Re: [bug-gettext] New Gettext contributor

[Miguel Ángel]

Daiki Ueno <address@hidden> writes: 
> Miguel Ángel <address@hidden> writes:
> It seems not yet, see my private mail.

Yes, I saw it. How couldn't i see there is no sign in the pdf? :(

> > * Glade bugs fixes and argument parser leaks fixes (three patches)
> >   L * GtkBuilder support (three patches)
> >       L * GSettings support (two patches, yet need docs and tests)
> >           L * update gnome examples -> gtk+-3.0 (several patches not
> > uploaded)
> 
> Those patches really depend on each other?

Yes and no:

Glade bugs fixes and GtkBuilder are intimately related. GtkBuilder uses
the arglist_parser_* patch for Glade. Leaks in xgettext are the first
patch, but actually It is an unrelated patch, so it can be in a separate
topic.

GSettings and update gnome are not dependant, but the first needs expat
compatibility extraction from x-glade.c (or adding an unlikely related
file type to x-glade.c to parse also GSettings) and the second needs
GtkBuilder extraction and I do not think that we want to depend on
intltool in our examples. ;-)

It can be independent, but the merge (or rebase) is a real mess, I have
tried it, and I have used GtkBuilder for new Gtk examples.

> What's the current state of each patch?  Last time I looked at your
> patches posted here, they were work-in-progress and github commits
> were more up to date.  Are they now stable?

I did a couple of minor changes since I have posted them, but GtkBuilder
patches has not changed since 2013-02-06 and Glade and xgettext ones
since 2013-01-31. They are tested with make check and with each package
with NLS I build at my main machine.

> > * libasprintf bug fixes (two patches)
> > * libgettextpo bug fix (one patch)
> 
> Those patches are small and I don't think they need separate remote
> branches.

I think libasprintf patches should be four, because we should add
-version-info flag at linking time (now is missing). First add
-version-info for buggy version, apply the patches and bump
-version-info. The default operator= can be at a different offset of the
class definition than new one, so this would be an ABI change.

I was going to send them (arglist_parser, Glade2, GtkBuilder and
libasprintf) but I have to do a question before, do I have to update
ChangeLog date? I mean, which date (patch date or the commit one) must
be there?

And about libgettextpo trailing backslash: It could be a documented
feature and not a bug, comments are welcome. But there is an error when
you execute this with the attached file:
xgettext --add-comments -o - Test.java | msgen -

because line

#. I am a bad guy and want to abuse xgettext #: Test.java:4

should be

#. I am a bad guy and want to abuse xgettext \?
#: Test.java:4

with ? being any valid character.

Happy hacking! :)
Miguel

Test.java
Description: Text Data