Releases/Fennec 1.1/BuildNotes
From MozillaWiki
< Releases
Contents
Build Engineers
aki - bug 566428
Signed-off Revision(s)
| Build # | mozilla-1.9.2 | mobile-1.1 |
| rc1 build 1 | 2cb83e106cf7 | 4a51427bed7c |
| rc1 build 2 | 91e79f7e8b96 | 4a51427bed7c |
| rc1 build 3 | 91e79f7e8b96 | 5c06d8f81a1a |
Tags
On releases/mozilla-1.9.2:
| Build # | Tag | Moz Changeset | Mobile Changeset |
| rc1 build 1 | GECKO1925_20100602_RELBRANCH | 2cb83e106cf7 | 4a51427bed7c |
| FENNEC_1_1rc1_BUILD1 | 7275c58eba49 | c6ad008b32ad | |
| rc1 build 2 | FENNEC_1_1rc1_BUILD2 | 91e79f7e8b96 | c6ad008b32ad |
| rc1 build 3 | FENNEC_1_1rc1_BUILD3 | 91e79f7e8b96 | 5c06d8f81a1a |
Build data
| Build # | Type | Build ID | Build machine |
| rc1 build 1 | Maemo4 | ||
| Linux | 20100602181702 | moz2-linux-slave35 | |
| Mac | 20100602184129 | moz2-darwin9-slave49 | |
| Windows | 20100602185328 | win32-slave59 | |
| rc1 build 2 | Maemo4 | 20100603130022 | moz2-linux-slave40 |
| Linux | 20100603125742 | moz2-linux-slave43 | |
| Mac | 20100603131037 | moz2-darwin9-slave30 | |
| Windows | 20100603123420 | mw32-ix-slave22 | |
| rc1 build 3 | Maemo4 | 20100614152856 | moz2-linux-slave37 |
| Linux | 20100614142409 | mv-moz2-linux-ix-slave02 | |
| Mac | 20100614152130 | moz2-darwin9-slave30 | |
| Windows | 20100615115055 | mw32-ix-slave23 |
Notes
Build 1
- Wished we had a clobberer for mobile releases bug 537426
- Axel landed the l10n json update
- Landed a release config update which contains changesets and turns mergeLocales to False (no English strings in l10n repacks)
- Previous prep visible in bug 566428
- Reconfiged pm02
- Started the automation with:
buildbot sendchange --username=aki --master=localhost:9010 --branch=releases/mobile-1.1 "Fennec 1.1rc1 build1" doit
Tag
Timed out after 1800 seconds during nl clone. Following the docs, but editing l10n-changesets_mobile-1.1.json .
- Copied l10n-changesets_mobile-1.1.json to l10n-changesets_mobile-1.1.json.orig
- Edited a copy of l10n-changesets_mobile-1.1.json (aki.json, softlinked to l10n-changesets_mobile-1.1.json) to remove the already tagged locales (ar through nb-NO)
- Tried a "force build" of the mobile_tag builder, which failed on ar
- Reconfiged pm02
- Tried another "force build" of the mobile_tag builder
- That worked; reverted l10n-changesets_mobile-1.1.json changes
- Followed the docs on creating a dummy tag_factory for tagging in release_mobile_master.py
- Reconfiged pm02
- Restarted automation with
buildbot sendchange --username=aki --master=localhost:9010 --branch=releases/mobile-1.1 "Fennec 1.1rc1 build1" doit
- Started scheming on how to make this better
- While filling out the tags, noticed that bug 542413 bit me (further maemo release bumpfiles support) -- no milestone update
- Killed builds
- Manually updated the milestone in bug 566428 and tagged via
hg tag -f -r 7275c58eba49 FENNEC_1_1rc1_BUILD1 hg tag -f -r 7275c58eba49 FENNEC_1_1rc1_RELEASE hg push
- Verified dummy factory still there (wasn't, due to hg pull -u for bug 542413; replaced)
- Manually clobbered:
- moz2-linux-slave39:/builds/slave/mobile_source
- mv-moz2-linux-ix-slave14:/scratchbox/users/cltbld/home/cltbld/build/maemo-1.9.2-release
- mv-moz2-linux-ix-slave24:/builds/slave/mobile_linux-i686_desktop_build
- moz2-darwin9-slave57:/builds/slave/mobile_macosx-i686_desktop_build
- win32-slave46:/e/builds/moz2_slave/mobile_win32-i686_desktop_build
- Restarted automation via sendchange
- Gah, dummy step didn't take. Double checked again, reconfiged again after |rm *.pyc|
- sendchange #4
- fixed typo :( ... reconfig, sendchange #5
Source
No problems.
(Source built 2x due to restart; was fine both times. I nuked stage's build1 candidates dir in between to be sure.)
