Re: gnome-devel-docs on developer.gnome.org

[Shaun McCance <shaunm gnome org>]

On Fri, 2014-05-09 at 21:08 +0200, Ask Hjorth Larsen wrote:
> Hello everyone
>
> First of all I am very willing to implement a mode in gtxml
> specifically for this.
>
> Fixing the translator-credits issue is obviously very easy.
>
> I can make it so that gtxml will only complain about syntax errors,
> i.e. strings that are not accepted by the xml parser.  This requires
> that all msgids and msgstrs - except the header and translator-credits
> - can be assumed to be XML.
>
> The files also contain "media" links to images/videos with their md5
> hashes.  Perhaps a special check should be made for those, since they
> are probably not required to be legal XML.  Are there other classes of
> strings to be aware of?  Is all this acceptable?  Also, who could
> actually approve this and make it run on the server?

For any special strings that don't correspond directly to some content,
itstool always uses msgctxt "_". So at least for itstool-generated PO
files, you could ignore any messages with msgctxt "_".

> Right now gtxml can to various extents be "smarter" and detect more
> errors - e.g. whether unexpected XML tags are used.  But some
> translators tend to insert their own little xml things when they want,
> which is fine by itself, but can lead to false positives when
> attempting to be too smart.
>
> What happens with placeholder tags (such as _:ulink-1, _:ulink-2 and
> so on), how bad is it if those are messed up?

For any placeholder tags, itstool always uses the namespace prefix "_".
(It does an internal mapping of this to a namespace URI, which nobody
should bother caring about (and yes, a known limitation of itstool is
that it will do bad things if you use "_" as a namespace prefix in your
source documents.))

Messages with placeholders should always be well-formed. It would be
nice to check that the placeholder elements in the source message line
up with those in the translated message.

Other XML->PO tools could do things completely differently. xml2po does
not use a special msgctxt or namespace prefix for anything. But people
should stop using xml2po anyway. I think KDE has their own tool. They
should probably just use itstool too. I'm not sure what Okapi does, but
it's targeted at XLIFF anyway, and its PO exporter is a bonus feature
that I'm not sure many people use.

There are probably other tools I don't know about. I saw a svg2po once
somewhere. They should really just use itstool too.

Extra bonus side note: I've asked the gettext maintainers before about
adding a flag to identify XML content, like:

#, xml-fragment

It's been a while since I bugged them about that. Then tools like gtxml
could just look for messages with that flag.

--
Shaun