test: implement common.skipIfNoIpv6Localhost

[joyeecheung]

Refs: #15936 (comment)

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• commit message follows commit guidelines

Affected core subsystem(s)

test

[joyeecheung]

CI: https://ci.nodejs.org/job/node-test-pull-request/10770/

[joyeecheung]

Fixed lint errors. New CI: https://ci.nodejs.org/job/node-test-pull-request/10779/

cc @nodejs/testing

[mhdawson]

LGTM

[maclover7]

Nit: should this change to be a skipIf function like skipIf32Bits and skipIfInspectorDisabled? Mostly just to fit to the same style.

[joyeecheung]

@maclover7 Yeah that's what I originally named this..I think at that point I thought the semantics of this did not align with other skipIf* functions but now looking at it again I don't find any problems? Anyway I'll rebase and rename it.

[joyeecheung]

@mhdawson @maclover7 Rebased and renamed , PTAL.

CI: https://ci.nodejs.org/job/node-test-pull-request/11211/

[maclover7]

LGTM

[joyeecheung]

This is failing on raspberry pis since they don't have ipv6 support and the tests mustCall the callbacks. I'll fix it.

[joyeecheung]

Removed the common.mustCall since uh...the point of this PR is to skip the callbacks when ipv6 localhost is not available...¯\(ツ)/¯

CI: https://ci.nodejs.org/job/node-test-pull-request/11354/

[joyeecheung]

Rebased. New CI: https://ci.nodejs.org/job/node-test-pull-request/11372/

[refack]

given that // This test that the family option of https.get is honored. would it be correct to replace this with '::1' since this what the server bind to?

[joyeecheung]

@refack I think replacing it with a ip would skip the family option handling altogether. The family option is used when performing a lookup, so if the connection doesn't even need to lookup then it won't be tested. In net.js:

  // If host is an IP, skip performing a lookup
  var addressType = cares.isIP(host);
  if (addressType) {
    nextTick(self[async_id_symbol], function() {
      if (self.connecting)
        internalConnect(self, host, port, addressType, localAddress, localPort);
    });
    return;
  }

....

  var dnsopts = {
    family: options.family,
    hints: options.hints || 0
  };

[refack]

Let's think is this is not overkill?

[refack]

Let's think is this is not overkill?

AFAICT those two test can skip the DNS calls, so if they work with '::1' I's rather have that.
(the working assumption is that the test harness should be K.I.S.S. as much as possible.)

[joyeecheung]

@refack

AFAICT those two test can skip the DNS calls, so if they work with '::1' I's rather have that.
(the working assumption is that the test harness should be K.I.S.S. as much as possible.)

Uh..those tests are testing DNS calls, family is passed to the lookup function.

[joyeecheung]

Ah, forgot to run the linter...also the arm compilation seems to be having a outage?

[refack]

Uh..those tests are testing DNS calls, family is passed to the lookup function.

I want to dig into that. let's see if we can refactor something around this.

[joyeecheung]

Closing in favor of #16976