question about `quit-char'

question about `quit-char'

[Drew Adams]

A recent message (can't find it now) that mentioned that `quit-char' is not
necessarily `C-g' prompted me to look through the docs for info about
`quit-char'. 

I found it only in the Elisp manual, node Input Modes, where `quit-char' is used
to name the 4th element of the list returned by `current-input-mode' (it is also
mentioned wrt `set-input-mode'; `set-quit-char' is not mentioned).

There is no variable or function `quit-char'.

I wonder what the recommendation or the intention is about referencing the quit
character. I have code, for instance, that does this for Icicle mode, where
`map' is a minibuffer map:

(define-key map [(control ?g)] 'foo)

Should I instead be doing something like this?

(define-key map (vector (cadr (cddr (current-input-mode)))) 'foo)

That seems to work OK, at least when the quit character is `C-g'. Dunno if it is
robust. If so, is it preferable in general?

If so, shouldn't such a recommendation be documented? And in that case shouldn't
we perhaps define a function `quit-char' that returns (cadr (cddr
(current-input-mode)))?

Re: question about `quit-char'

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Date: Tue, 21 Oct 2008 09:57:01 -0700
> 
> A recent message (can't find it now) that mentioned that `quit-char' is not
> necessarily `C-g' prompted me to look through the docs for info about
> `quit-char'. 

That was a message from Andreas Schwab, in response to Glenn's
suggestion of how to fix the bug I reported with

   "C-x RET c utf-8 RET C-x C-g"

> I found it only in the Elisp manual, node Input Modes, where `quit-char' is used
> to name the 4th element of the list returned by `current-input-mode' (it is also
> mentioned wrt `set-input-mode'; `set-quit-char' is not mentioned).

The ELisp manual is not yet up-to-date with Emacs 23; set-quit-char is
what you want.  Note that, for the change of quit-char to really work,
you need at least to remap keyboard-quit to the new key.  And there
are other keybindings that need to be remapped, if one _really_ wants
to replace C-g with something else.

> There is no variable or function `quit-char'.

As you found out, current-input-mode returns, among other things, the
current value of quit-char.

Re: question about `quit-char'

[Chong Yidong]

"Drew Adams" <drew.adams@oracle.com> writes:

> A recent message (can't find it now) that mentioned that `quit-char'
> is not necessarily `C-g' prompted me to look through the docs for info
> about `quit-char'.
>
> I found it only in the Elisp manual, node Input Modes, where
> `quit-char' is used to name the 4th element of the list returned by
> `current-input-mode' (it is also mentioned wrt `set-input-mode';
> `set-quit-char' is not mentioned).
>
> There is no variable or function `quit-char'.

The Fset_quit_char command was introduced during the multi-tty merge.
Apparently, no one ever bothered to make sure that the rest of Emacs
works when quit_char is not C-g: there are many, many places in the Lisp
sources that assume C-g quits.

So I think this feature should be removed.

This will involve eliminating the 4th (QUIT) element in the return value
of current-input-mode, the 4th argument of set-input-mode, and the
function set-quit-char.  As far as I can tell, nothing in the rest of
Emacs uses these.

Does anyone object?

Re: question about `quit-char'

[Eli Zaretskii]

> From: Chong Yidong <cyd@stupidchicken.com>
> Date: Tue, 21 Oct 2008 13:21:49 -0400
> Cc: emacs-devel@gnu.org
> 
> The Fset_quit_char command was introduced during the multi-tty merge.

That's true, but one could set quit-char in older Emacsen as well: you
could do that with a call to set-input-mode.

> Apparently, no one ever bothered to make sure that the rest of Emacs
> works when quit_char is not C-g: there are many, many places in the Lisp
> sources that assume C-g quits.

This was always the case, so nothing changed lately.

> So I think this feature should be removed.
> 
> This will involve eliminating the 4th (QUIT) element in the return value
> of current-input-mode, the 4th argument of set-input-mode, and the
> function set-quit-char.  As far as I can tell, nothing in the rest of
> Emacs uses these.
> 
> Does anyone object?

I do.  I don't see why we should remove one of the oldest features in
Emacs, just because it's complicated to use it.  Certainly not now.

Re: question about `quit-char'

[Lennart Borgman (gmail)]

> The Fset_quit_char command was introduced during the multi-tty merge.
> Apparently, no one ever bothered to make sure that the rest of Emacs
> works when quit_char is not C-g: there are many, many places in the Lisp
> sources that assume C-g quits.
> 
> So I think this feature should be removed.
> Does anyone object?

Is there anything to win from removing it? Is not the changes that need
to be made to make it work simple?

I think it might be good to keep it if we want to make emulation where
another quit char might fit better.

Re: question about `quit-char'

[Chong Yidong]

Eli Zaretskii <eliz@gnu.org> writes:

>> Apparently, no one ever bothered to make sure that the rest of Emacs
>> works when quit_char is not C-g: there are many, many places in the Lisp
>> sources that assume C-g quits.
>
> This was always the case, so nothing changed lately.
>
>> So I think this feature should be removed.
>> 
>> This will involve eliminating the 4th (QUIT) element in the return value
>> of current-input-mode, the 4th argument of set-input-mode, and the
>> function set-quit-char.  As far as I can tell, nothing in the rest of
>> Emacs uses these.
>> 
>> Does anyone object?
>
> I do.  I don't see why we should remove one of the oldest features in
> Emacs, just because it's complicated to use it.  Certainly not now.

Okay.  I didn't realize that current-input-mode has been around for a
while.

Probably the right thing to do, at this point, is to simple assume that
\C-g is the same as quit-char.  So Glenn's patch in bug#1205 is OK.

After the release, we can add a `quit-char' function and change the Lisp
code to use it.

RE: question about `quit-char'

[Drew Adams]

> > I found it only in the Elisp manual, node Input Modes, 
> > where `quit-char' is used to name the 4th element of the
> > list returned by `current-input-mode' (it is also mentioned
> > wrt `set-input-mode'; `set-quit-char' is not mentioned).
> 
> The ELisp manual is not yet up-to-date with Emacs 23; set-quit-char is
> what you want.

Uh, why is `set-quit-char' what I want? I want to bind the quit character to a
particular command in a particular keymap.

And the doc string for `set-quit-char' suggests that it is only for ttys (and my
application is only for use with a window mgr):

 "This function only has an effect on the controlling
 tty of the Emacs process."

(Though I admit that that sentence is not very clear to me.)

> Note that, for the change of quit-char to really work,

I don't want to change the quit character. I want to bind it to a particular
command in a particular keymap (in a particular minor mode).

> you need at least to remap keyboard-quit to the new key.

No, my use case involves minibuffer keymaps, where `keyboard-quit' is not the
command bound to the quit char by default.

> And there are other keybindings that need to be remapped, if
> one _really_ wants to replace C-g with something else.

I don't want to replace C-g with anything else. I want to make sure that
whatever the quit char is it gets bound to my command.

> > There is no variable or function `quit-char'.
> 
> As you found out, current-input-mode returns, among other things, the
> current value of quit-char.

Yes, the question is whether there should be a function that returns just the
quit char.

RE: question about `quit-char'

[Drew Adams]

> >> So I think this feature should be removed....
> >
> > I do.  I don't see why we should remove one of the oldest 
> > features in Emacs, just because it's complicated to use it.
> > Certainly not now.
> 
> Okay.  I didn't realize that current-input-mode has been
> around for a while. Probably the right thing to do, at
> this point, is to simple assume that \C-g is the same as
> quit-char. So Glenn's patch in bug#1205 is OK.
> 
> After the release, we can add a `quit-char' function and 
> change the Lisp code to use it.

C'mon guys, you've changed the thread.

My question is about whether I should be binding the quit character or `C-g',
and if the former, what the best way to do that is. Is the way I mentioned a
good one?

Please see my original message for the questions I'd like answers to.

I am not trying to set the quit character to something other than C-g, and I am
not interested here about the Emacs 23 changes wrt the quit character. Please
start a different thread for such stuff.

Re: question about `quit-char'

[Chong Yidong]

"Drew Adams" <drew.adams@oracle.com> writes:

> C'mon guys, you've changed the thread.
>
> My question is about whether I should be binding the quit character or `C-g',
> and if the former, what the best way to do that is. Is the way I mentioned a
> good one?

Just bind to C-g:

  (define-key map "\C-g" 'foo-bar)

We do this all over the lisp/ tree anyway, and this isn't going to
change before the release.

Re: question about `quit-char'

[Lennart Borgman (gmail)]

Chong Yidong wrote:
> "Drew Adams" <drew.adams@oracle.com> writes:
> 
>> C'mon guys, you've changed the thread.
>>
>> My question is about whether I should be binding the quit character or `C-g',
>> and if the former, what the best way to do that is. Is the way I mentioned a
>> good one?
> 
> Just bind to C-g:
> 
>   (define-key map "\C-g" 'foo-bar)
> 
> We do this all over the lisp/ tree anyway, and this isn't going to
> change before the release.


Wouldn't it be better for Drew to write a function using his own recipe
and use that? Then it would still work if the whole quit-char thing will
be implemented later.

(The crucial difference relative to your view is that Drew distribute
his libraries separately from Emacs and I believe he wants them to work
in both new and old versions of Emacs. If not there will be a lot of
extra trouble for him and the users of his libraries.)

RE: question about `quit-char'

[Drew Adams]

> > My question is about whether I should be binding the quit 
> > character or `C-g', and if the former, what the best way
> > to do that is. Is the way I mentioned a good one?
> 
> Just bind to C-g:
>   (define-key map "\C-g" 'foo-bar)
> We do this all over the lisp/ tree anyway, and this isn't going to
> change before the release.

Got it. Thanks.

But I still don't understand, I guess. What does the question have to do with
the release? I'm not asking about any Emacs 23 or future features, AFAIK. 

Users in any release can presumably use input methods and change the quit
character from C-g, no? And if they do that, then my hard-coded binding of C-g
won't pick up their key, right?

I thought I was asking a general, release-independent question. What am I
missing?

RE: question about `quit-char'

[Drew Adams]

> > Just bind to C-g:
> >   (define-key map "\C-g" 'foo-bar)
> > 
> > We do this all over the lisp/ tree anyway, and this isn't going to
> > change before the release.
> 
> Wouldn't it be better for Drew to write a function using his 
> own recipe and use that? Then it would still work if the whole
> quit-char thing will be implemented later.
> 
> (The crucial difference relative to your view is that Drew distribute
> his libraries separately from Emacs and I believe he wants them to
> work in both new and old versions of Emacs. If not there will be a
> lot of extra trouble for him and the users of his libraries.)

See my reply to Chong Yidong.

I don't think I'm asking anything that is dependent on Emacs 23 or any other
release. The question is about whether and how to bind the quit character
instead of binding C-g (hard-coding).

Re: question about `quit-char'

[Eli Zaretskii]

> From: Chong Yidong <cyd@stupidchicken.com>
> Cc: drew.adams@oracle.com,  emacs-devel@gnu.org
> Date: Tue, 21 Oct 2008 13:45:46 -0400
> 
> Probably the right thing to do, at this point, is to simple assume that
> \C-g is the same as quit-char.  So Glenn's patch in bug#1205 is OK.

I don't mind, but let's leave a comment there about assuming that
quit-char is C-g.  It will help us when we revisit this issue later
on.

Re: question about `quit-char'

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Cc: <emacs-devel@gnu.org>
> Date: Tue, 21 Oct 2008 11:10:05 -0700
> 
> Uh, why is `set-quit-char' what I want?

It's what you want if you want to set the value of quit-char.

Re: question about `quit-char'

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Cc: "'Eli Zaretskii'" <eliz@gnu.org>, <emacs-devel@gnu.org>
> Date: Tue, 21 Oct 2008 12:15:29 -0700
> 
> But I still don't understand, I guess. What does the question have to do with
> the release? I'm not asking about any Emacs 23 or future features, AFAIK. 

To support changing quit-char correctly, we need changes that probably
will not happen before the release.  Until then, you can safely assume
that quit-char will be C-g, especially on windowed displays.

> Users in any release can presumably use input methods

I believe you mean input-mode.  Input methods have nothing to do with
quit-char.

> and change the quit character from C-g, no? And if they do that,
> then my hard-coded binding of C-g won't pick up their key, right?

Yes, but many other places will not obey a change in quit-char, so you
are in good company.

