Re: Give people with crashes more of a clue
26 years, 1 month
Robin S. Socha
Robin S. Socha
* Hrvoje Niksic <hniksic(a)iskon.hr> writes:
> isn't it pointless to refer people to a PROBLEMS file that's not there
> anymore? If we aren't already shipping PROBLEMS with XEmacs, maybe we
> should start.
Would it make any sense at all to not call the file PROBLEMS but
PROBLEMS.<your_os> or even README.<your_os>, and advertise it in
the splash buffer/startup.el and the Help menu/menubar-items.el? I
remember seeing intimidated users that only knew KDE. Yes, I know
they should be using the RPMs.
--
Robin S. Socha
Big. Black. Beautified. <http://socha.net/XEmacs/>
- 1 participants
- 0 comments
Gutter Crash [was Re: 21.2-b19 "Sneaky Bastard 19991102" crashes when using toolbar for Ediff control buffer]
26 years, 1 month
Jan Vroonhof
Jan Vroonhof
[No it is not week-end here but I have 21.2 on our new linux box]
ALTMARK(a)de.ibm.com writes:
> 5. Moved the mouse pointer out of the Ediff control buffer. Crash ...
> #4 0x807cf7e in assert_failed ()
> #5 0x8112604 in start_end_of_last_line ()
> #6 0x8114f69 in pixel_to_glyph_translation ()
Ok, here is what happens.
The ediff window is very small. Thus if you run with the buffer tabs,
the buffer tabs take all the place of the window, so the buffers window
isn't actually displayed. Only the modeline is.
(Side issue: This window really shouldn't have tabs)
(Side issue: Why is the size of the gutters not taken into account to
set the frame size?)
Then you move the mouse outside. The mouse code tries to find the
nearest point in the buffer (pixel_to_glyph_translation), because the
displayed parts of the buffer are not near, it thinks the last
displayed line is closest and thus calls end_of_last_line (which is
really start_end_of_last_line).
These kinds of computation are really done in redisplay and thus
start_end_of_last_line calls point_in_line_start_cache to update the
cache that says where the display lines begin.
As the window really isn't displayed the display structures are empty
and thus point_in_line_start_cache does a quick minimal redisplay
(without actually updating the screen), that is the call to
update_line_start_cache on line 7254. As the cache is empty it hits
this code.
while (startp < old_lb || low_bound == -1)
{
int ic_elt;
Bufpos new_startp;
regenerate_window (w, startp, point, CMOTION_DISP);
update_internal_cache_list (w, CMOTION_DISP);
regenerate_window does that redisplaying.. However only the modeline
is displayed. update_internal_cache_list then copies all info form
the redisplay structures to a temporary working copy of the line start
cache. However as there aren't any real buffer lines displayed there
is no info to copy and the cache stays empty. That turns to be a known
problem
/* If this assert is triggered then regenerate_window failed
to layout a single line. That is not supposed to be
possible because we impose a minimum height on the buffer
and override vertical clip when we are in here. */
/* #### Ah, but it is because the window may temporarily
exist but not have any lines at all if the minibuffer is
real big. Look into that situation better. */
if (!Dynarr_length (internal_cache))
{
if (old_lb == -1 && low_bound == -1)
{
updating_line_start_cache = 0;
w->line_cache_validation_override--;
return;
}
and thus update_line_start_cache returns without doing anything and
leaving the cache empty. Then point_in_line_start_cache returns "-1"
because the value of window-start couldn't be find in the buffer.
This then triggers the abort.
The question is now how to solve it. Should we enforce the conditions
that part of a window should always be displayed? Should we remove the
abort and return a sensible guess? Should generate_window be fixed to
always layout a line regardless of whether it fixed?
Additionally, I see no gutter code in pixel_to_glyph_translation, is
that correct? Andy?
Jan
- 2 participants
- 1 comments
I thought someone had made widget/custom use native widgets?
26 years, 1 month
William M. Perry
William M. Perry
I don't see anything but the older pixmap-based buttons. Updated all my
packages, etc, using the latest CVS. Am I misremembering?
-bp
- 2 participants
- 1 comments
compiler warnings mess
26 years, 1 month
Stef Epardaud
Stef Epardaud
hi,
i decided to rebuild on cvs today, and although it did build, the
compilation gave me loads of warnings, and since i was really bored, i
took a look at them.
while i fixed (what i call 'fix' ... ;) some of them and produce some
patches, the dozens others remain and i cannot fix them because i don;'t
understand them, they seem plain wrong.
those i could fix were some with unused variables (forgotten somehow)
and some with using ints before initialization that i explicitly set to
0, and some chars used as integers casting problem.
but sometimes it just complains about ints used before initialization
and even setting them to any value does not fix the warning, sometimes
the initialization is a macro in the declaration and although it _is_
initialized, it just does not get it.
to me it seems broken or i gave it a wrong option (and it is by default
because i did not give any option)
and in the end i even wondered if my 10-20 warnings fix were valid
because the ones remaining seemd unjustified.
what should i do, or is it normal to get dozens of :
window.c: In function `window_loop':
window.c:2409: warning: `w' might be used uninitialized in this function
window.c:2411: warning: `next_window' might be used uninitialized in
this function
window.c:2412: warning: `last_window' might be used uninitialized in
this function
window.c:2413: warning: `frame' might be used uninitialized in this
function
window.c:2420: warning: `devcons' might be used uninitialized in this
function
window.c:2420: warning: `concons' might be used uninitialized in this
function
window.c:2443: warning: `obj' might be used uninitialized in this
function
window.c:2445: warning: `obj' might be used uninitialized in this
function
window.c:2477: warning: `obj' might be used uninitialized in this
function
window.c:2478: warning: `obj' might be used uninitialized in this
function
window.c:2551: warning: `obj' might be used uninitialized in this
function
window.c:2571: warning: `obj' might be used uninitialized in this
function
window.c:2647: warning: `obj' might be used uninitialized in this
function
window.c:2688: warning: `obj' might be used uninitialized in this
function
(example)
in a lot of files ?
