bug#11716: 24.1; `substitute-command-keys' does not respect `defalias'

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

> > (defun foo () "foo" (interactive) (message "FOO"))
> > (defalias 'bar 'foo)
> > (global-set-key "\C-d" 'foo)
> > (substitute-command-keys "\\[bar]")
> >  
> > The last sexp returns "M-x bar".  It should return "C-d".
> 
> No.  C-d is not bound to bar.

No what?  I didn't say it was.

C-d is, _in effect_, that is, from a user perspective, bound to `bar'.  It is
not bound to `bar' but the effect is that it invokes the code that `bar' runs.

`substitute-command-keys' is used most often in doc strings and such,
communicating to users what they can use to invoke particular commands etc.

It is far more useful to let a user know that s?he can use C-d than it is to
just punt and say "M-x bar".

If a programmer explicitly wants to choose "M-x bar" over letting users know
they can use C-d, s?he can use "M-x bar" and not "\\[bar]".

And yes, the user CAN use C-d, because of the binding of C-d to `foo' and the
defalias of `bar' to `foo'.  What's important here is what users can do, not
whether `bar' is actually bound to C-d.

Do you have a counter use case?  A typical use of `substitute-command-keys' (or
even an atypical case) where fixing this would cause a problem?  A case where
the user should see a distinction between `foo' and `bar' wrt key bindings?