bug#16553: 24.3.50; `file-truename' returns a cons? (wrong-type-argument stringp (...))

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

I ran into this when using my setup, not emacs -Q

Debugger entered--Lisp error: (wrong-type-argument stringp (require . fit-frame))
  file-name-nondirectory((require . fit-frame))
  file-name-sans-extension((require . fit-frame))
  help-fns--autoloaded-p(message-tab 
                         "d:/Emacs-24-2014-01-23/share/emacs/24.3.50/lisp/gnus/message.el")
  describe-function-1(message-tab)
  describe-key("	" 1 nil)
  call-interactively(describe-key nil nil)
  command-execute(describe-key)

I do redefine `describe-function-1', but I don't see why that should be
involved here.  It seems to me that the call to `help-fns--autoloaded-p'
is valid, and from then on only vanilla code is involved, AFAIK.

`help-fns--autoloaded-p' does this, at the start:

(let* ((file (file-name-sans-extension (file-truename file)))

From the above backtrace, I can only guess that `file-truename' applied
to that file name must be returning a cons.  However, if I use `M-:' to
evaluate:

(file-truename "d:/Emacs-24-2014-01-23/share/emacs/24.3.50/lisp/gnus/message.el")

then I get the same absolute file name back - no cons.  So I'm puzzled.

In GNU Emacs 24.3.50.1 (i686-pc-mingw32)
 of 2014-01-23 on ODIEONE
Bzr revision: 116129 monnier@iro.umontreal.ca-20140123150141-qopqqhpm8jqo8a18
Windowing system distributor `Microsoft Corp.', version 6.1.7601
Configured using:
 `configure --prefix=/c/Devel/emacs/binary --enable-checking=yes,glyphs
 'CFLAGS=-O0 -g3' LDFLAGS=-Lc:/Devel/emacs/lib
 CPPFLAGS=-Ic:/Devel/emacs/include'