I see instances of win7-ix machines on try running reftest, and reftest-unaccel. we should ensure we reduce our load on the win7-ix machines as much as reasonably possible- and make sure jobs are equal between production and try.

I see jittest and jsreftest as well

Tests that are currently running on Windows 7 hardware machines.

Tests that are currently running on Windows 7 VMs.

It follows that we have 21 tests that are running both on HW and VMs (all of them on try): Windows 7 32-bit try opt test mochitest-other Windows 7 32-bit try opt test jsreftest Windows 7 32-bit try opt test crashtest Windows 7 32-bit try opt test jittest Windows 7 32-bit try opt test reftest-no-accel Windows 7 32-bit try opt test reftest Windows 7 32-bit try opt test mochitest-media-e10s Windows 7 32-bit try opt test reftest-e10s Windows 7 32-bit try opt test mochitest-browser-screenshots Windows 7 32-bit try opt test mochitest-clipboard Windows 7 32-bit try opt test mochitest-clipboard-e10s Windows 7 32-bit try debug test mochitest-other Windows 7 32-bit try debug test jsreftest Windows 7 32-bit try debug test crashtest Windows 7 32-bit try debug test jittest Windows 7 32-bit try debug test reftest-no-accel Windows 7 32-bit try debug test reftest Windows 7 32-bit try debug test reftest-e10s Windows 7 32-bit try debug test mochitest-media-e10s Windows 7 32-bit try debug test mochitest-clipboard Windows 7 32-bit try debug test mochitest-clipboard-e10s

Patch to remove Win 7 HW try tests that also run on VMs.

Tested the patch on dev-master2 and worked without issues. Attached the tests diff file.

Clipboard is the exception in that list, because although it runs on both the jobs on VM do not work and are hidden (and are the reason the suite exists, because those tests don't work on VMs).

Oh, not the only exception. The four things we're still running on hardware on integration branches and mozilla-central, and that must still run on hardware on try as well, are mochitest-clipboard, mochitest-clipboard-e10s, mochitest-other and mochitest-media-e10s.

Regarding those 4 tests, we have the following: - mochitest-clipboard, mochitest-clipboard-e10s: VM tests are generally failing, hidden in TH - mochitest-other: VM tests are mostly green, yet they are also hidden in TH - mochitest-media-e10s: VM tests are generally failing, but they are not hidden in TH I also checked the rest of the VM tests and they all look fine, but some of them will *not* run by default (as their correspondent HW ones do) if 'Windows 7' is not explicitly specified as platform: - reftest-no-accel - reftest - reftest-e10s @jmaher: any thoughts on this?

the reason to keep jobs duplicated is for theoretical work to green them up. ideally we will fix the issues with clipboard in firefox and the mochitest-clipboard* jobs will be green; there are 2 tests which need attention in mochitest-other, and mochitest-media-e10s needs some attention for a few tests before moving it over. I am fine with all of those tests hidden by default on try (we should make mochitest-media-e10s hidden by default for win7-vm on try only). we really should have reftest* jobs running by default without requiring 'windows 7'. thanks for bringing that up.

(In reply to Alin Selagea [:aselagea][:buildduty] from comment #11) > Regarding those 4 tests, we have the following: > - mochitest-clipboard, mochitest-clipboard-e10s: VM tests are generally > failing, hidden in TH > - mochitest-other: VM tests are mostly green, yet they are also hidden > in TH > - mochitest-media-e10s: VM tests are generally failing, but they are not > hidden in TH Update on this: - mochitest-clipboard, mochitest-clipboard-e10s: nothing changed - VM tests are generally failing, hidden in TH - mochitest-other: it was split in [a11y, chrome] on all platforms in bug 1338610 - a11y: only running on VMs now (bug 1342202), not hidden in TH - chrome: VM tests are green, but hidden in TH - mochitest-media-e10s: VM tests are green this time, not hidden in TH Should the green ones run on AWS only?

all fully green tests should be AWS only- right now that is not mochitest-chrome, but it is: mochitest-a11y, and from the comment mochitest-media-e10s. I would like to ensure that mochitest-media-e10s has 50 retriggers with <=3 failures total, otherwise it might be too intermittent for running on AWS.

(In reply to Joel Maher ( :jmaher) from comment #14) > I would like to ensure that mochitest-media-e10s has 50 retriggers with <=3 > failures total, otherwise it might be too intermittent for running on AWS. I did a push to try to see how things go for multiple runs of mochitest-media-e10s on Windows 7 VMs: https://treeherder.mozilla.org/#/jobs?repo=try&revision=4c861bc93933e51a6abe88eb9c4c26dd4868230d&exclusion_profile=false&filter-searchStr=Windows%207%20VM As it can be noticed, there's a single failure for "Windows 7 VM opt" (out of 50 runs) and none for "Windows 7 VM debug" (again, out of 50 runs). That sounds like we're ready to stop running this test on hardware :-)

Found in triaging. @aselagea: is there anything else to do here or can we close this?

Most Windows 7 unittests are already in TC, so I don't think this is something to worry about anymore.