Axel Hecht Mozilla in Your Language

January 31, 2007

last call for FOSDEM presentations

Filed under: Mozilla — Axel Hecht @ 4:08 am

I need to hand in the schedule for the Mozilla Developers Room at FOSDEM in by the end of the week, so I’m making a last call for presentations. If you want to give a talk, please add yourself on the wiki.

January 29, 2007

Wasting time in Eclipse

Filed under: Eclipse,Mozilla — Axel Hecht @ 5:13 am

I’ve just spent the second half day trying to get the aOW plugin to work. Honestly, just to install its dependencies. So instead of trying to find out if aOW could help me in my work, I spent some quality time with each and every UE bug that the plugin manager in eclipse has. Like, claiming missing dependencies without giving any clue how to resolve them. Googling for one of my missing deps actually claims that it’s not missing at all, but conflicting. And IBM’s calisto all package just totally broke down and just conflicts with no further explanation at all. All of this of course with numerous slow downloads, loads of dialogs, and no chance to anyhow keep the work you already gone through.

If we ever really rely on extension dependencies, we should have a close look at Eclipse’ plugin manager and just do the opposite.

January 12, 2007

doing jars, baby

Filed under: Mozilla — Axel Hecht @ 12:45 pm

So I successfully hacked my python stuff to create jars, and to do so for python 2.5, too. Testing it in the MozillaBuildRC1, here’s time make -C browser chrome for the current code and my new hack:

real    0m21.812s
user    0m20.731s
sys     0m11.761s

vs.

real    0m10.375s
user    0m9.028s
sys     0m5.158s

from scratch. The numbers don’t change significantly if you run them again, which is to be expected, as most files are preprocessed and overwritten anyway.

January 11, 2007

python coming to a build near you

Filed under: L10n,Mozilla — Axel Hecht @ 3:41 pm

As a heads-up, I’m making some progress to use a single python process per jar.mn in bug 361583. AFAICT, I have all the odd ends figured out to merely run one python script per jar.mn, instead of starting a two perl processes per jar.mn, and one per preprocessed entry, plus two calls for zip.

I had to port preprocessor.pl and make-jars.pl over from single scripts to modules. Porting them to python looked like the smaller change at the time, even though I needed to code quite a bit to actually get ‘!’ working in the preprocessor. Damn python for using not to negate expressions. And I had to subclass zipfile.ZipFile to implement overwriting entries in zip files. That seems to work, now I need to hook it up to the jar maker, which currently only does flat chrome. Oh, and I need to make that code readable for people that don’t have direct access to my brain.The good news is, it seems to be blazing fast, my test suite contains a copy browser/base right now, and the current realchrome target takes some 4 seconds on my windows machine, the new one takes less than a second.

Oh, and yes, this is really a step on my path to require python and just python for localizers to create langpacks, and then repackage builds, too.

From a requirements point of view, I’ll be testing this at least on python 2.3 and python 2.4, I don’t have python 2.5 installed on one of my machines just yet. Which I should, as in particular the zip stuff seems to have changed in the backbone.

The Preprocessor.py is attached to the bug, so if you want to have a look, be my guest.

Powered by WordPress