1 -*- coding: utf-8; mode: text; -*-
3 Copyright (C) 2007-2024 Free Software Foundation, Inc.
4 See the end of the file for license conditions.
6 From README.multi-tty in the multi-tty branch.
7 Some of this information may be out of date.
13 The following is a (sadly incomplete) list of people who have
14 contributed to the project by testing, submitting patches, bug
15 reports, and suggestions. Thanks!
17 Bernard Adrian <bernadrian@free.fr>
18 ARISAWA Akihiro <ari@mbf.ocn.ne.jp>
19 Vincent Bernat <bernat@luffy.cx>
20 Han Boetes <han@mijncomputer.nl>
21 Francisco Borges <borges@let.rug.nl>
22 Damien Cassou <damien.cassou@laposte.net>
23 Robert J. Chassell <bob@rattlesnake.com>
24 Romain Francoise <romain@orebokech.com>
25 Ami Fischman <ami@fischman.org>
26 Noah Friedman <friedman@splode.com>
27 Friedrich Delgado Friedrichs <friedel@nomaden.org>
28 Samium Gromoff <_deepfire@mail.ru>
29 Mikhail Gusarov <dottedmag@dottedmag.net>
30 Eric Hanchrow <offby1@blarg.net>
31 IRIE Tetsuya <irie@t.email.ne.jp>
32 Yoshiaki Kasahara <kasahara@nc.kyushu-u.ac.jp>
33 Bas Kok <nekkobassu@yahoo.com>
34 Jurej Kubelka <Juraj.Kubelka@email.cz>
35 David Lichteblau <david@lichteblau.com>
36 Richard Lewis <rtf@jabble.com>
37 mace <mace@kirjakaapeli.lib.hel.fi>
38 Suresh Madhu <madhu@cs.unm.edu>
39 Xavier Mallard <zedek@gnu-rox.org>
40 Istvan Marko <mi-mtty@kismala.com>
41 Ted Morse <morse@ciholas.com>
42 Gergely Nagy <algernon@debian.org>
43 Dan Nicolaescu <dann@ics.uci.edu>
44 Kalle Olavi Niemitalo <kon@iki.fi>
45 Mark Plaksin <happy@mcplaksin.org>
46 Frank Ruell <stoerte@dreamwarrior.net>
47 Tom Schutzer-Weissmann <trmsw@yahoo.co.uk>
48 Joakim Verona <joakim@verona.se>
49 Dan Waber <dwaber@logolalia.com>
52 Richard Stallman was kind enough to review an earlier version of my
59 It still needs to be ported to Windows/Mac/DOS. Both multiple
60 tty device support and simultaneous X and tty frame support works
61 fine. Emacsclient has been extended to support opening new tty and X
62 frames. It has been changed to open new Emacs frames by default.
64 Tested on GNU/Linux, Solaris 8, FreeBSD and OpenBSD.
68 * GTK support. If you compile your Emacs with the GTK
69 toolkit, some functionality of multi-tty may be lost. In
70 particular, you may get crashes while working on multiple X
71 displays at once. Previous releases of GTK had limitations
72 and bugs that prevented full-blown multi-display support in
73 Emacs. (GTK crashed when Emacs tries to disconnect from an
74 X server.) Things are much improved in the current GTK
75 version, but if you do experience crashes in libgtk, try
76 compiling Emacs with the Lucid toolkit instead.
78 * The single-kboard mode.
80 If your multi-tty Emacs session seems to be frozen, you
81 probably have a recursive editing session or a pending
82 minibuffer prompt (which is a kind of recursive editing) on
83 another display. To unfreeze your session, switch to that
84 display and complete the recursive edit, for example by
85 pressing C-] ('abort-recursive-edit').
87 I am sorry to say that currently there is no way to break
88 out of this "single-kboard mode" from a frozen display. If
89 you are unable to switch to the display that locks the
90 others (for example because it is on a remote computer),
91 then you can use emacsclient to break out of all recursive
94 emacsclient -e '(top-level)'
96 Note that this (perhaps) unintuitive behavior is by design.
97 Single-kboard mode is required because of an intrinsic Emacs
98 limitation that is very hard to eliminate. (This limitation
99 is related to the single-threaded nature of Emacs.)
101 I plan to implement better user notification and support for
102 breaking out of single-kboard mode from locked displays.
107 To test the multi-tty feature, start up the Emacs server with the
113 and then (from a shell prompt on another terminal) start emacsclient
115 emacsclient -t /optional/file/names... (for a tty frame)
116 emacsclient /optional/file/names... (for an X frame)
118 (Make sure both emacs and emacsclient are multi-tty versions.)
119 You'll hopefully have two fully working, independent frames on
120 separate terminals. The new frame is closed automatically when you
121 finish editing the specified files (C-x #), but delete-frame (C-x 5 0)
122 also works. Of course, you can create frames on more than two tty
125 Creating new frames on the same tty with C-x 5 2 (make-frame-command)
126 works, and behaves the same way as in previous Emacs versions. If you
127 exit emacs, all terminals should be restored to their previous states.
132 I think the best way to use the new Emacs is to have it running inside
133 a disconnected GNU screen session, and always use emacsclient for
134 normal work. One advantage of this is that not a single keystroke of
135 your work will be lost if the display device that you are using
136 crashes, or the network connection times out, or whatever. (I had an
137 extremely unstable X server for some time while I was developing these
138 patches, and running Emacs this way has saved me a number of M-x
139 recover-session invocations.)
141 I use the following two bash scripts to handle my Emacs sessions:
143 -------------------------------------------------------connect-emacs--
145 # Usage: connect-emacs <name> <args>...
147 # Connects to the Emacs instance called NAME. Starts up the instance
148 # if it is not already running. The rest of the arguments are passed
154 if [ -z "$name" ]; then
155 echo "Usage: connect_emacs <name> <args>..." >&2
158 preload-emacs "$name" wait
159 /usr/bin/emacsclient.emacs-multi-tty -s "$name" "$@"
160 ----------------------------------------------------------------------
162 -------------------------------------------------------preload-emacs--
164 # Usage: preload-emacs <name> [<waitp>]
166 # Preloads the Emacs instance called NAME in a detached screen
167 # session. Does nothing if the instance is already running. If WAITP
168 # is non-empty, the function waits until the server starts up and
169 # creates its socket; otherwise it returns immediately.