thanx for making me smarter by answering :)
--
# Stef Epardaud, # There is no limit to the power of computing ...
# Java Defeater # ... except men maybe ?
# Rotterdam # Lunatech Research,
# NETHERLANDS # soon we will drop the searching for the finding...
# #
- 4 participants
- 3 comments
GPM mouse tracking issues solved, kindof
26 years, 1 month
William M. Perry
William M. Perry
I finally figured out a nice clean way to stick the GPM event into the
queue, and now everything works, including mouse-tracking when while doing
selection, so you get good incremental feedback and all that stuff.
I removed the static-ness of enqueue_command_event(). I can't use
unread-command-events, because mouse-motion events screw that up. With
this approach, everything works beautifully.
My question is, is this a completely bad way to achieve my goals? I want
to insert arbitrary events into the data stream from outside event-stream.c
or any of the event-XXX.c files. I could find no other way to do it.
I hope you don't get this twice ben, I wasn't sure if you were on the
normal xemacs-beta list, or just xemacs-review.
-Bill P.
- 2 participants
- 1 comments
Re: GPM mouse tracking issues solved, kindof
26 years, 1 month
Jan Vroonhof
Jan Vroonhof
wmperry(a)aventail.com (William M. Perry) writes:
> I removed the static-ness of enqueue_command_event(). I can't use
> unread-command-events, because mouse-motion events screw that up. With
> this approach, everything works beautifully.
>
> My question is, is this a completely bad way to achieve my goals? I want
> to insert arbitrary events into the data stream from outside event-stream.c
> or any of the event-XXX.c files. I could find no other way to do it.
Well. enqueue_command_event() is already accessible from outside of
event-stream.c by means of the wrappers for misc_user and magic_eval
events.
The biggest thing I have against this is that mouse_motion events are
not command-events. The over solution would be to maintain your own
que for stuff you want to insert and then put your own wrappers in
event_stream's
int (*event_pending_p) (int);
void (*next_event_cb) (struct Lisp_Event *);
slots. Then you set these to drain your own queue first and call the
original if empty.
Jan
- 1 participants
- 0 comments
Give people with crashes more of a clue
26 years, 1 month
Robert Pluim
Robert Pluim
I've finally cracked, I've had it with skip-syntax-backwards crashes
on gnu.emacs.xemacs. Polite (for once) patch attached. Recommended for
every version of XEmacs in the universe.
1999-11-05 Robert Pluim <rpluim(a)bigfoot.com>
* emacs.c (shut_down_emacs): Point users to PROBLEMS file
Index: emacs.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/emacs.c,v
retrieving revision 1.92
diff -u -w -u -r1.92 emacs.c
--- emacs.c 1999/11/02 04:17:14 1.92
+++ emacs.c 1999/11/05 17:20:06
＠＠ -2262,12 +2262,16 ＠＠
("Your files have been auto-saved.\n"
"Use `M-x recover-session' to recover them.\n"
"\n"
+ "If you have access to the PROBLEMS file that came with your\n"
+ "version of XEmacs, please check to see if your crash is described\n"
+ "there, as there may be a workaround available.\n"
#ifdef INFODOCK
- "Please report this bug by selecting `Report-Bug' in the InfoDock\n"
- "menu.\n"
+ "Otherwise, please report this bug by selecting `Report-Bug'\n"
+ "in the InfoDock menu.\n"
#else
- "Please report this bug by running the send-pr script included\n"
- "with XEmacs, or selecting `Send Bug Report' from the help menu.\n"
+ "Otherwise, please report this bug by running the send-pr\n"
+ "script included with XEmacs, or selecting `Send Bug Report'\n"
+ "from the help menu.\n"
"As a last resort send ordinary email to `crashes(a)xemacs.org'.\n"
#endif
"*MAKE SURE* to include the information in the command\n"
--
The above are my opinions,
and my opinions only.
- 1 participants
- 0 comments
21.2.19-sneaky: file-variables/enable-local-variables/gutter error
26 years, 1 month
Mark Thomas
Mark Thomas
This bug report will be sent to the XEmacs Development Team,
not to your local site managers!!
Please write in English, because the XEmacs maintainers do not have
translators to read other languages for them.
In XEmacs 21.2 (beta19) "Sneaky Bastard 19991102" [Lucid] (i686-pc-linux, Mule) of Fri Nov 5 1999 on ulysses.jprc.com
configured using `configure --prefix=/home/xemacs/debug --with-png --with-mule'
Please describe exactly what actions triggered the bug
and the precise symptoms of the bug:
When
(1) I start XEmacs with the -vanilla switch,
(2) I'm running under X with gutters visible,
(3) I set enable-local-variables to 'ask, and
(4) I type C-x C-f /tmp/foo.el where the entire contents of
/tmp/foo.el is:
-----cut-----
;; Local Variables: ***
;; foo: bar ***
;; End: ***
-----cut-----
XEmacs gives the warning:
(1) (local-variables/warning) File local-variables error: Missing
colon in local variables entry
and does not set the local variable foo. If I turn off the gutters,
or run in TTY mode, or leave enable-local-variables set to t, the
local-variable section behaves normally, i.e., foo gets set to bar.
