As discussed in this weeks's mobile meeting.

So to clarify armv6 tests were removed from branches < 32 in bug 1033507 (the ones that ran on tegras). So this is a request to remove the armv6 builds and tests that run on emulators that were implemented in bug 1020970. So that the only builds and tests that run on emulators will be 2.3 tests against 2.2 builds (which will be renamed to 2.3 builds in bug 1042838), and armv6 will be removed from branches < 32 Blassey is this a correct interpretation of the request?

Looks right to me

builder for build master

diff for test master The issue with this patch is that Armv6 builds can be disabled for < 32 and it looks as expected. For the tests, they were only enabled starting with 33 anyways so disabling them for < 32 disables them all

tbpl patch

Comment on attachment 8461928 [details] [diff] [review] bug1042835.patch Review of attachment 8461928 [details] [diff] [review]: ----------------------------------------------------------------- r+ for all bits but the builder name landings. I'd prefer we get "support" in the various places: * TBPL * Slave Health * Wait Times Report * Any I'm Missing before we do that, and its tangential to this actual bug (we can even cheat and have TBPL show as 2.3 even if the official buildername is 2.2) ::: mozilla-tests/mobile_config.py @@ +1645,5 @@ > + } > + BRANCHES[name]['platforms']['android-armv6']['ubuntu64_vm_armv6_mobile'] = { > + 'opt_unittest_suites': [], > + } > + nit: whitespace EOL ::: mozilla/config.py @@ +1225,5 @@ > 'product_name': 'firefox', > 'unittest_platform': 'android-opt', > 'app_name': 'browser', > 'brand_name': 'Minefield', > + 'base_name': 'Android 2.3 %(branch)s', nit: can we land this in a separate patch, with coordination to all-the-places we do "builder name" munging. This is a good idea, just not one I want to see without the various bits adjusted/ready first. @@ +1295,5 @@ > 'android-armv6': { > 'product_name': 'firefox', > 'unittest_platform': 'android-armv6-opt', > 'app_name': 'browser', > + 'base_name': 'Android 2.3 Armv6 %(branch)s', same @@ +1426,5 @@ > 'enable_nightly': False, > 'product_name': 'firefox', > 'app_name': 'browser', > 'brand_name': 'Minefield', > + 'base_name': 'Android 2.3 Debug %(branch)s', same

Comment on attachment 8462584 [details] [diff] [review] bug1042835tbpl.patch Review of attachment 8462584 [details] [diff] [review]: ----------------------------------------------------------------- This patch breaks Android !armv6 builds on TBPL. Also given comment 8 I think the name change pieces should be done in another bug? :-)

patch without builder name renames bug 1042838 has been reeopened to address builder name renames

So it turns out that since Android armv6 builds are disabled on release we also need to update the release templates otherwise chkconfig on some of the masters fail. test-masters.sh passes with this patch

Comment on attachment 8464047 [details] [diff] [review] bug1042835-3.patch Review of attachment 8464047 [details] [diff] [review]: ----------------------------------------------------------------- Looks good to me, but since this is release stuff and I've been out of the loop there for some time I'd like an additional review from a release person. (incase there are other things needing to be coped with for dropping this platform)

I've adjusted the fennec_push.sh script by making a copy for beta and release cases https://hg.mozilla.org/build/braindump/rev/2eb6e78db94b https://hg.mozilla.org/build/braindump/rev/46c013d28104 Doc update: https://wiki.mozilla.org/index.php?title=Release%3ARelease_Automation_on_Mercurial%3AUpdates_through_Shipping&diff=1001161&oldid=1001160 The release one is unchanged, but we can do that on bug 1036571.

Review of this is coming, back can somebody further clarify why we're removing armv6 from Gecko 32 and less? I thought we were shipping armv6 off of mozilla-esr31 now. I'm sure there's a good reason, but I don't understand why we'd run any armv6 builds or tests on any other branches given that.

(In reply to Ben Hearsum [:bhearsum] from comment #15) > Review of this is coming, back can somebody further clarify why we're > removing armv6 from Gecko 32 and less? I thought we were shipping armv6 off > of mozilla-esr31 now. I'm sure there's a good reason, but I don't understand > why we'd run any armv6 builds or tests on any other branches given that.

