How global is a define-global-minor-mode mode?

How global is a define-global-minor-mode mode?

[Lennart Borgman]

I am trying to use define-global-minor-mode, see below. I have 
customized the global mode to be t, but when I open new buffers the 
minor mode is not set in them. Is this a bug in define-global-minor-mode?


(define-minor-mode html-site-mode
  "Adds a menu for easy access of setting site, uploading etc."
  :init-value nil
  :lighter nil
  :keymap html-site-mode-map
  :group 'html-site)

(defvar html-site-mode-off-list
  '(nxhtml-mode))

(define-global-minor-mode html-site-global-mode html-site-mode
  (lambda()
    (html-site-mode 1)
    (when t ;buffer-file-name
      (unless (memq major-mode html-site-mode-off-list)
        (html-site-mode 1))))
  :group 'html-site)

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    I am trying to use define-global-minor-mode, see below. I have 
    customized the global mode to be t, but when I open new buffers the 
    minor mode is not set in them. Is this a bug in define-global-minor-mode?

My impression is that the global mode defined with
define-global-minor-mode does not affect all buffers to be created in
the future.  It only catches buffers created by visiting files.

If so, I think what is needed is to add this to the documentation of
define-global-minor-mode.

    Globally enabling the mode also affects buffers subsequently created
    by visiting files, but not buffers subsequently created in other ways.

Can anyone confirm that this is right?

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

[I sent this message a few days ago but did not get a response.
Could we get the discussion moving again?]

    I am trying to use define-global-minor-mode, see below. I have 
    customized the global mode to be t, but when I open new buffers the 
    minor mode is not set in them. Is this a bug in define-global-minor-mode?

My impression is that the global mode defined with
define-global-minor-mode does not affect all buffers to be created in
the future.  It only catches buffers created by visiting files.

If so, I think what is needed is to add this to the documentation of
define-global-minor-mode.

    Globally enabling the mode also affects buffers subsequently created
    by visiting files, but not buffers subsequently created in other ways.

Can anyone confirm that this is right?

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Richard Stallman wrote:
> [I sent this message a few days ago but did not get a response.
> Could we get the discussion moving again?]
>
>     I am trying to use define-global-minor-mode, see below. I have 
>     customized the global mode to be t, but when I open new buffers the 
>     minor mode is not set in them. Is this a bug in define-global-minor-mode?
>
> My impression is that the global mode defined with
> define-global-minor-mode does not affect all buffers to be created in
> the future.  It only catches buffers created by visiting files.
>
> If so, I think what is needed is to add this to the documentation of
> define-global-minor-mode.
>
>     Globally enabling the mode also affects buffers subsequently created
>     by visiting files, but not buffers subsequently created in other ways.
>
> Can anyone confirm that this is right?
>   

I believe the problem is that the setup of the hook functions is not 
done as expected. I have my global minor mode set to t in customize, but 
I do not found the global mode hook functions in the hook after startup. 
(The global minor mode is t.)

I think Drew said that the documentation for define-global-minor-mode 
needs some enhancement. I think that too, especially something could be 
mentioned about the difference using :global t with define-minor-mode.

BTW, it looks to me that there is no way to catch creation of new buffers.

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    I believe the problem is that the setup of the hook functions is not 
    done as expected. I have my global minor mode set to t in customize, but 
    I do not found the global mode hook functions in the hook after startup. 
    (The global minor mode is t.)

That sounds like a bug.  Can you debug it?  Can you provide a
_precise_ self-contained test case so someone else can debug it?

    I think Drew said that the documentation for define-global-minor-mode 
    needs some enhancement. I think that too, especially something could be 
    mentioned about the difference using :global t with define-minor-mode.

Maybe it does, but not based on a bug.  We should fix the bug,
and document the correct behavior.

