Closed Bug 1413737 Opened 7 years ago Closed 7 years ago

Make l10n single-locale repacks work with artifact builds

Categories

(Taskcluster :: General, defect)

defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: nalexander, Unassigned)

References

Details

Right now, they don't, for reasons that are very difficult to understand. Just try |mach try fuzzy --full| and select, say, nightly-l10n-android-api-16-nightly-45/opt. You'll get builds failing with Automation Error: Can't checkout https://hg.mozilla.org/build/tools! [log…] due to the robustcheckout plugin not being found. For example, see https://treeherder.mozilla.org/#/jobs?repo=try&revision=1a10cd5a10da700e80b749c1a63b90eb467b4c61&selectedJob=141481006. I think that what's happening is that the artifact bits are somehow running _more_ mozharness actions than the full builds do (I have try builds out to test this, but they're not done yet). Neither gps nor I can understand why the robustcheckout setup would be wrong, but there it is.
Component: Build Config & IDE Support → General
Product: Firefox for Android → Taskcluster
There are quite a few things about l10n repacks and artifact builds that don't work out aside from this, tracked in bug 1387485.
Blocks: 1387485
From diffing logs, I see: WARNING - Try syntax not found in: . INFO - Artifact build requested by try push. INFO - Build variant has `artifact_build_variant_in_try`: "api-16-artifact". INFO - Using artifact build variant "api-16-artifact". INFO - Updating self.config with the following from /builds/worker/workspace/build/src/testing/mozharness/scripts/../configs/builds/releng_sub_android_configs/64_api_16_artifact.py: INFO - {'base_name': 'Android armv7 api-16+ %(branch)s Artifact build', INFO - 'build_type': 'api-16-opt-artifact', INFO - 'multi_locale_config_platform': 'android', INFO - 'src_mozconfig': 'mobile/android/config/mozconfigs/android-api-16/nightly-artifact', INFO - 'stage_platform': 'android-api-16', INFO - 'tooltool_manifest_src': 'mobile/android/config/tooltool-manifests/android/releng.manifest'} INFO - Updating volatile_config to include default_actions from /builds/worker/workspace/build/src/testing/mozharness/scripts/../configs/builds/releng_sub_android_configs/64_api_16_artifact.py. INFO - Actions in volatile_config are now: ['get-secrets', 'clobber', 'clone-tools', 'checkout-sources', 'setup-mock', 'build', 'upload-files', 'sendchange', 'multi-l10n', 'update'] So this is intentional. I have no idea what the configuration is really supposed to be... I'll cull it back down to the "real" list: [get-secrets build multi-l10n update]
(In reply to Axel Hecht [:Pike] from comment #1) > There are quite a few things about l10n repacks and artifact builds that > don't work out aside from this, tracked in bug 1387485. Point, but this isn't for the _repack_ task to be an artifact (--disable-compile-environment) configuration, it's for the _repacked_ task to be an artifact configuration. The underlying en-US APK, for Android. I'm not aware of any reason that won't work. Pike, are you aware of anything busted repacking artifact APKs?
Flags: needinfo?(l10n)
Looking at your pushes, I try to understand what you're heading for. You want a BA as base build for the l10n repack instead of the N? TBH, I wondered, but never wondered enough as to why Callek chose to build Nightly builds on try for l10n try jobs, that might be related? And no, never tried to kick off an l10n repack on top of a package from an artifact build.
No longer blocks: 1387485
Flags: needinfo?(l10n)
This seems to affect more than just artifact builds; the MinGW builds in beta currently are failing with this error. Example: https://treeherder.mozilla.org/#/jobs?repo=mozilla-beta&revision=1c336e874ae849a9f0b6d12bb2858c4af5120779&filter-searchStr=mingw&selectedJob=141844795
I'm going to close this ticket; I think it doesn't make sense. We can't produce N builds with artifact mode right now, 'cuz we need to do multi-locale stuff that artifact mode doesn't support. For the next poor sucker trying to work on (Android) single-locale repacks, you can: - |mach try| to push to try, waiting for the (slow) underlying N job to complete - push follow-up to try using |hg push try|, so that no jobs get scheduled - edit the Nx repack Task Cluster job to use the GECKO_HEAD_REV from the follow-up |hg push| to try That's a hell of a lot faster than triggering a new N each cycle. It just breaks down when you're editing the job definition itself, i.e., changing the taskgraph decision task to add toolchains or change scopes.
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.