Once bug 557027 lands, we should be able to build all of our Fennec builds (both desktop and mobile) without the xulrunner build. The mozconfig should look a lot like the Firefox mozconfigs, just: ac_add_options --enable-application=mobile plus whatever other options we need.

The desktop builds built fine. Need to test: Maemo4 builds, release builds, Maemo4 l10n repacks, n810/n900 talos and unit tests

(broken w/out this patch)

Comment on attachment 440334 [details] [diff] [review] Configs to remove xulrunner Multi-locale builds + l10n repacks tested.

Comment on attachment 440336 [details] [diff] [review] potentially fix no-xulrunner maemo repacks Axel -- our Maemo5 builds already build xulrunner+fennec together (meaning platform.ini, not xulrunner/platform.ini). We're switching our Maemo4 builds, and all future Mobile builds, to be the same -- this should allow us to do so without breaking repacks.

Comment on attachment 440334 [details] [diff] [review] Configs to remove xulrunner - packageGlobList=['-r', 'mobile/dist/*.tar.bz2', - 'xulrunner/dist/*.tar.bz2'], + packageGlobList=['-r', 'dist/*.tar.bz2', + 'dist/*.tar.bz2'], The second .tar.bz2 should be .zip

Comment on attachment 440893 [details] [diff] [review] fix packageGlobLists looks good.

Comment on attachment 440886 [details] [diff] [review] buildbotcustom -- fix bugs found in build/repack testing > self.addStep(ShellCommand( > command=['wget', >- WithProperties("%(download_dir)s/xulrunner-%(milestone)s.en-US.linux-gnueabi-arm.tests.tar.bz2"), >- '-O', 'xulrunner-tests.tar.bz2'], >+ WithProperties("%(download_dir)s/fennec-%(milestone)s.en-US.linux-gnueabi-arm.tests.zip"), >+ '-O', 'fennec-tests.zip'], This will try to download fennec-1.9.2a3pre.en-US.... instead of fennec-1.1a1pre.en-US.... We could either parse the fennec version out of application.ini? or we could use sendchanges with two files and the UnittestPackagedBuildFactory.

Comment on attachment 440336 [details] [diff] [review] potentially fix no-xulrunner maemo repacks r=me, assuming that: - we don't have xulrunner builds as an option (if we do, should this be conditional?) - that you have a right time to land this to not break repack before we make the switch.

Comment on attachment 440909 [details] [diff] [review] milestone -> version >@@ -5225,26 +5222,26 @@ class MaemoBuildFactory(MobileBuildFacto > if self.multiLocale: > # In the multi-locale scenario we build and upload the single-locale > # before the multi-locale. This packageGlob will be used to move packages > # into the "en-US" directory before uploading it and later on the > # multi-locale overwrites it in addMultiLocaleSteps(...) >- self.packageGlob = "mobile/dist/*.tar.bz2 mobile/mobile/*.deb mobile/dist/deb_name.txt" >+ self.packageGlob = "dist/*.tar.bz2 mobile/*.deb dist/deb_name.txt" > self.compareLocalesRepo = self.getRepository(compareLocalesRepoPath) > self.compareLocalesTag = compareLocalesTag > self.addStep(ShellCommand, Should this include dist/*.tests.zip (or dist/*.zip)? r+ if not including zip files is the desired behaviour. >@@ -5401,18 +5395,18 @@ class MaemoBuildFactory(MobileBuildFacto > self.addPackageSteps(multiLocale=True, packageTests=True) >- self.packageGlob="mobile/dist/fennec*.tar.bz2 mobile/mobile/fennec*.deb " + \ >- "mobile/dist/deb_name.txt xulrunner/dist/*.tar.bz2" >+ self.packageGlob="dist/fennec*.tar.bz2 mobile/fennec*.deb " + \ >+ "dist/deb_name.txt dist/fennec*.zip" Should we generialize to *.tar.bz2, *.zip and *.deb now that the xulrunner* things aren't being built? I don't have strong opinions. >@@ -347,18 +347,18 @@ class MaemoTalosFactory(BuildFactory): > 'ts': 60, > 'tp4': 90, > 'tdhtml': 60, > 'tsvg': 60, > 'twinopen': 60, > 'tsspider': 60, > 'tgfx': 60, > }, >- talosTarball='http://mobile-master.mv.mozilla.com/maemo/talos.tar.bz2', >- pageloaderTarball='http://mobile-master.mv.mozilla.com/maemo/pageloader.tar.bz2', >+ talosTarball='http://staging-mobile-master.build.mozilla.org/maemo/talos.tar.bz2', >+ pageloaderTarball='http://staging-mobile-master.build.mozilla.org/maemo/pageloader.tar.bz2', nit: this should be passed in so that production-mobile-master is not dependent on staging-mobile-master being up. With the *rw configs + factories replacing this code soon I don't really know if this is something we need to spend time on. lets move the n810s to the new configs instead of fixing this :) >@@ -23,16 +23,17 @@ from buildbotcustom.steps.unittest impor ... > command=['wget', >- WithProperties("%(download_dir)s/xulrunner-%(milestone)s.en-US.linux-gnueabi-arm.tests.tar.bz2"), >- '-O', 'xulrunner-tests.tar.bz2'], >+ WithProperties("%(download_dir)s/fennec-%(version)s.en-US.linux-gnueabi-arm.tests.zip"), >+ '-O', 'fennec-tests.zip'], > workdir=self.base_dir, This looks good. In the long term, I think we should fix bug 557336 and use that factory as a base instead.

