GTK+ 4.0 Targeted For Its Initial Release This Fall, GTK+ 5.0 Development To Follow

[schmidtbag]

That's because of Google's versionning way.
They're going rolling-release without saying it, because everyone would understand that it's a bug-prone versioning way.
And unfortunately, a lot of people are folling them in their maddness.

If it is actually a rolling release, where GTK4 software can effortlessly be ported to GTK5, then I see no problem in what they're doing. But that's a big "if".

[Vistaus]

I don't think porting to GTK 4 will start until the release of GTK 4.8. This is when GTK4 will be declared stable. Given their current schedule of a release every 6 months it would take until fall 2020 for GTK devs to recommend GTK4 for third party developers.
Together with GTK 4.8 there will also be a release of GTK5, for which same things apply. It would take until 5.8 until they recommend GTK5 to third-party developers.
That's how GTK's new release cycle is supposed to work.
Less or even no breaking changes from X.8 on, because the breaking changes will happen on the Y.0 branch instead.

The plan is that third-party devs will only start porting to a new branch after it has been declared stable with the X.8 release, which means you'll start working on a port ~2 years after the first version of a new branch came out.

So the short answer is yes, it will take a long time to port to GTK4.
However, it should take less than the 5 years for XFCE. A new branch will become stable much faster than GTK3 did, so there will be less overall changes from one branch to the next.

Also, AFAIK GTK3 refactored too much from GTK2 (that is: you couldn't port the code 1-on-1), while GTK4 mostly adds a lot of new widgets which one could use but doesn't have to and new backends. So sure, most apps won't be ported overnight from 3 to 4, but it should take a lot less time given the fact that the core changes are less significant than from 2 to 3.

[UpsetingFact]

If it is actually a rolling release, where GTK4 software can effortlessly be ported to GTK5, then I see no problem in what they're doing. But that's a big "if".

Considering how relatively "well" rolling releases are doing "good" concerning stability, there will be bugs.

No one's perfect, so it should be wise to avoid problems as much as possible.

[Luke]

Also, AFAIK GTK3 refactored too much from GTK2 (that is: you couldn't port the code 1-on-1), while GTK4 mostly adds a lot of new widgets which one could use but doesn't have to and new backends. So sure, most apps won't be ported overnight from 3 to 4, but it should take a lot less time given the fact that the core changes are less significant than from 2 to 3.

Actually, GSK (GTK+ Scene Graph Kit) https://en.wikipedia.org/wiki/GTK%2B_Scene_Graph_Kit was a major, major backend change that while it does not change using GTK's own widgets introduces wholesale breakage of custom widgets. Thus, a lot of the work that Firefox and Libreoffice did to port to GTK3 would have had to be thrown away and started over if GSK had been merged into any GTK3 branch. This is why it was held past the GTK 3.22 release, and GTK then branched to the pre-GTK4 versions in Master.

There are also major barriers to porting the GNOME forks to GTK4, beginning with maintaining desktop icon support. Both the new split-out nemo-desktop and caja (as did nautilus-desktop) use a canvas for drawing the desktop that GNOME claims would be very difficult to port to GTK4. GNOME decided to drop desktop icon support from Nautilus entirely for this reason. Neither Cinnamon nor MATE can do this, as their users are those who want a traditional desktop.
Thus to port either one to GTK4 may require a clean-sheet reimplementation of the entire desktop canvas. You can't use a gnome-shell extension to render the desktop when not running gnome-shell, though I suppose a gnome-shell extension might be with considerable effort back-ported to Cinnamon, which is based on a far older version of GNOME.

Not sure whether dropping code replaced by GSK is the reason for this or dropping something else, but either way this is a major barrier to either GNOME fork porting to GTK4.

[Vistaus]

So the short answer is yes, it will take a long time to port to GTK4.
However, it should take less than the 5 years for XFCE. A new branch will become stable much faster than GTK3 did, so there will be less overall changes from one branch to the next.

However, if your project is written in C, you can actually semi-automate the porting from GTK3 to 4 thanks to Timm Bäder: https://github.com/baedert/autoport