173 screendir="/var/run/screen/S-$USER"
174 if [ "${XDG_RUNTIME_DIR+set}" ]; then
175 serverdir="$XDG_RUNTIME_DIR/emacs"
177 serverdir="${TMPDIR-/tmp}/emacs$UID"
179 emacs=/usr/bin/emacs-multi-tty # Or wherever you installed your multi-tty Emacs
181 if [ -z "$name" ]; then
182 echo "Usage: preload_emacs <name> [<waitp>]" >&2
186 if [ ! -e "$screendir"/*."$name" ]; then
187 if [ -e "$serverdir/$name" ]; then
188 # Delete leftover socket (for the wait option)
189 rm "$serverdir/$name"
191 screen -dmS "$name" "$emacs" -nw --eval "(setq server-name \"$name\")" -f server-start
193 if [ ! -z "$waitp" ]; then
194 while [ ! -e "$serverdir/$name" ]; do sleep 0.1; done
196 ----------------------------------------------------------------------
198 I have the following in my profile to have two instances automatically
199 preloaded for editing and email:
204 It is useful to set up short aliases for connect-emacs. I use the
207 alias edit="connect-emacs editor"
209 alias et="connect-emacs editor -t"
210 alias gnus="connect-emacs gnus"
216 ** See if 'tty-defined-color-alist' needs to be terminal-local.
217 Update: Dan says it should be, so convert it.
219 ** Mikhail Gusarov suggest to add a hook akin to
220 'after-make-frame-functions' that is called whenever Emacs connects
221 to a new terminal. Good idea!
223 ** emacsclient -t on the console does not work after su. You have to
224 use non-root accounts or start as root to see this.
230 *ERROR*: Could not open file: /dev/tty1
232 The tty can be opened as /dev/tty by emacsclient, but not by Emacs.
233 This seems to be a serious problem. Currently my only idea is to
234 bring back the ugly pty proxy hack from the initial versions of
235 multi-tty. Suggestions would be appreciated.
237 Update: we could change emacsclient to pass its open file
238 descriptor to the Emacs process. Unfortunately, this requires a
239 new Lisp-level Emacs API, and as file descriptors are not otherwise
240 exported to Lisp, this approach seems at least as ugly as the pty
243 ** lisp/vc.el depends on the terminal type during load time.
244 'vc-annotate-color-map' is one example that needs to be fixed.
246 ** Understand how 'quit_throw_to_read_char' works, and fix any bugs
249 ** See if getcjmp can be eliminated somehow. Why does Emacs allow
250 asynchronous input processing while it's reading input anyway?
252 ** 'delete-frame' events are handled by 'special-event-map'
253 immediately when read by 'read_char'. This is fine but it prevents
254 higher-level keymaps from binding that event to get notified of the
257 Sometimes it would be useful for Lisp code to be notified of frame
258 deletions after they have happened, usually because they want to
259 clean up after the deleted frame. Not all frame-local states can
260 be stored as a frame parameter. For example,
261 'display-splash-screen' uses 'recursive-edit' with a special keymap
262 override to create its buffer---and it leads to all kinds of
263 nastiness if Emacs stays in this recursive edit mode after the
264 frame containing the splash screen is deleted. Basically, the
265 splash-screen implementation wants to throw out of the recursive
266 edit when the frame is deleted; however, it is not legal to throw
267 from 'delete-frame-functions' because 'delete-frame' must not fail.
268 (Introducing 'delete-frame-after-functions' would not help either
269 because 'delete-frame' may not fail at that time either.)
271 Currently 'fancy-splash-screens' installs a
272 'delete-frame-functions' hook that sets up a timer to exit the
273 recursive edit. This is an adequate solution, but it would perhaps
274 be better to have something like a 'frame-deleted' event that could
275 be bound in the normal way.
277 ** Trouble: 'setenv' doesn't actually set environment variables in the
278 Emacs process. This defeats the purpose of the elaborate
279 'server-with-environment' magic around the 'tgetent' call in
282 ** (Possibly) create hooks in struct device for creating frames on a
283 specific terminal, and eliminate the hackish terminal-related frame
284 parameters (display, tty, tty-type).
289 ** Decide whether to keep the C implementation of terminal parameters,
290 or revert to the previous, purely Lisp code. It turned out that
291 local environments do not need terminal parameters after all.
293 ** Move Fsend_string_to_terminal to term.c, and declare get_named_tty
294 as static, removing it from dispextern.h.
295 Move fatal to emacs.c and declare it somewhere.
297 ** Search for 'suspend-emacs' references and replace them with
298 'suspend-frame', if necessary. Ditto for 'save-buffers-kill-emacs'
299 vs. 'save-buffers-kill-display'.
301 ** Emacs crashes when a tty frame is resized so that there is no space
302 for all its windows. (Tom Schutzer-Weissmann)
304 ** Report GTK multi-display problems to GTK maintainers. For extra
307 Currently you can not connect to new X displays when you compile
308 Emacs with GTK support. If you want to play around with GTK
309 multidisplay (and don't mind core dumps), you can edit src/config.h
310 and define HAVE_GTK_MULTIDISPLAY there by hand.
312 https://gitlab.gnome.org/GNOME/gtk/issues/221
314 Update: Han reports that GTK+ version 2.8.9 almost gets display
315 disconnects right. GTK will probably be fully fixed by the time
316 multi-tty gets into the trunk.
318 Update: I am still having problems with GTK+ 2.8.10. I have the
319 impression that the various multidisplay fixes will only get
320 released in GTK+ 2.10.
322 ** Audit 'face-valid-attribute-values' usage in customize and
323 elsewhere. Its return value depends on the current window system.
