Review Request 123494: Create DesktopView before connecting signals from the activity consumer

David Edmundson david at davidedmundson.co.uk
Sat Apr 25 10:52:20 UTC 2015



> On April 25, 2015, 10:27 a.m., David Edmundson wrote:
> > Ship it * 1000
> 
> David Edmundson wrote:
>     Wait. How can this make any difference?
>     
>     DBus slots all happen in the main thread's event loop, which means we must have finished all this function before that.
>     
>     The only way this could change the behaviour is if something in new DesktopView() is calling creating a new event loop or calling qApp->processEvents()
>     
>     do you still have the entire trace?

Well, I was right

ActivitiesSync::currentActivity

        while (activities->serviceStatus() == KActivities::Consumer::Unknown) {
            QCoreApplication::instance()->processEvents();
        }
        

Your QML code is creating an activities model which calls this rather questionable segment of code.
It's possible Kicker is too.

You can push your change if you want a quick fix, it won't do any harm, but the real fix will be replacing this method. 

Blocking DBus calls should be made with the normal sensible blocking DBus API (or not done at all) not doing this processing all sorts of things out of whack.


- David


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/123494/#review79491
-----------------------------------------------------------


On April 25, 2015, 3:56 a.m., Bhushan Shah wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/123494/
> -----------------------------------------------------------
> 
> (Updated April 25, 2015, 3:56 a.m.)
> 
> 
> Review request for Plasma and Marco Martin.
> 
> 
> Repository: plasma-workspace
> 
> 
> Description
> -------
> 
> This avoids the crash situation where something have created the activity consumer earlier and currentActivityChanged signal is emitted before view is created.
> 
> ```
> #0  PlasmaQuick::View::setContainment (this=0x0, cont=0x16a9fc0) at /home/bshah/aur/plasma-framework-git/src/plasma-framework/src/plasmaquick/view.cpp:243
> #1  0x0000000000462097 in StandaloneAppCorona::currentActivityChanged (this=0x7772a0, newActivity=...)
>     at /home/bshah/aur/plasma-workspace-git/src/plasma-workspace/shell/standaloneappcorona.cpp:192
> #2  0x00007ffff32383c9 in QMetaObject::activate(QObject*, int, int, void**) () from /usr/lib/libQt5Core.so.5
> #3  0x00007ffff71f9763 in KActivities::Consumer::currentActivityChanged (this=0x741aa0, _t1=...) at /home/bshah/aur/kactivities-git/src/build/src/lib/core/moc_consumer.cpp:216
> #4  0x00007ffff71f94dd in KActivities::Consumer::qt_static_metacall (_o=0x741aa0, _c=QMetaObject::InvokeMetaMethod, _id=0, _a=0x7fffffffa570)
>     at /home/bshah/aur/kactivities-git/src/build/src/lib/core/moc_consumer.cpp:103
> #5  0x00007ffff32383c9 in QMetaObject::activate(QObject*, int, int, void**) () from /usr/lib/libQt5Core.so.5
> #6  0x00007ffff71f91d6 in KActivities::ActivitiesCache::currentActivityChanged (this=0x77c900, _t1=...) at /home/bshah/aur/kactivities-git/src/build/src/lib/core/moc_activitiescache_p.cpp:353
> #7  0x00007ffff71ee4f7 in KActivities::ActivitiesCache::setCurrentActivity (this=0x77c900, activity=...)
>     at /home/bshah/aur/kactivities-git/src/kactivities/src/lib/core/activitiescache_p.cpp:297
> ```
> 
> I hit this crash when creating recent media backend for pmc. First backendsmodel is initalized which in turn loads the recentmedia backend and hence creates the activity consumer first. Afterwards when signal is connected it calls currentActivityChanged slot and fails to set containment on null view.
> 
> 
> Diffs
> -----
> 
>   shell/standaloneappcorona.cpp efa5c91 
> 
> Diff: https://git.reviewboard.kde.org/r/123494/diff/
> 
> 
> Testing
> -------
> 
> works fine
> 
> 
> Thanks,
> 
> Bhushan Shah
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/plasma-devel/attachments/20150425/0edd0e5f/attachment.html>


More information about the Plasma-devel mailing list