Recent keystrokes:
M-( e DEL s e t q SPC e n a b l e - l o c a l - v a
r i a b l e s SPC ' a s k right C-j C-x C-f / t m p
/ f o o . e l RET y M-: f o o RET M-x r e p o r t -
e m a c s - b TAB RET
Recent messages (most recent first):
Loading mail-abbrevs...done
Loading mail-abbrevs...
Loading emacsbug...done
Loading emacsbug...
Symbol's value as variable is void: foo
Set local variables as specified at end of foo.el? (y or n) Yes
Set local variables as specified at end of foo.el? (y or n)
Loading efs-cu...done
Loading efs-cu...
- 1 participants
- 0 comments
Another GPM patch...
26 years, 1 month
William M. Perry
William M. Perry
This supercedes the last one. I didn't have time this morning to
reorganize the C code and write a lisp-level interface for gpm-minor-mode
to make it more flexible, but you can now paste from other linux consoles
into XEmacs. It's kind of a hack, since the paste code in the linux kernel
writes directly to the console. If anyone can think of a better way to do
it, please let me know. This is what I do right now:
1) Hook up a filedesc_input_stream to DEVICE_INFD() of the device.
2) Make the filedescriptor non-blocking. I really hate doing this, but
there was no other way (that I could find) to make Lstream_read()
correctly bail out after reading all the ioctl-inspired data.
3) call { char c = 3; ioctl(fd, TIOCLINUX, &c); } - this actually copies
the pasted data into the virtual console's input buffer.
4) Read from the lstream until we stop getting data (<= 0 bytes read), and
write the data to a resizing_buffer_stream.
5) flush the buffer stream and make a string out of it
6) kill the streams
The reason I dislike #2 so much is that there is no
set_descriptor_blocking() function call, but in my testing under linux, it
has not proved fatal to leave the descriptor in non-blocking mode.
There was also a patch needed to mouse.el - it would just call (yank) if
you were not on a window system. I don't know how many other areas like
this there are lying around.
Unfortunately, I can't see an easy way to make the 'own_selection' method
work with the ioctl's that linux provides right now. The ioctl requires
the starting and ending positions on screen that you want to copy to the
cut buffer, which is nothing like the current XEmacs selection model.
With this latest patch, I can cut & paste from a virtual console to XEmacs,
and can also select & paste from within emacs. The selection is a little
weird, since you don't see the highlighting while the mouse is dragging,
but the old GPM code behaved in the same way.
The mouse-cursur flickering is still there as well. Anybody have a clue on
how to make this better? I could provide a lisp-function to call
GPM_DRAWPOINTER and push a magic eval even somewhere, but that seems kind
of icky, and I don't think there are any ways to do this type of event
mucking outside of event-Xt.c or event-unixoid.c - everything in there is
static.
I'm not sending this to xemacs-patches until a few more people can try it
out. I'd love to get the flicker problem fixed sometime tonight, so if
anybody has any ideas, lemme know.
-Bill P.
Index: lisp/mouse.el
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/lisp/mouse.el,v
retrieving revision 1.19.2.6
diff -u -r1.19.2.6 mouse.el
--- mouse.el 1999/09/02 08:47:55 1.19.2.6
+++ mouse.el 1999/11/05 15:03:12
＠＠ -88,7 +88,9 ＠＠
the X cutbuffer. A mark is pushed, so that the inserted text lies
between point and mark."
(interactive)
- (if (not (console-on-window-system-p))
+ (if (and (not (console-on-window-system-p))
+ (and (featurep 'gpm)
+ (not gpm-minor-mode)))
(yank)
(push-mark)
(if (region-active-p)
Index: src/console-tty.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/console-tty.c,v
retrieving revision 1.18.2.5
diff -u -r1.18.2.5 console-tty.c
--- console-tty.c 1999/10/24 03:48:31 1.18.2.5
+++ console-tty.c 1999/11/05 15:03:15
＠＠ -38,9 +38,6 ＠＠
#ifdef FILE_CODING
#include "file-coding.h"
#endif
-#ifdef HAVE_GPM
-#include "gpmevent.h"
-#endif
DEFINE_CONSOLE_TYPE (tty);
DECLARE_IMAGE_INSTANTIATOR_FORMAT (nothing);
＠＠ -129,10 +126,6 ＠＠
#endif /* MULE */
tty_con->terminal_type = terminal_type;
tty_con->controlling_process = controlling_process;
-
-#ifdef HAVE_GPM
- connect_to_gpm (con);
-#endif
if (NILP (CONSOLE_NAME (con)))
CONSOLE_NAME (con) = Ffile_name_nondirectory (tty);
Index: src/console-tty.h
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/console-tty.h,v
retrieving revision 1.8.2.2
diff -u -r1.8.2.2 console-tty.h
--- console-tty.h 1999/02/17 20:03:57 1.8.2.2
+++ console-tty.h 1999/11/05 15:03:15
＠＠ -41,9 +41,6 ＠＠
struct tty_console
{
int infd, outfd;
-#ifdef HAVE_GPM
- int mouse_fd;
-#endif
Lisp_Object instream, outstream;
Lisp_Object terminal_type;
Lisp_Object controlling_process;
＠＠ -203,9 +200,6 ＠＠
unsigned int is_stdio :1;
};
-#ifdef HAVE_GPM
-#define CONSOLE_TTY_MOUSE_FD(c) (CONSOLE_TTY_DATA (c)->mouse_fd)
-#endif
#define CONSOLE_TTY_DATA(c) CONSOLE_TYPE_DATA (c, tty)
#define CONSOLE_TTY_CURSOR_X(c) (CONSOLE_TTY_DATA (c)->cursor_x)
#define CONSOLE_TTY_CURSOR_Y(c) (CONSOLE_TTY_DATA (c)->cursor_y)
Index: src/device-tty.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/device-tty.c,v
retrieving revision 1.9
diff -u -r1.9 device-tty.c
--- device-tty.c 1998/05/27 20:42:00 1.9
+++ device-tty.c 1999/11/05 15:03:15
＠＠ -38,10 +38,6 ＠＠
#include "syssignal.h" /* for SIGWINCH */
-#ifdef HAVE_GPM
-#include <gpm.h>
-#endif
-
#include <errno.h>
Lisp_Object Qinit_pre_tty_win, Qinit_post_tty_win;
＠＠ -155,15 +151,6 ＠＠
CONSOLE_TTY_DATA (con)->width = width;
CONSOLE_TTY_DATA (con)->height = height;
-#ifdef HAVE_GPM
- /* We need to tell GPM how big our screen is now
- ** I am pretty sure the GPM library will get incredibly confused
- ** if you try to connect to more than one mouse-capable device,
- ** so I don't think it will cause any more damage in that case.
