test: parallel/test-net-socket-local-address frequently fails on freebsd

[bnoordhuis]

It seems to time out, e.g. https://jenkins-iojs.nodesource.com/job/node-test-commit-other/198/nodes=freebsd101-64/console - grep for 'not ok'.

not ok 522 - test-net-socket-local-address.js
  ---
  duration_ms: 60.74
  ...

Refs #2095, /cc @rmg

[rmg]

The test looks really straight forward to me, but I was the one that wrote it.. @bnoordhuis do you have any ideas on this? I'm not familiar enough with FreeBSD to know what kinds of things to even look for :-(

[thefourtheye]

If we have access to that box, we can debug it, right?

[bnoordhuis]

/cc @nodejs/platform-freebsd @nodejs/build - are the tests running inside a jail?

[jbergstroem]

Tests are running inside a jail, yes.

Edit: I'll have a look.

[jbergstroem]

Just a quick update: I haven't been able to reproduce this locally on the machine. I've run the parallel suite 5 times and the specific test 100 times; not really the frequency we're seeing here: https://jenkins-iojs.nodesource.com/computer/iojs-voxer-freebsd101-64-1/builds

[jbergstroem]

This will probably "fix" our issues: nodejs/build#228

[brendanashworth]

I believe this was fixed in b061e3a, so closing.

[jbergstroem]

Additionally, as of yesterday we don't run tests in jails any longer.

[Trott]

Just to make sure the problem really was the jail and that the issue is resolved, I've been trying to run a stress test for it on FreeBSD to confirm, but the build doesn't work so the tests don't run. See, for example: https://ci.nodejs.org/job/node-stress-single-test/34/nodes=freebsd102-64/console Not sure if that's a problem with the current FreeBSD build (which I believe has undergone some changes today and may still have a few checkboxes to check off before it's really done) or if the problem is in the stress test job itself (in which case: /cc @joaocgreis ).

That said, I do believe the problem is resolved. (And if it fails again, it will be easy enough to mark it as flaky and go back to square one.)