Hi,
I'm running into a strange issue with eof() tests (if it matters, on a GnuTLS session.read object). In 0.3.5, the eof(session) call returns fairly instantly for a session on both true and false. In 0.4, there's a ~15 second delay when eof(session.read) is true. The docs say If the stream is not yet exhausted, this function will block to wait for > more data if necessary, and then return "false”. But it appears that something is different here between 0.3.5 and 0.4. This has an impact for GnuTLS.jl and everything that depends on it (such as Requests), and it causes very long delays at the end of a GET request that uses TLS. Any pointers? I'm sorry there's no code, but it's a very complicated test case.