(In reply to comment #13) > Should this include dist/*.tests.zip (or dist/*.zip)? r+ if not including zip > files is the desired behaviour. It currently breaks if dist/*.zip is included. It was there in an earlier iteration, during tests. > Should we generialize to *.tar.bz2, *.zip and *.deb now that the xulrunner* > things aren't being built? I don't have strong opinions. Let's generalize on make upload :) > >+ talosTarball='http://staging-mobile-master.build.mozilla.org/maemo/talos.tar.bz2', > >+ pageloaderTarball='http://staging-mobile-master.build.mozilla.org/maemo/pageloader.tar.bz2', > > nit: this should be passed in so that production-mobile-master is not dependent > on staging-mobile-master being up. With the *rw configs + factories replacing > this code soon I don't really know if this is something we need to spend time > on. lets move the n810s to the new configs instead of fixing this :) We do pass this in, or this would have been broken all this time :) These are defaults. > This looks good. In the long term, I think we should fix bug 557336 and use > that factory as a base instead. Ok.

Comment on attachment 440909 [details] [diff] [review] milestone -> version (r+ per comments)

Comment on attachment 440893 [details] [diff] [review] fix packageGlobLists http://hg.mozilla.org/build/buildbot-configs/rev/e31f5013cffe

Comment on attachment 440909 [details] [diff] [review] milestone -> version http://hg.mozilla.org/build/buildbotcustom/rev/835ed8d422a2

Blocked by a pm02 restart =\

I seem to be unable to push to mobile-browser now (possibly due to committer perm changes?): akira-sasakis-macbook:mobile-browser asasaki$ hg push pushing to ssh://hg.mozilla.org/mobile-browser searching for changes remote: abort: could not lock repository /repo/hg/mozilla/mobile-browser: Permission denied abort: unexpected response: empty string Repacks and multi-locale builds will fail til this goes in :(

Comment on attachment 440336 [details] [diff] [review] potentially fix no-xulrunner maemo repacks http://hg.mozilla.org/mobile-browser/rev/05834514b444

Comment on attachment 440336 [details] [diff] [review] potentially fix no-xulrunner maemo repacks >diff --git a/locales/Makefile.in b/locales/Makefile.in >-GECKO_PLATFORM_INI_PATH="$(STAGEDIST)/xulrunner/platform.ini" >+GECKO_PLATFORM_INI_PATH="$(STAGEDIST)/platform.ini" > FENNEC_APPLICATION_INI_PATH="$(STAGEDIST)/application.ini" Since it is still possible to actually build fennec using XR, I think this patch should be: ifdef LIBXUL_SDK GECKO_PLATFORM_INI_PATH="$(STAGEDIST)/xulrunner/platform.ini" else GECKO_PLATFORM_INI_PATH="$(STAGEDIST)/platform.ini" endif

Fixing the locales/Makefile.in patch. a) I'm assuming that the OSX location will be similar, even though we don't currently have OSX repacks, and b) we'll need these changes landed in mobile-1.1 as well, assuming this passes r?.

Comment on attachment 441882 [details] [diff] [review] add ifdef LIBXUL_SDK to makefile All l10n repacks will be broken until this lands in releases/mobile-1.1 .

pushed m-b: http://hg.mozilla.org/mobile-browser/rev/0d986e7f659c pushed m-1.1: http://hg.mozilla.org/releases/mobile-1.1/rev/936d3030b49f