The roads I take...

Here's a summary of SeaMonkey/Mozilla-related work I've done in week 26/2009 (June 22 - 28, 2009):

• SeaMonkey Build/Release Harness:
  Followup work on getting SeaMonkey release automation (also has wiki documentation now) from testing into production to actually work is continuing. The main bug now has the SeaMonkey-specific configuration files attached for review, I tried to make repack independent of ChatZilla or venkman being enabled (doesn't correctly work yet, though), filed a bug and (trivial) patch for pretty names in Windows installer and added shipped-locales files to both trunk and the 2.0a3 release tag (release automation needs to fetch that file from the last release to determine what to generate updates for). The removed-files.in fixup I wrote up after analyzing update verification logs from the release harness has landed, which should also make us ready for shipping static builds in nightly updates.
  I did a small update to the patch for DMG unpackaging in the buildsystem and updated the patch for repack factory abstraction once again for some bitrot from buildbotcustom changes.
• Release Process:
  SeaMonkey 1.1.17 has been released on Monday, containing a good number of security fixes compared to 1.1.16. I continued uploading contibuted builds as they came in.
• Download Preferences:
  As I start to see more and more people being confused about the behavior preference on the download panel not working I started the work on updating the download preference panel to reflect changes from the download manager switch. This will focus on making the behavior and location prefs work correctly so that 2.0b1 will behave itself, more improvements can be done in later followups building on this.
• Bug Triage, Support Mails, Start Page:
  I spent some time looking at bugs that were changed again after the large NEW->UNCONFIRMED change, trying to get actions to happen where possible, or deciding to WONTFIX them in a number of cases (such decisions need module owners or Council members in most cases).
  Also, I finally came around to work the backlog of support mails I had in a subfolder of my inbox (I don't feel responsible for support, so I push them in there, but try to at least give some reply with pointers when I come around to it - might take weeks to months though). I ended up writing 70 replies in 4 hours, taking 3:30 minutes per mail, mostly with standardized replies, sometimes one or two sentences containing more specific help.
  The "start" page on the SeaMonkey project website has been warning people that their alpha/beta builds were outdated when they were older than four weeks which is not the best idea in the light of e.g. Alpha 3 being four months old and still "current" right now. I finally came around to modifying the page to restrict this warning to nightlies and doing a special warning for non-current alpha/betas.
• German L10n:
  Fixed some ChatZilla and general SeaMonkey strings to keep the de locale green.
• Various Discussions:
  Checkin for packaged tests uploading glitch, DEL and other keys in download manager, OpenWebCamp Vienna, Parallels VM adjustments, FF 3.5 release preparations, Scheduling of Mozilla's Weekly Update Meeting, thundertab restore and SeaMonkey, etc.

I have a very strong opinion about how successful our planning of SeaMonkey 2.0 Beta 1 in sync with Thunderbird 3.0 Beta 3 went - and believe me, there's nothing positive on it. Since we defined a freeze for our first alpha, we have learned that only a definitely scheduled freeze date will bring people to pick up speed and concentrate on the really needed stuff for that release. Sure, there's a lot of stuff to do in general, but usually only a few items that really need to go into a release. Most people fail to deliver on those things unless there are deadlines for making it happen. Thunderbird 3.0 Beta 3 is a glaring failure in that kind of scheduling, and with making SeaMonkey 2.0 Beta 1 dependent on that milestone, we ended up with a miserable performance on scheduling and delivering ourselves. If we had know how long that short delay would take, we might have done an Alpha 4 just before the download manager landing and still would be ready to wrap up the beta right now.
In any case, I'll propose uncoupling this beta from the Thunderbird cycle, set L10n and code freezes to happen soon and deliver SeaMonkey 2.0 Beta 1 within the next few (meaning really few) weeks, independently if Thunderbird ships a Beta in the fifth month after their most recent one or not.
We'll set freeze dates at this week's SeaMonkey Status Meeting and follow up with posts on the relevant groups and lists.

Here's a summary of SeaMonkey/Mozilla-related work I've done in week 25/2009 (June 15 - 21, 2009):

• SeaMonkey Build/Release Harness:
  After lots and lots of hours spent testing and fixing things, I could get the automated release harness for SeaMonkey to actually work, producing a bogus test-only "2.0a4" from current code, the resulting patches for buildbotcustom factory abstraction, dealing with brand names in update tools, DMG unpackaging in build and same in tools are all awaiting reviews.
  As a followup, I added a few files to be removed on update so that resulting complete or partial updates (yes, we can do the latter now for releases) end up as clean as they should be.
  I also updated the patch for repack factory abstraction for some bitrot, attached a patch for a small unit test packaging followup and made package-compare run in comm-central builds.
• Release Process:
  Continued the SeaMonkey 1.1.17 release process towards a planned public release in sync with Thunderbird 2.0.0.22, containing a good number of security fixes compared to 1.1.16.
• Misc Work:
  Did some changes in how my SeaMonkey development website retrieves and stores data: Weekly Bugzilla stats are now requested more efficiently using the microsummary ctype instead of full lists to retrieve the count and the results are kept in a DB, once I have more data and time, I'll make more than 3 weeks available on a separate page. Also, SeaMonkey 1.x update notification ping stats are now stored in a DB on the server, still need to write a report on those.
  I also fixed the Thunderbird and SeaMonkey nightly file paths on www.m.o/developer.
• Various Discussions:
  DEL and other keys in download manager, OpenWebCamp Vienna, Bugzilla improvements, NEW->UNCONFIRMED mass change, etc.

