Seth says that we will need desktop (x86) builds of Maemo for testers/localizers.
er, builds of Fennec.
Thanks for filing this bug. I cant remembered all details of when Seth and I talked; was just doing nightly builds enough, or was there also a need for per-checkin builds?
We should get the full monty for l10n. I would expect that to be easier to set up, too. I can't speak for the en-US desktop builds, for l10n we only need nightlies of those right now. Whether those should support depend builds, too, is a question for the fennec guys.
Also, I assume these are linux desktop builds? Or are we able to build win32/osx versions of desktop Fennec? I didn't find any instructions on desktop Fennec builds with a quick glance at the wiki; does anyone have some? (Manual steps, mozconfig)
(In reply to comment #4) > Also, I assume these are linux desktop builds? Or are we able to build > win32/osx versions of desktop Fennec? Yep. All 3 platforms are built pretty much the same way. I use a slightly different mozconfig for OS X, but Win32 and Linux is the same. > I didn't find any instructions on desktop Fennec builds with a quick glance at > the wiki; does anyone have some? (Manual steps, mozconfig) https://wiki.mozilla.org/Mobile/Build/Fennec has some details
Reassigning to jhford.
(In reply to comment #6) > Created an attachment (id=380872) [details] > mozconfig that worked on a manual build In general, I'm glad you're using --enable-tests, but do we need that for desktop builds? Might be faster to not enable the tests.
I'd actually be curious as to how Fennec unit/perf tests would do on our standard test images/machines as compared to the devices. That's certainly beyond the scope of this bug, however, and if the Mobile team doesn't think that's needed we can certainly disable them easily.
Builds and repacks are now working for Linux-i686. I am going to clean up and submit for review tomorrow morning.
Created attachment 382455 [details] [diff] [review] buildbot-configs Include a Linux i686 build of Fennec and a nightly repack.
Created attachment 382456 [details] [diff] [review] buildbotcustom Add a MobileDesktop build and repack factory
Created attachment 382458 [details] [diff] [review] buildbot-configs redux accidentally included an incorrect mozconfig with previous patch. sorry for the bugspam.
Comment on attachment 382458 [details] [diff] [review] buildbot-configs redux Nit: does a multi-line build name work? Not sure if you tested it this way or if this was a later change. >+MOBILE_BRANCHES['mobile-trunk']['platforms']['linux-i686']['base_name'] ='Linux >+Fennec Desktop mozilla-central' Nit: let's just do 'l10n' instead of 'build/l10n' to match the scratchbox layout. Spoiler: you may be working on changing these soon anyway, though not necessarily for l10n =) >+MOBILE_BRANCHES['mobile-trunk']['platforms']['linux-i686']['base_l10n_workdir'] = 'build/l10n' Nit: let's play a game of "spot the typo": >+ print 'platform %s is not linux-gnueabi-armi or linux-i686' % platform Also I'm not sure if it was from this morning's checkins or what, but I had issues applying this patch: patch: **** malformed patch at line 22: diff -r 18156ce9a107 mozilla2-staging/mobile_config.py
Comment on attachment 382456 [details] [diff] [review] buildbotcustom looks good. also, great job.
Created attachment 382585 [details] [diff] [review] buildbot-configs Fixed bitrot and the three nits Aki had.
Comment on attachment 382585 [details] [diff] [review] buildbot-configs revision 1201:6401e8060e93, 1202:4bc6ce43788a
Comment on attachment 382456 [details] [diff] [review] buildbotcustom revision 334:fa752dffb214
Seems to work in production, there was an issue with one specific slave (moz2-linux-slave07) regarding the mozilla-central checkout. I have clobbered all the linux fennec desktop builders and check tomorrow on the output
It seems like the DNS issue has been resolved and there are now the correct number of Fennec-i686 builds compared to Fennec-gnueabi-arm builds.