Re: Steps to get to GTK+ 3.0

["Havoc Pennington" <hp pobox com>]

Hi,

On Wed, Jun 4, 2008 at 2:37 PM, BJörn Lindqvist <bjourne gmail com> wrote:
> Regardless, gtk+ 3.0 is a long-term project, probably with a first
> release sometime in 2010 or so. Embedded developers wont want to pick
> it up before 2012.

AIUI this is NOT the case. My understanding of the Imendio proposal is
that GTK 3.0 *only* seals widgets and makes them private, removes
stuff that is *already* deprecated, but contains no features and does
not replace and deprecate anything new.

ABI/API changes that are *semantic* are being moved to GTK 4.0 at this
point, because the proposal is to make 3.0 a fast release that only
hides stuff. Hiding stuff still won't allow changes in API semantics
to be made compatibly.

Maybe 2010 is a fast release that only adds G_SEAL, but if so that's
kind of depressing... I would hope a 3.0 that only did that could come
out by end of this year or early 2009, maybe at the same time as the
next 2.x release even. The whole argument is that to start on 3.x
series feature work first the ABI must be sealed. Delaying feature
work to 2010 would certainly suck.

A competing proposal was to first try to address some features without
breaking ABI (Paul Davis had a nice list a bit further up in the
thread, I'll append it here), deprecate a lot more stuff, then do a
GTK 3.0 after that. But that isn't the approach that was chosen.

Disclaimer: I was not at the summit nor at the IRC meetings, this is
just my impression from the list and other forums.

Havoc

Paul's quote:
> much bigger issues for me have been the lack of a good, supported, fast
> canvas widget, the lack of non-resize-crazed label widgets, font sizing
> difficulties and the lack of DnD from treeviews to other widgets. i
> don't believe fixing any of these is particularly central to 3.0.

I have the same understanding as Paul that this stuff would be
post-3.0. And ABI/API breaks that change semantics rather than
implementation - like fixing the GDK API to be less weirdly Xlib-like,
or making a scene graph central to the toolkit - would be in 4.0.