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Richard Stallman wrote:
>     I believe the problem is that the setup of the hook functions is not 
>     done as expected. I have my global minor mode set to t in customize, but 
>     I do not found the global mode hook functions in the hook after startup. 
>     (The global minor mode is t.)
>
> That sounds like a bug.  Can you debug it?  Can you provide a
> _precise_ self-contained test case so someone else can debug it?
>   

Create a file tg.el in your emacs load-path:

  (define-minor-mode tg-mode ""
    :lighter "  **TG** ")

  (define-global-minor-mode tg-globalized-mode tg-mode
    (lambda()
      (tg-mode 1)))

Eval it and customize tg-globalized-mode permanently to t.

Start Emacs again:

    emacs
    M-x load-library tg
    C-h v tg-globalized-mode
       => Its value is t
    C-h v tg-mode
       => Its value is nil

Then try

     Mx tg-globalized-mode
       => Tg-Globalized mode disabled
     M-x tg-globalized-mode
       => Tg-Globalized mode enabled
     C-h v tg-mode
       => Its value is t

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

It looks like define-global-minor-mode simply fails to do anything
to arrange to handle customization of the globalized mode's variable.

Would someone please work on this?

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Richard Stallman wrote:
> It looks like define-global-minor-mode simply fails to do anything
> to arrange to handle customization of the globalized mode's variable.
>
> Would someone please work on this?
>   

I can't do that now, however I have a suggestion. 
define-globalized-minor-mode takes care of the nasty technical details, 
but maybe it could still be easier to use. If we assume that the minor 
mode to globalize is defined with define-minor-mode we know how to turn 
it on and off. How about something like this? :

   :turn-on-if (lambda() ....)
   :turn-off-if (lambda() ...)

where the two functions should used for turning on/off the minor mode in 
each buffer when the globalized mode is turned on/off.

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

I'd rather not think about changes in this now,
other than what is necessary to get it working correctly
with the custom mechanism.

Re: How global is a define-global-minor-mode mode?

[Johan Bockgård]

Richard Stallman <rms@gnu.org> writes:

> My impression is that the global mode defined with
> define-global-minor-mode does not affect all buffers to be created
> in the future. It only catches buffers created by visiting files.

It also affects buffers that use a non-fundamental mode. It does this:

     ;; Setup hook to handle future mode changes and new buffers.
     (if ,global-mode
         (progn
           (add-hook 'after-change-major-mode-hook
     		',MODE-enable-in-buffers)
           (add-hook 'find-file-hook ',MODE-check-buffers)
           (add-hook 'change-major-mode-hook ',MODE-cmhh))

-- 
Johan Bockgård

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    It also affects buffers that use a non-fundamental mode. It does this:

Thanks.  I will update that point in the Lisp manual.

Re: How global is a define-global-minor-mode mode?

[Chong Yidong]

Lennart Borgman <lennart.borgman.073@student.lu.se> writes:

> I am trying to use define-global-minor-mode, see below. I have
> customized the global mode to be t, but when I open new buffers the
> minor mode is not set in them. Is this a bug in
> define-global-minor-mode?

Seems to be working fine.  Your recipe is a little vague, but here is
a simple example:

1.  Evaluate the following in *scratch* buffer:

  (define-minor-mode foo-mode
   "Foo."
   nil
   " foo"
   :group 'foo)

  (define-global-minor-mode html-site-global-mode foo-mode
    (lambda () (foo-mode 1))
    :group 'foo)

2. M-x customize-option RET foo RET
3. Click on "Toggle" (global-foo-mode is turned on)
4. Click on "Set for Current Session"
5. C-x C-f asdf RET (open a new file)
6. Observe the presence of the " foo" lighter in the modeline.

There is one subtlety, which is that if you save the custom option for
future sessions, it won't activate foo-mode in future sessions unless
the above code defining foo-mode and global-foo-mode is already
loaded.  But I don't think that's necessary a bug; it's a flaw but not
one that's easily avoidable.

In short, I don't think there's anything that needs to be done here.

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    There is one subtlety, which is that if you save the custom option for
    future sessions, it won't activate foo-mode in future sessions unless
    the above code defining foo-mode and global-foo-mode is already
    loaded.  But I don't think that's necessary a bug; it's a flaw but not
    one that's easily avoidable.

That is the bug.  It can be fixed somehow, I'm sure.

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

>> I am trying to use define-global-minor-mode, see below. I have
>> customized the global mode to be t, but when I open new buffers the
>> minor mode is not set in them. Is this a bug in
>> define-global-minor-mode?

> Seems to be working fine.  Your recipe is a little vague, but here is
> a simple example:

> 1.  Evaluate the following in *scratch* buffer:

>   (define-minor-mode foo-mode
>    "Foo."
>    nil
>    " foo"
>    :group 'foo)

>   (define-global-minor-mode html-site-global-mode foo-mode
>     (lambda () (foo-mode 1))
>     :group 'foo)

> 2. M-x customize-option RET foo RET
> 3. Click on "Toggle" (global-foo-mode is turned on)
> 4. Click on "Set for Current Session"
> 5. C-x C-f asdf RET (open a new file)
> 6. Observe the presence of the " foo" lighter in the modeline.

> There is one subtlety, which is that if you save the custom option for
> future sessions, it won't activate foo-mode in future sessions unless
> the above code defining foo-mode and global-foo-mode is already
> loaded.  But I don't think that's necessary a bug; it's a flaw but not
> one that's easily avoidable.

Isn't it the case that it works if the minor mode function corresponding to
the variable is autoloaded?  Otherwise, IIRC you need to add a :require
to the definition of the minor mode.


        Stefan

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    Isn't it the case that it works if the minor mode function corresponding to
    the variable is autoloaded?  Otherwise, IIRC you need to add a :require
    to the definition of the minor mode.

Could define-global-minor-mode do that automatically?

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

>     Isn't it the case that it works if the minor mode function
>     corresponding to the variable is autoloaded?  Otherwise, IIRC you need
>     to add a :require to the definition of the minor mode.

> Could define-global-minor-mode do that automatically?

Not really: if the global minor mode is autoloaded, then :require is
a nuisance, so to do it automatically, define-global-minor-mode would have
to know whether it's autoloaded or not.  And in most/all cases it's
autoloaded anyway and if not, it should.


        Stefan

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

>     Isn't it the case that it works if the minor mode function
>     corresponding to the variable is autoloaded?  Otherwise, IIRC you need
>     to add a :require to the definition of the minor mode.

    Not really: if the global minor mode is autoloaded, then :require is
    a nuisance,

I don't follow.  Could you explain why it is a nuisance?
And how does this relate to the question of whether the globalized
mode is autoloaded?

		so to do it automatically, define-global-minor-mode would have
    to know whether it's autoloaded or not.  And in most/all cases it's
    autoloaded anyway and if not, it should.

Are you proposing that every use of define-global-minor-mode should
have an autoload cookie?

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

>> Isn't it the case that it works if the minor mode function
>> corresponding to the variable is autoloaded?  Otherwise, IIRC you need
>> to add a :require to the definition of the minor mode.

>     Not really: if the global minor mode is autoloaded, then :require is
>     a nuisance,

> I don't follow.  Could you explain why it is a nuisance?

The :require thingy is a risk: if you move the definition of the mode to
another file, then the user's customizations will be broken (it happened
with global-font-lock-mode).

> And how does this relate to the question of whether the globalized
> mode is autoloaded?

If it's autoloaded, then the :require is not necessary because the handling
of the autoload cookie will make sure that the minor mode function will be
called when the user has customized the mode (and will thus trigger loading
the relevant file).

> 		so to do it automatically, define-global-minor-mode would have
>     to know whether it's autoloaded or not.  And in most/all cases it's
>     autoloaded anyway and if not, it should.

> Are you proposing that every use of define-global-minor-mode should
> have an autoload cookie?

As a general rule, yes.  I'm sure we can come up with counter examples, tho,
so there's no point trying to enforce it.


        Stefan

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
>>> Isn't it the case that it works if the minor mode function
>>> corresponding to the variable is autoloaded?  Otherwise, IIRC you need
>>> to add a :require to the definition of the minor mode.
> 
>>     Not really: if the global minor mode is autoloaded, then :require is
>>     a nuisance,
> 
>> I don't follow.  Could you explain why it is a nuisance?
> 
> The :require thingy is a risk: if you move the definition of the mode to
> another file, then the user's customizations will be broken (it happened
> with global-font-lock-mode).

I have lost my way here, but I did some tests to try to understand. If I 
use defcustom with a :set function this will be called when the elisp 
file is loaded. Why can't something similar be done for 
define-global-minor-mode also? It is a global mode and then :set is 
passed to defcustom, or?

BTW should not the name be define-globalized-minor-mode as was suggested 
before?

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

>>>> Isn't it the case that it works if the minor mode function
>>>> corresponding to the variable is autoloaded?  Otherwise, IIRC you need
>>>> to add a :require to the definition of the minor mode.
>> 
>>> Not really: if the global minor mode is autoloaded, then :require is
>>> a nuisance,
>> 
>>> I don't follow.  Could you explain why it is a nuisance?
>> 
>> The :require thingy is a risk: if you move the definition of the mode to
>> another file, then the user's customizations will be broken (it happened
>> with global-font-lock-mode).

> I have lost my way here, but I did some tests to try to understand. If I use
> defcustom with a :set function this will be called when the elisp file is
> loaded. Why can't something similar be done for define-global-minor-mode
> also? It is a global mode and then :set is passed to defcustom, or?

I don't understand your question.  What do you suggest exactly and for
what purpose?

> BTW should not the name be define-globalized-minor-mode as was
> suggested before?

I believe so, yes.  Feel free to implement this change,


        Stefan

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
>>>>> Isn't it the case that it works if the minor mode function
>>>>> corresponding to the variable is autoloaded?  Otherwise, IIRC you need
>>>>> to add a :require to the definition of the minor mode.
>>>> Not really: if the global minor mode is autoloaded, then :require is
>>>> a nuisance,
>>>> I don't follow.  Could you explain why it is a nuisance?
>>> The :require thingy is a risk: if you move the definition of the mode to
>>> another file, then the user's customizations will be broken (it happened
>>> with global-font-lock-mode).
> 
>> I have lost my way here, but I did some tests to try to understand. If I use
>> defcustom with a :set function this will be called when the elisp file is
>> loaded. Why can't something similar be done for define-global-minor-mode
>> also? It is a global mode and then :set is passed to defcustom, or?
> 
> I don't understand your question.  What do you suggest exactly and for
> what purpose?

I previously used defcustom variables to manage something similar to 
define-globalized-mode MY-GLOBAL-MODE MY-MODE. This worked ok. I then 
used the :set function to  turn on the minor mode MY-MODE.

Using define-minor-mode there is no way to do something similar. There 
is no code that is run when the library file is loaded, even if the 
minor mode is global and that is in my opinion a bug (or at least a very 
inconvenient inconsistency).

Actually define-minor-mode says that "other keywords will be passed to 
defcustom if the minor mode is global". Should not this mean that :set 
should be passed to defcustom and therefore be run when loading the 
library file if the minor mode as been customized to be on? I can not 
see that :set is run however.

On the other hand one may wonder why not the body or the hook of the 
minor mode is run when loading the library. I can see the purpose of 
this (that you should be able to load a library without running any 
code), but there are a few inconsistenceis here (compared with 
defcustom). I think this inconsistencies should be addressed before 
trying to fix define-globalized-minor-mode. But I am quite a bit unsure 
about how to handle the problem with body/hook/:set for a define-minor-mode.

Below are some test code I have used in my own library. This code should 
give some messages when loading the library they are put in. Please test 
it with the defcustom and define-minor-mode set.


(defvar html-site-temp-mode-hook nil)
(add-hook 'html-site-temp-mode-hook
           (lambda()
             (message "html-site-temp-mode-hook, html-site-temp-mode=%s" 
html-site-temp-mode)))

(define-minor-mode html-site-temp-mode
   "dummy"
   :init-value nil
   :lighter nil
   :global t
   :keymap (let ((m (make-sparse-keymap)))
             (define-key m [f11] (lambda() (interactive) (message 
"html-site-mode-temp f11 here")))
             m)
   :set (lambda(sym val)
          (message ":set html-site-temp-mode: sym=%s, val=%s" sym val)
          (set-default sym val))
   (message "body html-site-temp-mode=%s, current-buffer=%s" 
html-site-mode (current-buffer)))

(defcustom html-site-temp nil
   "doc"
   :type 'boolean
   :set (lambda(sym val)
          (message ":set html-site-temp: sym=%s, val=%s" sym val)
          (set-default sym val)))

(message "html-site.el loaded, html-site-global-mode=%s, 
html-site-temp=%s, html-site-temp-mode=%s"
          html-site-global-mode
          html-site-temp
          html-site-temp-mode
          )


>> BTW should not the name be define-globalized-minor-mode as was
>> suggested before?
> 
> I believe so, yes.  Feel free to implement this change,
> 
> 
>         Stefan

As I use to say: I do not want to touch the CVS this close to the 
release since I am unfamiliar with cvs. I would be glad if you changed 
the name.

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

> I previously used defcustom variables to manage something similar to
> define-globalized-mode MY-GLOBAL-MODE MY-MODE. This worked ok. I then used
> the :set function to  turn on the minor mode MY-MODE.

> Using define-minor-mode there is no way to do something similar. There is no
> code that is run when the library file is loaded, even if the minor mode is
> global and that is in my opinion a bug (or at least a very inconvenient
> inconsistency).

Yes, there is code that runs the minor-mode function if it is global and has
been set.  This *should* work.  So please tell us what you tried that failed
to work.

> Actually define-minor-mode says that "other keywords will be passed to
> defcustom if the minor mode is global". Should not this mean that :set
> should be passed to defcustom and therefore be run when loading the library
> file if the minor mode as been customized to be on? I can not see that :set
> is run however.

Try to look at the macro expansion of (define-minor-mode foo-mode
"Foo" :global t) and you'll see it's got a :set.

> Please test it with the defcustom and define-minor-mode set.

I don't know what the above sentence means.

> (define-minor-mode html-site-temp-mode
>   "dummy"
>   :init-value nil
>   :lighter nil
>   :global t
>   :keymap (let ((m (make-sparse-keymap)))
>             (define-key m [f11] (lambda() (interactive) (message
> "html-site-mode-temp f11 here")))
>             m)
>   :set (lambda(sym val)
>          (message ":set html-site-temp-mode: sym=%s, val=%s" sym val)
>          (set-default sym val))
>   (message "body html-site-temp-mode=%s, current-buffer=%s" html-site-mode
> (current-buffer)))

Passing a :set here is likely to lead to surprises.

> As I use to say: I do not want to touch the CVS this close to the release
> since I am unfamiliar with cvs.

It's never too late to learn.  You can start by posting a patch generated
from "cvs diff", at which point "cvs commit -m <message>" should be all
that's needed.

> I would be glad if you changed the name.

So would I, but I don't have much time,


        Stefan

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
> Yes, there is code that runs the minor-mode function if it is global and has
> been set.  This *should* work.  So please tell us what you tried that failed
> to work.

Just put the code I sent in a previous message in a file, set customize 
the defcustom and the define-minor-mode variables to t. Save this 
customization and make sure the file is loaded when you restart Emacs. 
Read the messages in the *Message* buffer to see what code from those 
two variables were run.

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

>> Yes, there is code that runs the minor-mode function if it is global and has
>> been set.  This *should* work.  So please tell us what you tried that failed
>> to work.

> Just put the code I sent in a previous message in a file,

As mentioned, this code uses a :set which is a bad idea.  So I see no point
in trying that out before you try out a simpler code.  When simple code
doesn't do what one wants, it's clearly a bug.  When convoluted code behaves
strange, it's never really clear whether it's a bug in Emacs or in the
convoluted code.

> set customize the defcustom and the define-minor-mode variables to t.

How?  Where?  Those details are very important.

> Save this customization and make sure the file is loaded when you restart
> Emacs.

Which file?  How do you ensure it's loaded?  Again, those details
are very important.


        Stefan

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
>> I previously used defcustom variables to manage something similar to
>> define-globalized-mode MY-GLOBAL-MODE MY-MODE. This worked ok. I then used
>> the :set function to  turn on the minor mode MY-MODE.
> 
>> Using define-minor-mode there is no way to do something similar. There is no
>> code that is run when the library file is loaded, even if the minor mode is
>> global and that is in my opinion a bug (or at least a very inconvenient
>> inconsistency).
> 
> Yes, there is code that runs the minor-mode function if it is global and has
> been set.  This *should* work.  So please tell us what you tried that failed
> to work.


I think I am beginning to understand what is happening now because of 
the trick with macroexpand you mentioned. I felt rather confused, but I 
beleive that to a not too small extent depends on my inability to see a 
clear thought behind what should actually be run when for defcustom and 
define-minor-mode.

My initial problem was that a define-global-minor-mode did not work as I 
wanted it too. The define-minor-mode were not turned on at startup of 
Emacs even though I had customized the define-global-minor-mode to true.

However it worked if I turned off and on the define-global-minor-mode 
again. And it had worked before with my a little bit more low tech 
solution where I used a defcustom and the :set function instead.

I thought that the problem was more of the type a typo in 
define-global-minor-mode, but I did not look very much at it. Now since 
the problem remains I took a closer look at it and did some tests.

I was surprised about the incongruencies I found. I have attached a test 
file that can show them. More about that later.

Now a bit about the incongruencies (a little bit simplified picture to 
make it more clear):

- When a defcustom is set through custom (the GUI interface or 
customize-set-variable) the :set function is run.

- That is true also for the implicit defcustom created by a 
define-minor-mode.

- However for define-minor-mode you can also change the value of the 
implicit defcustom by using the mode function instead. In this case the 
:set function is not run.

- On the other hand when you use the mode function the body of the 
define-minor-mode is run.

- And then this body is in turned not run when you use set the implicit 
defcustom through customize.

I think this has led to the trouble we have seen with 
define-global-minor-mode.


*** My suggestion is this: Let the :set function and the body of a 
define-minor-mode be the same thing. I believe this will make it much 
easier for users and programmers.


Now back to the tests. If the above was a bit hard to grasp then maybe 
the test will help. To run the tests first load this file and customize 
the defcustom, define-minor-mode and define-global-minor-mode, all to t. 
Then restart Emacs and load the file again and now look in the *Message* 
buffer for the result.

Re: How global is a define-global-minor-mode mode?

[Lennart Borgman (gmail)]

[-- Attachment #1: Type: text/plain, Size: 399 bytes --]

Lennart Borgman (gmail) wrote:

> Now back to the tests. If the above was a bit hard to grasp then maybe 
> the test will help. To run the tests first load this file and customize 
> the defcustom, define-minor-mode and define-global-minor-mode, all to t. 
> Then restart Emacs and load the file again and now look in the *Message* 
> buffer for the result.



And here are the tests, sorry ... ;-)

[-- Attachment #2: defcustom-or-minor.el --]
[-- Type: text/plain, Size: 1935 bytes --]

(message "\n\n>>>>>>> defcustom-or-minor.el starting loading, temp-global-mode=%s" 
         (if (boundp 'temp-global-mode)
             temp-global-mode
           "not boundp"))

(message "\n****** Declarations: *********")

(define-minor-mode temp-mode
  "dummy"
  :init-value nil
  :lighter nil
  :global t
  :group 'temp-test
  :initialize 'custom-initialize-set
  :set (lambda(sym val)
         (message "  temp-mode :set => sym=%s, val=%s" sym val)
         (set-default sym val))
  (message "  temp-mode body => temp-mode=%s, current-buffer=%s" temp-mode (current-buffer)))

(define-global-minor-mode temp-global-mode temp-mode
  (lambda()
    ;;(message "here %s" (substring (buffer-name) 0 1))
    (when (cond
           ((member (substring (buffer-name) 0 1) '(" " "*" "."))
            nil)
           (t t))
      (message "  temp-global-mode turn on called")
      (temp-mode 1)))
  :group 'temp-test
  :initialize 'custom-initialize-set
  :set (lambda(sym val)
         (message "  temp-global-mode :set => sym=%s, val=%s" sym val)
         (set-default sym val)))

