fix: revert changes to test output for uncaught errors

[bartlomieju]

Alternative to #15230 as suggested in #15230 (comment)

CC @dsherret @nayeemrmn

[nayeemrmn]

I suggest removing this loop which causes the change in output (... cancelled and num failures):

deno/cli/tools/test.rs

Lines 1156 to 1164 in 45c4903

	for desc in tests.read().values() {
	if desc.origin == origin {
	sender.send(TestEvent::Result(
	desc.id,
	TestResult::Cancelled,
	0,
	))?
	}
	}

and also any place TestResult::Cancelled is referenced which would become dead code.

[nayeemrmn]

Or I guess changing the test is valid as well... out of curiosity has anyone been able to reproduce the flaky unexpected output locally? Running repeatedly on main (debug), 0f6b455 on canary linux (release), 0f6b455 on canary windows (release), I still haven't seen it. Seems to only happen on windows fastci.

[bartlomieju]

Or I guess changing the test is valid as well... out of curiosity has anyone been able to reproduce the flaky unexpected output locally? Running repeatedly on main (debug), 0f6b455 on canary linux (release), 0f6b455 on canary windows (release), I still haven't seen it. Seems to only happen on windows fastci.

No, I haven't been really able to reproduce the problem locally (M1, both debug and release). I'm more than open to changing the test output to avoid the flakiness.

[dsherret]

LGTM. I haven't been able to reproduce it on linux or windows after ~40mins of repeating the test in multiple test runners.