The roads I take...

Here's a short summary of SeaMonkey/Mozilla-related work I've done in week 25/2011 (June 20 - 26, 2011):

• Mozilla work / crash-stats:
  Made instrumentation for finding out something that is nearer to the real crash rate of the 5.0 final build than what calculation on a single throttle rate could tell.
  Investigated how Socorro finds it crash rates to know what numbers of mine to compare to and saw how convoluted that code currently is.
  Looked at 5.0 crash data to see if anything sticks out.
  Attended the new Firefox release launch broadcast.
  Looked at my Flash hang reports and saw the trend I saw in the first reports go on. Will need to do further investigation there soon.
  More discussions on JS crashes, trying to work the JS team to get to instrumentation and fixes there.
  Met with the CrashKill team to discuss my proposals for Q3 priorities we see in Socorro work and forwarded that draft to the Socorro team.
  Engaged in discussions on a GC fix that should reduce Firefox memory usage and out-of-memory crashes, but has some risk for being fast-tracked into Aurora and Beta.
  As always, looked into "explosive"/rising crashes with my experimental stats.
• SeaMonkey Build & Release:
  Helped Callek to get 2.2 Betas moving forward.
  Checked in my patch for removing branch hacks from comm-central.
• SeaMonkey 2.1 UI:
  I had a number of discussions on SeaMonkey 2.1 code I worked on, including the new zoom backend support and Data Manager - bugs and ideas have been brought forward, other people might work on improvements to the work I started there, which is a good sign after all.
• SeaMonkey L10n:
  Worked on more L10n sign-offs for SeaMonkey 2.2 Beta, more and more locales are joining the train.
• German Mozilla Community:
  I pushed the new design for the German Planet Mozilla, as we now have mozilla.de in that new design as well.
• Various Discussions/Topics:
  New Firefox process in the Enterprise (including the actual discussion this is mocking), Linux 3.0 Mozilla build bustage, more SeaMonkey 2.1 release feedback, MemShrink and memory reporting, user volume on Aurora and Beta, MeeGo N900 CE, N9/Harmattan announcement, Mozilla sponsorship for SotM-EU, getting SSL to run on a first domain on my own web server, etc.

A lot of things happened this week, including good and bad messaging in the press for Mozilla, as we held the promise to go for faster release cycles. There is a lot of hope, but also fear, uncertainty, and doubt out there, both in our community and the wider public. We need to address that, but I think we can. I'm excited about the new process because we have an unprecedented time of only stability and polish work on every release when they are in Aurora and even more in Beta stage, and we can really concentrate on doing good crash analysis and catching of nasty bugs there. For this release, we weren't yet confident to call it the most stable release of Firefox ever, but following this new process, we will soon be able to say that based on real data we have. I'm really looking forward to that and to being part of this great story.
And to everyone who still has fears, uncertainties and doubts about it, try to think about how you could embrace this new process of a lot of small steps and only after careful analysis and weighing against taking huge leaps with more time, where you are more prone to stumble when you finally move. Let's try to stay positive and think about how we can make things better on the base of what we have now instead of how a new thing could make some part of the world collapse. Let's stay positive and let's make this world more modern and replace doing revolutions after long times of stagnation with evolving more continuously instead. I believe that could help everyone

Inspired by dolske's great post on Enterprise integration for Firefox (thanks for mentioning my work!) I have been thinking and found that I have all the means to present just how well the new Firefox release process fits into the Enterprise:



That said, I really think it fits even into other enterprises and we should help them come along on that great ride into the future and our ongoing voyage on the Mozilla mission.

(P.S.: Thanks to dolske for the great train of thought and to Paramount for the great DVD set of the recent Star Trek movie that I could use for the picture.)

Technology. The current frontier. These are the voyages of June 21st, it's ongoing mission to find strange new releases, new devices, and new software. To boldly go where no day has gone before...


