[Kbabel] Full Gettext support in KBabel

[Nicolas Goutte]

On Sunday 23 October 2005 22:11, Chusslove Illich wrote:
> > [: Nicolas Goutte :]
> > Well, for KDE 3.5, it is enough that the check can be disabled by the
> > user. (That is not an emergency.)
>
> Right, I forgot that it can already be turned off in settings :)
>
> > As written, msgfmt without any check parameter checks little. So a
> > rather robust loading code should be able to do as well.
>
> The perculiarity with catalog manager is that if it gets more robust than
> msgfmt, that is a disadvantage rather than an advantage. It is surely an
> advantage for the editor, because editor will save it without critical
> errors, but catalog manager shouldn't swallow a file that msgfmt wouldn't.
> Or vice versa.

I think that I understand what you mean but I do not know if we have the 
luxury of being able to do it. 

First: a PO file syntax remains a PO file syntax. And as it is a rather simple 
syntax, it is difficult to do wrong when parsing it. (I would tell probably 
something else if we were discussing about XLIFF.)

But also somewhere we have to draw a limit. I try to place the limit to what I 
have seen in practice on the l10n module in KDE SVN. For example, that is why 
conflict markers will be a reason for "disqualification". (I do not think 
that it is in the interest of any translator, of any team or of any project 
to consider them right.)

Also the errors that we mean here are really basic ones. They are *not* of the 
kind of mismatched %something between msgid and msgstr but more errors like 
an "unknown keyword".

>
> The point being, catalog manager should at least tell us how the msgfmt
> will react to a file. Anything further I would consider an extra feature
> (like a robust code doing its best to read a file, for the sake of
> overview, with an extra note of whether msgfmt likes it or not).

Perhaps the (manual) solution is to run the statistic function but I do not if 
it returns errors to the user. (I have never tried.)

Also perhaps we had misunderstood us. My goal for the catalog manager is not 
to read it whatever it contains. (The editor tries that, for helping to 
recover data.)

Propably what you want is that we disable the "recovered error" facility and 
consider a recoverable error being a full error. I do not mind (especially 
that a correct header is important).

>

(...)

Have a nice day!