bug#48153: 28.0.50; minor mode keymaps should not override keymap given to read-from-minibuffer

[Gregory Heytings]

> Second of all, in order to get the key map bindings to 
> read-from-minibuffer overridden on emacs 27, it requires quitting from 
> switch-to-buffer, and then C-x b C-h b again to see that C-k is bound to 
> `bad`. In other words, on emacs 27, one has to pop up the minibuffer 
> twice. On emacs 28, just opening the minibuffer the first time will see 
> the key bindings overridden. I guess this is why I say this is a 
> regression on emacs 28 but you said you couldn't reproduce it. I think 
> both versions are wrong, it's just on emacs 28, it's wrong in a more 
> obvious way.

Okay, now I finally see what you mean (and I can reproduce it).  Indeed 
this is a recent change in the handling of minibuffers, the first time the 
minibuffer is entered is not anymore different from the other times.

> In any case, I'm not sure if the minor mode maps should override the key 
> bindings in the keymap given to read-from-buffer.

Why would they not?  The minibuffer behaves, in that respect, like any 
other buffer.  Note that they do so only when the minor mode is enabled in 
the minibuffer.  With the example init file, using

(define-global-minor-mode test-global-mode test-mode (lambda () (or 
(minibufferp) (test-mode 1))))

works as you want it with both Emacs 27 and 28.