bug#20841: 25.0.50; Turning off `which-function-mode' does not remove its last effect from the minibuffer

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

> > Subject line says it all.
> 
> I'm unable to reproduce this on the trunk.  I see that there have been
> changes to the mode in the last year, but skimming the changes, I don't
> see anything in the "switch the mode off" path...
> 
> Are you still seeing this?

Yes, I'm still seeing it.  I turn it off and I still see things
like `[Documentation]' and `[???]' in the mode line of existing
buffers.

When this is true of *Help*, for instance, i.e., it remains
in the mode line after I've toggled it off, if I then ask
for something else to be displayed in *Help* (e.g. do `C-h f'),
then it is removed from the mode line.

I see this in my setup, but my setup does nothing with
`which-function-mode'.  One guess is that the bug might
manifest itself only if you have a standalone minibuffer.

Another guess might be (since ediff uses a separate control
frame (?)) that it is similar to whatever is behind this
comment in the code:

;; FIXME: Why does ediff require special support?

Another guess is that it might have to do with
`mode-line-misc-info'.

After I've turned the mode off I see that the value of
variable `which-func-mode' remains `t' in buffers, even
though the global value has been changed to `nil'.

I don't have time to try to debug this.
Close it if no one else has the time.  I no longer
turn the mode on at all, to avoid the annoyance.