[#6712] Use HostnameEndpoint with Agent for HTTPS

[glyph]

https://twistedmatrix.com/trac/ticket/6712

[glyph]

This is probably going to fail the nodeps builders, since wrapClientTLS depends pretty intimately on SSL actually existing for SSL tests. I'll fix the computation of the skips soon, but it's otherwise in good shape for a review.

[codecov-io]

Current coverage is 91.04% (diff: 100%)

Merging #624 into trunk will decrease coverage by 0.06%

@@              trunk       #624   diff @@
==========================================
  Files           837        837          
  Lines        146006     146069    +63   
  Methods           0          0          
  Messages          0          0          
  Branches      12910      12912     +2   
==========================================
- Hits         133033     132990    -43   
- Misses        10729      10815    +86   
- Partials       2244       2264    +20   

Powered by Codecov. Last update cfa0537...4b3a5ca

[itamarst]

In general looks good (I skimmed the tests a bit more than earlier code). However, since the goal was to make Agent work with IPv6, perhaps a test that actually does that should be added?

[glyph]

Thanks @itamarst !

Since you didn't specify a "next action" I'm going to interpret this as "add a test that verifies IPv6 connectivity for Agent, then land".

The reason I didn't begin with such a test is that, as it turns out, there's no new code here specifically for IPv6, so there was no opportunity to write a test that failed. Previously there was no way to mock out name resolution, and once name resolution is mocked out, the handling of an IPv6 string vs. an IPv4 string is entirely internal to HostnameEndpoint.

That said, those are just reasons that a unit test was not required in TDD discipline; a little integration test to make sure nothing blows up once an IPv6Address gets in the mix would not be amiss here. I'll add and merge.

[itamarst]

(It did seem like TDD would be tricky, yeah, since code changed so much.)