My understanding is that the TTY code should be doing the same checking of dispatch_event_queue and quit_check_signal_tick_count as the event-Xt code, since emacs_tty_next_event and the event-stream code pays attention to the former, and the signal handler to the latter. The below makes my TTY-only build react more crisply (but doesn’t make VM startup on decent-size mailboxes as fast as on a build with the XT event loop, even on a TTY--I need to look further into why that is.) http://mid.gmane.org/17597.11630.343389.787670＠parhasard.net is a similar problem, but more clear-cut. I suspect the two stem from an oversight at the same point in time. I’d appreciate a look from someone who knows their way around the event-handling code. src/ChangeLog addition: 2006-12-08 Aidan Kehoe <kehoea(a)parhasard.net&gt; * event-tty.c: * event-tty.c (emacs_tty_event_pending_p): * event-tty.c (reinit_vars_of_event_tty): Pay attention to the dispatch event queue, and input pending signals in emacs_tty_event_pending_p. Makes pure TTY builds more responsive. XEmacs Trunk source patch: Diff command: cvs -q diff -u Files affected: src/event-tty.c =================================================================== RCS Index: src/event-tty.c =================================================================== RCS file: /pack/xemacscvs/XEmacs/xemacs/src/event-tty.c,v retrieving revision 1.17 diff -u -r1.17 event-tty.c --- src/event-tty.c 2006/08/04 20:55:04 1.17 +++ src/event-tty.c 2006/12/08 13:10:02 ＠＠ -45,6 +45,8 ＠＠ extern int mswindows_is_blocking; #endif +static int last_quit_check_signal_tick_count; +  /************************************************************************/ /* timeout events */ ＠＠ -84,22 +86,76 ＠＠ static int emacs_tty_event_pending_p (int how_many) { + Lisp_Object event; + int tick_count_val; + + /* Cf. the comments on emacs_Xt_event_pending_p in event-xlike-inc.c . */ + if (!how_many) { EMACS_TIME sometime; - /* see if there's a pending timeout. */ + + /* (1) Any pending events in the dispatch queue? */ + if (!NILP(dispatch_event_queue)) + { + return 1; + } + + /* (2) Any TTY or process input available? */ + if (poll_fds_for_input (non_fake_input_wait_mask)) + return 1; + + /* (3) Any timeout input available? */ EMACS_GET_TIME (sometime); if (tty_timer_queue && EMACS_TIME_EQUAL_OR_GREATER (sometime, tty_timer_queue->time)) return 1; + } + else + { + /* HOW_MANY > 0 */ + EVENT_CHAIN_LOOP (event, dispatch_event_queue) + { + if (command_event_p (event)) + { + how_many--; + if (how_many <= 0) + return 1; + } + } - return poll_fds_for_input (non_fake_input_wait_mask); } + + tick_count_val = quit_check_signal_tick_count; + + /* Checking in_modal_loop here is a bit cargo-cultish, since its use is + specific to builds with a window system. */ + if (!in_modal_loop && + (last_quit_check_signal_tick_count != tick_count_val)) + { + last_quit_check_signal_tick_count = tick_count_val; - /* #### Not right! We need to *count* the number of pending events, which - means we need to have a dispatch queue and drain the pending events, - using drain_tty_devices(). */ - return poll_fds_for_input (tty_only_mask); + /* We need to drain the entire queue now -- if we only drain part of + it, we may later on end up with events actually pending but + detect_input_pending() returning false because there wasn't + another SIGIO. */ + event_stream_drain_queue (); + + if (!how_many) + return !NILP (dispatch_event_queue); + + EVENT_CHAIN_LOOP (event, dispatch_event_queue) + { + if (command_event_p (event)) + { + how_many--; + if (how_many <= 0) + return 1; + } + } + } + + return 0; } static void ＠＠ -304,6 +360,8 ＠＠ tty_event_stream->drain_queue_cb = emacs_tty_drain_queue; tty_event_stream->create_io_streams_cb = emacs_tty_create_io_streams; tty_event_stream->delete_io_streams_cb = emacs_tty_delete_io_streams; + + last_quit_check_signal_tick_count = 0; } void -- Santa Maradona, priez pour moi! _______________________________________________ XEmacs-Patches mailing list XEmacs-Patches(a)xemacs.org http://calypso.tux.org/cgi-bin/mailman/listinfo/xemacs-patches