https://issues.apache.org/bugzilla/show_bug.cgi?id=52476
--- Comment #18 from William A. Rowe Jr. <[email protected]> 2012-01-23 04:29:19 UTC --- So, there is no special-case filter... I had modified the logic that was unusable within the 'public', simple-yet-insufficient ap_process_connection, which would otherwise do; ap_update_vhost_given_ip(c); rc = ap_run_pre_connection(c, csd); if (rc != OK && rc != DONE) { c->aborted = 1; } if (!c->aborted) { ap_run_process_connection(c); } Note that the event, simple and winnt mpm's all have custom implementations of that not-altogether useful sequence of instructions. So mpm/winnt/child.c worker_main() reimplements this logic for those cases where we must inject an initial bucket of data, pulled from the accepted socket by AcceptEx. Really there is little special about Acceptfilter none; it bypasses that magic, the initial brigade is empty and obtains one socket bucket. Now the only idea I have left is that somewhere in the code path else /* (accf = 0) e.g. 'none' */ where we handle accept(), play some setsockopt games, and call apr_os_sock_make apr and mpm_winnt disagree on the default state of that new apr_socket_t. That is about the last explanation which would address these interesting timing observations, that apr is not blocking, but httpd assumes that it is. -- Configure bugmail: https://issues.apache.org/bugzilla/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
