Closed Bug 1546732 Opened 8 months ago Closed 8 months ago

windows/aarch64 - enable jittest

Categories

(Testing :: General, task)

task
Not set

Tracking

(firefox68 fixed)

RESOLVED FIXED
mozilla68
Tracking Status
firefox68 --- fixed

People

(Reporter: egao, Assigned: egao)

References

Details

(Whiteboard: [arm64:m3])

Attachments

(2 files)

Enable jittest for windows10-aarch64 on try and mozilla-central.

Disable any tests that fail as part of this work.

Try server should continue to require the --full command line argument in order to schedule tasks.

Assignee: nobody → egao
Type: defect → task
Component: CPPUnitTest → General
No longer depends on: 1546728
Pushed by egao@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/dc42ae4e7a3a
enable jittest for windows10-aarch64 r=gbrown
Status: NEW → RESOLVED
Closed: 8 months ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla68

This did not start running jittest on central. I think that's because of:

https://searchfox.org/mozilla-central/rev/d143f8ce30d1bcfee7a1227c27bf876a85f8cede/taskcluster/ci/test/compiled.yml#95

...which makes me wonder, should we run SM instead?

Flags: needinfo?(egao)

(In reply to Geoff Brown [:gbrown] (pto April 26-29) from comment #4)

This did not start running jittest on central. I think that's because of:

https://searchfox.org/mozilla-central/rev/d143f8ce30d1bcfee7a1227c27bf876a85f8cede/taskcluster/ci/test/compiled.yml#95

...which makes me wonder, should we run SM instead?

Checked the base regex (?=windows).*(?!-ccov) with regex tester, and it does catch this example string windows10-aarch64. What if I make the regex more specific? That may be not worth the amount of hours spent on it.

Are Jit and SpiderMonkey the same thing?

Flags: needinfo?(egao) → needinfo?(gbrown)

I know very little about the SpiderMonkey tests.

:jmaher, :cpeterson -- windows/aarch64 jittest is ready for mozilla-central, but we have noticed that jittest is not normally run on Windows -- a comment says it is redundant with SpiderMonkey. Can you shed light on that? Ultimately we want to know: Do we want to run windows/aarch64 jittest, and do we want to run windows/aarch64 SM?

Flags: needinfo?(jmaher)
Flags: needinfo?(gbrown)
Flags: needinfo?(cpeterson)

running jittest on win/aarch64 allows us to gain arm64 coverage. The plan is:
win/aarch64:

  • jittest
  • jsreftest

android/arm7:

  • jittest
  • jsreftest

the goal here is to have good coverage for arm7 and arm64.

Flags: needinfo?(jmaher)

(In reply to Geoff Brown [:gbrown] (pto April 26-29) from comment #6)

:jmaher, :cpeterson -- windows/aarch64 jittest is ready for mozilla-central, but we have noticed that jittest is not normally run on Windows -- a comment says it is redundant with SpiderMonkey. Can you shed light on that? Ultimately we want to know: Do we want to run windows/aarch64 jittest, and do we want to run windows/aarch64 SM?

Sean, why are the jittests "redundant with SM(p)" [1] on Windows? Is this true for windows/aarch64?

https://searchfox.org/mozilla-central/rev/d143f8ce30d1bcfee7a1227c27bf876a85f8cede/taskcluster/ci/test/compiled.yml#95

Flags: needinfo?(cpeterson) → needinfo?(sstangl)
Whiteboard: [arm64:m3]
Pushed by egao@mozilla.com:
https://hg.mozilla.org/integration/autoland/rev/500cdbe0977e
enable jittest on try, mozilla-central for windows10-aarch64 r=gbrown
See Also: → 1549763

Sean, why are the jittests "redundant with SM(p)" [1] on Windows? Is this true for windows/aarch64?

If I'm reading js/src/devtools/automation/autospider.py correctly, the default options of SM(p) run all of "jstests", "jittest", "jsapitests", and "checks".

Based on that, I think it would be true for windows/aarch64 also, if we turn on SM(p).

Flags: needinfo?(sstangl)
Blocks: 1552051
You need to log in before you can comment on or make changes to this bug.