Releases/Firefox 23.0b4/BuildNotes
- Who: coop
- Tracking bug: bug 891165
- Bugs hit:
- bug 891128 - Multiple services are offline in SCL3
Doing a release? Here's a list of things you should consider whether or not you need to do. Note that a chemspill requires the same steps as a regular release of the same type. Be sure to take notes on manual steps and any problems you encounter.
-
Ship the l10n milestone DESKTOP MOBILE RELEASE BETA -
Submit to Ship It DESKTOP MOBILE ESR BETA RELEASE -
Run pushsnip DESKTOP ESR BETA RELEASE -
Update bouncer links DESKTOP BETA RELEASE -
Post-release tasks DESKTOP ESR BETA RELEASE
Recovery from colo outage
The colo failed *right* in the middle of the kickoff, so we ended up with some repos tagged and some not. I tried various permutations of |hg tag --remove| and changing whitespace in the release configs so the versions could be successfully. I eventually ended up running:
cd buildbotcustom hg up -r production-0.8 hg tag --remove FIREFOX_23_0b4_RELEASE FENNEC_23_0b4_RELEASE FIREFOX_23_0b4_BUILD1 FENNEC_23_0b4_BUILD1 hg push cd ../buildbot-configs # Add extra whitespace line to end of release-fennec-mozilla-beta.py and # release-firefox-mozilla-beta.py. # Committed those changes to both the default and production branches hg up -r production hg tag --remove FIREFOX_23_0b4_RELEASE FENNEC_23_0b4_RELEASE FIREFOX_23_0b4_BUILD1 FENNEC_23_0b4_BUILD1 hg push
This got us past the build repo tagging portion of the automation, but then we hit an ISE 500 error on hg while trying to reconfig.
I performed a manual reconfig of the scheduler and build masters, and then Nick and I went looking for examples of sendchange commands to run. I found this:
PYTHONPATH=. python ../tools/buildbot-helpers/release_sanity.py \ --dryrun \ -u coop \ -V 23.0b4 \ --branch mozilla-beta \ --build-number=1 \ --release-config release-firefox-mozilla-beta.py \ --release-config release-fennec-mozilla-beta.py \ --skip-reconfig \ localhost:9001
This worked as a dryrun, but when I tried to run it without --dryrun, the command failed. In hindsight, it was probably just because of the change from --who to --username.
Nick came up with the following sendchange:
buildbot sendchange -u lsblakk@mozilla.com \ --master buildbot-master36.srv.releng.scl3.mozilla.com:9301 \ --branch releases/mozilla-beta \ -p products:firefox \ -p script_repo_revision:FIREFOX_23_0b4_RELEASE \ release_build
This sendchange did work. It would have worked for Fennec too, if I had also changed the script_repo_revision on first submission. This is what I ran for Fennec in the end:
buildbot sendchange -u lsblakk@mozilla.com \ --master buildbot-master36.srv.releng.scl3.mozilla.com:9301 \ --branch releases/mozilla-beta \ -p products:fennec \ -p script_repo_revision:FENNEC_23_0b4_RELEASE \ release_build
Android build failure
Ran out of space in make buildsymbols:
/tools/android-ndk-r8c/toolchains/arm-linux-androideabi-4.6/prebuilt/linux-x86/bin/arm-linux-androideabi-objcopy:./dist/bin/libxul.so.dbg[.debug_info]: No space left on device /tools/android-ndk-r8c/toolchains/arm-linux-androideabi-4.6/prebuilt/linux-x86/bin/arm-linux-androideabi-objcopy:./dist/bin/libxul.so.dbg: No space left on device
Build was auto-retried and was successful.
The initial build had this in the clean_old_builds step:
purge_actual: '10.53GB' purge_target: '10GB'
The successful build was on the same slave and had:
purge_actual: '10.69GB' purge_target: '10GB'