This is surely an interesting day. Nokia has just presented the Linux-powered Nokia N9 with a completely new UI and it's surely a very slick device, interesting UI concept, and no matter if system-wise this midway point between Maemo5 and full-fledged MeeGo can be called "MeeGo" legally, having a mass-market phone out there that comes with a fully open "real Linux" is awesome.
The N900, which feels old, slow and clunky nowadays, has a damn good successor - even though the keyboard-attached N950 version has been blocked by carriers and is only available as a "loaned" dev kit to people creating N9 apps. I hope to see that N9 device out there soon, and perhaps it's done well enough that the absence of the keyboard can be taken, but I'd really need to test it for that. Also, I hope that enough of the UI stuff can be opened enough that MeeGo proper can ship it as well. Until all that clears up, I'll keep testing the MeeGo N900 Community Edition, which is shaping up nicely as well. Hopefully open-software phones have a future with all those moves (and I surely hope other vendors will chime in as well, as Nokia can't be fully trusted in that way any more).

But there's much more: I just saw Mozilla people on the US West Coast join IRC at 5am their time and start their work day - Firefox 5 is going public today as the first one off our new release process. While it doesn't ship a lot in terms of new features, the big thing here is that it kickstarts the new process that will get us new Firefox releases every 6-12 weeks that are easy to update to because they don't have a ton of new stuff but still a number of nice features. This time, CSS Animations are probably the only larger thing (next to performance improvements), and most users won't notice them yet, unless they look for some demos. But, the important point is that they're ready and so we can ship them to hundreds of millions of people, not needing to wait for a major version coming in a year or so. It's (going) out there, right now!
This is also the first release I have been there in "Crash Scene Investigation" for its whole cycle, and we learned a lot about a number of things in this cycle, including that we need to attract more people to the Aurora and Beta channels to get even better data, but also that there are some classes of crashes and hangs we need to take a closer look at, and we are doing that. All in all, our beta numbers of Firefox 5 have been quite good, we expect it to be at least as stable as Firefox 4.0.1, probably somewhat better.

In addition to this, Mozilla is shipping the probably last security update to Firefox 3.6, Thunderbird ships a security update for 3.1, and, very importantly, Firefox 5 also ships for Android and the before-mentioned N900 (maemo5) today, right at the same time with the desktop Firefox!

Not enough, though: The SeaMonkey team has just finished up building the first beta of SeaMonkey 2.2 and will ship that to its beta testers later today. This version has the same web-facing features as Firefox 5 (including CC Animations) and the security fixes shipped in other versions today, as well as a number of smaller fixes to SeaMonkey code, some of which have been found since 2.1 has been released.
I'm informed that the team will try to ship the final 2.2 release as soon as possible in the next weeks, hoping that this first beta will do well in testing.

And it's surely possible that this day has even more in store, it's not over yet!

Engage!

Here's a short summary of SeaMonkey/Mozilla-related work I've done in week 24/2011 (June 13 - 19, 2011):

• Mozilla work / crash-stats:
  Looked at data of new 5.0 beta builds, but there was a problem with the update system so uptake was a bit slow. Still, the more users we have on beta, the better the data looks.
  Did run some investigation on hangs and Flash versions and preliminary could find that the 10.3 versions of Flash seem to be more prone to hang than older versions. This also would explain why 5.0 has a higher hang rate than 4.0.1, as users on beta are more likely to have newer Flash versions installed.
  Discussed JS PGO a lot, and even though I still am pretty sure it doesn't make us crashier, we'll probably end up turning it off on Aurora as well and first look into more GC and MethodJIT crash instrumentation before we flip that switch again.
  As always, looked into "explosive"/rising crashes with my experimental stats.
• SeaMonkey Build & Release:
  Proposed for us to go directly to 2.2 with major updates from 2.0 and therefore double up on pushing 2.2 out soon, Callek and others seem to agree.
  Helped getting aurora builds on track, among other things by updating the automated update system (AUS) server we use, so it recognizes by-channel settings.
  When I first made trunk updates going, I introduced a typo that broke updates but I solved that fast after it got reported.
  For 2.2b1, I helped getting the initial update config files in place so that we ship automated updates to all 2.1 (beta and final) users that are on the beta channel.
  On the build system side, I did a patch on removing the Mozilla-branch-based switches from comm-central.
• SeaMonkey L10n:
  Sent another message to the L10n list and accepted more L10n sign-offs for 2.2 beta, most locales we shipped in 2.1 seem to be on track to get into 2.2 as well - at least I hope to hear back soon from those that aren't accepted in yet.
• Themes:
  I spent a number of additional hours working, and finally, could update my personal theme page to have the new 2.1 versions of the EarlyBlue and LCARStrek themes announced! Reviews on addons.mozilla.org are still pending, but I did upload them there as well - both work with SeaMonkey 2.1 and probably well enough with the newer versions as well, and in addition, LCARStrek now supports Firefox 4 (and probably newer ones) as well!
• Various Discussions/Topics:
  Lists across multiple domains in Data Manager, hg precommit hooks for bug numbers in commit messages, update to Linux 3.0-rc kernels and seeing compile problems now due to that and a different bug, SeaMonkey 2.1 release feedback, MemShrink and memory reporting, errors and assertions, beta build turnaround speed, MeeGo N900 CE testing, sponsorship for SotM-EU, etc.