(Killed mobile_source build 20 when it was triggered from a sendchange in the build/repack steps)
Build/Repack
- Maemo4: ran into bug 567827 which killed the build before the multilocale repack portion.
- Linux: no problems
- Mac: no problems
- Win32: no problems
Maemo4 needs to succeed during a sendchange since repacks are triggered from it. The desktop builds have no repacks to worry about.
So, to fix maemo4, I:
- edited mobile_config.py MOBILE_SLAVES to point maemo4 to config.LINUX_VMS (bug 567827 kills make buildsymbols on ix boxes only).
- linux+mac were done and win32 putting away happily, so I added a dummy build_factory to the |for platform in enUSDesktopPlatforms:| loop in release_mobile_master.py
- reconfiged
- sendchange #6?
- removed maemo4 from candidates/1.1rc1-candidates/build1/ just to be sure
- Gah, noticed that bug 566311 isn't fixed on m-1.9.2; flagged a1.9.2.5?
Partner Builds
- Didn't trigger for some reason :P
- Force built, worked.
Deb Repo Creation
Push Live
Copy bits to releases/
Create new .install files
Create bouncer entries
Build 2
- Wished we had a clobberer for mobile releases bug 537426
- clobbered mv-moz2-linux-ix-slave22, moz2-linux-slave42 manually (tag/source, maemo4 build). The desktop builders should be ok.
- Keeping the same l10n-changesets from build1 yesterday afternoon.
- Landed a release config update which contains changesets, build# bump, relbranch overrides.
- Forced maemo4 to build on a VM:
diff -r faa11e08ea8d mozilla2/mobile_config.py
--- a/mozilla2/mobile_config.py Thu Jun 03 11:39:17 2010 -0700
+++ b/mozilla2/mobile_config.py Thu Jun 03 12:30:10 2010 -0700
@@ -15,7 +15,8 @@
SBOX_HOME = '/scratchbox/users/cltbld/home/cltbld'
MOBILE_SLAVES = {
- 'maemo4': config.SLAVES['linux'],
+# 'maemo4': config.SLAVES['linux'],
+ 'maemo4': config.LINUX_VMS,
'maemo5-gtk': config.SLAVES['linux'],
'maemo5-qt': config.SLAVES['linux'],
'linux-i686': config.SLAVES['linux'],
- Reconfiged pm02, pmm 8011
- Started the automation with:
buildbot sendchange --username=aki --master=localhost:9010 --branch=releases/mobile-1.1 "Fennec 1.1rc1 build2" doit
Tag
No problems. (yay!)
Er, noticed that this needed to be revision c6ad008b32ad:
- killed the builds, clobbered mv-moz2-linux-ix-slave15 (source), moz2-linux-slave37 (maemo4 build), moz2-linux-slave35 (tag), stage (1.1rc1-candidates/build2) manually.
- clobber == ssh in, mv /builds/slave/mobile_* /builds/slave/foo, rm -rf /builds/slave/foo (tag, source)
- clobber == ssh in, mv /scratchbox/users/cltbld/home/cltbld/build/*release /builds/slave/foo, rm -rf /builds/slave/foo (maemo4 build)
- landed release config change for posterity
- tagged c6ad008b32ad manually
cd mobile-1.1 hg pull -u hg up -r GECKO1925_20100602_RELBRANCH hg tag -r c6ad008b32ad -f FENNEC_1_1rc1_BUILD2 hg tag -r c6ad008b32ad -f FENNEC_1_1rc1_RELEASE hg outgoing | more hg push
- added dummy factory for tagging in release_mobile_master.py
- reconfig
- sendchange #2:
buildbot sendchange --username=aki --master=localhost:9010 --branch=releases/mobile-1.1 "Fennec 1.1rc1 build2 take2" doit
Source
No problems. s: moz2-linux-slave37
Build/Repack
- Maemo4: No problems. [!]
- Linux: No problems.
- Mac: No problems.
- Win32: No problems.
L10n waiting on slaves... we're down to 2 l10n slaves for linux on pm02. To speed things up, I:
builders.append({
'name': '%s_repack' % platform,
# 'slavenames': MOBILE_L10N_SLAVES['maemo4'],
'slavenames': branchConfig['platforms']['linux']['slaves'],
and reconfiged.
Afterwards, the l10n repacks went well, except for fr which hung (seems like a lot of hg-related jobs hung on multiple machines at the same time). I killed that, force-built it but it died on get enUS source again (I think from a partial clone). Clobbered the dir on mv-moz2-linux-ix-slave16 and re-forced by going to master, going to builder "maemo4_repack" and setting:
#set name and 3 properties locale fr l10n_revision FENNEC_1_1rc1_RELEASE en_revision FENNEC_1_1rc1_RELEASE
and that worked.
Partner Builds
- First attempt timed out during repack step. Attempted "rebuild" from buildbot.
