Unhandled error [AssertionError] from child_process.js

[srlowe]

Just got an unhandled error from child_process.js (Node v4.2.0)

AssertionError: UNHANDLED! false == true
  File "internal/child_process.js", line 440, in Pipe.channel.onread
  File "internal/child_process.js", line 686, in handleMessage
  File "events.js", line 172, in ChildProcess.emit
  File "events.js", line 92, in emitTwo
  File "internal/child_process.js", line 464, in ChildProcess.<anonymous>

Perhaps related to this issue? : #3341

[bnoordhuis]

Is that with a patched node version? I'm sure we don't throw that particular AssertionError anywhere.

[srlowe]

This node version was installed via apt-get, using the Nodesource binary distributions. We haven't modified anything on top of that.

[bnoordhuis]

Can you report it to Nodesource in that case?

[srlowe]

I was under the impression that these were simply builds of the unmodified node source. Are you suggesting that they are different?

[bnoordhuis]

They must be. The canonical source doesn't have asserts in that location and the stack trace is formatted differently as well.

[srlowe]

It's not this line?

464:     assert(Array.isArray(target._handleQueue));

(I'll check with nodesource about modifications - that would be concerning if true)

[bnoordhuis]

I guess the stack trace threw me off. Normal stack traces show the throw site first, here it's last.

Core doesn't throw that "UNHANDLED" error message though. Are you using modules that monkey-patch node core? If the answer is no, it must be the binary itself.

[srlowe]

Ah sorry, it's our logging system adding that UNHANDLED, and some changes recently (some 3rd party modules) seem to have changed the stacktrace order. I'll make sure to post standard stacktraces next time.

[evanlucas]

Is this issue still occurring?

[srlowe]

@evanlucas According to our logs this last happened on 29th Oct.

@bnoordhuis We're not using any modules that monkey patch node core. (Sorry didn't notice that part of your post before)

[cjihrig]

@bgSosh some code to reproduce would be best. It looks like you haven't bumped into this in a while. If you run into it again and can reproduce it, I'll open this back up.

[srlowe]

@cjihrig I'm not able to reproduce it in code unfortunately. I'll report back here if it happens again.

[srlowe]

@cjihrig I've had more occurrences of this this week, so I'd appreciate if you could re-open this.

[r-52]

@bgSosh I've reopened this issue - have you found some code that could help to reproduce this error?

[srlowe]

@romankl - I've no way of reproducing this reliably, but I believe this is connected with #3072 , which is easy to reproduce:

Just start two instances of a node app on the same port (an app using cluster). Along with the expected EADDRINUSE error, this error appears.

[jasnell]

Is there reason to keep this open?

[srlowe]

@jasnell - it's still happening almost daily.

[cjihrig]

I'm going to close this again, as there is still no way to reproduce it, and may have even been fixed by now. If you come up with a reproduction, by all means, post it here. If this is tied in with #3072, then we can track it there.