NoMethod error in fork_supervisor upon terminating processes

[Darhazer]

Since we upgraded to 0.6.0 our error reporting system started producing:

Undefined method `>' for nil (NoMethodError)

          if (terminated_fork = forks.delete(pid)) && !status.exited? || status.exitstatus > 0

The code was introduced in #277

I suppose the code is meant to be

if (terminated_fork = forks.delete(pid)) && (!status.exited? || status.exitstatus > 0)

[rosa]

Oh, yes, indeed! Do you want to open a PR with that change?

We haven't run into this error, and I didn't notice it in different tests... I think it's because in all my tests, a terminated_fork is found, and either status.exited? is false, so status.exitstatus is not evaluated, or status.exited? is true and then status.exitstatus is present.

I imagine you're running into this error because in your case forks.delete(pid) returns nil. But I'm not sure how that could happen 😕 How can you have a child process that was forked without having the pid tracked? 😕 Do you have more details on where you're getting this error?

[Darhazer]

I don't know how and why this is happening, and since I don't know how to reproduce it and how to write a test for it, I didn't submit a PR. So if I figure out how to reproduce it at least in a test, I'm happy to open PR. Any tips?

All I can say is we are using Heroku and we started getting the above mentioned error in Sentry

[rosa]

No worries, I'll take care of it. I'm not sure how to write a test for this either right now as I don't know how it can happen 😅

[rosa]

This one should be fixed now as part of #308, and shipped in version 0.7.0.

[Darhazer]

Thank you, @rosa