324 Replace static initializers using it with runtime functions. For
325 example, custom's buttons are broken on non-initial device types.
327 ** Possibly turn off the double C-g feature when there is an X frame.
328 C.f. (emacs)Emergency Escape.
330 ** frames-on-display-list should also accept frames.
332 ** Consider the 'tty-type' frame parameter and the 'display-tty-type'
333 function. They serve the exact same purpose. I think it may be
334 a good idea to eliminate one of them, preferably 'tty-type'.
336 ** The handling of lisp/term/*.el, and frame creation in general, is a
337 big, big mess. How come the terminal-specific file is loaded by
338 tty-create-frame-with-faces? I don't think it is necessary to load
339 these files for each frame; once per terminal should be enough.
340 Update: lisp/term/*.el is not loaded repeatedly anymore, but
341 faces.el still needs to be cleaned up.
343 ** Fix frame-set-background-mode in this branch. It was recently
344 changed in CVS, and frame.el in multi-tty has not yet been adapted
345 for the changes. (It needs to look at
346 default-frame-background-mode.) (Update: maybe it is fixed now;
349 ** I think '(set-)terminal-local-value' and the terminal parameter
350 mechanism should be integrated into a single framework.
352 (Update: '(set-)terminal-local-value' is now eliminated, but the
353 terminal-local variables should still be accessible as terminal
354 parameters. This also applies to 'display-name' and similar
357 ** Add the following hooks: after-delete-frame-hook (for server.el,
358 instead of delete-frame-functions),
359 after-delete-terminal-functions, after-create-terminal-functions.
361 ** BULK RENAME: The 'display-' prefix of new Lisp-level functions
362 conflicts with stuff like 'display-time-mode'. Use 'device-'
363 or 'terminal-' instead. I think I prefer 'terminal-'.
365 It turns out that most of the offending Lisp functions were defined
366 in the trunk. Therefore, compatibility aliases should be defined
367 for the following names:
369 display-color-cells terminal-color-cells
370 display-color-p terminal-color-p
371 display-graphic-p terminal-graphic-p
372 display-grayscale-p terminal-grayscale-p
373 display-images-p terminal-images-p
374 display-mm-height terminal-mm-height
375 display-mm-width terminal-mm-width
376 display-mouse-p terminal-mouse-p
377 display-multi-font-p terminal-multi-font-p
378 display-multi-frame-p terminal-multi-frame-p
379 display-pixel-height terminal-pixel-height
380 display-pixel-width terminal-pixel-width
381 display-pixels-per-inch terminal-pixels-per-inch
382 display-planes terminal-planes
383 display-popup-menus-p terminal-popup-menus-p
384 display-save-under terminal-save-under
385 display-screens terminal-screens
386 display-supports-face-attributes-p terminal-supports-face-attributes-p
387 display-visual-class terminal-visual-class
388 framep-on-display framep-on-terminal
389 frames-on-display-list frames-on-terminal-list
391 The following functions were introduced in the multi-tty branch, and
392 were renamed without aliases:
394 delete-display delete-terminal
395 display-controlling-tty-p controlling-tty-p
396 display-list terminal-list
397 display-live-p terminal-live-p
398 display-name terminal-name
399 display-tty-type tty-type
400 frame-display frame-terminal
401 selected-display selected-terminal
403 ** The single-keyboard mode of MULTI_KBOARD is extremely confusing
404 sometimes; Emacs does not respond to stimuli from other keyboards.
405 At least a beep or a message would be important, if the single-mode
406 is still required to prevent interference. (Reported by Dan
409 Update: selecting a region with the mouse enables single_kboard
410 under X. This is very confusing.
412 Update: After discussions with Richard Stallman, this will be
413 resolved by having locked displays warn the user to wait, and
414 introducing a complex protocol to remotely bail out of
415 single-kboard mode by pressing C-g.
417 Update: Warning the user is not trivial to implement, as Emacs has
418 only one echo area, shared by all frames. Ideally the warning
419 should not be displayed on the display that is locking the others.
420 Perhaps the high probability of user confusion caused by
421 single_kboard mode deserves a special case in the display code.
422 Alternatively, it might be good enough to signal single_kboard mode
423 by changing the modelines or some other frame-local display element
424 on the locked out displays.
426 Update: In fact struct kboard does have an echo_string slot.
428 ** The session management module is prone to crashes when the X
429 connection is closed and then later I try to connect to a new X
432 #0 0xb7ebc806 in SmcGetIceConnection () from /usr/X11R6/lib/libSM.so.6
433 #1 0x080e6641 in x_session_check_input (bufp=0xbf86c9c0) at xsmfns.c:144
434 #2 0x080d3bbc in XTread_socket (device=0xa722ff8, expected=1, hold_quit=0xbf86ca90) at xterm.c:7037
435 #3 0x080fa404 in read_avail_input (expected=1) at keyboard.c:6696
436 #4 0x080fa4ca in handle_async_input () at keyboard.c:6900
437 #5 0x080d51fa in x_term_init (display_name=162628899, xrm_option=0x0, resource_name=0x857068c "emacs") at xterm.c:10622
438 #6 0x080d920e in x_display_info_for_name (name=162628899) at xfns.c:3975
439 #7 0x080d92f9 in check_x_display_info (object=1) at xfns.c:274
440 #8 0x080d97b8 in Fx_create_frame (parms=151221485) at xfns.c:3016
441 #9 0x0815bf72 in Ffuncall (nargs=2, args=0xbf86ceec) at eval.c:2851
443 I installed a workaround to prevent this. The X session manager is
444 only contacted when the very first display in the Emacs session is
445 an X display. Also, x_delete_display() on this display aborts
446 session management, and XTread_socket only calls
447 x_session_check_input when it is called for the display that the
448 session was opened on. While this does not really fix the bug, it
449 makes it much less frequent, because session manager support will
450 not normally be enabled when Emacs can survive the shutdown of the
453 See if xsmfns.c should be updated.
455 ** Hunt down display-related functions in frame.el and extend them all
456 to accept display ids.
