Checking format specifiers

["Martin v. Löwis"]

Wolfgang Bangerth wrote:
> Or it's this: while many messages have c-format, many also have the fuzzy 
> flag. Take this broken one from da.po:17470:
>     #: cp/typeck.c:1516
>     #, fuzzy, c-format
>     msgid "invalid application of `%s' to a member function"
>     msgstr "ugyldig brug af 'this' i ikke-medlemsfunktion"
> msgfmt --check doesn't barf on this. It does, however, if the fuzzy flag is 
> removed. 

That is correct behaviour. The fuzzy flag might be added in response to 
adding a new % field to the msgid - then the msgstr becomes fuzzy. If 
msgfmt would reject all such messages, the translator would need to 
update all fuzzy messages after an update, which is not feasible.

Instead, all projects ignore (or should ignore) fuzzy messages when
delivering catalogs. Then any errors in fuzzy messages (which may go
beyond mere format errors) become irrelevant. If gcc delivers fuzzy
translations to users, I recommend that this is changed ASAP.

I also worry that your script gives too many false alarms.

So I revert my earlier acceptance - unless it ignores fuzzy messages, it 
is likely unusable, and should not be automatically invoked by the TP robot.

Regards,
Martin