Plasma port to Wayland

Martin Graesslin mgraesslin at kde.org
Wed Nov 5 10:01:27 UTC 2014


On Wednesday 05 November 2014 10:37:59 Sebastian Kügler wrote:
> Hi Pier Luigi,
> 
> Thanks for the updates, such a high-level overview is much appreciated. 
Some
> question inline below.
> 
> On Monday 03 November 2014 12:45:14 Pier Luigi Fiorini wrote:
> > As some of you might know I've been working on the Plasma port to 
Wayland.
> > 
> > I would like to give you some pointers where the on going effort is.
> > 
> > Last month I've been super busy with other stuff, but I plan to resume my
> > work on Wayland during November.
> > 
> > You can see it in action on my compositor, Green Island:
> > 
> > http://imgur.com/8F4Emax
> > http://imgur.com/Q69CB2v
> 
> Awesome. :)
> 
> > Here's the components I've been working on:
> >   - kwayland
> > 
> > git clone -b plasma kde:clones/kwayland/pierluigifiorini/plasma.git
> > 
> > The plasma branch contains patches to add Plasma protocols.
> > 
> > The wlscanner branch makes use of the new find module for wayland-
scanner
> > that I recently made; it can be merged as soon as a new ECM version 
comes
> > out, see https://git.reviewboard.kde.org/r/120656/
> > 
> > All is currently rebased on top of master.
> > 
> >   - kwindowsystem
> > 
> > git clone -b wayland 
kde:clones/kwindowsystem/pierluigifiorini/wayland.git
> > 
> >   - plasma-framework
> > 
> > git clone -b wayland
> > kde:clones/plasma-framework/pierluigifiorini/wayland.git
> > 
> >   - plasma-workspace
> > 
> > git clone -b wayland
> > kde:clones/plasma-workspace/pierluigifiorini/wayland.git
> > 
> > Ported ksplashqml and plasmashell (panels and desktop view); I'll reread
> > the patch when I get back to this stuff.
> > Review is not started: blocked by the review of my changes to kwayland.
> > Currently rebased on top of v5.1.0
> 
> It would probably make sense to rebase on master, so the delta becomes a 
bit
> smaller. I think it's time to start shipping cookies to Martin Graesslin to
> keep him happy during extended reviews. :D

pointing out the obvious: if I'm the bottleneck in doing reviews than we have 
a serious problem for any Wayland work.

> 
> Otherwise, we're waiting for the ECM to be merged? The RR has a shipit, so I
> guess that can go in, and we can proceed with merging other patches in. Do
> you need any help with that?

pointing out the second obvious: we have a dependency problem for merging 
all the patches. The protocols are in my opinion not yet in a state that we 
can submit to ABI stability for all of KF5 - after all it's a new technology 
for us and we have to be careful. Then obviously KWayland cannot become a 
framework with the current compiler requirements which in turn means that 
KWindowSystem cannot depend on it. This needs thinking and discussion on 
the framework mailing list. At the moment I don't consider KWayland as suited 
for a framework yet.

There were a few showstoppers in the review request for KWayland last time I 
looked at it. E.g. it implied a Qt 5.4 dependency due to using QtWayland. That 
is currently not possible and we need a better solution for combining with 
QtWayland. The patches are in that area too hackish as the API becomes not 
obvious which parts need to be used with QtWayland and which with 
KWayland's own wayland connection. And the patches were completely 
missing unit tests when I last looked at them which is a must requirement for 
getting anything into KWayland.

Overall I don't want us to rush things. There is no problem with these things 
taking time and living in a branch. We have not won anything if we hurry now 
to notice we created a problem in a few months. As said: I think submitting 
the protocols at this point in time to stability is too early. It's good that 
we have the protocols, but we need to give them a few more rounds of review 
and thinking. We also should look at them and think about what makes sense 
to upstream into Wayland. Once we are 100 % sure that these protocols are 
the way to go, we can submit to them.

This also means that people experienced with how the Plasma<->windowing 
system interaction worked on X11 should have a look at the protocol, that 
people maintaining windowing system relevant areas should look at them.

Sorry, for being the one hitting the break ;-)

Cheers
Martin


More information about the Plasma-devel mailing list