(defcustom temp-defcustom nil
  "doc"
  :type 'boolean
  :group 'temp-test
  :set (lambda(sym val)
         (message "  temp-defcustom :set => sym=%s, val=%s" sym val)
         (set-default sym val)))


(message "\n****** Running tests: *********")

(message "*** (customize-set-variable 'temp-defcustom t) =>")
(customize-set-variable 'temp-defcustom t)

(message "\n*** (customize-set-variable 'temp-mode t) =>")
(customize-set-variable 'temp-mode t)
(message "*** (temp-mode 1) =>")
(temp-mode 1)

(message "\n*** (customize-set-variable 'temp-global-mode t) =>")
(temp-global-mode 1)
(message "*** (temp-global-mode 1) =>")
(customize-set-variable 'temp-global-mode t)

(message "\n<<<<<<<<< defcustom-or-minor.el loaded: temp-global-mode=%s, temp-mode=%s, temp-defcustom=%s\n\n"
         temp-global-mode
         temp-mode
         temp-defcustom)

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

> (define-minor-mode temp-mode
>   "dummy"
>   :init-value nil
>   :lighter nil
>   :global t
>   :group 'temp-test
>   :initialize 'custom-initialize-set
>   :set (lambda(sym val)
>          (message "  temp-mode :set => sym=%s, val=%s" sym val)
>          (set-default sym val))
>   (message "  temp-mode body => temp-mode=%s, current-buffer=%s" temp-mode (current-buffer)))

I do not care to support a define-minor-mode with a :set.  And same thing
about a :initialize.  The same holds for define-global-minor-mode of course.

Please take a big step back and tell me what you want to do, rather than
try to ask how to get your ugly-hack to work.


        Stefan

Re: How global is a define-global-minor-mode mode?

[Chong Yidong]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> (define-minor-mode temp-mode
>>   "dummy"
>>   :init-value nil
>>   :lighter nil
>>   :global t
>>   :group 'temp-test
>>   :initialize 'custom-initialize-set
>>   :set (lambda(sym val)
>>          (message "  temp-mode :set => sym=%s, val=%s" sym val)
>>          (set-default sym val))
>>   (message "  temp-mode body => temp-mode=%s, current-buffer=%s" temp-mode (current-buffer)))
>
> I do not care to support a define-minor-mode with a :set.  And same thing
> about a :initialize.  The same holds for define-global-minor-mode of course.
>
> Please take a big step back and tell me what you want to do, rather than
> try to ask how to get your ugly-hack to work.

I concur.  This should not be a FOR-RELEASE bug.

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

> *** My suggestion is this: Let the :set function and the body of
> a define-minor-mode be the same thing. I believe this will make it much
> easier for users and programmers.

Huh, check the code: the :set does nothing more than run the body of the
define-minor-mode.


        Stefan

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    Now back to the tests. If the above was a bit hard to grasp then maybe 
    the test will help. To run the tests first load this file and customize 
    the defcustom, define-minor-mode and define-global-minor-mode, all to t. 
    Then restart Emacs and load the file again and now look in the *Message* 
    buffer for the result.

That seems to be what define-minor-mode produces now, when the minor
mode is global.

This cannot be done for a minor mode that is normally local
because customizing a variable is always global.

Re: How global is a define-global-minor-mode mode?

[Richard Stallman]

    >     Not really: if the global minor mode is autoloaded, then :require is
    >     a nuisance,

    > I don't follow.  Could you explain why it is a nuisance?

    The :require thingy is a risk: if you move the definition of the mode to
    another file, then the user's customizations will be broken (it happened
    with global-font-lock-mode).

It seems that you are saying the risk is that the :require has to
specify the file name, so if you move the definition to another file,
the :require argument will be wrong.

If the :require is generated automatically, perhaps it would never be
wrong.

    > And how does this relate to the question of whether the globalized
    > mode is autoloaded?

