Fixes the call to setsockopt(SO_UPDATE_ACCEPT_CONTEXT) on Windows (#1598) #1604
Conversation
0e0d33f
to
e588f9d
Compare
src/platform/windows/win_tcpconn.c
Outdated
| @@ -260,7 +260,13 @@ tcp_close(void *arg) | |||
| if (!c->closed) { | |||
| c->closed = true; | |||
| if (!nni_list_empty(&c->recv_aios)) { | |||
| CancelIoEx((HANDLE) c->s, &c->recv_io.olpd); | |||
There was a problem hiding this comment.
So the problem is that closing the socket didn't up the waiting receivers? Is that the situation? (A similar problem exists with UNIX and there some unfortunate workarounds for it in the POSIX side of the code.)
There was a problem hiding this comment.
FINALLY!
After your comment about closing the socket, I have checked the result of shutdown() and it also was raising an error (WSAENOTCONN).
After searching this error, it came from an issue on the call to setsockopt(SOL_SOCKET, SO_UPDATE_ACCEPT_CONTEXT) after AcceptEx.
From the docs, this function requires as a parameter the listener socket, not a boolean value.
Now it closes the connections correctly.
I have rebased this PR with these new changes, so it should be ready for review.
There was a problem hiding this comment.
Excellent. Strange behavior in Windows around this, but clearly what I had was a bug.
From the Windows docs, this option must receive as parameters the listener socket.
e588f9d
to
2adb27e
Compare
From the docs, this function requires as a parameter the listener socket, not a boolean value.
With this change,
shutdownnow can start the socket closing process correctly.fixes #1598