cc @ziransun

https://dev.azure.com/web-platform-tests/wpt/_build/results?buildId=52438&view=results should be an Edge Dev run with this commit.

The Edge run finished but there was an error (I suspect py3-caused) that resulted in it getting no 'revision' tag, which means wpt.fyi won't process it. Unfortunately such errors are deliberately swallowed. I've uploaded a patch that will hopefully log the error, and set off another run - https://dev.azure.com/web-platform-tests/wpt/_build/results?buildId=52472&view=results

Yep, vcs.py grabs the output of git rev-parse --show-cdup via subprocess.check_output, then compares it to "\n", which fails since the actual output is b"\n". I don't understand yet why this appears to work on Linux on Py3 (given our previous success wpt.fyi uploads), but for now I'm going to just stick a fix in here and trigger another Edge Dev run: https://dev.azure.com/web-platform-tests/wpt/_build/results?buildId=52525&view=results

I don't understand yet why this appears to work on Linux on Py3 (given our previous success wpt.fyi uploads)

https://community-tc.services.mozilla.com/tasks/f0VA2hcISQClpnai1ahO-w/runs/0/logs/live/https%3A%2F%2Fcommunity-tc.services.mozilla.com%2Fapi%2Fqueue%2Fv1%2Ftask%2Ff0VA2hcISQClpnai1ahO-w%2Fruns%2F0%2Fartifacts%2Fpublic%2Flogs%2Flive.log appears to confirm that it shouldn't, but yet we've definitely had successful uploads from Linux+Py3 in the past. Once that run is done I'll pull down both the artifacts for it (to see the run_info) and look at the wpt.fyi logs (to see if and how it passes the processor).

Moving back to Windows, the latest Edge Dev Py 3.8.5 run fails at the end with:

Traceback (most recent call last):
  File "D:\a\1\s\_venv3\Lib\site-packages\mozlog\structuredlog.py", line 274, in _handle_log
    handler(data)
  File "D:\a\1\s\_venv3\Lib\site-packages\mozlog\handlers\base.py", line 74, in __call__
    formatted = self.formatter(data)
  File "D:\a\1\s\_venv3\Lib\site-packages\mozlog\handlers\base.py", line 42, in __call__
    return self.inner(item)
  File "D:\a\1\s\_venv3\Lib\site-packages\mozlog\reader.py", line 74, in __call__
    return handler(data)
  File "D:\a\1\s\tools\wptrunner\wptrunner\formatters\wptreport.py", line 73, in suite_end
    return json.dumps(self.results) + "\n"
  File "C:\hostedtoolcache\windows\Python\3.8.5\x64\lib\json\__init__.py", line 231, in dumps
    return _default_encoder.encode(obj)
  File "C:\hostedtoolcache\windows\Python\3.8.5\x64\lib\json\encoder.py", line 199, in encode
    chunks = self.iterencode(o, _one_shot=True)
  File "C:\hostedtoolcache\windows\Python\3.8.5\x64\lib\json\encoder.py", line 257, in iterencode
    return _iterencode(o, 0)
  File "C:\hostedtoolcache\windows\Python\3.8.5\x64\lib\json\encoder.py", line 179, in default
    raise TypeError(f'Object of type {o.__class__.__name__} '
TypeError: Object of type bytes is not JSON serializable

So there are some bytes somewhere in self.results I guess.

Success, we have Py3.8.5 + Edge Dev results! https://wpt.fyi/results/?run_id=640430002

https://wpt.fyi/results/?q=is%3Adifferent&run_id=626930002&run_id=640430002 is the diff to the nearest run (2df9cdd has two commits that aren't in the RHS, but should be minor). FYI @ziransun .

Doing another Edge run is blocked on #25923

I have pushed e2af840 to triggers/edge_dev. The parent commit is c960a24, aka https://wpt.fyi/results/?run_id=714260003

Will update with a wpt.fyi diff link once the run is done.

This is looking good! I've successfully run two runs, the second of which had to uninstall and reinstall py3.8 and did so fine. I think we're in a good place to turn this into a CL that we could land once we wanted to go to Py3.8 (e.g. by getting rid of most of the debug logging, etc, will do that today).

Thanks @mustjab for your work here - very much appreciated!