On Thu Apr 17 19:59:06 2025 +0000, Dmitry Timoshkov wrote:
The patch doesn't really break the todos, although it may look like it. Rather the affected tests are broken because it seems that they will fail even when called with a valid socket handle, what will break them is the next condition if (!overlapped && !ret_size) { SetLastError( WSAEFAULT ); return -1; } So, it seems that it's better to set wrong last error value than completely fail send().
Poor phrasing on my part, sorry. My point is that better yet would be to match native behaviour, and not add any new todos.