Re: msgfmt -C gets confused by existing fuzzy translations

[Nils Philippsen]

Hi,

On Sun, 2010-08-29 at 12:15 +0200, Bruno Haible wrote:
> Jens Petersen wrote:
> > I would like to report a problem raised by
> > Nils Philippsen in <https://bugzilla.redhat.com/show_bug.cgi?id=621946>:
> > 
> > "When trying to update older versions of po files with translations 
> > submitted to
> > a newer version I found that "msgfmt -C/--compendium <new_po_file>
> > <old_po_file> <project_pot_file>" gets confused by fuzzy translations in the
> > old po file and doesn't use the updated translations from the new po file 
> > for
> > these msgids.
> > ...
> > Expected results:
> > When using the compendium file and the existing old translation file, it 
> > should
> > use the translation found in the compendium file for all msgids, even those
> > marked fuzzy.
> 
> Your expectations are wrong. Translations from the compendium file are not 
> used
> for fuzzy entries. Rationale: The fuzzy translation in the old PO file is a
> translation for a similar English string in the *same* PO file. The 
> translation
> from the compendium is for a similar English string in a *different* PO file.
>
> You can expect that translations from the same PO file will use the same
> vocabulary/terms than translations from different PO files. (E.g. a "switch"
> would translate to "Schalter" in a program for house design, but to "Switch"
> in a program for network configuration.)

ahh, that's probably because I used -C in a different way than
originally intended, namely to merge new translations that are made in
the master branch of a project into an older version (using SCMs to
merge PO files is rather unpleasant). In this case, I'd rather have the
non-fuzzy translation for the same msgid from the compendium file (which
really is the same file, just from the up-to-date master branch) than
the fuzzy translation in the older branch.

> If you want to ignore fuzzy entries from the PO file (because somehow they
> are less reliable what you have in the compendium), you can play tricks with
> 'msgattrib' and 'msgcat':
> 
>     $ msgattrib --no-fuzzy old-zh_TW.po \
>       | msgmerge -C compendium-zh_TW.po - system-config-services.pot \
>       | msgcat --use-first - old-zh_TW.po \
>       > system-config-services-zh_TW.po
>     $ msgfmt -cvo /dev/null system-config-services-zh_TW.po
>     57 translated messages.
>
> The msgattrib command creates a temporary PO files where the fuzzies are
> missing. The msgmerge program then searches the compendium in first priority.
> The msgcat program then adds back the fuzzies in those places where the
> compendium did not provide a translation.

I see, but I don't think I'll remember this if I get into this situation
again...

Would you accept a patch which lets users specify that msgmerge should
prefer the compendium file (maybe using a different term than
"compendium")? Perhaps rather using mostly the same code, but having a
"--update-file" option which acts like "-C" except for the differences I
described?

Nils
-- 
Nils Philippsen      "Those who would give up Essential Liberty to purchase 
Red Hat               a little Temporary Safety, deserve neither Liberty
address@hidden       nor Safety."  --  Benjamin Franklin, 1759
PGP fingerprint:      C4A8 9474 5C4C ADE3 2B8F  656D 47D8 9B65 6951 3011