Weekly Status Report, W36/2010

Here's a summary of SeaMonkey/Mozilla-related work I've done in week 36/2010 (September 6 - 12, 2010):

• Releases:
  Shipped SeaMonkey 2.0.7 on Tuesday but pulled updates soon after for a startup crash seen by a number of users, which stops them from even launching the application or getting further updates, so we went for rather safe than sorry and stopped updating the other half of users still on 2.0.6 and not yet having applied the update. We will create a 2.0.8 release soon with fixes for the crasher and then will deploy updates straight to that one.
  Given that, I worked on fixing one issue related to the crash, but there's a second one pending that affects Firefox and Thunderbird on the same platform version as well. I also readied us to take a mail signature fix along with that update.
• Build Machines:
  Just when that requirement landed, I got note that we need to install YASM on Windows slaves and upgrade other systems, so I set the time aside to do that, and after some investigation on how Firefox release engineering did that, could complete it successfully on all our machines. IT also once again was very fast and helpful when one slave decided to go missing on the network. Thanks for that!
• Automated tests:
  I landed the test for reusing empty tabs that I created the week before. Apart from that, I didn't get to look into a lot in that category, but Callek, IanN, Aqualon, and sgautherie have been quite busy there, which is nice to see, as the oranges come down to manageable numbers now.
• Places:
  After some more reviewing, the fast bookmarking button has been accepted and I could land it.
• Doorhanger notifications:
  I did another followup patch, this time for lightweight themes installation, but given that Neil seems to have decided to rewrite the system, I'm seriously thinking of throwing the towel on the whole stack of bugs and patches, unless he'll backport all this work to Firefox as well.
• UA string:
  After doing a Council vote on including a Firefox token by default in our UA string, and that vote running positive, I got reviews and checked this in.
• Data Manager:
  After addressing Neil's last super-review comments, I uploaded a new version 1.0.2 to AMO with those improvements.
  Ian came around with some review comments on Data Manager, but from a first glimpse, either his build has problems or my code does. Need to find out about that.
• OpenSearch:
  After a number of additional hours spent on it, OpenSearch support has a patch up for review, but it looks like some points of my work are controversial and it will take some time to agree on something there, I fear. I filed a few followup bugs for things this first patch doesn't cover.
• Various Discussions:
  SeaMonkey Development Meeting, visit to Bay Area and Mozilla HQ, building with libxul and/or "fatlibxul", organizational future for SeaMonkey, more prompts converted to doorhangers, JavaScript speed and JaegerMonkey, message account wizard, Firefox UI work, etc.

We had a few messages recently about JavaScript speed and SeaMonkey, and it's quite fitting to those that the new "JaegerMonkey" method tracing just landed on the developement "trunk" and our nightlies this weekend. Out of personal interest and for calming people who fear we wouldn't work on speed, I did some runs of the popular SunSpider benchmark today. While the SeaMonkey 1.1.19 versions with which we announced the end of the line for the 1.x series comes in at 17501ms in that benchmark on my machine, a current SeaMonkey 2.0.x build comes in at 1738ms with the JS debugger enabled (default) and 1402ms with it disabled, which is a huge improvement - but just the baseline for current work. Now, today's pre-2.1 SeaMonkey build again just takes about a third of that baseline time and lands at 477ms - and the JaegerMonkey work isn't even finished yet. I think we are moving in the right direction.