Since bug 1403322 landed, jsreftests seem to run on nearly every push (at least on some platform). I believe they should only run when there are files modified under js/src or js/public. https://treeherder.mozilla.org/#/jobs?repo=autoland&filter-searchStr=jsreftest&fromchange=e188cb9fe8f5f3e73e06a13cdebde190b142d432&tochange=54015b4eb2ed3b79d5513c0a7547bae586dc6715 (test-verify and jittest seem okay this time!)

In task-graph.json for a recent push on autoland: https://public-artifacts.taskcluster.net/ARQoxkKHRwSSa16BAa68Jw/0/public/task-graph.json "APPUn87gTY-7gI8fQIGFpA": { "attributes": { ... "unittest_flavor": "jsreftest", "unittest_suite": "reftest", "unittest_try_name": "jsreftest" }, "label": "test-android-4.3-arm7-api-16/debug-jsreftest-15", "optimization": { "skip-unless-schedules-or-seta": [ "reftest", "android" ] }, yet python/mozbuild/mozbuild/schedules.py has INCLUSIVE_COMPONENTS = [ ... 'jsreftest', ] so I think the fix here is to make this suite `jsreftest`. I don't really understand the ramifications of that kind of change, though. Do you? Would something break? I think it just ends up in the task.extra. Maybe Joel knows if that would cause issues?

I don't know for sure, but we could validate on try. Reftest is a harness and there are 3 main test types: reftest, crashtest, jsreftest. jsreftest is the most unique of the 3. What I don't know is how taskcluster and mozharness play into decisions with suites and specific names. The only concern I might have here is that we could have odd try scheduling quirks with the old syntax vs fuzzy syntax. In short, we should be able to fix any fallout :)

[task 2018-01-12T14:40:15.614Z] desktop_unittest.py: error: no such option: --jsreftest-suite I think what I will do is add a parameter to test descriptions that can override the scheduling component -- so usually it's the suite, but can be set to something else. In this case, to 'jsreftest' :)

Comment on attachment 8942189 [details] Bug 1429236: put jsreftests in their own scheduling component; https://reviewboard.mozilla.org/r/212448/#review218206 while I like the name chose, the try push failed: desktop_unittest.py: error: no such option: --jsreftest-suite please ensure that there is tests for Android as well, that is treated differently.

How is Android different?

different harness scripts- cli args might be handled differently

Thanks, I'll include those in the next try push

Comment on attachment 8942189 [details] Bug 1429236: put jsreftests in their own scheduling component; https://reviewboard.mozilla.org/r/212448/#review219978 slight change in direction, looks good,

We're sorry, Autoland could not rebase your commits for you automatically. Please manually rebase your commits and try again. hg error in cmd: hg rebase -s 699c807f3f6b50e70a6c3ee1a9ae1a679f2882a7 -d 7438fea5bc44: rebasing 443284:699c807f3f6b "Bug 1429236: put jsreftests in their own scheduling component; r=jmaher" (tip) merging taskcluster/ci/test/reftest.yml merging taskcluster/taskgraph/transforms/tests.py warning: conflicts while merging taskcluster/ci/test/reftest.yml! (edit, then use 'hg resolve --mark') unresolved conflicts (see hg resolve, then hg rebase --continue)

Pushed by dmitchell@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/9c1e9d7bec5f put jsreftests in their own scheduling component; r=jmaher