
> Could you give an example please? I dont feel good to review again via git
> diff before commit but if something was not seen by our review process, I
> can try for it to eliminate this kind of typo
Then please ignore this. I often find things this way, for example, in
my recent update in the compendium for French I noticed that a path
was wrong (see commit d55bfd64fdc4c12cf3d4d3ad0a4d9cf3939c5895 for
min-002-occurences.po. Of course, this is far from perfect and having
a good review process is much better.
> > 3. I'm not sure about your systematics, but I do notice that quite a
> > few man pages are close to complete or close to reach the 80%
> > threshhold. Once they are at 80% they are shipped to users. Of
> > course, sometimes you do not want them to be shipped (unless 100%),
> > maybe because of problematic strings, so this might be deliberate.
> > However, when prioritizing, looking at
> > https://manpages-l10n-team.pages.debian.net/manpages-l10n/debian-unstable-fr.html
> > (and similarly for all other distros like Arch, Fedora, ...) you
> > can notice that several man pages just lack one or two strings to
> > reach 80% (and using the compendium, see 2., the number of these
> > pages maybe reduced even further).
> >
> > I regularly build all man pages and provide you a list of all pages
> > between 70%-79% below. In case you would like to receive this list
> > regularly, I can try to set this up (but the web listing above
> > should cover this nicely as well).
>
> Many thanks. So far I just go forward following the alphabetic order.
> Because I want to avoid the risk of duplicate work with other translators,
> without needing to use an ITT message to the list. It would require a good
> coordination process. But why not, according to what others think.
You have a luxury problem, in German we are only two doing the
translations. And we simply divided the pool, using the "Last
translator" field. So maybe some "pre sharing" might be a suggestion?
> > 4. Check for similar strings or easy unfuzzy after upstream updates.
> > When looking at French pages, I often noticed that there were quite
> > a few strings which are trivially to handle. Sometimes upstream
> > just removed or added spaces after full stops, corrected a link or
> > a typo ...
>
> Indeed, unfortunately. It completely disappointed me last months to review
> man1 due to this. A pain! Thanks to Jean-Pierre who tries to catch up this.
I share your pain. I know projects which like to reformat their man
pages about once a year. And even the regular updates can be quite
time consuming, e.g. for systemd.
> > Another example from today: wget.1.po was checked in today,
> > unfortunately not completely translated. However, one missing
> > string was just a typo away from an already translated string and
> > others were trivially to handle (e.g. version numbers), so I could
> > almost complete the part of OpenSUES Leap as well, even without
> > French knowledge.
> >
> > I noticed (and sometimes handled) similiar issues in the past as well,
> > but doing this more systematically of course requires some time. And
> > since I don't speak French, I can only perform trivial updates and some
> > easy fixes (if I knew French), i.e. "low hanging fruits" I have to
> > skip.
>
> What is your approach? Reviewing this manually is really a pain for me. Any
> suggestion to reduce the needed work is welcome.
Well, I'm not sure which editor you use. Some po editors can handle
fuzzy strings quite nicely, I was told. I'm quite old school myself
and I use vim (with po extensions) and run plain "diff" on the
previous and the current string, May some day #611544 is resolved
(dreaming).
Greetings
Helge
--
Dr. Helge Kreutzmann debian@helgefjell.de
Dipl.-Phys. http://www.helgefjell.de/debian.php
64bit GNU powered gpg signed mail preferred
Help keep free software "libre": http://www.ffii.de/
Attachment:
signature.asc
Description: PGP signature