bug#25109: 25.1.50; `symbol-file' is sometimes lying

[Michael Heerdegen]

Hi,

emacs -Q:

(symbol-file 'pcase)
  ==> "pcase"

(require 'subr-x)
  ==> subr-x

(symbol-file 'pcase)
  ==> "/usr/local/share/emacs/25.1.50/lisp/emacs-lisp/subr-x.elc"

`pcase' is not (re)defined in subr-x.

Seems the reason for the wrong result is that the entry

   (require . pcase)

in subr-x's entry in `load-history' is mistakenly treated as a
definition of the symbol "pcase".

Dunno if there are other cases like this (or unlike this).  If we can't
be sure that `symbol-file' returns the correct value in any case, we
should say in the documentation that it works only heuristically.  But I
guess we just need to be more accurate.


TIA,

Michael.



In GNU Emacs 25.1.50.10 (x86_64-pc-linux-gnu, GTK+ Version 3.22.3)
 of 2016-11-22 built on drachen
Repository revision: 3f2b497cdddcf1f124f1c9f94354d5f457a4070f
Windowing system distributor 'The X.Org Foundation', version 11.0.11804000
System Description:     Debian GNU/Linux testing (stretch)