Microsoft.Extensions.Caching.Memory.Tests hanging in Mono after test execution finishes

[jaredpar]

Builds

Build	Pull Request	Test Failure Count
#629156	Rolling	1
#629314	Rolling	1
#629997	Rolling	1
#631423	Rolling	1
#632868	Rolling	1
#632967	Rolling	1
#633095	Rolling	1
#634699	Rolling	1

Configurations

• net5.0-Windows_NT-Release-x64-Mono_release-Windows.81.Amd64.Open
• netcoreapp5.0-Windows_NT-Release-x64-Mono_release-Windows.10.Amd64.Server19H1.Open
• netcoreapp5.0-Windows_NT-Release-x64-Mono_release-Windows.10.Amd64.ServerRS5.Open
• netcoreapp5.0-Windows_NT-Release-x64-Mono_release-Windows.81.Amd64.Open

Helix Logs

Build	Pull Request	Console	Core	Test Results	Run Client
#629156	Rolling	console.log			run_client.py
#629314	Rolling	console.log			run_client.py
#629997	Rolling	console.log			run_client.py
#631423	Rolling	console.log			run_client.py
#632868	Rolling	console.log			run_client.py
#632967	Rolling	console.log			run_client.py
#633095	Rolling	console.log			run_client.py
#634699	Rolling	console.log			run_client.py

[ghost]

Tagging subscribers to this area: @directhex
Notify danmosemsft if you want to be subscribed.

[jaredpar]

@steveisok @akoeplinger FYI

[akoeplinger]

@jaredpar all of the libicui18n.so.50: cannot open shared object file: No such file or directory errors are happening on a single PR, which is working on integrating the ICU shim so those failures are "expected" with the current state of the PR: #35790

The remaining two errors look like hangs since test execution seems to finish: https://helix.dot.net/api/2019-06-17/jobs/cfe744d2-90d6-4002-9304-aa45ea40efd0/workitems/Microsoft.Extensions.Caching.Memory.Tests/files/console.85791f98.log

We should remove the icu failures from this issue.

[safern]

Yeah, the icu failures are related to the changes of that PR, I confirmed it with @thaystg. We should definitely remove it from here.

The other 2 failures seem like a weird issue as the tests did finish, maybe loop in @MattGal here see if he has any insight there as xunit did finish.

[MattGal]

Taking a look.

[MattGal]

It's a timeout - from the log and where it stops, it looks like the xunit call never actually returns, otherwise it'd have various other things logged.

[safern]

So it seems like the test is barely finishing before the timeout but there is no timeout for the reporting steps, right?

[safern]

But XUnit reports that the test finished in 20s which is very weird.

[MattGal]

All that means here is that you made it to the point of writing that to the console. Something ... whether it be an infinite loop in a finalizer, a dialog that pops up that DHandler doesn't know how to click, or otherwise... is preventing that Xunit invocation from finishing.

[safern]

Updated the issue description, it seems like from 50 builds it has happened a lot more times than just 2.

All that means here is that you made it to the point of writing that to the console. Something ... whether it be an infinite loop in a finalizer, a dialog that pops up that DHandler doesn't know how to click, or otherwise... is preventing that Xunit invocation from finishing.

Yeah I just thought about that, @akoeplinger seems like a runtime hang? Seems like it only happens on Windows.

[safern]

Seeing this too often. So disabling this on: #36083

[akoeplinger]

@lateralusX seems to be a Windows only issue, would you mind trying to repro?