Open matthiaskrgr opened 2 years ago
@rustbot label +O-windows
This is causing a lot of spurious failures, as evidenced by the numerous "mentioned this issue" items in this issue. Should this test be ignored temporarily until the underlying problem is identified and fixed? cc @rust-lang/infra
@rustbot modify labels: +O-windows-msvc
I'm on PTO but I'll put this on my list of things to look into when I get back next week. In the mean time, disabling the test seems reasonable to me.
The test is now ignored, so there shouldn't be any more spurious failures. However, the underlying issue isn't fixed yet.
The comment in the test seems potentially relevant:
If dbghelp.dll
is essentially returning an arbitrary symbol name when the test passes, perhaps it's (non-deterministically?) choosing to give up and not return any symbol in the cases where the test fails.
We could modify the normalize-stderr-test
regex replacement to just remove the whole "5: ..." line if it exists.
The
ui\panics\panic-short-backtrace-windows-x86_64.rs
test seems to randomly fail onx86_64-msvc-2
sometimes.https://pipelines.actions.githubusercontent.com/DhU3eyB11Ei4viT99vzh0CpKzg7REqx1Nup2tB9LzmxBcoHdbh/_apis/pipelines/1/runs/14339/signedlogcontent/31?urlExpires=2021-12-16T12%3A15%3A54.9212371Z&urlSigningMethod=HMACV1&urlSignature=%2FsTx%2F9cdlfikD4rjevyZM9XXK2s%2BqXIVMXzWYgCm%2BpI%3D