Today I decided to do something about it and came up with awful XSLT (I believe it’s impossible to write XSLT that doesn’t make your eyes bleed) that converts the Apple documentation into a nice devhelp book you can browse. I can also get the function/keyword index from individual reference files (e.g. NSOpenGLContext) quite easily. Now I just need to figure out how to list all files automatically and how to organize the results better (all-in-one and class-per-book don’t quite work for 700+ html files.)

First there was massive breakageimprovement to support multiple stages in clutter. It’s really good news for like clutter-gtk, allowing you to have multiple clutter embeds in an application. The bad news was that it changed the backend API and broke OSX build.

Somehow somewhere in the middle of relocating myself to London I managed to get the OSX backend back working again. But not before the shiny new cogl, another big rewrite landed, and broke the build again.

Good news

Little hacking later cogl is building again and multistage works on OSX. Once those, and a couple of other minor patches have been reviewed and committed clutter can be built on OSX again.

On top of that there’s a patch to make it possible to build clutter on OSX without gtk+ using instead Core Image (Quartz) to load the textures. Should make it a little bit easier to bootstrap the building process. It’s not as big a deal as one might think as clutter is in MacPorts now.

Bad news

The bad news is that with all the patches to make clutter build on OSX again it can produce weird artefacts or plain crash. I was assuming it had something to do with NPOT textures but after poking around with OpenGL Profiler I’m not so sure anymore. Using OpenGL Profiler can crash some clutter test programs, even the ones before cogl rewrite and multistage, with invalid memory access. Seems to happen only with the native driver and some combination of extensions and whatnot.

My totally unfounded guess is that I’ve gotten some reference counting wrong or missing locking somewhere, but so far any attempts to tweak things haven’t worked and I’m running out of ideas. Help?

Since leaving Nokia I’ve had the pleasure of meeting the team and hunting an apartment in London (still can’t quite believe the prices!) Found myself a nice-ish place in Fulham where I’ll be moving this week. My mobile number will cease to function, send me a note if you want my new number.

The bureaucracy involved with moving to a different country is interesting. Fortunately as EU citizen I don’t need a visa, or basically any permit or notification to move to another country in EU, but the dead tree mess is still annoying. Basically in order to get a bank account you need a UK address, in order to get an address you need a UK bank account. Fortunately relocation agency was able to help there somewhat, though not with the 18 page lease I had to print and scan.

But still, I couldn’t get phone line activated as BT can only handle local VISA… weird, need to get some local to pay initial deposit. And you can’t even order any reasonably priced Internet connection without a phone line. I’m already feeling disabled for the lack of Internet.

Anyway, this is going to be an exciting year.

This marks an end of era for me for I’ve been working for Nokia almost a decade (had my interview in Nokia House May ‘98 — Hi Patrik!) latter half of which I spent in a great team, excited to be able to help Nokia evolve into something new and wonderful. Unfortunately lately I’ve lost the excitement and work has become too much of an energy drain for me. The energy I feel like I’ve been putting in isn’t paying off and I’m unable to sustain it any longer. To some extent I had already given up which I find an alarming sign.

I’m fully confident Quim and Ari will carry on making maemo an exciting platform to work with. It’s just time for us to part ways.

I’m going to miss some people I’ve grown to know, others not so much, and some I’ll be seeing again soon enough. There’s an exciting opportunity I’m following. More on that later.

Now jobi invited me to mugshot and all I can think of is “Not another site registration.” I think I’ll pass. Nothing against jobi or anyone in mugshot or facebook or myspace, but enough is enough.

gtk+3 bling etc.

I made my presentation on maemo and gtk+ describing our current situation with our patches as well as some pain points we’ve found during our time with gtk+. We continued the discussion about theming separately with several people, trying to bring up all the different needs on the table as toolkit developers simply aren’t currently aware of them all. Benjamin Berg took the ball and is pinging everyone to update the wiki

The discussion around gtk+3 was surprisingly aligned. The consensus seems to be that current gtk+ is a dead end in terms of developing important new features. Most of the time in reviewing patches nowadays goes to trying to guess whether it will break some applications. Making any changes to the core of gtk+ simply isn’t feasible. gtk+3 should be designed in such a way that it would be more easy to modify.

As, again, no one really has all the requirements for more modern toolkit so lots of experimentation, with clutter, lowfat, HippoCanvas, etc. is strongly encouraged. And then next GUADEC the results could be evaluated and discussed.

gtk+2

gtk+2 is still actively maintained. Things scheduled to be considered for 2.14: offscreen rendering, filechooser API, press and hold, GVFS (now would be a good time to review the GVFS API.) Extended layout SoC project is progressing nicely, natural size requisition is working already so we might want to have a peek and maybe solve some of our remaining problems.

online desktop

Apparently controversial topic that raised quite some discussion. In my opinion seamless integration with network services would make a lot of sense with mobile device like n800. I’m thinking the device should be just a big cache for things like Flickr.

Oh, whatever happened to NFlick again…?

Hiker

Hiker is about UI-less services, so no gtk+ on this level. Alarm and notification services are a potential area for collaboration, not only in mobile space but also on the desktop. One thing that sounded interesting was the Bundle Manager which allows creating Mac style application bundles, applications reside only in one directory as opposed to spread out in /bin,/share,/lib,… and that directory can be easily moved around.

But then again, we have apt and dependency resolution so it might not that interesting.

Hildon

One point to take home from Practical Project Maintenance: Reduce friction!

We need to make Hildon an easy project to approach if we are to attract outside developers. So jhbuild moduleset you can use on average desktop, integration with gdm or startx rather than running 10 different commands to start up the session, normal l10n practices, transparent processes, etc… Entry points to the project need to be clear.