Things are moving fast in Mozilla land - while SeaMonkey just released 2.1 last week, we've been using that one to get "2.3a2" builds from aurora to work and update correctly and get ready for doing the first and potentially last beta for 2.2, which should go for a release within the next 2 weeks or so.

And Firefox is releasing 5.0 tomorrow! I'm excited to see how it does on crashes, as this is the first release on which I've been around in "Crash Scene Investigation" the whole cycle. So far we are looking good, but we still have way fewer users on Aurora and Beta than we would like to have for comprehensive pre-release crash analysis, so the first days of sending it out to users will give us a better picture. Given that changes compared to Firefox 4 are small and the beta numbers we have so far are looking really good, I expect this to be a smooth release in terms of stability - but I'm standing by to help figure things out in case that will not be the case.

Here's a short summary of SeaMonkey/Mozilla-related work I've done in week 23/2011 (June 6 - 12, 2011):

• Mozilla work / crash-stats:
  More investigation of Firefox 5.0b2 vs. 5.0b3, but mostly waiting for data on 5.0b5 to come to a conclusion about this upcoming release. We still have too few people on the Beta and Aurora channels to make good and swift assessments of crashiness, though.
  Still concerned about the hang volume on 5.0, we'll need to put more investigation into this.
  Came to the conclusion that turning off PGO for JS between 5.0b2 and 5.0b3 probably did not change the crash volume, but only shifted crash locations back to well-known signatures (affects mostly GC and MethodJIT crashes).
  Found a significant decrease of hangs on trunk some time ago, which in turn led to realizing that the problem it fixed affected Aurora and the fix needed to land there as well.
  Of course, looked into "explosive"/rising crashes with my experimental stats.
  Worked with the bmo and Socorro people to push and verify the new Crash Signature field in Bugzilla.
  Made sure the bugs targeted at the Socorro 1.9 milestone are not forgotten as that release is dropped.
  Started work on Q3 priorities/goals to hand to the Socorro team.
• Jökulsárlón Download Manager
  The new 0.4 version I submitted last week just got review and is now available on AMO!
• SeaMonkey Build & Release:
  I helped Callek with a small things to get 2.1 final done and out the door, including the fix of a glitch in update generation.
  Got a decision and implemented the version bump on post-2.1 trees.
• SeaMonkey L10n:
  Accepted a few L10n sign-offs for 2.2 beta.
• German L10n:
  Updated inspector L10n, readied language packs for ChatZilla and venkman for AMO, updated trunk localization.
  Localized the 2.1 announcement, corrected localized relnotes, and made 2.1 public for German as well.
• Themes:
  Did some more work on LCARStrek for Firefox 4, this is starting to look good.
• Various Discussions/Topics:
  Ad/malware crashes, non-granted reviews on FF search code and changed dataman invocation, SeaMonkey 2.1 release feedback, mozilla-inbound, multi-process tabs, MemShrink, beta build turnaround speed, MeeGo N900 DE becoming Community Edition (CE), making desktop layout more fitting for full-HD resolution, etc.

The big news for this week surely is that SeaMonkey 2.1 has finally been released! The last few weeks have been a bit bumpy, with various packaging and installer issues, and it's still a bit unclear if all of them have been resolved, but we've finally done it. I was more involved in the final shipping than what I should, but Callek's TODO list is piling up and I could help with a few small tasks that made things run more smoothly and faster overall, so I did that. This is a somewhat emotional moment for me, as I have considered 2.1 "my last SeaMonkey release" for some time, but I want to address this in a separate post (hope I find time for it).

And in terms of my "Crash Scene Investigation" work, things are shaping up nicely, from closely watching Firefox 5 via tooling improvements to working on goals for the next quarter, a lot of items are moving - and that's always a good sign.

