It's definately wrong to start ap_bprintf() on a connection that's got an

error, or one which is not buffered.

It's wrong for ap_bflush() to return 0 when B_EOUT is set -- B_EOUT can
be set by timeout(), and it seems pointless for code everywhere to have
to test for B_EOUT when they can simply test for a ap_bflush() failure.


git-svn-id: https://svn.apache.org/repos/asf/httpd/httpd/branches/1.3.x@81638 13f79535-47bb-0310-9956-ffa450edef68