- */
- gpm_mx = width;
- gpm_my = height;
-#endif
for (tail = DEVICE_FRAME_LIST (d);
!NILP (tail);
tail = XCDR (tail))
Index: src/emacs.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/emacs.c,v
retrieving revision 1.82.2.32
diff -u -r1.82.2.32 emacs.c
--- emacs.c 1999/11/02 02:41:45 1.82.2.32
+++ emacs.c 1999/11/05 15:03:23
＠＠ -1088,6 +1088,10 ＠＠
syms_of_eldap ();
#endif
+#ifdef HAVE_GPM
+ syms_of_gpmevent ();
+#endif
+
/* Now create the subtypes for the types that have them.
We do this before the vars_*() because more symbols
may get initialized here. */
＠＠ -1481,6 +1485,10 ＠＠
#ifdef HAVE_LDAP
vars_of_eldap ();
+#endif
+
+#ifdef HAVE_GPM
+ vars_of_gpmevent ();
#endif
/* Now initialize any specifier variables. We do this later
Index: src/event-Xt.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/event-Xt.c,v
retrieving revision 1.41.2.14
diff -u -r1.41.2.14 event-Xt.c
--- event-Xt.c 1999/10/10 11:45:15 1.41.2.14
+++ event-Xt.c 1999/11/05 15:03:30
＠＠ -2034,9 +2034,6 ＠＠
{
Lisp_Object console;
int infd;
-#ifdef HAVE_GPM
- int mousefd;
-#endif
if (CONSOLE_X_P (con))
return; /* X consoles are automatically selected for when we
＠＠ -2044,22 +2041,6 ＠＠
infd = event_stream_unixoid_select_console (con);
XSETCONSOLE (console, con);
select_filedesc (infd, console);
-#ifdef HAVE_GPM
- /* On a stream device (ie: noninteractive), bad things can happen. */
- if (EQ (CONSOLE_TYPE (con), Qtty)) {
- mousefd = CONSOLE_TTY_MOUSE_FD (con);
- /* We check filedesc_to_what_closure[fd] here because if you run
- ** XEmacs from a TTY, it will fire up GPM, select the mouse fd, then
- ** if you run gnuattach to connect to another TTY, it will fire up
- ** GPM again, and try to reselect the mouse fd. GPM uses the same
- ** fd for every connection apparently, and select_filedesc will
- ** fail its assertion if we try to select it twice.
- */
- if ((mousefd >= 0) && !filedesc_to_what_closure[mousefd]) {
- select_filedesc (mousefd, console);
- }
- }
-#endif
}
static void
＠＠ -2067,9 +2048,6 ＠＠
{
Lisp_Object console;
int infd;
-#ifdef HAVE_GPM
- int mousefd;
-#endif
if (CONSOLE_X_P (con))
return; /* X consoles are automatically selected for when we
＠＠ -2077,15 +2055,6 ＠＠
infd = event_stream_unixoid_unselect_console (con);
XSETCONSOLE (console, con);
unselect_filedesc (infd);
-#ifdef HAVE_GPM
- /* On a stream device (ie: noninteractive), bad things can happen. */
- if (EQ (CONSOLE_TYPE (con), Qtty)) {
- mousefd = CONSOLE_TTY_MOUSE_FD (con);
- if (mousefd >= 0) {
- unselect_filedesc (mousefd);
- }
- }
-#endif
}
/* read an event from a tty, if one is available. Returns non-zero
Index: src/event-unixoid.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/event-unixoid.c,v
retrieving revision 1.9.2.1
diff -u -r1.9.2.1 event-unixoid.c
--- event-unixoid.c 1999/03/07 15:25:17 1.9.2.1
+++ event-unixoid.c 1999/11/05 15:03:30
＠＠ -41,10 +41,6 ＠＠
#include "sysproc.h" /* select stuff */
#include "systime.h"
-#ifdef HAVE_GPM
-#include "gpmevent.h"
-#endif
-
/* Mask of bits indicating the descriptors that we wait for input on.
These work as follows:
＠＠ -83,12 +79,6 ＠＠
Lisp_Object console;
XSETCONSOLE (console, con);
-
-#ifdef HAVE_GPM
- if (fd == CONSOLE_TTY_MOUSE_FD (con)) {
- return handle_gpm_read (event,con,fd);
- }
-#endif
nread = read (fd, &ch, 1);
if (nread <= 0)
Index: src/frame-tty.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/frame-tty.c,v
retrieving revision 1.8.2.1
diff -u -r1.8.2.1 frame-tty.c
--- frame-tty.c 1998/12/05 16:55:48 1.8.2.1
+++ frame-tty.c 1999/11/05 15:03:31
＠＠ -33,10 +33,6 ＠＠
#include "events.h"
-#ifdef HAVE_GPM
-#include <gpm.h>
-#endif
-
/* Default properties to use when creating frames. */
Lisp_Object Vdefault_tty_frame_plist;
＠＠ -86,33 +82,6 ＠＠
call1 (Qinit_post_tty_win, FRAME_CONSOLE (f));
}
-#ifdef HAVE_GPM
-static int
-tty_get_mouse_position (struct device *d, Lisp_Object *frame, int *x, int *y)
-{
- Gpm_Event ev;
- int num_buttons;
-
- num_buttons = Gpm_GetSnapshot(&ev);
- *x = ev.x;
- *y = ev.y;
- *frame = DEVICE_SELECTED_FRAME (d);
- return (1);
-}
-
-static void
-tty_set_mouse_position (struct window *w, int x, int y)
-{
- /* XXX
- I couldn't find any GPM functions that set the mouse position.
- Mr. Perry had left this function empty; that must be why.
- karlheg
- */
-}
-
-#endif
-
-
/* Change from withdrawn state to mapped state. */
static void
tty_make_frame_visible (struct frame *f)
＠＠ -223,10 +192,6 ＠＠
CONSOLE_HAS_METHOD (tty, init_frame_1);
CONSOLE_HAS_METHOD (tty, init_frame_3);
CONSOLE_HAS_METHOD (tty, after_init_frame);
-#ifdef HAVE_GPM
- CONSOLE_HAS_METHOD (tty, get_mouse_position);
- CONSOLE_HAS_METHOD (tty, set_mouse_position);
-#endif
CONSOLE_HAS_METHOD (tty, make_frame_visible);
CONSOLE_HAS_METHOD (tty, make_frame_invisible);
CONSOLE_HAS_METHOD (tty, frame_visible_p);
Index: src/gpmevent.c
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/gpmevent.c,v
retrieving revision 1.5
diff -u -r1.5 gpmevent.c
--- gpmevent.c 1998/03/31 20:11:47 1.5
+++ gpmevent.c 1999/11/05 15:03:31
＠＠ -1,5 +1,28 ＠＠
-/* William Perry 1997 */
+/* GPM functions
+ Copyright (C) 1997 William M. Perry <wmperry(a)gnu.org>
+ Copyright (C) 1999 Free Software Foundation, Inc.
+
+This file is part of XEmacs.
+
+XEmacs is free software; you can redistribute it and/or modify it
+under the terms of the GNU General Public License as published by the
+Free Software Foundation; either version 2, or (at your option) any
+later version.
+
+XEmacs is distributed in the hope that it will be useful, but WITHOUT
+ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+for more details.
+
+You should have received a copy of the GNU General Public License
+along with XEmacs; see the file COPYING. If not, write to
+the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+Boston, MA 02111-1307, USA. */
+/* Synched up with: Not in FSF. */
+
+/* Authors: William Perry */
+
#include <config.h>
#include "lisp.h"
#include "console.h"
＠＠ -8,11 +31,17 ＠＠
#include "events.h"
#include "events-mod.h"
#include "sysdep.h"
+#include "commands.h"
+#include "lstream.h"
#ifdef HAVE_GPM
#include "gpmevent.h"
#include <gpm.h>
+/* This is commented out in commands.h, I'm not sure why */
+extern Lisp_Object Vunread_command_events;
+extern Lisp_Object Vunread_command_event;
+
#if (!defined(__linux__)) /* possible under xterm */
#define KG_SHIFT 0
#define KG_CTRL 2
＠＠ -21,94 +50,383 ＠＠
#include <linux/keyboard.h>
#endif
-int
-handle_gpm_read (struct Lisp_Event *event, struct console *con, int fd)
+Lisp_Object Vgpm_minor_mode;
+
+DEFUN ("receive-gpm-event", Freceive_gpm_event, 0, 2, 0, /*
+Run GPM_GetEvent().
+This function is the process handler for the GPM connection.
+*/
+ (ignore1, ignore2))
{
- Gpm_Event ev;
- int modifiers = 0;
- int type = -1;
- int button = 1;
-
- if (!Gpm_GetEvent(&ev))
- return 0;
-
- event->timestamp = 0;
- event->channel = CONSOLE_SELECTED_FRAME (con);
-
- /* Whow, wouldn't named defines be NICE!?!?! */
- modifiers = 0;
-
- if (ev.modifiers & 1) modifiers |= MOD_SHIFT;
- if (ev.modifiers & 2) modifiers |= MOD_META;
- if (ev.modifiers & 4) modifiers |= MOD_CONTROL;
- if (ev.modifiers & 8) modifiers |= MOD_META;
-
- if (ev.type & GPM_DOWN)
- type = GPM_DOWN;
- else if (ev.type & GPM_UP)
- type = GPM_UP;
- else if (ev.type & GPM_MOVE) {
- type = GPM_MOVE;
- GPM_DRAWPOINTER(&ev);
- }
+ Gpm_Event ev;
+ int modifiers = 0;
+ int type = -1;
+ int button = 1;
+ Lisp_Object fake_event;
+ struct Lisp_Event *event = NULL;
+ struct gcpro gcpro1;
+ static int num_events;
+
+ if (!Gpm_GetEvent(&ev))
+ {
+ warn_when_safe (Qnil, Qcritical, "Gpm_GetEvent failed");
+ return(Qzero);
+ }
+
+ GCPRO1(fake_event);
+
+ num_events++;
+
+ fake_event = Fmake_event (Qnil, Qnil);
+ event = XEVENT(fake_event);
+
+ event->timestamp = 0;
+ event->channel = Fselected_frame (Qnil); /* CONSOLE_SELECTED_FRAME (con); */
+
+ /* Whow, wouldn't named defines be NICE!?!?! */
+ modifiers = 0;
+
+ if (ev.modifiers & 1) modifiers |= MOD_SHIFT;
+ if (ev.modifiers & 2) modifiers |= MOD_META;
+ if (ev.modifiers & 4) modifiers |= MOD_CONTROL;
+ if (ev.modifiers & 8) modifiers |= MOD_META;
+
+ if (ev.type & GPM_DOWN)
+ {
+ type = GPM_DOWN;
+ }
+ else if (ev.type & GPM_UP)
+ {
+ type = GPM_UP;
+ }
+ else if (ev.type & GPM_MOVE)
+ {
+ type = GPM_MOVE;
+ }
+
+ if (ev.buttons & GPM_B_LEFT)
+ {
+ button = 1;
+ }
+ else if (ev.buttons & GPM_B_MIDDLE)
+ {
+ button = 2;
+ }
+ else if (ev.buttons & GPM_B_RIGHT)
+ {
+ button = 3;
+ }
+
+ switch (type) {
+ case GPM_DOWN:
+ case GPM_UP:
+ event->event_type =
+ type == GPM_DOWN ? button_press_event : button_release_event;
+ event->event.button.x = ev.x;
+ event->event.button.y = ev.y;
+ event->event.button.button = button;
+ event->event.button.modifiers = modifiers;
+ break;
+ case GPM_MOVE:
+ event->event_type = pointer_motion_event;
+ event->event.motion.x = ev.x;
+ event->event.motion.y = ev.y;
+ event->event.motion.modifiers = modifiers;
+ default:
+ /* This will never happen */
+ break;
+ }
+
+ /* Handle the event */
+ /* Let's try sticking it on the unread events queue */
+ if (!command_event_p (fake_event))
+ {
+ /* unread_command_events does not like motion events */
+ Fdispatch_event (fake_event);
+ }
+ else if (NILP (Vunread_command_events))
+ {
+ Vunread_command_events = Fcons (fake_event, Qnil);
+ }
+ else
+ {
+ Lisp_Object arg[2];
+ arg[0] = Vunread_command_events;
+ arg[1] = fake_event;
- if (ev.buttons & GPM_B_LEFT)
- button = 1;
- else if (ev.buttons & GPM_B_MIDDLE)
- button = 2;
- else if (ev.buttons & GPM_B_RIGHT)
- button = 3;
-
- switch (type) {
- case GPM_DOWN:
- case GPM_UP:
- event->event_type =
- type == GPM_DOWN ? button_press_event : button_release_event;
- event->event.button.x = ev.x;
- event->event.button.y = ev.y;
- event->event.button.button = button;
- event->event.button.modifiers = modifiers;
- break;
- case GPM_MOVE:
- event->event_type = pointer_motion_event;
- event->event.motion.x = ev.x;
- event->event.motion.y = ev.y;
- event->event.motion.modifiers = modifiers;
- default:
- return 0;
- }
- return 1;
+ Vunread_command_events = Fappend (2,arg);
+ }
+
+ GPM_DRAWPOINTER(&ev);
+
+ UNGCPRO;
+
+ return (Qzero);
}
-void
-connect_to_gpm (struct console *con)
+static void turn_off_gpm (void)
{
- /* Only do this if we are running after dumping and really interactive */
- if (!noninteractive && initialized) {
- /* We really only want to do this on a TTY */
- CONSOLE_TTY_MOUSE_FD (con) = -1;
- if (EQ (CONSOLE_TYPE (con), Qtty)) {
- Gpm_Connect conn;
- int rval;
-
- conn.eventMask = GPM_DOWN|GPM_UP|GPM_MOVE;
- conn.defaultMask = GPM_MOVE;
- conn.minMod = 0;
- conn.maxMod = ((1<<KG_SHIFT)|(1<<KG_ALT)|(1<<KG_CTRL));
-
- rval = Gpm_Open (&conn, 0);
- switch (rval) {
- case -1: /* General failure */
- break;
- case -2: /* We are running under an XTerm */
+ Fdelete_process (build_string ("gpm"));
Gpm_Close();
- break;
- default:
- set_descriptor_non_blocking (gpm_fd);
- CONSOLE_TTY_MOUSE_FD (con) = gpm_fd;
- }
- }
+ Vgpm_minor_mode = Qnil;
+
+ /* Remove the TTY mouse methods */
+ /* This is breaking abstraction a bit, but... */
+ tty_console_methods->get_mouse_position_method = NULL;
+ tty_console_methods->set_mouse_position_method = NULL;
+ tty_console_methods->own_selection_method = NULL;
+ tty_console_methods->get_foreign_selection_method = NULL;
+}
+
+#ifdef TIOCLINUX
+static Lisp_Object
+tty_get_foreign_selection (Lisp_Object selection_symbol, Lisp_Object target_type)
+{
+ /* This function can GC */
+ struct device *d = decode_device (Qnil);
+ int fd = DEVICE_INFD (d);
+ char c = 3;
+ Lisp_Object output_stream;
+ Lisp_Object terminal_stream;
+ Lisp_Object output_string;
+ struct gcpro gcpro1,gcpro2,gcpro3;
+
+ GCPRO3(output_stream,terminal_stream,output_string);
+
+ /* The ioctl() to paste actually puts things in the input queue of
+ ** the virtual console, so we need to trap that data, since we are
+ ** supposed to return the actual string selection from this
+ ** function.
+ */
+
+ /* I really hate doing this, but it doesn't seem to cause any
+ ** problems, and it makes the Lstream_read stuff further down
+ ** error out correctly instead of trying to indefinitely read from
+ ** the console.
+ **
+ ** There is no set_descriptor_blocking() function call, but in my
+ ** testing under linux, it has not proved fatal to leave the
+ ** descriptor in non-blocking mode.
+ **
+ ** William Perry Nov 5, 1999
+ */
+ set_descriptor_non_blocking (fd);
+
+ /* We need two streams, one for reading from the selected device,
+ ** and one to write the data into. There is no writable version
+ ** of the lisp-string lstream, so we make do with a resizing
+ ** buffer stream, and make a string out of it after we are
+ ** done.
+ */
+ output_stream = make_resizing_buffer_output_stream ();
+ terminal_stream = make_filedesc_input_stream (fd, 0, -1, LSTR_BLOCKED_OK);
+ output_string = Qnil;
+
+ /* We should arguably use a specbind() and an unwind routine here,
+ ** but I don't care that much right now.
+ ** ### FIXME?
+ */
+ if (NILP (output_stream) || NILP (terminal_stream))
+ {
+ /* Should we signal an error here? */
+ goto out;
+ }
+
+ if (ioctl (fd, TIOCLINUX, &c) < 0)
+ {
+ /* Could not get the selection - eek */
+ UNGCPRO;
+ return (Qnil);
+ }
+
+ while (1)
+ {
+ Bufbyte tempbuf[1024]; /* some random amount */
+ int size_in_bytes = Lstream_read (XLSTREAM (terminal_stream),
+ tempbuf, sizeof (tempbuf));
+
+ if (size_in_bytes <= 0)
+ {
+ /* end of the stream */
+ break;
+ }
+
+ Lstream_write (XLSTREAM (output_stream), tempbuf, size_in_bytes);
+ }
+
+ Lstream_flush (XLSTREAM (output_stream));
+
+ output_string = make_string (resizing_buffer_stream_ptr (XLSTREAM (output_stream)),
+ Lstream_byte_count (XLSTREAM (output_stream)));
+
+ Lstream_delete (XLSTREAM (output_stream));
+ Lstream_delete (XLSTREAM (terminal_stream));
+
+ out:
+ UNGCPRO;
+ return (output_string);
+}
+
+static Lisp_Object
+tty_selection_exists_p (Lisp_Object selection)
+{
+ return (Qt);
+}
+#endif /* TIOCLINUX */
+
+#if 0
+static Lisp_Object
+tty_own_selection (Lisp_Object selection_name, Lisp_Object selection_value)
+{
+ /* There is no way to do this cleanly - the GPM selection
+ ** 'protocol' (actually the TIOCLINUX ioctl) requires a start and
+ ** end position on the _screen_, not a string to stick in there.
+ ** Lame.
+ **
+ ** William Perry Nov 4, 1999 */
+}
+#endif
+
+/* This function appears to work once in a blue moon. I'm not sure
+** exactly why either. *sigh*
+** William Perry Nov 4, 1999
+*/
+static int
+tty_get_mouse_position (struct device *d, Lisp_Object *frame, int *x, int *y)
+{
+ Gpm_Event ev;
+ int num_buttons;
+
+ memset(&ev,'\0',sizeof(ev));
+ num_buttons = Gpm_GetSnapshot(&ev);
+
+ if (!num_buttons)
+ {
+ /* This means there are events pending... */
+ return(-1);
}
+ *x = ev.x;
+ *y = ev.y;
+ *frame = DEVICE_SELECTED_FRAME (d);
+ return (1);
}
+static void
+tty_set_mouse_position (struct window *w, int x, int y)
+{
+ /* XXX
+ I couldn't find any GPM functions that set the mouse position.
+ Mr. Perry had left this function empty; that must be why.
+ karlheg
+ */
+}
+
+DEFUN ("gpm-minor-mode", Fgpm_minor_mode, 0, 1, 0, /*
+Start accepting GPM mouse events iff ARG is positive.
+*/
+ (arg))
+{
+ Gpm_Connect conn;
+ int rval;
+ Lisp_Object gpm_process;
+ Lisp_Object gpm_filter;
+
+ if (NILP (arg))
+ {
+ /* Toggle current GPM mode */
+ if (!NILP (Vgpm_minor_mode))
+ {
+ /* Need to turn it off */
+ turn_off_gpm ();
+ return (Qnil);
+ }
+ }
+ else
+ {
+ int i;
+
+ CHECK_INT (arg);
+
+ i = XINT (arg);
+
+ if (i < 0)
+ {
+ turn_off_gpm ();
+ return (Qnil);
+ }
+ }
+
+ if (!NILP (Vgpm_minor_mode))
+ {
+ error ("Already connected to GPM");
+ }
+
+ if (noninteractive)
+ {
+ error ("Can't connect to GPM in batch mode.");
+ }
+
+ /* Install the mouse position methods for the TTY console type */
+ CONSOLE_HAS_METHOD (tty, get_mouse_position);
+ CONSOLE_HAS_METHOD (tty, set_mouse_position);
+
+#ifdef TIOCLINUX
+ CONSOLE_HAS_METHOD (tty, get_foreign_selection);
+ CONSOLE_HAS_METHOD (tty, selection_exists_p);
+#if 0
+ CONSOLE_HAS_METHOD (tty, own_selection);
#endif
+#endif /* TIOCLINUX */
+
+ conn.eventMask = GPM_DOWN|GPM_UP|GPM_MOVE;
+ conn.defaultMask = GPM_MOVE;
+ conn.minMod = 0;
+ conn.maxMod = ((1<<KG_SHIFT)|(1<<KG_ALT)|(1<<KG_CTRL));
+
+ rval = Gpm_Open (&conn, 0);
+
+ switch (rval) {
+ case -1: /* General failure */
+ break;
+ case -2: /* We are running under an XTerm */
+ Gpm_Close();
+ break;
+ default:
+ set_descriptor_non_blocking (gpm_fd);
+ Vgpm_minor_mode = Qt;
+ gpm_process = connect_to_file_descriptor (build_string ("gpm"), Qnil,
+ make_int (gpm_fd),
+ make_int (gpm_fd));
+
+ if (!NILP (gpm_process))
+ {
+ Fprocess_kill_without_query (gpm_process, Qnil);
+ XSETSUBR (gpm_filter, &SFreceive_gpm_event);
+ set_process_filter (gpm_process, gpm_filter, 1);
+ }
+ else
+ {
+ Gpm_Close();
+ Vgpm_minor_mode = Qnil;
+ }
+ }
+ return(Vgpm_minor_mode);
+}
+
+void vars_of_gpmevent (void)
+{
+ Fprovide (intern ("gpm"));
+
+ DEFVAR_LISP ("gpm-minor-mode", &Vgpm_minor_mode /*
+Non-nil when gpm-minor-mode is active. Call `gpm-mode' to toggle.
+*/ );
+ Vgpm_minor_mode = Qnil;
+}
+
+void syms_of_gpmevent (void)
+{
+ DEFSUBR (Freceive_gpm_event);
+ DEFSUBR (Fgpm_minor_mode);
+}
+
+#endif /* HAVE_GPM */
Index: src/gpmevent.h
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/gpmevent.h,v
retrieving revision 1.3
diff -u -r1.3 gpmevent.h
--- gpmevent.h 1998/03/31 20:11:47 1.3
+++ gpmevent.h 1999/11/05 15:03:31
＠＠ -1,7 +0,0 ＠＠
-#ifndef _HAVE_GPM
-#define _HAVE_GPM
-
-int handle_gpm_read (struct Lisp_Event *event, struct console *con, int fd);
-void connect_to_gpm (struct console *con);
-
-#endif
Index: src/symsinit.h
===================================================================
RCS file: /usr/CVSroot/XEmacs/xemacs/src/symsinit.h,v
retrieving revision 1.31.2.13
diff -u -r1.31.2.13 symsinit.h
--- symsinit.h 1999/10/30 03:00:51 1.31.2.13
+++ symsinit.h 1999/11/05 15:03:32
＠＠ -148,6 +148,7 ＠＠
void syms_of_window (void);
void syms_of_xselect (void);
void syms_of_eldap (void);
+void syms_of_gpmevent (void);
/* Initialize the console types (dump-time but for reinit_). */
＠＠ -368,6 +369,7 ＠＠
void vars_of_xselect (void);
void reinit_vars_of_xselect (void);
void vars_of_eldap (void);
+void vars_of_gpmevent (void);
/* Initialize specifier variables (dump-time only). */
- 1 participants
- 0 comments
Re: Reworked GPM support...
26 years, 1 month
William M. Perry
William M. Perry
Jan Vroonhof <vroonhof(a)math.ethz.ch> writes:
> wmperry(a)aventail.com (William M. Perry) writes:
>
> > Here is a reworked version of the GPM support. This is equivalent
> > functionality-wise to the current implementation, but does not sink its
> > teeth into all the console and event loops like the old one.
>
> Cool... The patch is mangled however. It has broken lines and there is
> the tab vs space problem.
Tabs are your friend. :)
> > -----------------
> > `gpm-mode' is a built-in function
> > (gpm-mode &optional ARG)
>
> How do you suppose this gets enabled? I would love for this to be still
> the default.
We could throw a call to it in syms_of_gpmevent() if you want.
> > - if ((mousefd >= 0) && !filedesc_to_what_closure[mousefd]) {
> > - select_filedesc (mousefd, console);
> > }
> > - }
> > -#endif
> > -}
>
> You really don't want to delete that brace.
But there's an extra one from up above - everything balances out, it just
looks a little odd. :)
> However if you want to progress from mere coolness to utter
> enlightenment then you should
>
> 0. make sure the case when we have frames on more than one VC works
> 1. make sure the xterm mouse stuff works.
There is no way to get this working using GPM, because it requires you to
use a GPM function to read characters from the terminal, and it deals with
cooking them into GPM events or really passing them back to you.
> 2. [Bonus points] include the xterm only file that libgpm supports in
> the source so that if you are not on Linux you still get xterm
> mouse support.
Now that Hrvoje made make-event actually useful, we could port xt-mouse.el
from Emacs. It works off of function-key-map, and in trivial testing right
now, it mostly works. Just a few functions missing in XEmacs (window-at,
coordinates-in-window-p), but that should be easy to work around.
> 3. Make XEmacs support the GPM selection protocol.
Where is this documented? I couldn't find anything in the GPM headers or
documentation.
-Bill P.
- 2 participants
- 2 comments