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

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

And in fact, if I do not redefine `describe-function-1', i.e., if I
evaluate the original definition, then I get the same error.

I did `debug-on-entry' for `help-fns--autoloaded-p' and then traced
things until the error was produced.  This is what I saw:

All of the calls to `file-truename' that I saw returned the right values.

There were multiple calls to `file-name-sans-extension', each of which,
except for the last, was passed an absolute file name and returned the
right value.  The last of these correct calls was this one:

Debugger entered--entering a function:
* file-name-sans-extension("d:/Emacs-24-2014-01-23/share/emacs/24.3.50/lisp/emacs-lisp/cl-indent.elc")

That returned, correctly,
"d:/Emacs-24-2014-01-23/share/emacs/24.3.50/lisp/emacs-lisp/cl-indent"

(None of the calls involved file handlers or symlinks, FWIW.)

The next thing I saw in the debugger was this bizarre call to
`file-name-sans-extension':

Debugger entered--entering a function:
* file-name-sans-extension((require . fit-frame))
* #[514...]
* apply(#[514...] (message-tab "d:/Emacs-24-2014-01-23/share/emacs/24.3.50/lisp/gnus/message.el"))
* 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)

That is, as soon as the previous call to file-name-sans-extension
returned, correctly (and I traced that call step by step, using `d'),
`d' in the debugger came up with what you see above.

I also tried `debug-on-entry' for `describe-function-1' (the vanilla
Emacs version) and then  `C-h f message-tab'.  Same thing.

My wild guess is that the problem comes somehow from using Cygwin (?).