Closed alexcrichton closed 1 year ago
Yup. Discovered the same thing yesterday. I was already working on an updated design. If it is causing issues for the CI, maybe we should temporarily comment the test out
Nah no worries about CI, this ended up uncovering a separate bug (linked above) which needs to be orthogonally fixed. Wanted to file this to keep it on record though!
I'm debugging a flaky test right now in Wasmtime's CI, and one interesting behavior I'm seeing is that there's a
blocking_receive
helper for UDP sockets which tries to wait in a loop, and I was going to add an upper limit to the loop to help time it out. This loop, however, never blocks. The socket is always ready for I/O, despite thereceive
call not actually returning anything. Still trying to figure out the latter, but for the former I believe one issue is that thesubscribe
for UDP sockets is either readable or writable readiness right now so if a socket is acting as a server then there's no way to block it for read readiness since it's always writable if you're not writing.