https://bugs.winehq.org/show_bug.cgi?id=55952
--- Comment #14 from Zeb Figura z.figura12@gmail.com --- (In reply to Esme Povirk from comment #13)
I don't think the tests using SocketResponder care about what data the SocketResponder gets, except when requestHandler consumes the data, but it does seem like the expectation is that data sent by SocketResponder can be consumed on the other end, even though it does a Shutdown and Close without verifying that the data has been read.
Given that I haven't seen this specific failure mode before, I think it's unlikely there's a race condition on Windows, but maybe that could be tested by inserting a sleep() in the right place in one of the tests.
The problem is the other way around—the *client* (i.e. the server) is sending data that the *responder* never reads. I don't think a race would change anything—IIRC you get TCP reset regardless of whether the OS has received and buffered the data or not.