Jeff Trawick wrote:
the deal was that some pipe handles were supposed to be set non-blocking, but they weren't
Maybe I'm confused, I understood this patch to set a timeout, they were non-blocking with an infinite timeout. No difference you might say ;-) But on win32 a blocking pipe once open can't be set to a timeout/nonblock. Ancient issue which causes still today 2 faults in testpipe.c, our actual apr_file_pipe_open needs such a flag, too. Bill