458 ** rif->flush_display_optional (NULL) calls should be replaced by a
461 ** The set-locale-environment hack (adding the DISPLAY option) should
462 be replaced with a clean design.
464 ** standard-display-table should be display-local.
465 standard-display-european should be display-local.
467 ** With iswitchb-default-method set to 'always-frame, only frames on
468 the current display should be considered. This might involve
469 extending 'get-buffer-window'.
471 ** Have a look at Vlocale_coding_system. Seems like it would be a
472 tedious job to localize it, although most references use it for
473 interfacing with libc and are therefore OK with the global
476 Exceptions found so far: x-select-text and
479 ** Have a look at fatal_error_hook.
481 ** Have a look at set_frame_matrix_frame.
483 ** I think tip_frame should be display-local.
485 ** Check display reference count handling in x_create_tip_frame.
487 ** make-frame does not correctly handle extra parameters in its
490 (frame-parameter (make-frame (list (cons 'foobar 42))) 'foobar)
493 (This is likely an error in the CVS trunk.)
495 ** Gergely Nagy suggests that C-x # should only kill the current
496 frame, not any other emacsclient frame that may have the same file
497 opened for editing. I think I agree with him.
499 ** Very strange bug: visible-bell does not work on secondary
500 terminals in xterm and konsole. The screen does flicker a bit,
501 but it's so quick it isn't noticeable.
503 (Update: This is probably some problem with padding or whatnot on
504 the secondary terminals.)
506 ** Move baud_rate to struct display.
508 ** Implement support for starting an interactive Emacs session without
509 an initial frame. (The user would connect to it and open frames
510 later, with emacsclient.)
512 ** Implement Mac/Windows/DOS support. Many XXX comments mark things
513 that probably need updating, ChangeLogs will help in spotting
514 changes to X specific files that may need porting.
516 ** Do a grep on XXX and ?? for more issues.
518 ** flow-ctrl.el must be updated.
520 ** Fix stuff_char for multi-tty. Doesn't seem to be of high priority.
525 (ex-TODO items with explanations.)
527 -- Introduce a new struct for terminal devices.
529 (Done, see struct tty_output. The list of members is not yet
532 -- Change the bootstrap procedure to initialize tty_list.
534 (Done, but needs review.)
536 -- Change make-terminal-frame to support specifying another tty.
538 (Done, new frame parameters: 'tty' and 'tty-type'.)
540 -- Implement support for reading from multiple terminals.
542 (Done, read_avail_input tries to read from each terminal, until one
543 succeeds. MULTI_KBOARD is not used. Secondary terminals don't send
546 (Update: They do, now.)
548 (Update2: After enabling X, they don't.)
550 -- other-frame should cycle through the frames on the 'current'
553 (Done, by trivially modifying next_frame and prev_frame.)
555 -- Support different terminal sizes.
559 -- Make sure terminal resizes are handled gracefully. (Could be
562 (Done. We don't get automatic SIGWINCH for additional ttys,
565 -- Extend emacsclient to automatically open a new tty when it connects
568 (Done. It's an ugly hack, needs more work.)
570 -- Redisplay must refresh the topmost frame on *all* terminals, not
571 just the initial terminal.
573 (Done, but introduced an ugly redisplay problems. Ugh.)
575 -- Fix redisplay problems.
577 (Done; it turned out that the entire Wcm structure must be moved
578 inside tty_output. Why didn't I catch this earlier?)
580 -- Provide a way for emacsclient to tell Emacs that the tty has been
583 (Done, simply forward the SIGWINCH signal.)
585 -- Each keypress should automatically select the frame corresponding
586 to the terminal that it was coming from. This means that Emacs
587 must know from which terminal the last keyboard event came from.
589 (Done, it was quite simple, the input event system already
590 supported multiple frames.)
592 -- Fix SIGIO issue with secondary terminals.
594 (Done, emacsclient signals Emacs after writing to the proxy pseudo
595 terminal. Note that this means that multi-tty does not work with
598 (Update: This is bullshit. There is a read_input_waiting function,
599 extend that somehow.)
601 (Update of update: The first update was not right either, extending
602 read_input_waiting was not necessary. Secondary ttys do seem to
603 send signals on input.)
605 (Update^3: Not any more.)
607 -- Make make-terminal-frame look up the 'tty' and 'tty-type' frame
608 parameters from the currently selected terminal before the global
613 -- Put all cached terminal escape sequences into struct tty_output.
614 Currently, they are still stored in global variables, so we don't
615 really support multiple terminal types.
617 (Done. It was not fun.)
619 -- Implement sane error handling after initialization. (Currently
620 emacs exits if you specify a bad terminal type.) The helpful error
621 messages must still be provided when Emacs starts.
625 -- Implement terminal deletion, i.e., deleting local frames, closing
626 the tty device and restoring its previous state without exiting
629 (Done, but at the moment only called when an error happens during
630 initialization. There is a memory corruption error around this
631 somewhere.) (Update: now it is fully enabled.)
633 -- Implement automatic deletion of terminals when the last frame on
634 that terminal is closed.
638 -- Restore tty screen after closing the terminal.
640 (Done, we do the same as Emacs 21.2 for all terminals.)
642 -- 'TERM=dumb src/emacs' does not restore the terminal state.
646 -- C-g should work on secondary terminals.
648 (Done, but the binding is not configurable.)
650 -- Deal with SIGHUP in Emacs and in emacsclient. (After this, the
651 server-frames may be removed from server.el.)
653 (Done, nothing to do. It seems that Emacs does not receive SIGHUP
654 from secondary ttys, which is actually a good thing.) (Update: I
655 think it would be a bad idea to remove server-frames.)
657 -- Change emacsclient/server.el to support the -t argument better,
658 i.e. automatically close the socket when the frame is closed.
660 (Seems to be working OK.)
662 -- Fix mysterious memory corruption error with tty deletion. To
663 trigger it, try the following shell command:
665 while true; do TERM=no-such-terminal-definition emacsclient -h; done
667 Emacs usually dumps core after a few dozen iterations. (The bug
668 seems to be related to the xfreeing or bzeroing of
669 tty_output.Wcm. Maybe there are outside references to struct Wcm?
