This branch attempts to resolve https://github.com/rustls/rustls-ffi/issues/336. I say "attempts" because I've had little luck trying to force the flake to happen in CI (even when hacking up the build tasks to just run multiple instances of the Windows/MacOS tests that seem to catch the flake most often).
I think this is the most obvious reason why we'd see the flakes we see, but it's hard to say conclusively :-)
tests: fix wait_tcp_port format string.
The success-case print for the wait_tcp_port helper wasn't formatting its message as an f-string, instead printing the literal {host} and {port} placeholders.
tests: wait for tcp port for all test types.
The run_mtls_client_tests and run_mtls_client_crl_tests were being run without calling wait_tcp_port first. I suspect this is the reason we sometimes see the connect tests fail in CI on some platforms with output about the connection being refused.
Description
This branch attempts to resolve https://github.com/rustls/rustls-ffi/issues/336. I say "attempts" because I've had little luck trying to force the flake to happen in CI (even when hacking up the build tasks to just run multiple instances of the Windows/MacOS tests that seem to catch the flake most often).
I think this is the most obvious reason why we'd see the flakes we see, but it's hard to say conclusively :-)
tests: fix wait_tcp_port format string.
The success-case
printfor thewait_tcp_porthelper wasn't formatting its message as an f-string, instead printing the literal{host}and{port}placeholders.tests: wait for tcp port for all test types.
The
run_mtls_client_testsandrun_mtls_client_crl_testswere being run without callingwait_tcp_portfirst. I suspect this is the reason we sometimes see the connect tests fail in CI on some platforms with output about the connection being refused.