Archive for March, 2009

Not only have I not stopped working on WebKit, but I even touched the debian/copyright file :-/

Update: and now I even uploaded a new xulrunner package...

Since I am expected to spend more than half my packaging time updating the debian/copyright file, I am hereby orphaning nspr, nss, iceape and xulrunner. I am also stopping work on webkit and iceweasel, but they don't end up in the orphan state since they are comaintained.

Good luck to my fellow developpers. (And sorry, sincerely)

Update: As I do realize writing while being pissed doesn't help making the motives right, and apparently, some people have seen this as an extortion, let me make things clearer:

I was starting to work on xulrunner 1.9.1 when the discussion about the copyright files came up. It will require a significant amount of time, and while Noah Slater's opinion alone wouldn't really have carried me that far, despite me saying so because I got pissed by his words, 2 messages from Jörg Jaspert (the only ones he posted so far in the thread, by the way) did make it clear that my work on xulrunner 1.9.1 was going to be a waste of time, which I already lack to properly handle the bugs in my maintained packages, let alone keeping the copyright file up-to-date.

As I am obviously unable to handle the amount of work required to maintain big packages, as drawing new blood in the mozilla team has always failed so far, I just prefer to stop than to over-overflow. Call it extortion to get people in the mozilla team if you want, I'm fine with the notion.

I've been thinking to stop working on big packages for nearly a year already, but never mentionned it but to a few people in a few occasions. I couldn't resolve myself to do it, though I did reduce the amount of time I spend on these packages (I was overflown, a year ago). I just found an excuse to actually do it.

I must say I feel awkward now, and I still don't know if I will be able to keep this resolution.

As for the new copyright file format, with full licensing information and copyright holders list, I *did* try, on a significantly smaller piece of software than the mozilla packages, namely on Webkit, which is not really small either but still 6 times less files than xulrunner. I must say I hate to have to list copyright holders and file names with a passion, and the amount of time it takes. It is the main reason why there wasn't more uploads of WebKit svn snapshots in the archive...

Last but not least, thanks for the nice comments.

You may have noticed, or not, but WebKit 1.1.3, which has been released a few days ago, is available in experimental. The great news is that we now have a real maintenance team, because I now have a co-maintainer, who actually did most of the work getting the 1.1.x releases in shape for experimental.

Now, some JavaScript performance figures, as I have been doing for most WebKit releases I uploaded to the archive, with a recap of previous episodes:

• svn 27674: 3932.0ms
• svn 31842: 2612.2ms
• release 1.0.1: 1658.4ms
• release 1.1.3: 1333.2ms

All there tests have obviously been run on the same machine, under the same conditions. The machine in question is my x86-64 laptop, which means all these test are with a x86-64 binary.

This also means the last release, 1.1.3, doesn't take advantage of the Just In Time JavaScript compiler, which is only available on x86 binaries.

With the x86 binaries under a x86 personality chroot, I get under one second:

• release 1.1.3, with JIT, x86: 985.4ms

But, in the last few days, I've been working on getting JIT first to build and then to work on x86-64 linux, and with the help of folks on the webkit-dev list, that just happened. And the result is just... wow.

• release 1.1.3, with JIT, x86-64: 623.0ms

(Note that a few tests are actually slower than on x86)

That's so many times faster than what we had a year ago that it's almost unbelievable.

Expect the next upstream release, planned for some time in the near future, to have my patches applied. In the meanwhile, you can get them here and here.