This bug requires parsing the .json changesets file And then passing in the explicit revisions to the mozharness commands for locales, via the task graph.

This comes as a surprise to me. How is this intended to pick up changes from the l10n dashboard, and how does the timing look?

Comment on attachment 8838130 [details] Bug 1337825 - Jamun: Change android-api-15-{l10n,nightly} to l10n bumped file Oops, some tweak are still needed https://tools.taskcluster.net/task-group-inspector/#/D3YOpqk6QkiGHAMC3jEmyg/G3g88SHAS4aaLLDaLneYCA?_k=o3gey8 . Even though I correctly pass the revision, I need to point it to the correct repo.

:pike - bug 1339641 and bug 1337645. All in-tree.

Here's how it worked on jamun: * https://tools.taskcluster.net/task-group-inspector/#/KS9VknC0Sq6ls8Xp3XKfqA * https://treeherder.mozilla.org/#/jobs?repo=jamun&revision=a0d4507034847c53a939fb39f011e3cd43e9dba7 We're good for review this time :)

Comment on attachment 8838123 [details] Bug 1337825 - Schedule l10n repacks for fennec with specified locale revisions using an intree changesets file https://reviewboard.mozilla.org/r/113124/#review114584 Once we have l10n-changesets.json on mozilla-central and mozilla-aurora, we'll want this patch on both of them. r+ with nits; let me know if you want me to change the platforms list in l10n-changesets.json. ::: taskcluster/docs/attributes.rst:135 (Diff revision 1) > of relevant locales for the platform. > > +all_locales_with_changesets > +=========================== > + > +Contains a dict of l10n changesets, mappyed by locales (same as in ``all_locales``). *mapped ::: taskcluster/taskgraph/transforms/l10n.py:144 (Diff revision 1) > locales = [] > + > + with open(locales_file, mode='r') as f: > - if locales_file.endswith('json'): > + if locales_file.endswith('json'): > - # Release process uses .json for locale files sometimes. > - raise NotImplementedError("Don't know how to parse a .json locales file") > + all_locales = json.load(f) > + single_locales = {locale: data['revision'] for locale, data in all_locales.items() if 'android' in data['platforms']} We'll probably have to deal with this hardcode at some point. Would you prefer I put android-api-15 and android-x86 in the platforms list? We can do that. Also, do you want to set `locales` here? We don't have to specify `multi_locales` so that change would simply this block. ::: taskcluster/taskgraph/transforms/l10n.py:157 (Diff revision 1) > > +def _remove_ja_jp_mac_locale(locales): > + # ja-JP-mac is a mac-only locale, but there are no mac builds being repacked, so just omit it unconditionally > + return { > + locale: revision for locale, revision in locales.items() if locale != 'ja-JP-mac' > + } If we do standardize on the json, we would eliminate ja-JP-mac from non-mac platforms by the `platforms` key/value pair. It's harmless for now.

Comment on attachment 8838130 [details] Bug 1337825 - Jamun: Change android-api-15-{l10n,nightly} to l10n bumped file https://reviewboard.mozilla.org/r/113126/#review114630 ::: testing/mozharness/configs/multi_locale/jamun_android-armv6.json:5 (Diff revision 2) > +{ > + "work_dir": ".", > + "log_name": "multilocale", > + "objdir": "obj-firefox", > + "locales_file": "build/mobile/android/locales/maemo-locales", here too. ::: testing/mozharness/configs/multi_locale/jamun_android-x86.json:5 (Diff revision 2) > +{ > + "work_dir": ".", > + "log_name": "multilocale", > + "objdir": "obj-firefox", > + "locales_file": "build/mobile/android/locales/maemo-locales", and here. ::: testing/mozharness/configs/multi_locale/jamun_android.json:5 (Diff revision 2) > +{ > + "work_dir": ".", > + "log_name": "multilocale", > + "objdir": "obj-firefox", > + "locales_file": "build/mobile/android/locales/maemo-locales", We'll want to a) point this at `l10n-changesets.json`, and b) specify `"locales_platform": "android-multilocale",` per https://reviewboard.mozilla.org/r/112934/diff/1#index_header If you want to do that here, great. If you want me to update these files when I get the rest of bug 1339706 ready, let me know. ::: testing/mozharness/configs/single_locale/jamun_android-api-15.py:15 (Diff revision 2) > + "objdir": "obj-l10n", > + "is_automation": True, > + "buildbot_json_path": "buildprops.json", > + "force_clobber": True, > + "clobberer_url": "https://api.pub.build.mozilla.org/clobberer/lastclobber", > + "locales_file": "%s/mobile/android/locales/all-locales" % MOZILLA_DIR, I think we can specify the l10n-changesets file here, and specify the locales_platform as `android` or `android-api-15` depending whether we change the platforms list. However, since we're explicitly passing in the `--locales`, I believe this is ignored.