Comment on attachment 8464047 [details] [diff] [review] bug1042835-3.patch Review of attachment 8464047 [details] [diff] [review]: ----------------------------------------------------------------- The changes to the release configs look fine, but shouldn't we be removing it from Beta too? ::: mozilla-tests/mobile_config.py @@ +1617,5 @@ > } > > # bug 1020970 Schedule all Android 2.3 armv6 tests, except mochitest-gl, > # on all trunk trees and make them ride the trains > + # Loop removes it from any branch that gets beyond here Did you really mean to remove the "for name, branch in items_before" loop here? The block it's now a part of is for things that are Gecko 32 or higher, which I think is the opposite of what you want. @@ +1622,5 @@ > + for platform in branch['platforms']: > + if not platform in PLATFORMS: > + continue > + if not platform == ('android-armv6'): > + continue If you're only looking for a single platform there's no reason to have this check - just don't use a loop. However, I'm also unsure why you're using slave platform adjustment. Is there a reason you're not just del'ing the entire platform branch BRANCHES[name]['platforms'] as in config.py?

(In reply to Ben Hearsum [:bhearsum] from comment #15) > Review of this is coming, back can somebody further clarify why we're > removing armv6 from Gecko 32 and less? I thought we were shipping armv6 off > of mozilla-esr31 now. I'm sure there's a good reason, but I don't understand > why we'd run any armv6 builds or tests on any other branches given that. We're end of life'ing support for ARMv6. Our end of life plan is to stop shipping new features to ARMv6 users after 31, but to continue shipping security fixes to them for 6 months. So, the 31esr builds are to provide those security fixes.

Comment on attachment 8464047 [details] [diff] [review] bug1042835-3.patch (In reply to Brad Lassey [:blassey] (use needinfo?) from comment #18) > (In reply to Ben Hearsum [:bhearsum] from comment #15) > > Review of this is coming, back can somebody further clarify why we're > > removing armv6 from Gecko 32 and less? I thought we were shipping armv6 off > > of mozilla-esr31 now. I'm sure there's a good reason, but I don't understand > > why we'd run any armv6 builds or tests on any other branches given that. > > We're end of life'ing support for ARMv6. Our end of life plan is to stop > shipping new features to ARMv6 users after 31, but to continue shipping > security fixes to them for 6 months. So, the 31esr builds are to provide > those security fixes. OK, so this bug summary isn't quite right. We should stop doing armv6 builds/tests everywhere except esr31. Kim, I think this means you'll need blocks like this win64 one rather thans items_before/items_at_least: https://github.com/mozilla/build-buildbot-configs/blob/master/mozilla-tests/config.py#L1913

We don't currently have anything running on an ESR branch for either tegra or Emulator tests on Arvm6. Or builds. My understanding is that how we'll deliver this is being worked out in bug 1042128. So thus I'll remove all of the Armv6 builds and tests. If we have an ESR branch created as a result of the work in bug 1042128, appropriate jobs can be added. Does this sound okay?

wfm

This disables all armv6 builds and tests and all branches. I left the platform definitions in for ESR work in bug 1036571

builder diff for test master also running test-masters.sh was clean

Comment on attachment 8465635 [details] [diff] [review] bug1042835-4.patch Review of attachment 8465635 [details] [diff] [review]: ----------------------------------------------------------------- other than mostly nits and two small issues/concerns this looks good. Unless something more substantive changes r+ ::: mozilla-tests/mobile_config.py @@ +1401,5 @@ > + 'remote_extras': ANDROID_UNITTEST_REMOTE_EXTRAS, > + 'tegra_android-armv6': { > + 'opt_unittest_suites': [], > + 'debug_unittest_suites': [], > + }, nit whitespace EOL (two places above) ::: mozilla/config.py @@ +1433,5 @@ > 'multi_locale_script': 'scripts/multil10n.py', > 'tooltool_manifest_src': 'mobile/android/config/tooltool-manifests/android/releng.manifest', > }, > 'android-armv6': { > + 'enable_nightly': False, nit: whitespace EOL @@ +2485,5 @@ > branches = BRANCHES.keys() > branches.extend(ACTIVE_PROJECT_BRANCHES) > for branch in branches: > + if 'android-armv6' in BRANCHES[branch]['platforms']: > + del BRANCHES[branch]['platforms']['android-armv6'] huh comment above this says "remove the following block after gcc-4.5 switch. which sounds like this could get accidentally removed if done as-is. Would rather its own block with own comment. ::: mozilla/release-fennec-mozilla-beta.py @@ +125,5 @@ > > releaseConfig['multilocale_config'] = { > 'platforms': { > 'android': > + 'multi_locale/release_mozilla-beta_android.json', nit: whitespace EOL

patch with review comments included

in production and verified on tbpl

something(s) here went to production today

I noticed that there are no Android jobs on esr31 and saw this bug referenced adding armv6 builds [1] # bug 1042835 Disable armv6 builds and tests everywhere apart from esr31 [2] [1] https://treeherder.mozilla.org/#/jobs?repo=mozilla-esr31&filter-searchStr=android [2] http://hg.mozilla.org/build/buildbot-configs/file/default/mozilla/config.py#l2855