else
{
- if (smtp_out != NULL && !f.disable_delay_flush)
+ if (smtp_out && !f.disable_delay_flush)
mac_smtp_fflush();
#if !defined(NO_POLL_H) && defined (POLLRDHUP)
HDEBUG(D_acl) debug_printf_indent("delay cancelled by peer close\n");
}
#else
- /* It appears to be impossible to detect that a TCP/IP connection has
- gone away without reading from it. This means that we cannot shorten
- the delay below if the client goes away, because we cannot discover
- that the client has closed its end of the connection. (The connection
- is actually in a half-closed state, waiting for the server to close its
- end.) It would be nice to be able to detect this state, so that the
- Exim process is not held up unnecessarily. However, it seems that we
- can't. The poll() function does not do the right thing, and in any case
- it is not always available.
- */
+ /* Lacking POLLRDHUP it appears to be impossible to detect that a
+ TCP/IP connection has gone away without reading from it. This means
+ that we cannot shorten the delay below if the client goes away,
+ because we cannot discover that the client has closed its end of the
+ connection. (The connection is actually in a half-closed state,
+ waiting for the server to close its end.) It would be nice to be able
+ to detect this state, so that the Exim process is not held up
+ unnecessarily. However, it seems that we can't. The poll() function
+ does not do the right thing, and in any case it is not always
+ available. */
while (delay > 0) delay = sleep(delay);
#endif