With the Mesa fix Kristian recently posted the Qt drawing code for Wayland works well enough to draw the analogclock widget! Here’s hoping this is just the first of many Qt based clocks that will run on Wayland in the future.
Finally sat down with Mario to review his patches and ideas for making our OML sync implementation more accurate. He needs very accurate time stamping for his applications, and needs it in every configuration. He’s implemented patches to the DRM core, radeon and intel drivers to dramatically increase the accuracy and precision of our vblank timestamps. And we discussed a way to get accurate timestamps and necessary information for buffer swaps implemented with a blit. In short, we plan to have the kernel generate an event when a swap blit completes, and include info on the current scanline and all the usual timestamp fields. Hopefully Mario will find some time to post his series to dri-devel in the next few weeks so we can get it some wider exposure.
Had a nice night out yesterday (and this morning!) with various XDS attendees. Details omitted to shield the guilty, suffice it to say we met some real characters and had a lot of fun. Hope to go out to a nice restaurant tonight before flying back home tomorrow morning (will be nice to get some sleep finally on the plane, this has been a busy and mostly sleepless conference!).
I’m here in Toulouse, France (a charming example of an old French city) for XDS 2010, which is going well so far. The venue is good, we have a relatively large number of attendees compared to some of the recent X conferences we’ve had, and there has been good discussion in many of the talks and of course in the hallways
Kristian and I have been working more on the Qt Wayland port, and have basic apps like analogclock drawing. I’m working on fixing the Qt Wayland code to support more complete drawing so that other apps run; I hope to have sliders and other examples going by next week. At that point we can start the process of building libmeegotouch on top of the Qt Wayland environment and seeing what it will take to get some sample MeeGo apps running.
The ultimate goal of the work is to remove the X server (which acts as an obtrusive middle man in handset configurations especially) from the graphics stack for certain environments. We felt MeeGo was a good target because its apps are relatively new and shouldn’t have many X dependencies, and the philosophy of MeeGo is consistent with what we want to achieve. Overall, we hope for a reduction in the number of context switches (down from 3 or more to get stuff on to the screen with X and a compositor to 2 with Wayland), a reduction in the amount of code and memory needed to run the environment, and a massive simplification of the architecture, which should allow for faster and more robust development.
more to come (off to get drinks with krh, DrJakob, ickle, rib and others now)