bug#45234: Annotation function vs. Affixation function

[Juri Linkov]

> On current master the results of the new annotation function are passed to
> the affixation function in minibuffer-completion-help:
>
>     (when ann-fun
>       (setq completions
>             (mapcar (lambda (s)
>                       (let ((ann (funcall ann-fun s)))
>                         (if ann (list s ann) s)))
>                     completions)))
>     (when aff-fun
>       (setq completions
>             (funcall aff-fun completions)))
>
> If the annotation function returns a result it's packed into a list
> which should be an unexpected format for the affixation function?
> Maybe when both are defined only the affixation function should be used?

The current implementation assumes that the affixation function
can be smart enough to recognize that completions were modified
by a previous call of the annotation function.

If you can demonstrate that this is not always the case, then
the implementation could be changed, there is enough time before
the next release, after that it would be harder to change this.