    If it's autoloaded, then the :require is not necessary because the handling
    of the autoload cookie will make sure that the minor mode function will be
    called when the user has customized the mode (and will thus trigger loading
    the relevant file).

I don't see how that is the case.  If the minor mode is autoloaded
then calling the minor mode function will load the file.  But how would
this autoload cause the function to be called in the first place?

    > Are you proposing that every use of define-global-minor-mode should
    > have an autoload cookie?

    As a general rule, yes.  I'm sure we can come up with counter examples, tho,
    so there's no point trying to enforce it.

It might be an ok solution.

But do you have any objection to using :require if we make sure it
never gets the wrong file?

Re: How global is a define-global-minor-mode mode?

[Chong Yidong]

Richard Stallman <rms@gnu.org> writes:

>     >     Not really: if the global minor mode is autoloaded, then :require is
>     >     a nuisance,
>
>     > I don't follow.  Could you explain why it is a nuisance?
>
>     The :require thingy is a risk: if you move the definition of the mode to
>     another file, then the user's customizations will be broken (it happened
>     with global-font-lock-mode).
>
> It seems that you are saying the risk is that the :require has to
> specify the file name, so if you move the definition to another file,
> the :require argument will be wrong.
>
> If the :require is generated automatically, perhaps it would never be
> wrong.

The only way for Emacs to know which source file to :require, for a
third-party package, is to look at the custom-set-variables
declaration in .emacs.  Therefore, the only way your suggestion can
work is for define-global-minor-mode to somehow put this information
in .emacs.  If you move the source file to a different location there
is no way for .emacs to be automatically updated.

So there's no sane way to handle this automagically.

Re: How global is a define-global-minor-mode mode?

[David Kastrup]

Chong Yidong <cyd@stupidchicken.com> writes:

> Richard Stallman <rms@gnu.org> writes:
>
>>     >     Not really: if the global minor mode is autoloaded, then
>>     >     :require is a nuisance,
>>
>>     > I don't follow.  Could you explain why it is a nuisance?
>>
>>     The :require thingy is a risk: if you move the definition of
>>     the mode to another file, then the user's customizations will
>>     be broken (it happened with global-font-lock-mode).
>>
>> It seems that you are saying the risk is that the :require has to
>> specify the file name, so if you move the definition to another
>> file, the :require argument will be wrong.
>>
>> If the :require is generated automatically, perhaps it would never be
>> wrong.
>
> The only way for Emacs to know which source file to :require, for a
> third-party package, is to look at the custom-set-variables
> declaration in .emacs.

Uh, wouldn't the file to require be load-file-name at the time
define-global-minor-mode is loaded or executed via autoload?

-- 
David Kastrup

Re: How global is a define-global-minor-mode mode?

[Chong Yidong]

David Kastrup <dak@gnu.org> writes:

>> The only way for Emacs to know which source file to :require, for a
>> third-party package, is to look at the custom-set-variables
>> declaration in .emacs.
>
> Uh, wouldn't the file to require be load-file-name at the time
> define-global-minor-mode is loaded or executed via autoload?

That breaks if you move a third-party package to another directory or
another file name.  More importantly, this kind of behavior is
different from similar existing Emacs mechanisms.  For instance, when
you write define-minor-mode and defcustom statements, it does not
automagically :require the current load-file.

I really don't see why it's a problem having to separately specify the
requisite load-file.

Re: How global is a define-global-minor-mode mode?

[Stefan Monnier]

>     The :require thingy is a risk: if you move the definition of the mode to
>     another file, then the user's customizations will be broken (it happened
>     with global-font-lock-mode).

> It seems that you are saying the risk is that the :require has to
> specify the file name, so if you move the definition to another file,
> the :require argument will be wrong.

> If the :require is generated automatically, perhaps it would never be
> wrong.

No, the :require is of course correct.  The problem is that when you set the
variable, the current :require gets copied into your .emacs file, and when
you later run an Emacs where that minor mode was moved to another file, your
.emacs still points to the old location and the startup fails to set the var
(and complains that the file doesn't define function bar-mode).


        Stefan