The "EventEmitter memory leak" issue with Node Test Runner

[sernaferna]

This might be more prevalent on Windows than other operating systems?

I get that watch mode is experimental so I can't expect it to work perfectly but I'm wondering if anyone has tips or tricks for getting it to work more consistently? 🙂

I find myself regularly kicking off my tests using combinations of --watch and --test-only and immediately having to Ctrl+c and restart until it starts a "clean" execution. Sometimes it works well right off the bat; sometimes I multitask and work on something else, constantly Alt+Tabbing back to VS Code to Ctrl+C and restart the tests over and over again until it works.

I'll note that I'm still on Node 20 because I want to be on the LTS version.

[avivkeller]

@nodejs/test_runner

[sernaferna]

I'm sensing this means I've posted to the wrong discussion group, but I can't find the test_runner one you're referring to; do you have a link?

[avivkeller]

No, it doesn’t mean that. I’ve pinged the team of people who have a better understanding of the issue at hand.

[sernaferna]

Ah, even better!

[juliangruber]

I'll note that I'm still on Node 20 because I want to be on the LTS version.

Have you tried reproducing this issue on Node 22?

[sernaferna]

Quick note to say that I haven't found the time to try Node 22 yet, however, I have found something that helps (that maybe was obvious but I missed): setting NODE_ENV to test makes things much more stable. (In my case I have a function I use to execute my tests programmatically, so I set it right at the beginning of that before calling run() from node:test.)

This seems to work much more reliably for most of the applications in my monorepo. (I have a React app for which I still get a ton of MaxListenersExceededWarning / EventEmitter memory leak detected issues, and I'm not sure why I'm having more issues with it than with the others, but for now I'm just avoiding watch mode when testing that app unless I have to...)

[sernaferna]

Apologies for the spam, but correction: Despite an afternoon in which everything seemed to be very stable, I haven't subsequently seen any change to stability, even with NODE_ENV set to test. So I'll have to set aside an afternoon someday when I have time to change everything to Node 22 and see what happens (before changing everything back).

[benjamingr]

I use the test runner and didn't run into this issue. If you make a self contained repro (a gh repo with minimal code to reproduce the issue) I am happy to take a look and try and fix whatever is causing this.

Bug reports are valuable :]