[Freewx-maint] wxWidgets GTK+ 3

[Olly Betts]

On Sat, Sep 24, 2016 at 07:42:54PM -0400, Scott Talbert wrote:
> On Sun, 25 Sep 2016, Olly Betts wrote:
> 
> >>I know GTK+ 3 has been discussed before, and I know you've had reservations,
> >>but what do you think about building for GTK+ 3 alongside GTK+ 2?  That way,
> >>individual dependencies could be migrated as they are ready.
> >
> >Parallel packages increases the maintenance work, and I don't have time to spare
> >for that currently - I already have enough things I want to get sorted
> >out before
> >the upcoming freeze.  This would also increase the maintenance work for some
> >dependent packages (wxpython, wxperl, wxglade, etc would likely need parallel
> >versions too).
> 
> I'm not sure how much extra maintenance it would be.

We'll have to deal with additional bug reports due to GTK+3-specific
issues.  And any change to one variant really needs to at least be
considered for the other (or if they're built from the same source
package, any change to fix an issue with one variant also needs testing
with the other).  That seems likely to be a significant extra load.

The alternative is to neglect one version, as tended to happen in the
past when we had parallel packages of different wx release series, but I
don't think that's good for anyone.

> Sure, it would take some work to get the parallel packages setup
> initially, but I'm not sure it would be that much work after it is
> already set up.

And indeed, there's also work just to set things up to start with.

> >If you really want to make a case for this, I think a list of concrete
> >benefits are
> >needed - all I can see in your email seems to amount to "some dependent
> >packages are ready".
> 
> I don't have a whole lot other than I think it would be better to
> have GTK+3 support sooner rather than later so that non-X11 backends
> (e.g., Wayland) can be supported.

https://wiki.gnome.org/Initiatives/Wayland/GTK%2B suggests "GTK+ already
has a functional Wayland backend, but several pieces of functionality
need to be completed before it is ready for prime time", which sounds
like it is not the sort of thing we want to rush into a Debian stable
release.

> >My preference is still to do the migration along with the next wx
> >version (so 3.2),
> >assuming that comes along in a timely fashion.  And if it doesn't look
> >like it'll
> >be out soon enough, then to start work on GTK+3 early in a Debian release cycle.
> >That would give us some hope of being able to migrate everything and drop the
> >GTK+2 packages so we can avoid having to support two parallel versions for a
> >whole stable release cycle.  And if the migration is problematic, issues can at
> >least get shaken out, and either fixed or affected dependent packages can revert
> >to the GTK+2 version before the freeze.
> 
> I'm not optimistic about 3.2 being any time soon.  It seems that
> upstream development has slowed recently - well, not necessarily
> development itself, but releases.  Ie, they've been talking about a
> 3.0.3 release "coming soon" for quite a while now and there doesn't
> seem to be a lot of movement.

It is a bit frustrating that there are plenty of changes in git
which could go in 3.0.3, but no release.  I've wondered if we should be
considering packaging the branch head.

The threads I read gave me the impression that the lack of interest in
releasing 3.0.3 from upstream was because the development focus was on
the 3.1 series.  But it's certainly true that upstream has a patchy
record on getting new release series out in a timely fashion.

> The other reason I'd propose parallel packages is that I think
> moving some applications to GTK+3 is going to take a bit of work
> (and thus, the work would not *have* to happen within a release
> cycle).

Maybe it'll take two Debian release cycles.  But if we start on
it early in a release cycle, it's less likely to take three.

Cheers,
    Olly