test: wait for stream finish when --test-force-exit

[avivkeller]

Fixes #54327.

This PR stores the reporter streams inside of TestsStream, and waits for them to finish before exitting the process.

[nodejs-github-bot]

Review requested:

• @nodejs/test_runner

[cjihrig]

This is introducing new public API that I don't think we want to support.

[avivkeller]

I can redo it with a symbol / _? Unless you know a better way to wait for the streams?

[cjihrig]

Technically, it doesn't need to be attached to the stream at all. For example, the reporterScope object is internal and could hold them (you would need to make sure it works with watch mode though).

[avivkeller]

IMO attaching to the stream makes the most sense, it's the.. well... stream.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.07%. Comparing base (7014e50) to head (2e11df6).
Report is 99 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #54953   +/-   ##
=======================================
  Coverage   88.06%   88.07%           
=======================================
  Files         652      652           
  Lines      183545   183558   +13     
  Branches    35854    35859    +5     
=======================================
+ Hits       161648   161670   +22     
+ Misses      15146    15143    -3     
+ Partials     6751     6745    -6     

Files with missing lines	Coverage Δ
lib/internal/test_runner/test.js	96.93% <100.00%> (-0.07%)	⬇️
lib/internal/test_runner/tests_stream.js	91.30% <100.00%> (+0.22%)	⬆️
lib/internal/test_runner/utils.js	64.05% <100.00%> (+0.20%)	⬆️

... and 25 files with indirect coverage changes

[avivkeller]

@cjihrig the issue was that I was originally appending stream, not destination.

[avivkeller]

It appears now it's waiting too long 😅 .

It looks like it's waiting for the schedule to complete, I wonder why? Maybe they are ref'd (edit: if I .unref() it, this no longer accurs).

[cjihrig]

Maybe they are ref'd (edit: if I .unref() it, this no longer accurs).

What exactly is ref'ed that you unref'ed? Unfortunately, --test-force-exit really only exists for people that refuse to clean up after their code.

A few things to look into:

• We might not need to do anything for destinations that are stdout/stderr.
• For file destinations, the test runner owns those. I don't think we actually close() those streams anywhere though. We probably should for this use case.
• When creating file streams, we should probably use the flush option.

[cjihrig]

@redyetidev are you still working on this?

I noticed you have 36(!) open PRs on the project. I think it would be in everyone's best interest to either close some, or focus on getting them merged. Having so many open PRs is a lot of cognitive overhead and prevents other people from fixing the issues.

[avivkeller]

No, someone else can take on the work. I'm aware I have many many open PRs, and I'm working on getting them closed in a way that benefits the entire community.

😅

I'm sorry for all my open PRs, I got a bit carried away, and your right that it takes a lot of effort to keep open. I've closed some, and I'll refrain from opening any more until I close/land some more. Thanks for your help :-)