test: fix flaky test-http-set-timeout-server

[santigimeno]

Checklist

• tests and code linting passes
• the commit message follows commit guidelines

Affected core subsystem(s)

test

Description of change

Use different port numbers for every connection to avoid rare
EADDRINUSE errors on SmartOS.

Fixes: #6197

[santigimeno]

Ci: https://ci.nodejs.org/job/node-test-pull-request/2289/

[mscdex]

On a semi-related note, is there a real reason why we don't just allow the OS to assign a free port number for tests instead of using common.PORT? I tried looking back through the archive and the commit that added common.PORT was just trying to unify the various ports used in tests.

Would it be better to modify tests to use a random free port to avoid potential issues like this?

/cc @nodejs/testing

[santigimeno]

Would it be better to modify tests to use a random free port to avoid potential issues like this?

@mscdex I think it is the correct solution. I'll update this test accordingly. The rest of the test suite will take more time. Thanks for the suggestion!

[santigimeno]

Added a new commit so random free ports are used. @mscdex is this what you had in mind? Thanks

[mscdex]

@santigimeno Yes.

[santigimeno]

Rebased into a single commit.

[santigimeno]

New CI run: https://ci.nodejs.org/job/node-test-pull-request/2290/

[jbergstroem]

Perhaps even expand how common.PORT works, making it more resilient against similar things? If we run parallel tests we for instance increment it with +1, but testing another, random, port would be even better (before exiting), no?

[Trott]

LGTM

[jasnell]

I think refactoring or improving common.PORT to handle this better would be a great idea, but in a separate PR. This LGTM

[santigimeno]

Rebased. New CI: https://ci.nodejs.org/job/node-test-pull-request/2312/

[santigimeno]

All is green except for one of the recurrent test failures in the pi2

[santigimeno]

Thanks for the reviews

[santigimeno]

Landed in 57f0595