ignore is bound to <lwindow> but <lwindow> is META

ignore is bound to <lwindow> but <lwindow> is META

[Lennart Borgman]

This is maybe w32 specific, I am not sure.

I have bound <lwindow> to META but `describe-function' says `ignore' is 
still bound to <lwindow>. What is wrong? Am I overlooking something?

Re: ignore is bound to <lwindow> but <lwindow> is META

[Stefan Monnier]

> I have bound <lwindow> to META but `describe-function' says `ignore' is
> still bound to <lwindow>.  What is wrong? Am I overlooking something?

What makes you think something's wrong?  The two notions of "bound" are
different, so I don't see any obvious conflict.


        Stefan

Re: ignore is bound to <lwindow> but <lwindow> is META

[Jason Rumney]

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

> This is maybe w32 specific, I am not sure.
>
> I have bound <lwindow> to META but `describe-function' says `ignore'
> is still bound to <lwindow>. What is wrong? Am I overlooking
> something?

What do you mean when you say you bound lwindow to META?

Re: ignore is bound to <lwindow> but <lwindow> is META

[Lennart Borgman]

Jason Rumney wrote:
> Lennart Borgman <lennart.borgman.073@student.lu.se> writes:
>
>   
>> This is maybe w32 specific, I am not sure.
>>
>> I have bound <lwindow> to META but `describe-function' says `ignore'
>> is still bound to <lwindow>. What is wrong? Am I overlooking
>> something?
>>     
>
> What do you mean when you say you bound lwindow to META?
>   
Sorry for my bad wordings. I do something like this:

  (when lwindow-meta
    (setq w32-lwindow-modifier 'meta)
    (setq w32-pass-lwindow-to-system nil))

This is in my w32-meta.el (that uses my patch for the low level keyboard 
hook, but that is another story). Maybe I just ought just to add

     (global-unset-key [(lwindow)])

at least to be more consistent? On the other hand I do not think that 
Emacs should say that a key is bound to anything if it is used as META. 
But I am a bit confused here so I appreciate explanations!

Re: ignore is bound to <lwindow> but <lwindow> is META

[Stefan Monnier]

> at least to be more consistent? On the other hand I do not think that
> Emacs should say that a key is bound to anything if it is used as META.

It's more general than that:  When Emacs says "foo is bound to M-c, <prior>,
<S-C-mouse-3>" all it means is that if you somehow manage to generate events
M-c, <prior> or <S-C-mouse-3>, then that'll call those functions.

But M-c may be remapped via key-translation-map to something else, so
hitting Meta+c may not actually call your function `foo'.

And the key that you think of as "prior" may not send the event <prior>.
Think of the RET vs `return' case.

Even after your (setq w32-lwindow-modifier 'meta), you may still be able to
generate `lwindow' event (using some other key(combo)).
         

        Stefan

Re: ignore is bound to <lwindow> but <lwindow> is META

[Lennart Borgman]

Stefan Monnier wrote:
>> at least to be more consistent? On the other hand I do not think that
>> Emacs should say that a key is bound to anything if it is used as META.
>>     
>
> It's more general than that:  When Emacs says "foo is bound to M-c, <prior>,
> <S-C-mouse-3>" all it means is that if you somehow manage to generate events
> M-c, <prior> or <S-C-mouse-3>, then that'll call those functions.
>
> But M-c may be remapped via key-translation-map to something else, so
> hitting Meta+c may not actually call your function `foo'.
>
> And the key that you think of as "prior" may not send the event <prior>.
> Think of the RET vs `return' case.
>
> Even after your (setq w32-lwindow-modifier 'meta), you may still be able to
> generate `lwindow' event (using some other key(combo)).
>          
>
>         Stefan
>   
Oh, I see. But is not that quite confusing? There might be many ways to 
generate an event (I actually do not know), but when I see "foo is bound 
to M-c, ..." I believe that when I hit M-c Emacs will call foo. I 
suspect most users will expect the same.

Is not key-translation-map a simple case? Can't this be checked in help?

Re: ignore is bound to <lwindow> but <lwindow> is META

[Stefan Monnier]

> Oh, I see.  But is not that quite confusing?

Yes, I wouldn't call it a feature.  But more generally, figuring out the
reverse mapping from a command (or an event) to its possible sources
(e.g. which physical key (with its name), or mouse movement) is between
difficult and impossible.

> There might be many ways to generate an event (I actually do not know),
> but when I see "foo is bound to M-c, ..." I believe that when I hit M-c
> Emacs will call foo.  I suspect most users will expect the same.

> Is not key-translation-map a simple case?

Partly.  It's relatively easy to check whether M-c will be affected or not
by key-translation-map, but it can be very hard to determine if
key-translation-map can translate some other event into M-c.

> Can't this be checked in help?

It could.  And it might get us one step closer.  But there are still many
more steps, some of which outside of Emacs's control and/or knowledge, and
additional steps get added every once in a while, sometimes without even
being aware of it.

Even without key-translation-map, the window-manager might catch the M-c
before Emacs sees it, or it might be bound via special-event-map, or ...


        Stefan