bug#43617: 27.1; Define-minor-mode keybindings not get precedence over global keymap

[Drew Adams <drew.adams@oracle.com>]

> Ah, it just works for "j" by accident -- define-minor-mode does not
> evaluate anything in the keymap form.

It should.  According to its doc string.

> So there's no bug here in define-minor-mode, but the
> here is wrong.  It should be:
> 
> (define-minor-mode pdf-continuous-scroll-mode-5
>   "Emulate continuous scroll with two synchronized buffers"
>   nil
>   " Continuous"
>   `((,(kbd "j") . (lambda () (interactive) (print "pushed j")))
>     (,(kbd "C-n") . (lambda () (interactive)(print "pushed C-n"))))
>   (print "toggled minor mode"))
> 
> So I'm closing this bug report.

I don't agree that there's no bug (IIUC).

I think there's either a doc bug (if you think
the current behavior is what we want - I don't)
or a behavior bug (if you think the doc's
described behavior is what we want - I do).

The doc string says that KEYMAP can be:

 an expression that returns either a keymap or
 a list of (KEY . BINDING) pairs where KEY and
 BINDING are suitable for `define-key'

The cons ((kbd "C-n") . 'foo) is exactly such a
(KEY . BINDING) pair - both KEY and BINDING are
suitable arguments for `define-key'.

Yes, it's true that the _result of evaluating_
(kbd ...) is ALSO an acceptable arg for
`define-key'.  But `define-key' doesn't _require_
its KEY arg to be, say, a string or vector.
`define-key' evaluates its arg.  And a list
(kbd ...) is an acceptable arg for `define-key'.

You can write

 (define-key map (kbd "C-n") 'foo)

Or you can write

 (define-key map "
" 'foo)

where that string with char ?\n (Control-J) is
the result of evaluating the sexp (kbd "C-n").

Or you can write

 (define-key map "\n" 'foo)

`define-minor-mode' should accept any expression
that `define-key' accepts for KEY.  In the case
of `C-n' that means (kbd "C-n"), "\n", and "
".