actix-test: allow the configuration of the TestServer address

[mpalmer]

PR Type

Feature

PR Checklist

• Tests for the changes have been added / updated.
• Documentation comments have been added / updated.
• A changelog entry has been made for the appropriate packages.
• Format code with the latest stable rustfmt.
• (Team) Label with affected crates and semver status.

Overview

This is useful if you're running (say) Selenium tests against a running TestServer, and the Selenium workers are Docker containers elsewhere in the network. Not a particularly common use case, perhaps, but one that I can attest happens every now and then.

[mpalmer]

CI failures appear to be the TlsConnectorService issues that are being addressed in #3350.

[robjtede]

thanks for the addition 👍🏻

looks like it needs a quick look at the new test failure

[mpalmer]

Those test failures are weird -- there's no immediately-obvious relationship to my changes. However, since I can't see any recent changes on master that could be to blame, I'll roll up my sleeves and dig in, I guess.

[mpalmer]

After some rummaging around, I'm fairly confident my changes are at fault, if only because the three failing tests are the ones that are calling actix_test::start, while the other tests are spawning a testing server using other mechanisms. Complicating the problem is that I can't repro the test failures locally, but I'll figure that out one way or another.

[mpalmer]

Test suite passes now. The problem appears to be that, for some mysterious reason, an IPv6 address in the actix-test-constructed URL is being passed in the resolver library, which proceeds to lose its mind. But only in CI -- it works fine for me locally, even when I hard-code an IPv6 listen address. 🤷

[robjtede]

The CI runners, on their VMs, probably have some crazy IPv6 things going on.

I think the quick fix is to use 127.0.0.1 as the default instead of localhost.

[robjtede]

Ahh I see that's actually what you did. Sorry I was on mobile earlier and didn't check the newest commit.