Thanks for your commit [subject above]. On 1/23/2012 5:59 AM, Jim Jagielski wrote: > How goes the Windows fixes? Any work at all done on them?
Yup. As Joe and co indicate on the list, several people have looked into the core network code, only to discover "here there be dragons". Outch! This is not windows specific. The only likely windows defect is misapplication of blocking/nonblock semantics by apr, which may be the result of apr_socket_make(), and this is being discussed on its bug ticket. Joe, Stephan and I are determining the best API solution to the specific problem set of injecting one initial data bucket from accept-filter like providers who have the first data block in hand at accept() time. Today that is only implemented on win32, AIUI, but is easily extended to the unix kernel/socket provider space. Again, this is being discussed on its bug ticket. I'm only aware of those who are working on this from that ticket, so if others have observations, please add them! https://issues.apache.org/bugzilla/show_bug.cgi?id=52476