Comment on attachment 8838513 [details] Bug 1337825 - Jamun: Change android-api-15-{l10n,nightly} to l10n bumped file Crap, I messed up with mozreview. This patch is attachment 8838130 [details] + review comments addressed. Carrying over r+. I addressed them all, with the exception of: > We'll want to a) point this at `l10n-changesets.json`, done > and b) specify `"locales_platform": "android-multilocale",` per https://reviewboard.mozilla.org/r/112934/diff/1#index_header I left this change for bug 1339706 > and specify the locales_platform as android or android-api-15 Like above, it should be done in bug 1339706

Latest changes tested against: * Jamun (where locale comes from l10n-changeset.json): https://tools.taskcluster.net/task-group-inspector/#/R8V_mbnuQoasxeuMzGwgaw * Date (where locales are from the all-locale file, and hence using "default" as revision for l10n repos): https://tools.taskcluster.net/task-group-inspector/#/A1mferJjTk-uWbkDRyJ6RA I'm gonna land attachment 8838123 [details] in central. This needs to ride aurora, so leaving the bug open. I won't land attachment 8838513 [details] until l10n-changesets.json is present in central and aurora. Keeping this bug open for this reason too.

Pushed by jlorenzo@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/c7c9f7712163 Schedule l10n repacks for fennec with specified locale revisions using an intree changesets file r=aki

sorry had to back out for flake 8 error like https://treeherder.mozilla.org/logviewer.html#?job_id=78299256&repo=mozilla-inbound

Pushed by jlorenzo@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/17d5fa614caa Schedule l10n repacks for fennec with specified locale revisions using an intree changesets file r=aki

Thank you for spotting this error :Tomcat. I just relanded a flake8-error-free revision[1]. [1] Tested locally with `mach lint -l flake8 -f treeherder`

Backout by cbook@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/d226bf0c5b39 Backed out changeset c7c9f7712163 for flake8 failure

Jamun now has the new l10n-changesets.json format, if you want to key off 'android-api-15' in platforms. I plan on having the new l10n-changesets.json in m-c and m-a earlyish next week, though I'm off Monday. https://hg.mozilla.org/projects/jamun/rev/c26b44d6efa1

Landed in aurora at: https://hg.mozilla.org/releases/mozilla-aurora/rev/0ef6ce3951925a1972396f61db87aedfb3c9004a

Discussed with Aki, we should also land the Jamun changes on central and aurora. I applied attachment 8838513 [details] on top of inbound and run all the linters[1]. No errors related to the patch were detected. I pushed the patch on inbound. [1] `./mach lint -f treeherder`

Pushed by jlorenzo@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/405ee7f97cfa Jamun: Change android-api-15-{l10n,nightly} to l10n bumped file r=aki

Attachment 8838513 [details] landed in aurora at https://hg.mozilla.org/releases/mozilla-aurora/rev/982aaad2e5d766031aa81e7f4366524bdae950f4 All the changes will now ride the trains. This bug can now be closed.