- That worked! :)
Deb Repo Creation
Worked ! :)
I:
ssh cltbld@staging-mobile-master.build cd /scratchbox/users/cltbld/home/cltbld rm -rf mozharness repos logs configs # This will get better I promise hg clone http://hg.mozilla.org/users/asasaki_mozilla.com/mozharness mozharness/scripts/signdebs.py --logLevel debug --multiLog --configFile deb_repos/1.1_release.json
Then I looked in the logs/, noted that the warning, error, critical, and fatal logs were size 0, and did a quick check of the raw log to verify that I didn't miss any error regexes:
cat logs/signdebs_raw.log | egrep -v '(^Running |^Moving )' | more
I then took an n810, then:
- connected to wifi
- opened App Manager
- uninstalled Firefox (and mozilla-runtime if necessary)
- removed all Mozilla* repos in Tools->App Catalog
- closed App Manager
- opened browser to http://ftp.mozilla.org/pub/mozilla.org/mobile/candidates/1.1rc1-candidates/repos
- browsed to a repo, clicked on an install file
- chose to open + install
Push Live
This is still accurate: https://wiki.mozilla.org/Release:Release_Automation_on_Mercurial:Mobile#Copying_files_from_candidates.2F_to_releases.2F
I've also updated some of the rest of that page with brief 1.1 stuff.
Copy bits to releases/
Create new .install files
Create bouncer entries
Build 3
- Wished we had a clobberer for mobile releases bug 537426
- Axel landed l10n changesets: 871067634bf4
- Landed config updates
- Forced maemo4 to build on a VM:
diff -r faa11e08ea8d mozilla2/mobile_config.py
--- a/mozilla2/mobile_config.py Thu Jun 03 11:39:17 2010 -0700
+++ b/mozilla2/mobile_config.py Thu Jun 03 12:30:10 2010 -0700
@@ -15,7 +15,8 @@
SBOX_HOME = '/scratchbox/users/cltbld/home/cltbld'
MOBILE_SLAVES = {
- 'maemo4': config.SLAVES['linux'],
+# 'maemo4': config.SLAVES['linux'],
+ 'maemo4': config.LINUX_VMS,
'maemo5-gtk': config.SLAVES['linux'],
'maemo5-qt': config.SLAVES['linux'],
'linux-i686': config.SLAVES['linux'],
- Reconfiged pm02. Skipping pmm8011 since my mozharness script handles deb updates for releases.
- Started the automation with:
buildbot sendchange --username=aki --master=localhost:9010 --branch=releases/mobile-1.1 "Fennec 1.1rc1 build3" doit
Tag
- Died on el, which is a new locale added in Axel's l10n changeset landing. I knew this, but didn't realize I'd have to manually create the relbranch.
- Brainstormed on how to fix this... by default create the relbranch if it doesn't exist on l10n repos only?
- Manually created the relbranch in el:
hg clone http://hg.mozilla.org/releases/l10n-mozilla-1.9.2/el cd el # edit .hg/hgrc to allow for default-push hg branch GECKO1925_20100602_RELBRANCH hg commit hg push
- Edited l10n-changesets_mobile-1.1.json to remove all locales before el
- Reconfiged; forced the tag builder to rebuild
- Gah, clobbered mv-moz2-linux-ix-slave16, forced the tag builder to rebuild.
- Tag factory needs more love =P
- Backed out l10n-changesets_mobile-1.1.json local changes, added dummy tag_factory in release_mobile_master.py, reconfiged
- Restarted automation with
buildbot sendchange --username=aki --master=localhost:9010 --branch=releases/mobile-1.1 "Fennec 1.1rc1 build3 post-tag" doit
Source
No problems. s: moz2-linux-slave37
Build/Repack
- maemo4: No problems.
- linux: No problems.
- macosx: No problems.
- win32: compilation error:
make[6]: *** No rule to make target `nsGB2312ToUnicodeV2.obj', needed by `ucvcn_s.lib'. Stop.
- However, a rebuild the next day worked. Need to investigate.
- L10n repacks: No problems.
Partner Builds
- Didn't trigger :P
- Force build worked.
Deb Repo Creation
- Landed a mozharness config update
- Ran it:
ssh cltbld@staging-mobile-master.build cd /scratchbox/users/cltbld/home/cltbld hg clone http://hg.mozilla.org/users/asasaki_mozilla.com/mozharness mozharness/scripts/signdebs.py --logLevel debug --multiLog --configFile deb_repos/1.1_release.json
Push Live
Copy bits to releases/
No problems
Create new .install files
Created new .install files. However, then I changed the firefox.install files back to chinook (only beta has fremantle/chinook split; release firefox.install should still point to chinook)
Create bouncer entries
Done, Firefox-Mobile-1.1rc1 and Firefox-Mobile-1.1rc1-N810
Re-done, Firefox-Mobile-1.1rc1 and Firefox-Mobile-1.1rc1-N900