With all the work I did put into that in the last weeks, I hope we will be able to actually ship the SeaMonkey 2.0 Beta 1 release off the automated harness which can produce L10n builds as well as partial updates, both of which we didn't have previously. Additionally, using that buildbot-based mechanism makes it much less manual work to get the release done. We still need to do a bit of discussion if we can live with using so-called "pretty" file names for the release (e.g. "SeaMonkey Setup 2.0 Beta 1.exe" instead of "seamonkey-2.0b1.en-US.installer.exe") and we also have no signing infrastructure for Windows builds, but we can fake it by copying unsigned builds to the final place and continue to ship unsigned releases as we did up to now.

Next to that, I'm realizing that a few areas of our project are pretty dormant and I can't find time myself to move them much forward myself. this especially concerns QA and marketing. We really would need people to help making progress there - if you can support us in any of them, please contact me!

Here's a summary of SeaMonkey/Mozilla-related work I've done in week 24/2009 (June 8 - 14, 2009):

• SeaMonkey Build Machines:
  Started working on the automated release harness for SeaMonkey, which needs a quite extensive list of new buildbotcustom factory abstractions, including some in the CC*Factory classes. I'm testing with my own Mercurial user for now while still waiting for seabld hg access.
  Followed through a Parallels update and upgraded our Mac VMs to 10.5.7, but that unfortunately didn't fix our network losses, application losses and VM crashes we're seeing esp. with Mac VMs there. Tickets are filed with Parallels, we hope to see a resolution soon.
• Release Process:
  Continued the SeaMonkey 1.1.17 release process towards a planned public release in sync with Thunderbird 2.0.0.22, containing a good number of security fixes compared to 1.1.16.
• Download Manager:
  The missing download icons on Mac are checked in now.
• Extension Updates:
  We're now sending the app version for extension update requests so that AMO can determine if the add-on is compatible with our version or not.
• SeaMonkey L10n:
  search-rdf.dtd is now included in the build, so that the advanced search sidebar works correctly again.
• German L10n:
  Updated the tree for a few mailnews changes and the exclusion of PalmSync as well as a ChatZilla change and a mozilla-central change to about:plugins, but only after verifying that the "merged L10n" SeaMonkey builds worked even without those changes, displaying the missing strings in English.
• Various Discussions:
  Bugzilla improvements, NEW->UNCONFIRMED mass change, Thunderbird API refactorings, Firefox 3.5 going for RCs, etc.

Our friends in the Thunderbird project are making nice steps towards cleaner APIs for the mail and news back- and frontends (this week it was folder display and thread pane, earlier they already improved the folder pane itself), it would be nice if someone could help bringing those improvements to SeaMonkey, which would make it easier to work on other code improvements - something you, my dear reader, could help with?

Here's a summary of SeaMonkey/Mozilla-related work I've done in week 23/2009 (June 1 - 7, 2009):

• SeaMonkey Build Machines:
  Updated the L10n repack factory abstraction patch for a syntax error, and while I was there, filed a bug and patch for moving comm-central repack factories into buildbotcustom.
  I also filed a bug for seabld hg access, which is a prerequisite for release automation, but didn't come around to actual work towards that goal.
• Release Process:
  As Thunderbird 2.0.0.22 is upcoming, I started the SeaMonkey 1.1.17 release process the same day the Gecko release branch was cut for those versions. People are encouraged to test the candidate builds. This release will contain a good number of security fixes compared to 1.1.16.
• Download Manager:
  I obviously missed that reusing toolkit icons works badly if they use differently arranged images for different platforms, so I needed to fix missing download icons on Mac and made us safer against toolkit changes in the process.
• Fishcam:
  Our famous Ctrl+Alt+F "easter egg" displaying the Amazing Fishcam website has been throwing up warnings for some time about that key combination potentially not being available on all platforms. When I read a discussion about a similar problem in Firebug, I also found a possibility to suppress the key warning, which I implemented. This is no important function, it's a fun thing, so we just don't care if there's some systems out there where the OS uses the key combination otherwise and it doesn't call the fishcam.
• SeaMonkey L10n:
  As I didn't come around to land the necessary string changes for German this week, I could test with its L10n builds that "L10n merge" is indeed working on the new "comm-1.9.1" buildbots: If some strings are missing, the L10n repackaging system inserts the original en-US strings and the localized nightlies end up actually working! Thanks to Axel for figuring out all the hard work, I just needed to turn this on.
• Various Discussions:
  Target milestones in the SeaMonkey Bugzilla product, waiting for Parallels fix, comm-central commit access, sr policy, Fennec L10n, ISP info for account autoconfig, Firefox 3.5 nearing finish line, etc.

