I am able to reproduce this with https://github.com/jaydenseric/graphql-upload/commit/3b7f034e1c2d084f314734377c4351c171b87a42 on macOS. It does not reproduce 100% of the time (possibly a race condition?) so I used a small bash script: #51314
Labels
invalid
Issues and PRs that are invalid.
I believe this is only related to
NODE_V8_COVERAGE
and not the test runner. Also note that the bash script does not use--test
. I've also heard from other folks that this can be reproduced without the test runner (ie, using node-tap on a different codebase).When
NODE_V8_COVERAGE
is present, the inspector is enabled at startup. When the process is getting ready to exit, we callEndStartedProfilers()
.I'm not yet sure if we are missing some cleanup step, but when the process hangs, we enter an infinite loop of flushing and adding new tasks:
At least some of the new tasks are coming from
v8::internal::Heap::PostFinalizationRegistryCleanupTaskIfNeeded()
(note that I added aprintf()
if you're curious about the top couple frames):@nodejs/v8 does anything immediately stand out to you?
Originally posted by @cjihrig in #49344 (comment)
The text was updated successfully, but these errors were encountered: