Problem with Emacs 21.3 pretest

Problem with Emacs 21.3 pretest

[Ehud Karni]

I have a strange problem with Emacs 21.3.
It works as expected on tty (emacs -nw) but fails on X and I can't 
debug it because I don't get any real information.

Here is the command line and its output:
 ehud@beta[pts/1]-~-1520% emacs-21.2.90.1 -q --no-site-file --debug-init
Attempt to modify read-only object
Exit 255 <---- this is reported by tcsh

The emacs-version is:
GNU Emacs 21.2.90.1 (i686-redhat-linux-gnu, X toolkit, Xaw3d scroll bars) of 2002-05-27 on beta

An important note: This is a dumped emacs which contains many elisp
packages (1122604 pure bytes used).

Ehud.


--
 Ehud Karni           Tel: +972-3-7966-561  /"\
 Mivtach - Simon      Fax: +972-3-7966-667  \ /  ASCII Ribbon Campaign
 Insurance agencies   (USA) voice mail and   X   Against   HTML   Mail
 http://www.mvs.co.il  FAX:  1-815-5509341  / \
 mailto:ehud@unix.simonwiesel.co.il          Better  Safe  Than  Sorry

Re: Problem with Emacs 21.3 pretest

[Richard Stallman]

    Here is the command line and its output:
     ehud@beta[pts/1]-~-1520% emacs-21.2.90.1 -q --no-site-file --debug-init
    Attempt to modify read-only object
    Exit 255 <---- this is reported by tcsh

Please run it under GDB with a breakpoint at pure_write_error
and send the backtrace.

Re: Problem with Emacs 21.3 pretest

[Ehud Karni]

On Mon, 27 May 2002 23:16:19 -0600 (MDT), Richard Stallman <rms@gnu.org> wrote:
>
>      ehud@beta[pts/1]-~-1520% emacs-21.2.90.1 -q --no-site-file --debug-init
>     Attempt to modify read-only object
>
> Please run it under GDB with a breakpoint at pure_write_error
> and send the backtrace.

OK. The trace allowed me to solve this error.

    (gdb) break pure_write_error
    Breakpoint 1 at 0x8113282
    (gdb) run
    Starting program: /usr/local/bin/emacs-21.3

    Breakpoint 1, 0x08113282 in pure_write_error ()
    (gdb) where
    #0  0x08113282 in pure_write_error ()
    #1  0x080d7b26 in parse_menu_item ()
    #2  0x08071b2f in single_menu_item ()
    #3  0x08071a44 in single_keymap_panes ()
    #4  0x08071c1e in single_menu_item ()
    #5  0x08071a44 in single_keymap_panes ()
    #6  0x08072d54 in single_submenu ()


The reason was these definitions (note the `defconst' of the keymap list):

(defconst EK-menu-mark-3
  '(keymap "EK Marks whole ops"
       (copy-mk menu-item " copy  marked "         copy-marked         :keys "F3"   )
       (move-mk menu-item " move  marked "         move-marked         :keys "S-F3" )
       (overlay menu-item "overlay marked"         overlay-marked      :keys "Ctrl-S-F9"   :enable (= type-mark-set 4))
       (dlt-mrk menu-item "delete  marked"         delete-marked       :keys "Ctrl-F9" )
       (-1 "--")
       (fill-mk menu-item "fill with char"         fill-marked         :keys "Ctrl-F8" )
       (shift-r menu-item "shift right marked"     shift-marked        :keys "Alt-S-+" :enable (/= type-mark-set 2))
       (shift-l menu-item "shift left marked "     shift-marked-left   :keys "Alt-S--" :enable (/= type-mark-set 2))
       (reflowm menu-item "reformat  marked"       reflow-marked       :keys "C-x R" )
   )
  "EK marks sub 3 - whole ops")