670 Why were these vars collected into a struct before multi-tty
673 (Done. Whew. It turned out that the problem had nothing to do
674 with hypothetical external references to Wcm, or any other
675 tty_output component; it was simply that delete_tty closed the
676 filehandles of secondary ttys twice, resulting in fclose doubly
677 freeing memory. Utterly trivial matter. I love the C's memory
678 management, it puts hair on your chest.)
680 -- Support raw secondary terminals. (Note that SIGIO works only on
681 the controlling terminal.) Hint: extend read_input_waiting for
682 multiple ttys and hopefully this will be fixed.
684 (Done, it seems to have been working already for some time. It
685 seems F_SETOWN does work, after all. Not sure what made it fail
686 earlier, but it seems to be fixed (there were several changes
687 around request_sigio, maybe one of them did it).
688 read_input_waiting is only used in sys_select, don't change
689 it.) (Update: After adding X support, it's broken again.)
690 (Update^2: No it isn't.) :-)
692 -- Find out why does Emacs abort when it wants to close its
693 controlling tty. Hint: chan_process[] array. Hey, maybe
694 noninterrupt-IO would work, too? Update: no, there is no process
697 (Done. Added add/delete_keyboard_wait_descriptor to
698 term_init/delete_tty. The hint was right, in a way.)
700 -- Issue with SIGIO: it needs to be disabled during redisplay. See if
701 fcntl kernel behavior could be emulated by emacsclient.
703 (Done. Simply disabled the SIGIO emulation hack in emacsclient.)
704 (Update: it was added back.) (Update^2: and removed again.)
706 -- server.el: There are issues with saving files in buffers of closed
707 clients. Try editing a file with emacsclient -f, and (without
708 saving it) do a delete-frame. The frame is closed without
709 question, and a surprising confirmation prompt appears in another
712 (Done. delete-frame now asks for confirmation if it still has
713 pending buffers, and modified buffers don't seem to be deleted.)
715 -- emacsclient.el, server.el: Handle eval or file open errors when
720 -- Make parts of struct tty_output accessible from Lisp. The device
721 name and the type is sufficient.
723 (Done, see frame-tty-name and frame-tty-type.)
725 -- Export delete_tty to the Lisp environment, for emacsclient.
727 (Done, see delete-tty.)
729 -- Get rid of the accessor macros in termchar.h, or define macros for
734 -- Move device-specific parameters (like costs) commonly used by
735 device backends to a common, device-dependent structure.
737 (Done. See struct display_method in termhooks.h.)
741 (Done. Well, it seems to be working.)
743 -- Allow simultaneous X and tty frames. (Handling input could be
744 tricky. Or maybe not.)
746 (Done. Allowed, that is. It is currently extremely unstable, to
747 the point of being unusable. The rif variable causes constant
748 core dumps. Handling input is indeed tricky.)
750 -- Rewrite multi-tty input in terms of MULTI_KBOARD.
752 (Done. In fact, there was no need to rewrite anything, I just
753 added a kboard member to tty_display_info, and initialized the
754 frame's kboard from there.)
756 -- Fix rif issue with X-tty combo sessions. IMHO the best thing to do
757 is to get rid of that global variable (and use the value in
758 display_method, which is guaranteed to be correct).
760 (Done, did exactly that. Core dumps during combo sessions became
761 much rarer. In fact, I have not yet met a single one.)
763 -- Add multi-tty support to talk.el.
767 -- Clean up the source of emacsclient. It is a mess.
769 (Done, eliminated stupid proxy-pty kludge.)
771 -- Fix faces on tty frames during X-tty combo sessions. There is an
772 init_frame_faces call in init_sys_modes, see if there is a problem
775 (Done, there was a stupid mistake in
776 Ftty_supports_face_attributes_p. Colors are broken, though.)
778 -- C-x 5 2, C-x 5 o, C-x 5 0 on an emacsclient frame unexpectedly
779 exits emacsclient. This is a result of trying to be clever with
780 delete-frame-functions.
782 (Fixed, added delete-tty-after-functions, and changed server.el to
785 -- Something with (maybe) multi-keyboard support broke function keys
786 and arrows on ttys during X+tty combo sessions. Debug this.
788 (I can't reproduce it, maybe the terminal type was wrong.)
790 -- Fix input from raw ttys (again).
792 (Now it seems to work all right.)
794 -- During an X-tty combo session, a (message "Hello") from a tty frame
795 goes to the X frame. Fix this.
797 (Done. There was a safeguard against writing to the initial
798 terminal frame during bootstrap which prevented echo_area_display
799 from working correctly on a tty frame during a combo session.)
801 -- If there are no frames on its controlling terminal, Emacs should
802 exit if the user presses C-c there.
804 (Done, as far as possible. See the SIGTERM comment in
805 interrupt_signal on why this seems to be impossible to solve this
808 -- During an X session, Emacs seems to read from stdin. Also, Emacs
809 fails to start without a controlling tty.
811 (Fixed by replacing the troublesome termcap display with a dummy
812 bootstrap display during bootstrap.
814 -- Do tty output through struct display, like graphical display
819 -- Define an output_initial value for output_method for the initial
820 frame that is dumped with Emacs. Checking for this frame (e.g. in
821 cmd_error_internal) is ugly.
823 (Done, breaking interactive temacs.)
825 -- The command "emacsclient -t -e '(delete-frame)'" fails to exit.
829 -- frame-creation-function should always create a frame that is on the
830 same display as the selected frame. Maybe frame-creation-function
831 should simply be removed and make-frame changed to do the right
834 (Done, with a nice hack. frame-creation-function is now frame-local.)
836 -- Fix C-g on raw ttys.
838 (Done. I disabled the interrupt/quit keys on all secondary
839 terminals, so Emacs sees C-g as normal input. This looks like an
840 overkill, because emacsclient has extra code to pass SIGINT to
841 Emacs, so C-g should remain the interrupt/quit key on emacsclient
842 frames. See the next entry why implementing this distinction would
845 -- Make sure C-g goes to the right frame with ttys. This is hard, as
846 SIGINT doesn't have a tty parameter. :-(
848 (Done, the previous change fixes this as a pleasant side effect.)
850 -- I have seen a case when Emacs with multiple ttys fell into a loop
851 eating 100% of CPU time. Strace showed this loop:
854 kill(30284, SIGIO) = 0
855 --- SIGIO (I/O possible) @ 0 (0) ---
856 ioctl(6, FIONREAD, [0]) = -1 EIO (Input/output error)
857 ioctl(5, FIONREAD, [0]) = -1 EIO (Input/output error)
858 ioctl(0, FIONREAD, [0]) = 0
859 sigreturn() = ? (mask now [])
860 gettimeofday({1072842297, 747760}, NULL) = 0
861 gettimeofday({1072842297, 747806}, NULL) = 0
862 select(9, [0 3 5 6], NULL, NULL, {0, 0}) = 2 (in [5 6], left {0, 0})
863 select(9, [0 3 5 6], NULL, NULL, {0, 0}) = 2 (in [5 6], left {0, 0})
864 gettimeofday({1072842297, 748245}, NULL) = 0
866 I have seen something similar with a single X frame, but have not
867 been able to reproduce it for debugging.
869 Update: This may have been caused by checking for nread != 0
870 instead of nread > 0 after calling read_socket_hook in
873 (Fixed. This was caused by unconditionally including stdin in
874 input_wait_mask in init_process_emacs. The select call in
875 wait_reading_process_input always returned immediately, indicating
876 that there is pending input from stdin, which nobody read.
878 Note that the above strace output seems to be an unrelated but
879 similar bug. I think that is now fixed.)
881 -- Exiting Emacs while there are emacsclient frames doesn't restore the
882 ttys to their default states.
884 (This seems to be fixed by some previous change.)
886 -- Allow opening an X session after -nw.
890 -- Fix color handling during tty+X combo sessions. (It seems that tty
891 sessions automatically convert the face colors to terminal colors
892 when the face is loaded. This conversion must happen instead on
893 the fly in write_glyphs, which might be problematic, as color
894 approximation is currently done in lisp (term/tty-colors.el).)
895 (Update: hm, colors seem to work fine if I start emacs with -nw and
896 then create an X frame. Maybe it's just a small buglet somewhere.)
898 (Seems to be fixed. The problem was in startup.el, it did not
899 initialize tty colors when the initial window system was
902 -- emacs -nw --eval '(y-or-n-p "Foobar")' segfaults. (Reported by
905 (Fixed, there was a keyboard initialization problem.)
907 -- Fix interactive use of temacs. There are face-related SEGVs, most
908 likely because of changes in realize_default_face, realize_face.
912 -- Don't exit Emacs when the last X connection fails during a
913 multi-display session.
917 -- Dan Nicolaescu noticed that starting emacsclient on the same
918 terminal device that is the controlling tty of the Emacs process
919 gives unexpected results.
923 -- Istvan Marko reported that Emacs hang on ttys if it was started
926 (Fixed. There was a bug in the multi-tty version of
927 narrow_foreground_group. tcsetpgrp blocks if it is called from a
928 process that is not in the same process group as the tty.)
930 -- emacsclient -t from an Emacs term buffer does not work, complains
931 about face problems. This can even lock up Emacs (if the recursive
932 frame sets single_kboard). Update: the face problems are caused by
933 bugs in term.el, not in multi-tty. The lockup is caused by
934 single_kboard mode, and is not easily resolvable. The best thing to
935 do is to simply refuse to create a tty frame of type 'eterm'.
937 (Fixed, changed emacsclient to check for TERM=eterm. The face
938 complaints seem to be caused by bugs in term.el; they are not
939 related to multi-tty.)
941 -- Find out the best way to support suspending Emacs with multiple
942 ttys. My guess: disable it on the controlling tty, but from other
943 ttys pass it on to emacsclient somehow. (It is (I hope) trivial to
944 extend emacsclient to handle suspend/resume. A 'kill -STOP' almost
947 (Done. I needed to play with signal handling and the server
948 protocol a bit to make emacsclient behave as a normal UNIX program
949 wrt foreground/background process groups.)
951 -- There is a flicker during the startup of 'emacs -nw'; it's as if
952 the terminal is initialized, reset and then initialized again.
953 Debug this. (Hint: narrow_foreground_group is called twice during
958 -- Robert Chassell has found serious copy-paste bugs with the
959 multi-tty branch. There seem to be redisplay bugs while copying
960 from X to a terminal frame. Copying accented characters do not
963 (Patch-124 should fix this, by changing the interprogram-*-function
964 variables to be frame-local, as suggested by Mark Plaksin
965 (thanks!). I think that the redisplay bugs are in fact not bugs,
966 but delays caused by single_kboard --> perhaps MULTI_KBOARD should
969 -- frame-creation-function was removed, which might be a bad idea.
970 Think up a compatible solution.
972 (It was an internal interface that may be changed when necessary.)
974 -- Change Lisp code not to (getenv "TERM"); use the 'tty-type' frame
975 parameter or the frame-tty-type function instead. (M-x tags-search
976 "TERM" helps with this.) Update: Actually, all getenv invocations
977 should be checked for multi-tty compatibility, and an interface
978 must be implemented to get the remote client's environment.
980 (Done. Only getenv calls in lisp/term/*.el were changed; other
981 calls should be mostly left as they are.)
983 -- Add an elaborate mechanism for display-local variables. (There are
984 already a few of these; search for 'terminal-local' in the Elisp
987 (Not needed. Display-local variables could be emulated by
988 frame-local variables.)
990 -- Emacs assumes that all terminal frames have the same locale
991 settings as Emacs itself. This may lead to bogus results in a
992 multi-locale setup. (E.g., while logging in from a remote client
993 with a different locale.)
994 (Update after new bugreport by Friedrich Delgado Friedrichs:
995 (at least) the structs terminal_coding and keyboard_coding in
996 coding.c must be moved to struct display, and the Lisp interface
997 [set-]keyboard-coding-system must be adapted for the change.)
999 (Fixed. Emacs now uses the locale settings as seen by the
1000 emacsclient process for server tty frames.)
1001 (Update: Not really; Vlocale_coding_system is still global.)
1003 -- Make 'struct display' accessible to Lisp programs. Accessor functions:
1005 (displayp OBJECT): Returns t if OBJECT is a display.
1006 => Implemented as display-live-p.
1008 (display-list): Returns list of currently active displays.
1011 (selected-display): Returns the display object of the selected frame.
1012 => Not strictly necessary, but implemented anyway.
1014 (frame-display FRAME): Returns the display object of FRAME.
1017 (display-frames DISPLAY): Returns a list of frames on DISPLAY.
1018 => Already implemented, see frames-on-display-list.
1020 (display-type DISPLAY): Returns the type of DISPLAY, as a
1021 symbol. (See 'framep'.)
1022 => Implemented as display-live-p.
1024 (display-device DISPLAY): Returns the name of the device that
1025 DISPLAY uses, as a string. (E.g: "/dev/pts/16", or
1027 => Implemented as display-name.
1031 See next issue why this is necessary.
1033 (Update: The consensus on emacs-devel seems to be to do this via
1034 integer identifiers. That's fine by me.)
1038 -- The following needs to be supported:
1045 The cleanest way to solve this is to allow multiple displays on the
1046 same terminal device; each new emacsclient process should create
1047 its own display. As displays are currently identified by their
1048 device names, this is not possible until struct display becomes
1049 accessible as a Lisp-level object.
1053 -- Support multiple character locales. A version of
1054 'set-locale-environment' needs to be written for setting up
1055 display-local settings on ttys. I think calling
1056 set-display-table-and-terminal-coding-system and
1057 set-keyboard-coding-system would be enough. The language
1058 environment itself should remain a global setting.
1060 (Done, by an ugly hack.)
1062 -- The terminal customization files in term/*.el tend to change global
1063 parameters, which may confuse Emacs with multiple displays. Change
1064 them to tweak only frame-local settings, if possible. (They tend
1065 to call define-key to set function key sequences a lot.)
1067 (Done, by making 'function-key-map' terminal-local (i.e., part of
1068 struct kboard). This has probably covered all the remaining problems.)
1070 -- xt-mouse.el needs to be adapted for multi-tty. It currently
1071 signals an error on kill-emacs under X, which prevents the user
1072 from exiting Emacs. (Reported by Mnemonikk on freenode.)
1077 -- Having {reset,init}_all_sys_modes in set-input-mode breaks arrow
1078 keys on non-selected terminals under screen, and sometimes on other
1079 terminal types as well. The other function keys continue to work
1080 fine. Sometimes faces on these screens become garbled.
1082 This only seems to affect displays that are of the same terminfo
1083 type as the selected one. Interestingly, in screen Emacs normally
1084 reports the up arrow key as 'M-o A', but after the above SNAFU, it
1085 complains about 'M-[ a'. UNIX ttys are a complete mystery to me,
1086 but it seems the reset-reinitialize cycle somehow leaves the
1087 non-selected terminals in a different state than usual. I have no
1088 idea how this could happen.
1090 Currently set-input-mode resets only the currently selected
1091 terminal, which seems to somehow work around the problem.
1095 Dan Nicolaescu <dann@ics.uci.edu> writes:
1096 > Some terminals have 2 modes for cursor keys: Application Mode where
1097 > the cursor keys transmit the codes defined in the terminfo entry, and
1098 > Cursor mode. Applications have to send the smkx and rmkx terminfo
1099 > strings to switch between the 2 modes. So Emacs (and emacsclient) have
1100 > to send smkx when initializing and rmkx when quitting (or on
1103 (I think patch-370 fixed this.)
1105 -- This long-standing bug (first reported by Han Boetes) seems to come
1106 and go all the time. It is time to track it down and fix it.
1111 # From another xterm:
1112 emacsclient -e '(y-or-n-p "Do you want me to crash? ")'
1113 # Notice how the answer ends up in the *scratch* buffer
1117 (Fixed in patch-414 after detailed analysis by Kalle Olavi Niemitalo.)
1119 -- normal-erase-is-backspace-mode in simple.el needs to be updated for
1120 multi-tty (rep. by Dan Waber). (The Delete key is broken on X
1123 (Fixed in patch-427.)
1125 -- I think keyboard-translate-table should be made terminal-local.
1127 (Done in patch-431.)
1129 -- The semantics of terminal-local variables are confusing; it is not
1130 clear what binding is in effect in any given time. See if
1131 current_kboard (or at least the terminal-local bindings exported to
1132 Lisp) might be changed to be tied to the selected frame instead.
1133 Currently, 'function-key-map' and 'key-translation-map' may be
1134 accessed reliably only using the hackish
1135 '(set-)terminal-local-value' functions.
1137 Perhaps there should be a difference between 'last-command' &co.
1138 and these more conventional configuration variables.
1139 (E.g. 'symbol-value' would use current_kboard to access
1140 'last-command', but SELECTED_FRAME()->display->kboard to get the
1141 value of 'function-key-map'.
1143 (Fixed in patch-434.)
1145 -- If the first key pressed on a new tty terminal is a function key,
1146 it is not recognized correctly. May be related to the bug below.
1148 (Seems to have been fixed as a side effect of patch-434. "The bug
1149 below" was the set-input-mode madness.)
1151 (Update: this bug was fixed for good in patch-449. It was tracked
1152 down to a bug in 'read_key_sequence': it failed to reinitialize its
1153 local function-key-map/key-translation-map references when it
1154 switched keyboards. I don't understand why did this bug only
1155 appear on brand new frames, though!)
1157 -- Disable connecting to a new X display when we use the GTK toolkit.
1159 (Disabled in patch-450.)
1161 -- Implement automatic forwarding of client environment variables to
1162 forked processes, as discussed on the multi-tty list. Terminal
1163 parameters are now accessible in C code, so the biggest obstacle is
1164 gone. The 'getenv_internal' and 'child_setup' functions in
1165 callproc.c must be changed to support the following variable:
1167 terminal-local-environment-variables is a variable defined in ...
1169 Enable or disable terminal-local environment variables.
1171 If set to t, 'getenv', 'setenv' and subprocess creation
1172 functions use the environment variables of the emacsclient
1173 process that created the selected frame, ignoring
1174 'process-environment'.
1176 If set to nil, Emacs uses 'process-environment' and ignores
1177 the client environment.
1179 Otherwise, 'terminal-local-environment-variables' should be a
1180 list of variable names (represented by Lisp strings) to look
1181 up in the client environment. The rest will come from
1182 'process-environment'.
1184 (Implemented in patch-461; 'terminal-getenv', 'terminal-setenv' and
1185 'with-terminal-environment' are now replaced by extensions to
1186 'getenv' and 'setenv', and the new 'local-environment-variables'
1189 (Updated in patch-465 to fix the semantics of let-binding
1190 'process-environment'. 'process-environment' was changed to
1191 override all local/global environment variables, and a new variable
1192 'global-environment' was introduced to have 'process-environment's
1195 (Updated in patch-466 to fix the case when two emacsclient sessions
1196 share the same terminal, but have different environment. The local
1197 environment lists are now stored as frame parameters, so the
1198 C-level terminal parameters are not strictly necessary any more.)
1200 -- 'Fdelete_frame' is called from various critical places where it is
1201 not acceptable for the frame deletion to fail, e.g. from
1202 x_connection_closed after an X error. 'Fdelete_frame' now protects
1203 against 'delete-frame-functions' throwing an error and preventing a
1204 frame delete. (patch-475)
1206 -- Fix set-input-mode for multi-tty. It's a truly horrible interface;
1207 what if we'd blow it up into several separate functions (with a
1208 compatibility definition)?
1210 (Done. See 'set-input-interrupt-mode', 'set-output-flow-control',
1211 'set-input-meta-mode' and 'set-quit-char'.) (patch-457)
1213 -- Let-binding 'overriding-terminal-local-map' on a brand new frame
1214 does not seem to work correctly. (See 'fancy-splash-screens'.)
1215 The keymap seems to be set up right, but events go to another
1216 terminal. Or is it 'unread-command-events' that gets Emacs
1217 confused? Investigate.
1219 (Emacs was confused because a process filter entered
1220 'recursive-edit' while Emacs was reading input. I added support
1221 for this in the input system.) (patch-489)
1223 -- I smell something funny around pop_kboard's "deleted kboard" case.
1224 Determine what are the circumstances of this case, and fix any
1225 bug that comes to light.
1227 (It happens simply because single_kboard's terminal is sometimes
1228 deleted while executing a command on it, for example the one that
1229 kills the terminal. There was no bug here, but I rewrote the whole
1230 single_kboard mess anyway.) (patch-489)
1232 -- Understand Emacs's low-level input system (it's black magic) :-)
1233 What exactly does interrupt_input do? I tried to disable it for
1234 raw secondary tty support, but it does not seem to do anything
1235 useful. (Update: Look again. X unconditionally enables this, maybe
1236 that's why raw terminal support is broken again. I really do need
1237 to understand input.)
1238 (Update: I am starting to understand the read_key_sequence->read-char
1239 ->kbd_buffer_get_event->read_avail_input->read_socket_hook path. Yay!)
1241 (Update: OK, it all seems so easy now (NOT). Input could be done
1242 synchronously (with wait_reading_process_input), or asynchronously
1243 by SIGIO or polling (SIGALRM). C-g either sets the Vquit_flag,
1244 signals a 'quit condition, or throws to 'getcjmp' when Emacs was
1245 waiting for input when the C-g event arrived.)
1247 -- Replace wrong_kboard_jmpbuf with a special return value of
1248 read_char. It is absurd that we use setjmp/longjmp just to return
1249 to the immediate caller.
1251 (Done in patch-500.)
1253 -- 'tool-bar-mode', 'scroll-bar-mode', 'menu-bar-mode' and
1254 'fringe-mode' are modes global to the entire Emacs session, not
1255 just a single frame or a single terminal. This means that their
1256 status sometimes differs from what's actually displayed on the
1257 screen. As a consequence, the Options | Show/Hide menu sometimes
1258 shows incorrect status, and you have to select an option twice for
1259 it to have any visible effect on the current frame.
1261 Change Emacs so that the status of the items in the Options |
1262 Show/Hide menu correspond to the current frame.
1264 (Done in patch-537.)
1266 -- The 'default-directory' variable should somehow be set to the
1267 cwd of the emacsclient process when the user runs emacsclient
1268 without file arguments. Perhaps it is OK to just override the
1269 directory of the *scratch* buffer.
1271 (Done in patch-539.)
1273 -- The borders on tooltip frames on X are messed up. More
1274 specifically, the frame's internal border (internal-border-width
1275 frame parameter) is not filled with the correct background color.
1277 It seems the border contents are drawn onto by the
1278 update_single_window call in 'x-show-tip'. After some debugging, I
1279 think the window's background color is not set up
1280 correctly---calling 'x_clear_area' fills the specified area with
1281 black, not light yellow.
1283 (Done in patch-544. A background_pixel field was defined both in
1284 struct frame and struct x_output, and Emacs got confused between
1288 This file is part of GNU Emacs.
1290 GNU Emacs is free software: you can redistribute it and/or modify
1291 it under the terms of the GNU General Public License as published by
1292 the Free Software Foundation, either version 3 of the License, or
1293 (at your option) any later version.
1295 GNU Emacs is distributed in the hope that it will be useful,
1296 but WITHOUT ANY WARRANTY; without even the implied warranty of
1297 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
1298 GNU General Public License for more details.
1300 You should have received a copy of the GNU General Public License
1301 along with GNU Emacs. If not, see <https://www.gnu.org/licenses/>.