Re: master 18b680cfd1: Fix bug#52467 by adding a new custom variable 'display-comint-buffer-action'

[martin rudalics]

> Or to add more semantics:
>
>    (pop-to-buffer "*shell*" '(nil . ((caller . comint))))

IIRC I once used the term 'label' for it.  It vanished soon.

>> which means that customizing 'display-buffer-alist' with that value on
>> Emacs 28 will report a mismatch but that's the only mishap that should
>> happen.
>>
>> We also could have 'display-buffer-assq-regexp' do
>>
>>        (when (or (and (stringp key)
>>                       (string-match-p key buffer-name))
>>                  (and (functionp key)
>>                       (funcall key buffer-name action))
>>                    (and (symbolp key)
>>                         (cadr (assq key (cdr action)))))
>
> That will shorten the above example to
>
>    (customize-set-variable
>     'display-buffer-alist
>     '((comint
>        display-buffer-same-window (nil))))
>
> ?

Yes.  Although we should be careful with the symbol.  'shell' would be
interpreted as a function to call via (funcall key buffer-name action).
We could wrap that in a 'condition-case' but what if the function has
some nasty side-effect before it errs out?  So maybe *comint* or
something similar ...

>> so we could do away with 'display-buffer-match-comint' but then we would
>> have to change the custom type of 'display-buffer-alist' - no great deal
>> either.  WDYT?
>
> The obvious downside is the need to tag hundreds of existing calls of
> display-buffer, pop-to-buffer, etc.

Just the ones we care about.  With our usual speed of progress in this
area I won't live long enough to see that happen.  Think of pearls like

          (select-window                        ; to return to
           (prog1 (selected-window)     ; WoMan window
             (select-window (display-buffer (current-buffer)))

martin