Originally coming as a request from the Socorro team, we have been pushing for some time to get a better solution for connecting bugs to crash reports/stats than the bug summary - in the form of a separate "Crash Signature" Field in bug reports.

It wasn't just messy up to now for Socorro (crash-stats) to parse the signatures out of headers, we also had a number of bugs where the constraints on the summary length caused problems with adding multiple signatures or a very long signature - not to speak of making summaries hard to read for humans. All that should get solved with the new special-purpose field that also allows greater length of its content.

You might have noticed that this "Crash Signature" field has turned up in many products of Mozilla's Bugzilla in the last days, below the tracking/status/blocking flags on the right above the comments.

The syntax for entering signatures is similar to what we did put into summaries so far: a signature is listed as [@ crash::signature(params) ] - for multiple signatures, we prefer to use a new line for every signature to make the field more readable, though.

We're migrating over the contents from summaries to the new field, so you don't need to copy those manually - some of that happened already, but the script didn't catch all necessary bugs the first time it ran. Also, the work on Socorro to use the new field has not landed on production yet. Both should be done soon (within days), and then we can fully use the new field. Until then, it's best to leave the field alone and keep using the summary (though for those cases where the summary is too small, populating the new field right away is safe).

In the future, i.e. once Socorro uses the new field, we'll also be able to make summaries more human-readable, e.g. "[meta] EnterMethodJIT crashes" or something similar will be far easier on the eyes for bug 595351.

Thanks to everyone who worked and is still working on implementing that!

Here's a short summary of SeaMonkey/Mozilla-related work I've done in week 22/2011 (May 30 - June 5, 2011):

• Mozilla work / crash-stats:
  We had a lot of discussions and looks into data of Firefox 5.0b2 vs. 5.0b3, esp. since we have turned off PGO for JS between those as we supposed that might help crash volume.
  My report on crash volume for different builds turned up that the volume of hangs is a lot higher on 5.0 than on 4.0.1, and I forwarded my concerns about that.
  As always, I continued to look into rising crashes with my experimental stats.
• Jökulsárlón Download Manager
  I submitted a new 0.4 version once I had done some code improvements, most of which I wanted to do for a while, and implemented observing the closeWhenDone pref that also exists on the toolkit download manager, as well as a new pref that stops constant flashing when a download is going on and we're set to an app tab on Firefox 4 or higher.
• (Tahoe) Data Manager:
  Some of the code improvements noted for Jökulsárlón also apply here the the Data Manager, so I did them there as well and also made a test more stable, submitted for the SeaMonkey version as a patch as well.
  I didn't really make progress on the web storage panel, but as I don't see how to get events for the others, I decided that dynamic updates there will be for DOM storage only for now.
• SeaMonkey Build & Release:
  I helped Callek somewhat to get 2.1 RC2 done, as we saw some strange woes there.
  For the future builds from the new release process, I put forward a proposal for detailed version numbers so we can move forward there. This will make builds from current comm-beta called "2.2" once they are generated, from comm-aurora "2.3a2" and from comm-central "2.4a1" - changes on all those are minimal compared to 2.1, but they catch up with Gecko feature and security updates.
• SeaMonkey L10n:
  Filed a bug to get sign-offs going on esp. the comm-beta branch so Callek can build a 2.2 Beta 1 as soon as 2.1 is done.
• German L10n:
  Got the SeaMonkey part of beta and aurora synched with the respective English versions, and also synched up SeaMonkey on trunk to the current state once again.
• Various Discussions/Topics:
  Build/test turnaround speeds, getting access to machine for custom reports (or not yet), beta build turnaround speed, user volumes on branches, channel switcher removal, removing microsummary support from SeaMonkey, MeeGo N900 DE, getting desktop to run with 3D support, Endeavour coming home for the last time, etc.

I seem to not succeed getting those reports out as timely as I should, and some other things I should or want to get done are not happening as well, even though I on the other hand get a number of things done. I guess the day just has too little time - or I'm spending too little of it in front of the computer... Well, with the current climate being relatively hot and humid here, more time staring at this screen would just be more straining to those eyes that seem to get tired easily in this setup anyhow. Maybe I'll find some other solution.

That said, SeaMonkey is nearing its 2.1 release, Firefox 5 seems to be shaping up reasonably well in terms of stability, and Socorro just shipped an update that incorporates a few things our team has put on the last priorities list, so I think I can be satisfied with what's going on in the areas I'm helping with!