Codecov Report

Attention: Patch coverage is 88.40580% with 8 lines in your changes missing coverage. Please review.

Project coverage is 94.02%. Comparing base (1ad365a) to head (49e6118).

@@            Coverage Diff             @@
##             main     #104      +/-   ##
==========================================
- Coverage   94.62%   94.02%   -0.61%     
==========================================
  Files          11       12       +1     
  Lines         725      786      +61     
==========================================
+ Hits          686      739      +53     
- Misses         35       42       +7     
- Partials        4        5       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Hey @korotin!

This is a great find, thanks a lot!
Indeed, the stop logic is not synchronized with task submission so it could happen that a new worker is launched (and thus workersWaitGroup.Add(1) called) while waiting for the wait group.

I think we can fix this by reusing the existing mutex on the pool struct rather than using a different one, which could introduce some additional latencies (in my experience, calls to lock a mutex are pretty expensive so it's best to avoid it as much as possible).
I started to explore a solution in this PR #105 but i'll need to work on it for some time before merging it.

My first attempt was to reuse pool's mutex but that didn't work well. We cannot guard waitgroup's calls with mutex because it may lead to deadlocks.

Anyway, I think it's best to close this one and continue discussion in your MR :)