While tabbed mail for SeaMonkey is getting reviewed slowly but surely, the Thunderbird guys are working on a lot of internal reworks and code improvements in their UI code (e.g. included in the easy mail search work), and I strongly believe that we should get those applied to SeaMonkey code as well as those built a good basis for future improvements on the UI as well as make it easier to port Thunderbird work to SeaMonkey and our work back to their product. This is nice stuff to work on for someone who doesn't know our code that well yet, as most is just applying their patches on our side and refining that following review comments. Some work for you to pick up?

Here's a summary of SeaMonkey/Mozilla-related work I've done in week 22/2009 (May 25 - 31, 2009):

• SeaMonkey Build Machines:
  Some more work on the new buildbots, enabling static builds for nightlies but not normal builders. This reduced download size between 3% and 10% depending on platform, I would be glad if someone could do some performance testing on nightlies from the new and old buildbots to get numbers on static vs. shared builds in that regard as well.
  PalmSync will be removed from the comm-central tree soon and doesn't work with static builds, so I removed it from the new buildbot config right away.
  I also cared that the work for uploading packaged tests done by Serge was enabled on the new buildbots - this now enables anyone to run our test suites, even without having a full build system installed.
  The patch for L10 repack factory abstraction was updated for some bitrot, but not changed significantly. While I fixed that bitrot, I found some trailing whitespace and decided to fix whitespace in all of factory.py while I was there.
  When I investigated package file changes for static builds, I realized the the new CCMercurialBuildFactory doesn't call package-compare and wrote a patch for it.
  And after updating our new config for running compare-locales, the L10n build cycles now do that additional check and might even try "L10n merging", i.e. trying to produce working builds even when the localization isn't complete by merging in en-US strings - but I have not proven this to work for SeaMonkey in any way yet.
• Static Nightly Builds:
  As said above, I enabled static nightlies on the new buildbot configuration, and therefore looked into open issues we still have with making this the official configuration. Therefore, I made a patch for missing icons on Windows and finally even did a large fixup for package files and the removed-files list so that those are easier to manage in the future and so that static builds ship the right files without too much noise (and remove shared libs when a shared build is upgraded to a static one).
• Download Manager:
  The really big story of this week is the download manager landing.
  My work on the new tree-based download manager UI and download progress window rewrite as well as a prefs followup could land right away with the main work for switching from the xpfe to the toolkit backend.
  As followups, I filed and landed patches for making an ifndef really work and fixing a Mac-only test, both oversights by Callek or me from earlier patches.
• SeaMonkey L10n:
  Finally, the Build SeaMonkey locales from Mercurial bug could be marked FIXED, as with the download manager landing, we finally have all strings in the appropriate locales/ directories and picked up correctly by the "source L10n" build/(re)package mechanisms, so that localized SeaMonkey builds should be fully working (as long as dashboard is green for the used locale and the strings don't have syntax errors).
• German L10n:
  New strings for download manager and mail archiving UI landed, de was the first locale to be green after the download manager landing.
• Various Discussions:
  Mail archiving UI for SeaMonkey, 1.8 branch landings and 1.1.17 planning, tracking OSX buildbot failures down to a "hardware" (Parallels) issue, waiting for Parallels fix, Qt build bustages, comm-central commit access, sr policy, etc.

We're seeing some small followup issues with download manager, but general feedback is quite good so far. We are definitely coming into beta state now, and current SeaMonkey 2 nightlies are in some ways more mature than stable 1.1.x releases. Tabbed mail is going through the review process and with the current patch, Lightning is usable. The future looks bright.
Well, it would look brighter if we had some schedules for beta or final. As we need to build stable milestones and releases upon a stable mailnews backend, we need to closely coordinate with Thunderbird, and that team has been pushing out their upcoming beta further and further while waiting on their feature of vastly improved search to land. That also has been pushing out our SeaMonkey 2.0 Beta 1 release. And with them adding a further beta, it now looks very clear to us that we will add a SeaMonkey 2.0 Beta 2 after this one as well. We unfortunately still need to wait for the clouds over the Thunderbird schedules to lift before we can come up with schedules for those milestones and releases on our side, but we count on being close to their betas and final in terms of dates.
Depending on the speed of the review process, tabbed mail will either make Beta 1 or be pushed to Beta 2, I hope we can take a few more of the Thunderbird 3 improvements in Beta 2. The current code is quite ready for Beta 1 right now, so it can only get better.
Oh, and we now know for sure that we will have fully localized builds for all major platforms right with the Beta 1 release - for the first time ever in suite history!