bug#52448: key-valid-p vs "C-M-x" and "ESC C-x"

bug#52448: key-valid-p vs "C-M-x" and "ESC C-x"

[Stefan Kangas]

AFAIU, we want exactly one valid way to write keybindings with
`key-valid-p'.  Should both of these be `key-valid-p'?

1.  (key-description (kbd "ESC C-x"))
    => "C-M-x"

    (key-valid-p "ESC C-x")
    => t

2.  (key-description (kbd "C-M-x"))
    => "C-M-x"

    (key-valid-p "C-M-x")
    => t

However, there is also this, and the comment at the top of keymap.c:

    (kbd "ESC C-x")
    => "^[^X"

    (kbd "M-C-x")
    => [134217752]

bug#52448: key-valid-p vs "C-M-x" and "ESC C-x"

[Lars Ingebrigtsen]

Stefan Kangas <stefan@marxist.se> writes:

> AFAIU, we want exactly one valid way to write keybindings with
> `key-valid-p'.  Should both of these be `key-valid-p'?
>
> 1.  (key-description (kbd "ESC C-x"))
>     => "C-M-x"
>
>     (key-valid-p "ESC C-x")
>     => t

That's what happens today, but Stefan M had a cunning plan to make
round-tripping these keystrokes (EST/TAB/RET) work better.

In that case

"ESC C-x"

and

"C-M-x"

would still both be valid, but refer to different things (that get
resolved to the same thing).

But we're not there today, so when writing keymaps one should use the
intended mnemonics.  For instance, it's

  "C-c C-i" #'gnus-info-find-node

and not

  "C-c TAB" #'gnus-info-find-node

because the "i" is for "info".

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#52448: key-valid-p vs "C-M-x" and "ESC C-x"

[Stefan Kangas]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Stefan Kangas <stefan@marxist.se> writes:
>
>> AFAIU, we want exactly one valid way to write keybindings with
>> `key-valid-p'.  Should both of these be `key-valid-p'?
>>
>> 1.  (key-description (kbd "ESC C-x"))
>>     => "C-M-x"
>>
>>     (key-valid-p "ESC C-x")
>>     => t
>
> That's what happens today, but Stefan M had a cunning plan to make
> round-tripping these keystrokes (EST/TAB/RET) work better.
>
> In that case
>
> "ESC C-x"
>
> and
>
> "C-M-x"
>
> would still both be valid, but refer to different things (that get
> resolved to the same thing).

So both will be valid even with the planned changes, and this is
notabug?

bug#52448: key-valid-p vs "C-M-x" and "ESC C-x"

[Lars Ingebrigtsen]

Stefan Kangas <stefan@marxist.se> writes:

> So both will be valid even with the planned changes, and this is
> notabug?

Yes, both will remain valid, but it's a bug that it's not round-tripping
properly, so the report might as well stay open until that's fixed.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

bug#52448: key-valid-p vs "C-M-x" and "ESC C-x"

[Stefan Kangas]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Yes, both will remain valid, but it's a bug that it's not round-tripping
> properly, so the report might as well stay open until that's fixed.

OK.

I'm linking the emacs-devel thread where round-tripping was discussed:

    https://lists.gnu.org/archive/html/emacs-devel/2021-11/msg00941.html