Re: question about `quit-char'

[Eli Zaretskii]

> Date: Tue, 21 Oct 2008 19:31:03 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: drew.adams@oracle.com, emacs-devel@gnu.org
> 
> > From: Chong Yidong <cyd@stupidchicken.com>
> > Date: Tue, 21 Oct 2008 13:21:49 -0400
> > Cc: emacs-devel@gnu.org
> > 
> > The Fset_quit_char command was introduced during the multi-tty merge.
> 
> That's true, but one could set quit-char in older Emacsen as well: you
> could do that with a call to set-input-mode.

Btw, Emacs 22 and before did allow changing quit-char on X, so I
believe the fact that this was disallowed during multi-tty merge is a
bug, or at least something that needs a good explanation.  The ELisp
manual still says that setting quit-char works on X as well, unlike
the `interrupt' argument to `set-input-mode'.

Does anyone know why this was disabled on X (and on other types of
graphic display)?

RE: question about `quit-char'

[Drew Adams]

> > But I still don't understand, I guess. What does the 
> > question have to do with the release? I'm not asking
> > about any Emacs 23 or future features, AFAIK. 
> 
> To support changing quit-char correctly, we need changes that probably
> will not happen before the release.  Until then, you can safely assume
> that quit-char will be C-g, especially on windowed displays.

I see; I didn't know that. Thx.

> > Users in any release can presumably use input methods
> 
> I believe you mean input-mode.  Input methods have nothing to do with
> quit-char.

Yes, sorry.

> > and change the quit character from C-g, no? And if they do that,
> > then my hard-coded binding of C-g won't pick up their key, right?
> 
> Yes, but many other places will not obey a change in quit-char, so you
> are in good company.

Got it. Thx.

Re: question about `quit-char'

[Chong Yidong]

Eli Zaretskii <eliz@gnu.org> writes:

> I don't mind, but let's leave a comment there about assuming that
> quit-char is C-g.  It will help us when we revisit this issue later
> on.

I've filed a wishlish bug in the bugtracker for this purpose.

RE: question about `quit-char'

[Stephen J. Turnbull]

Drew Adams writes:

 > I thought I was asking a general, release-independent
 > question. What am I missing?

An explanation of the semantics of the function you are binding.
Whether you "want" to bind to the quit character or to C-g depends on
whether or not the function has the semantics of quitting.

RE: question about `quit-char'

[Drew Adams]

>  > I thought I was asking a general, release-independent
>  > question. What am I missing?
> 
> An explanation of the semantics of the function you are binding.
> Whether you "want" to bind to the quit character or to C-g depends on
> whether or not the function has the semantics of quitting.

Totally irrelevant. What command I want to bind the key/char to and why, and
what that command does, are all irrelevant to the question I raised. Which has
already been answered, BTW. 

I asked a question about binding a command to the quit character, whatever it
might be, instead of assuming that the quit character is C-g. That's all.

What kind of "quitting" I might want the key to accomplish - or even whether its
action is "quitting" in any sense or blowing up balloons or painting fractal
coloring books, is irrelevant.

RE: question about `quit-char'

[Stephen J. Turnbull]

Drew Adams writes:

 > >  > I thought I was asking a general, release-independent
 > >  > question. What am I missing?
 > > 
 > > An explanation of the semantics of the function you are binding.
 > > Whether you "want" to bind to the quit character or to C-g depends on
 > > whether or not the function has the semantics of quitting.
 > 
 > Totally irrelevant.

Of course it's not irrelevant.  You asked whether to bind a command to
`quit-char' or `C-g', and the answer is "in principle, it depends on
if the command has quit semantics, in practice currently you may as
well just bind `C-g'".  Since you asked about `quit-char', respondents
seem to have assumed at first that you wanted to know something about
the appropriate way to bind a quit command to a key, and away we go.

 > What command I want to bind the key/char to and why, and what that
 > command does, are all irrelevant to the question I raised. Which
 > has already been answered, BTW.

True.  I'm trying to explain to you why a lot of bandwidth was spent
on a question that wasn't the one you wanted answered, to answer your
question "What am I missing?"  Your original question was ambiguous,
even misleading, and people answered the wrong interpretation.