Review Request 129204: Add toggle signal for applet de-/activation in order to fix non-closing launchers on Meta (and also on Active Screenedges)

[Kai Uwe Broulik]

> On Okt. 17, 2016, 8:26 vorm., David Edmundson wrote:
> > src/scriptengines/qml/plasmoid/appletinterface.cpp, line 151
> > <https://git.reviewboard.kde.org/r/129204/diff/1/?file=482310#file482310line151>
> >
> >     Because AppletInterface::activated is a signal, applets could connect to to do extra actions on external activation; we're not emitting that with this signal.
> >     
> >     and this isn't just hypothetical as kicker does.
> >     
> >     kicker/package/contents/ui/CompactRepresentation.qml:   
> >     plasmoid.activated.connect(function()
> >     
> >     (ironically to do toggling on the dash window)
> >     
> >     By introducing a new signal, this breaks existing applets.
> >     
> >     We could maybe simply add 
> >     if (activate) emit activated();
> >     
> >     and it'd work.
> >     
> >     Question:
> >     the current global shortcuts seem to toggle, and they only connect to activate(), how do they work?
> 
> David Edmundson wrote:
>     sorry, you literally have a sentence for that in your description. Ignore my last comment.

> the current global shortcuts seem to toggle, and they only connect to activate(), how do they work?

>From what I can tell, technically the Meta behavior is actually the correct one. As far as I understood, KGlobalAccel registers a "soft grab" on the global shortcuts. When they are pressed, X grabs them, and KGlobalAccel ungrabs them again; this causes the window to lose focus and it closes.

You can observe this when you e.g. change volume while any Plasma popup is open, behavior is the same. I don't know how it behaves on Wayland, though, and what we should do with the Meta key to improve the situation.


- Kai Uwe


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/129204/#review100067
-----------------------------------------------------------


On Okt. 17, 2016, 7:02 vorm., Roman Gilg wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/129204/
> -----------------------------------------------------------
> 
> (Updated Okt. 17, 2016, 7:02 vorm.)
> 
> 
> Review request for Plasma and Martin Gräßlin.
> 
> 
> Bugs: 367685
>     http://bugs.kde.org/show_bug.cgi?id=367685
> 
> 
> Repository: plasma-framework
> 
> 
> Description
> -------
> 
> # The new Meta-key support for launcher opening doesn't work for closing at the moment. My analysis of the problem was as follows:
> 
> - KWin calls Applet::activated() over DBus
> - Applet::activated() is connected to AppletInterface::activated()
> - AppletInterface::activated() is connected to setExpanded(true), which can only expand the launcher, but not the other way around
> 
> # Q: Why is Alt+F1 working though?
> A: The launchers seem to inherit the reimplemented function Dialog::focusOutEvent(..). Atleast when you delete line 1094 in dialog.cpp, which sets the visibility to false, it's not working anymore. Alt+F1 triggers the focusOutEvent(..) as a global shortcut.
> 
> # Q: Why is it working with the Dashboard though?
> A: The dashboard doesn't use the expanded feature of the plasmoid, but rather connects directly to the AppletInterface::activated() signal and shows/hides an independent widget when this signal gets triggered.
> 
> # Solution:
> Create new toggled() signal chain in KWin, Applet, AppletInterface, which tests the current expanded state and sets it afterwards accordingly to the opposite. This diff here in plasma-framework is the first one, which the others depend on. The others are on Phabricator:
> 
> - kwin: https://phabricator.kde.org/D3077
> - plasma-workspace: https://phabricator.kde.org/D3078
> - plasma-desktop: https://phabricator.kde.org/D3079
> 
> # Need feedback regarding:
> 
> - Should we remove the activated() signal chain? Is it used somewhere else than KWin?
> - Could a race condition occur if we deexpand the applet while at the same time setting the visibility to false through focusOutEvent()? My tests until now don't suggest it, but I haven't yet looked into it extensively.
> 
> 
> Diffs
> -----
> 
>   src/plasma/applet.h 89498ea 
>   src/scriptengines/qml/plasmoid/appletinterface.h a1e2cd7 
>   src/scriptengines/qml/plasmoid/appletinterface.cpp 1cd6934 
> 
> Diff: https://git.reviewboard.kde.org/r/129204/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Roman Gilg
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20161017/6ae79ead/attachment-0001.html>