Somehow the Windows tests are always failing lately on the same socket tests. If someone has the time and energy to look into this, please do!
FYI, can confirm this doesn't happen in my local win env - I'm going to set up an appveyor build on my acct so I can test reproducing and hopefully find a cause
🙏
this sometimes happens in circleci too, but lately it seems to happen consistently in appveyor. we could just mark this test as ^:flaky as a last resort though
the sort of funny part is, it is marked flaky, but since it's behind the env var, it doesn't execute as part of the flaky tests, only as part of the BABASHKA_SOCKET_REPL_TEST tests
aha!
I think the flakiness could be a timing thing... I put some longer notes at <https://gist.github.com/bobisageek/709d4a8b119d5947e74d559a7f4a8354>, but the tl;dr is I think the port re-use causes some undesired overlap, and I'd like to use different ports for each socket repl test; best case the flakiness goes away, worst case I'm definitely wrong about the cause 🙂
Good idea