bug#24543: 25.1; describe-symbol doesn't show face with same name as a variable

[Alex]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Alex <agrambot@gmail.com>
>> Date: Fri, 16 Jun 2017 15:54:45 -0600
>> 
>> > emacs -Q
>> > C-h o font-lock-comment-face RET
>> >
>> > Only the variable of this name is shown, not the face (which can be
>> > viewed with describe-face, of course).
>> >
>> > It appears that when the face only shares a name with a function, then
>> > both the function and face are shown. Only a variable shadows the face
>> > in describe-symbol.
>> 
>> This is because describe-symbol isn't returning (buffer-string) like
>> describe-{function, variable} are.
>> 
>> I've attached a patch fixing this problem.
>
> Thanks.  This LGTM, but doesn't the fact that describe-face's return
> value will change after this constitute an incompatible change?  And

I suppose so; would it be expected to be documented in NEWS?

> if so, could it break some of the users of this function out there?

Potentially, but I don't see that being an issue in this case. The
return value isn't documented, so users shouldn't be (and likely aren't)
relying on describe-face always returning nil. It makes more sense to
treat describe-face like describe-{function, variable}.

Though I'm not sure why those two procedures return (buffer-string) in
the first place. In any case, I figure it's better to leave those and
describe-symbol alone and just make describe-face behave similarly.