Between Linux and Anime

I know the pencils down was a week (plus?) ago, but I did a fresh upgrade of my system (to OpenSUSE 11.3, then to KDE SC 4.5.0) and then realized to my horror that there was obviously no way I could make a screencast or even take screenshots for this post until I’ve rebuilt trunk and plasma-mobile. So I frantically spent the past week doing exactly that, and now without further ado, the screencast:

My trunk kdebase is still a little screwy at the moment, in case you noticed the extreme ugliness of that context menu there. Also, somehow the sizing and position of the icons have gone a little off between the end of GSoC and now (Damn).

Anyway as you can see, nothing much has changed in the high-level ideas of the system tray I described in my previous post. Most of the work I put in since the mid-term had been in the way of improving the implementation strategy – cutting most of the harder ties between the system tray and the mobile shell, and getting the system tray to update its behavior based on loose events like resizing instead of specific events like Qt signals. All this of course to fit into the Plasma philosophy of creating loosely coupled independent parts – so that for instance we could put in a replacement tray without needing to change the shell code – since the shell only knows it is given a containment that it needs to resize – and we could also easily put my systray elsewhere as long as it knows to resize the tray the right way.

There is also of course noticeable change in the look and feel. The tray containment paints its own background now, and the icons now resize along with the containment instead of disappearing and reappearing at the ends of the animation – so that we don’t need extra signals to indicate when those ends occur. I’ve also thrown a couple of “fake” plasmoids in – a fake battery indicator and a fake service signal indicator – since both of these things apparently require hardware-specific treatment. Unlike the desktop system tray, the mobile version only supports the plasmoid and dbus system tray protocols – xembed isn’t supported. Iirc those aren’t resizable anyway :)

It’s been fun. I have always loved Plasma for the sheer sensibility and elegance of its software architecture, and its really nice to be given a chance to gain a deeper understanding of it and experience it for myself. It was also great to be forced to crunch through a bunch of documentation for all kinds of QGraphicsObject derivatives – and have a brief tantalizing tango with QML while I was at it. It was a great GSoC, and I’d like to thank my mentor Alexis and the ever helpful (some say omnipotent) Marco for making this possible for me.

Oh, and I plan to be hanging around of course. I forsee my school final year project crashing down on me tsunami-style very soon, but I still hope to get a couple of things done before 4.6 gets out. Will blog about those when the time comes ;)

Had been holding off posting on this until I have something reasonably pretty to show for it, but damn, said pretty thing sure took its time. Anyway, GSoC mid-term evaluations just ended, and I finally managed to hack up something that looks somewhat presentable. Here’s a short screencast of it.

Some key ideas of the current implementation:

• The tray lives in it’s own containment separate from the main activities/launchers.
• There is a passive “shrunken” form and an active “enlarged” form.
• The passive form is not interactive and click/tapping it simply switches it to the active form (with a simple QML animation thrown in). It shows a number of defined “always-show” applets and a limited number of the other applets, prioritized by recent activity.
• The active form shows all available applets in large, finger-friendly sizes, and is scrollable in case it doesn’t fit the screen (hello Plasma::ScrollWidget!). It behaves more or less like one would expect a systray to behave.

Not a lot to show for one month plus worth of development I admit :( This is actually my third implementation attempt. In the first attempt there were two “trays” – one small and one large, and taking hints from the comic plasmoid, I put the large one on a full-screen transparent QGraphicsView which covers the small one when activated – but this isn’t nice since it, in notmart’s words, relies on compositing being available to not look like crap. So the second implementation gets rid of the QGraphicsView and instead positions the large tray relative to the parent containment, but then there was another problem: we can only have one set of plasmoid applets running to save memory, and there was no way to display one set of applets in two trays. So I gutted the thing again and did this third attempt where there is only one tray which shrinks and expands accordingly… and that’s what you see in the screencast :)

This is far from finished of course – if nothing else that “cancel” button probably doesn’t quite belong there :P Iirc I should also be thinking about doing notifications for my GSoC, so there’s plenty more to do for the rest of the GSoC period. Fun stuff. Alright, think it’s time I wrapped this post up and go clean up my code and commit it – and pray I don’t get torn to bits by darktears and notmart for some terrible mistake or design flaw I made :)