(defconst EK-ops-menu
  (list 'keymap "EK ops"
        (list 'sav 'menu-item "saves/grep/tabs/shell"  EK-menu-save)
        (list 'lng 'menu-item "Languge/direction/mode" EK-menu-lang)
        '(-1  "--")
        '(-2  "--")
        (list 'sm1 'menu-item "EK Marks un/marking"    EK-menu-mark-1)
        (list 'sm2 'menu-item "EK Marks push/pop"      EK-menu-mark-2)
        '(-3        menu-item "--"                     nil             :visible  (mark-exist-chk))
        (list 'sm3 'menu-item "Whole marked area ops"  EK-menu-mark-3 ':visible '(mark-exist-chk))
        (list 'sm4 'menu-item "Translate in marked"    EK-menu-mark-4 ':visible '(mark-exist-chk))
        (list 'sm5 'menu-item "Marked area lines ops"  EK-menu-mark-5 ':visible '(mark-exist-chk))
        (list 'sm6 'menu-item "marked area as input"   EK-menu-mark-6 ':visible '(mark-exist-chk))
        '(-4  "--")
        '(compute  menu-item "compute expression"      compute     :keys "Alt-C" )
        '(cmp-wnd  menu-item "compare next window"     compare-windows )
        '(fix-col  menu-item "Fixed column movment"    col-move-fix-toggle :keys "Alt-X" :button (:toggle . (not position-col-var)))
  )
  "EK marks menu")

(define-key global-map [menu-bar EK_ops] (list 'menu-item "EK ops" EK-ops-menu))

When I changed all the `defconst' to `defvar' this problem disappeared.
I still wonder why does `parse_menu_item' change the values of my vars (constants).

I have other problems, I'm working on it.

Ehud.


--
 Ehud Karni           Tel: +972-3-7966-561  /"\
 Mivtach - Simon      Fax: +972-3-7966-667  \ /  ASCII Ribbon Campaign
 Insurance agencies   (USA) voice mail and   X   Against   HTML   Mail
 http://www.mvs.co.il  FAX:  1-815-5509341  / \
 mailto:ehud@unix.simonwiesel.co.il          Better  Safe  Than  Sorry

Re: Problem with Emacs 21.3 pretest

[Richard Stallman]

    When I changed all the `defconst' to `defvar' this problem disappeared.
    I still wonder why does `parse_menu_item' change the values of my vars (constants).

The change is due to caching the keyboard equivalents of these menu
items.  That is done so menu redisplay can be fast.

Re: Problem with Emacs 21.3 pretest

[Stefan Monnier]

>     When I changed all the `defconst' to `defvar' this problem disappeared.
>     I still wonder why does `parse_menu_item' change the values of my vars (constants).
> 
> The change is due to caching the keyboard equivalents of these menu
> items.  That is done so menu redisplay can be fast.

Note that I'm less and less sure that this caching is still important.
I have turned it off here (or rather I have changed the code to always
refresh the cache) and it's not noticeably slower (on my 266Mhz PII).

The worst case is with X-Symbol which has an enormous keymap (because
of two input methods implemented as keymaps: one in a menu and another
bound to C-=) where the menu takes about 1s to come up.


	Stefan

Re: Problem with Emacs 21.3 pretest

[Miles Bader]

"Stefan Monnier" <monnier+gnu/emacs@RUM.cs.yale.edu> writes:
> Note that I'm less and less sure that this caching is still important.
> I have turned it off here (or rather I have changed the code to always
> refresh the cache) and it's not noticeably slower (on my 266Mhz PII).

I agree that the `menu-item munging' is bad, it often seems to be the
cause of wierd `bugs'.

Couldn't we just use a (weak) hash table to do the caching, anyway?
Then it could still cache, but it wouldn't munge the input.

-Miles
-- 
Saa, shall we dance?  (from a dance-class advertisement)

Re: Problem with Emacs 21.3 pretest

[Richard Stallman]

    The worst case is with X-Symbol which has an enormous keymap (because
    of two input methods implemented as keymaps: one in a menu and another
    bound to C-=) where the menu takes about 1s to come up.

I would say that is serious.  Users expect menus to appear very quickly.

Re: Problem with Emacs 21.3 pretest

[Richard Stallman]

    Couldn't we just use a (weak) hash table to do the caching, anyway?

That idea seems worth trying.  (We did not have weak hash tables back
then.)  If it is fast enough (and I expect it is), it would be a big
increase in clarity and reliability.  Would you like to try
implementing it?