KDE's window manager KWin gets forked with 'KWinFT' to accelerate the development and better Wayland
Stick a fork in it! KDE's window manager KWin officially has a full fork with a new project called KWinFT, with an aim to support modern development practices and further expand Wayland support.
Announced by Roman Gilg, the same developer who became a contractor for Valve last year and part of that work was actually to improve KWin so it looks like this may have come as a result of that. What's interesting about KWinFT, is that it's supposed to be a "drop-in replacements for KDE's window manager KWin and its accompanying KWayland library" making it easy to get started with it.
Gilg said they did this because "Classic KWin can only be moved with caution, since many people rely on it in their daily computing and there are just as many other stakeholders" so they can push through more advanced changes and overhauls.
What's already in? According to Gilg:
- My rework of KWin's composition pipeline that, according to some early feedback last year, improves the presentation greatly on X11 and Wayland. Additionally a timer was added to minimize the latency from image creation to its depiction on screen.
- The Wayland viewporter extension was implemented enabling better presentation of content for example for video players and with the next XWayland major release to emulate resolution changes for many older games.
- Full support for output rotation and mirroring in the Wayland session.
The first public release was put out along with the announcement and if you use Manjaro Linux, it's already available in the unstable branch under the kwinft package. You can read the announcement here.
What are your thoughts on this? As someone who transitioned to KDE from the GNOME a while ago, anything that can further improve it sounds excellent.
Quoting: CatKillerAn experimental branch that tests things out before they get polished and moved upstream would be great. This, though, sounds like someone just got grumpy that their "revolutionary" Wayland ideas weren't immediately accepted by Kwin. I hope it turns into the former even though it's starting out as the latter.
I doubt that. Roman for sure wasn't happy with the development process of KWin, wanting a faster pace of development with less stakeholders (and there are quite a few in KWin), and without a fork it probably hardly would be used and tested.
It's a fork to get a different development model, which can disrupt things and make things not work. KWin with the focus on reliability hardly can do that. Though, that certain improvements once stable enough can find their way back to KWin. Wonder how complicated that'll be...
Also, huge dependency on Qt that Roman pointed out is a major problem.
I don't think the intention is to keep this fork separate forever. Once it matures, it can replace the original KWin if the upstream of course agrees on that.
Last edited by Shmerl on 17 April 2020 at 3:00 pm UTC
I do naturally hope that any usable improvements get picked up by KWin and the two projects merge back together when the goals of the fork have been achieved.
thanks for your interest in the project. So as I'm a gamer myself one thing I also want to aim for with KWinFT is to make the gaming experience in its Wayland session really top-notch.
I did work on that in the past already so it is normally usable nowadays without major issues but I want to really polish it out over time.
So it would be cool, even when you don't normally use KDE Plasma, if you could try it out from time to time and report any issues you might experience at https://gitlab.com/kwinft/kwinft/-/issues.
For example I found out when I played some SoSe recently that right-click-panning doesn't work on Wayland.
There are for sure more such small issues so getting in feedback on that from many people would be super helpful.
Thanks,
Roman
Last edited by subdiff on 17 April 2020 at 7:11 pm UTC
Quoting: subdiffThere are for sure more such small issues so getting in feedback on that from many people would be super helpful.
Thanks for trying to accelerate things and keeping gaming in mind!
I'll surely try testing it. How is subsurface clipping faring in KWinFT? That's the major issue that was blocking me from switching to Wayland Plasma session. Another one is lack of adaptive sync, but that was more of an uptream Wayland issue.
Looks like there is some progress on it, but it's slow: https://gitlab.freedesktop.org/wayland/wayland/issues/84
Last edited by Shmerl on 17 April 2020 at 7:36 pm UTC
Quoting: ShmerlQuoting: subdiffThere are for sure more such small issues so getting in feedback on that from many people would be super helpful.
Thanks for trying to accelerate things and keeping gaming in mind!
I'll surely try testing it. How is subsurface clipping faring in KWinFT? That's the major issue that was blocking me from switching to Wayland Plasma session. Another one is lack of adaptive sync, but that was more of an uptream Wayland issue.
Looks like there is some progress on it, but it's slow: https://gitlab.freedesktop.org/wayland/wayland/issues/84
I'm not aware at the moment of any specific subsurface clipping issues, but the last state I know of was that the subsurface implementation in Scene is still somewhat broken. I plan to look at this at some point in the future but it is at the moment not on my priority list, sorry. I would still urge you to open a ticket for it if you experience it when you test KWinFT. I will shift my priorities depending on how many people "like" issues.
Adaptive Sync is something I'm following passively at the moment until I find time to go in depth. I have a nice 144Hz FreeSync2 monitor standing around here, so I also have a personal interest in making this happen. :D
Quoting: subdiffI'm not aware at the moment of any specific subsurface clipping issues, but the last state I know of was that the subsurface implementation in Scene is still somewhat broken.
<...>
Adaptive Sync is something I'm following passively at the moment until I find time to go in depth. I have a nice 144Hz FreeSync2 monitor standing around here, so I also have a personal interest in making this happen.
Thanks!
The upstream issue for subsurface clipping is here: https://bugs.kde.org/show_bug.cgi?id=387313
And also some thread here: https://phabricator.kde.org/T10530
From the comments, it sounded like as serious issue that needed a deep redesign, rather than a minor fix.
I'll check how KWinFT compares in this regard and if anything is similarly broken, I'll open a separate ticket for it.
Last edited by Shmerl on 17 April 2020 at 8:07 pm UTC
See more from me