The continuing intermittent macOS CI build failures

[Numpsy]

We seem to still have an intermittent CI build failure on macOS, with errors about not being able to get the start time of a process -

https://github.com/fsprojects/FAKE/actions/runs/10654117014/job/29530030223#step:8:9395

Looking at the code, it seems to expect to fail to get the process start time sometimes, and responds by logging it and returning DateTime.Now -

FAKE/src/app/Fake.Core.Process/Process.fs

Line 470 in fac7295

Trace.traceFAKE "Error while retrieving StartTime of started process: %O" e

There are sometimes multiple instances of that error in the logs but the only one that seems to make it break is in the no dependencies hello world and casing #2314 test case at

FAKE/src/test/Fake.Core.IntegrationTests/SimpleHelloWorldTests.fs

Line 79 in fac7295

testCase "no dependencies hello world and casing #2314"

where it explicitly expects that there is no error output from the test process -

FAKE/src/test/Fake.Core.IntegrationTests/SimpleHelloWorldTests.fs

Line 95 in fac7295

stdErr.Trim() |> Expect.equal "empty expected" ""

I'm not sure why the error occurs (maybe it is just because the test process exits too quickly and is gone by the time the runner tries to get the time), so I 'm not sure if there is a bug or if the test is a bit too strict?
The error only seems to happen on macOS, though there is a different code path for Windows vs. macOS/Linux that might effect things there.

[Thorium]

Ok it seems the test is asserting that there is no output. But Fake.Core.Trace.TraceFAKE will be output. Now the question is, if proc.StartTime fails to get, is it just that this platform didn't work and we can ignore, or is it so important error that we should fail the test?

[Numpsy]

Now the question is, if proc.StartTime fails to get, is it just that this platform didn't work and we can ignore, or is it so important error that we should fail the test?

The way it catches the exceptions and returns DateTime.Now suggests an explicit decision to not treat it as an error, but I don't know any history behind the decision