Fennec 1.0 will be shipped off of mozilla-1.9.1. The sooner we can get these builds going and verified as working, the better.

Here's a teeny tiny little patch that you can review with one eye closed ;-) * I refactored mobile_config.py to look much like config.py, except with mobile-isms. I could potentially merge this into config.py though that could get a bit cluttered. * Since both m-c and 1.9.1 mobile builds use mobile-browser as the mobile repository, I'm renaming the upload milestone from mobile-browser to mobile-trunk and mobile-1.9.1, respectively. * L10n updated to match. Once we have at least one successful Maemo 1.9.1 build, we can set the mobile-1.9.1 enable_l10n to True, and it should Just Work. * Maemo and WinCE 1.9.1 builds are turned on. These are both broken in the compile step; we can work on fixing the code base once we have builds to detect the breakage. It will result in burning builds on the Mobile tree in the short term, however. To test on staging I had to hack DOWNLOAD_BASE_URL to point to staging-stage, since there is no latest-mobile-trunk directory on ftp.m.o as of yet. Patch one of two.

buildbotcustom side.

Comment on attachment 375399 [details] [diff] [review] add baseUploadDir to MobileBuildFactory and MobileNightlyRepackFactory I'm not sure I understand the need for baseUploadDir in the en-US builds - it looks like it's the same as mobileBranchName in all of our uses. Is there a reason you need it there? The l10n side looks fine.

Comment on attachment 375398 [details] [diff] [review] mobile_config refactor to enable branch builds > # most of the config is in an external file > import config > reload(config) > from config import * > import mobile_config > reload(mobile_config) >-from mobile_config import mobile_slaves >+from mobile_config import * Please don't import * from multiple files, it can easily end up being the source of weird behaviour due to naming conflicts. You probably want to 'from config import * as nightly_config', or something like that, a few lines up. >-change_source.append(HgPoller( >- hgURL=config.HGURL, >+m['builders'] = [] >+m['schedulers'] = [] >+m['change_source'] = [] >+m['status'] = [] >+ >+allMobileBranchBuilders = [] This name is a tad confusing. I read the "branch" here as "mozilla-1.9.1", which isn't right, of course. Any reason not to just use allMobileBuilders? >+ m['status'].append(TinderboxMailNotifier( >+ fromaddr="bootstrap@mozilla.com", itym buildbot@mozilla.com, not that it matters. >+ hg_all_locales_poller = HgAllLocalesPoller(hgURL = HGURL, >+ repositoryIndex=branch['l10n_repo_path'], >+ pollInterval=5*60, >+ ) Having another one of these makes me fear for our twistd.log's and more importantly, load on production-master and hg.m.o. Please be sure to keep an eye on production-master when you land this patch. >+ hg_all_locales_poller.parallelRequests = 1 >+ m['change_source'].append(hg_all_locales_poller) >+ >+ # change sources >+ m['change_source'].append(HgPoller( >+ hgURL=HGURL, >+ branch=branch['repo_path'], >+ pushlogUrlOverride='%s/%s/index.cgi/pushlog' % (HGURL, >+ branch['repo_path']), >+ pollInterval=1*60 >+ )) We've already got HgPollers for mozilla-central and mozilla-1.9.1 in master.cfg - no need for them here. The production part looks fine proviso the comments above. r=bhearsum with the excess HgPollers removed and imports changed. Hope to see the naming nits addressed, too. Good work/refactor, overall!

> Please don't import * from multiple files Fixed. Turns out I just need MOBILE_BRANCHES from mobile_config. >>+allMobileBranchBuilders = [] > > This name is a tad confusing. I read the "branch" here as "mozilla-1.9.1", > which isn't right, of course. Any reason not to just use allMobileBuilders? Maybe allMobileRepoBuilders? Changed to "mobileBuilders" but I'm not tied to it. > itym buildbot@mozilla.com, not that it matters. I got bootstrap@ from the l10n configs, but doesn't matter to me. Changed. > Having another one of these makes me fear for our twistd.log's and more > importantly, load on production-master and hg.m.o. Please be sure to keep an > eye on production-master when you land this patch. > ... > We've already got HgPollers for mozilla-central and mozilla-1.9.1 in > master.cfg - no need for them here. I think the l10n pollers are dups as well, since I'm just watching the same repos in l10n-central. I've removed the l10n and m-c/m-1.9.1 pollers, which will actually reduce the number of pollers that are live on production-master today. I've seen 1.9.1 and m-c checkins trigger builds on s-m2 with this patch. Pretty sure the nightly l10ns will trigger at 4am, but I'm leaving the patch on s-m2 tonight just to make sure.

Comment on attachment 375782 [details] [diff] [review] fix nits, remove dup change_sources Looks good! I'm very happy that we don't need the HgAllLocalesPoller!

l10n fennec builds on 1.9.1 are going to break as long as we don't have l10n.mk on that branch. Ted and I would prefer to first make Firefox use that on the trunk before porting it to the branch. Is doing that next week timely enough?

Comment on attachment 375399 [details] [diff] [review] add baseUploadDir to MobileBuildFactory and MobileNightlyRepackFactory After chatting on IRC yesterday this seems fine.

Comment on attachment 375399 [details] [diff] [review] add baseUploadDir to MobileBuildFactory and MobileNightlyRepackFactory Revision 270:47e119d3b744

Comment on attachment 375782 [details] [diff] [review] fix nits, remove dup change_sources Rev 1130:6194c02e4732

This is live. Axel: that's fine with me; I'd guess that's ok with the Mobile team? Both 1.9.1 builds are broken atm but knowing/tracking where they're broken is good.