Why are test runs failing without any error after adding this logger?

[AArnott]

I'm adding your test logger to my repo template for spreading around a bunch more repos:
AArnott/Library.Template#319

But it's causing all non-Windows build agents to fail without any error message like this.

Any idea why?

[Tyrrrz]

No clue, there seems to not be any error message produced in the process. Does removing IncludeAssets="runtime" do anything?

[Tyrrrz]

I have not used the --diag option so don't really know what to advise there, sorry

[AArnott]

I worked around the exit code 1 failure by suppressing the --diag switch when running in GitHub Actions on linux/mac. It always adds --logger GitHubActions when under GitHub Actions, however.

But when testing the annotations from the logger, I see it only works on Windows agents. I have arranged for unique test failures on each agent, but only the Windows agents' failures result in annotations, as can be seen here.

Is that a known issue?

[Tyrrrz]

No, it is not. Sounds like something is wrong with your setup.

Here's a working example for reference: https://github.com/Tyrrrz/CliWrap/actions/runs/12285513560

[AArnott]

It turns out that your logger doesn't record annotations on mac/linux when --collect "Code Coverage;Format=cobertura" or --settings some.runsettings is specified on the CLI as well. So in the end, I got it to work by adding a few more condition checks to avoid using those switches on GitHub Actions (while retaining them everywhere else).

So yay, I'm unblocked and can use your logger. But it would sure be nice to find out why it (or the test runner?) is so fragile.

[AArnott]

Well shoot. I'm going to have to roll this back because once my template repo merged into one of my real repos (with a lot of tests), the test step fails with a crash in the runner, with this message:

Data collector 'Blame' message: Data collector caught an exception of type 'System.ArgumentException': 'Process with an Id of 1980 is not running.'. More details: Blame: Creating hang dump failed with error...

I haven't been able to get a callstack for this exception. I really like the annotations that this logger offers, but I think I'll have to revisit and try again in a few months or longer, because this has introduced too much instability and complexity to my test scripts to try to make work.

[Tyrrrz]

I am using it with --settings some.runsettings myself. I also collect coverage, although not specifically with cobertura.

https://github.com/Tyrrrz/.github/blob/b37ad17e1525b7a27ab8aa9baa7c16a2e38d363c/.github/workflows/nuget.yml#L100-L108

If I did have to guess, I'd assume it's the --diag switch that's causing this.

[AArnott]

Thanks for reviewing.

I already took the --diag switch out (as was required to get past the "exit 1" with no error message problem. But even without that switch, and with changes to my code coverage and test settings switches to match yours, I still got the ArgumentException that crashes the test runner (and only on Windows IIRC) and fails the overall step in the workflow, once I applied it to a real repo with a decent set of tests.