[Bug-openmcl] file-length

[bryan o'connor]

i'm sure this won't be the last time file-length gets hacked
for ansi-compliance..

here's my current version based off your changes:

(defun file-length (stream)
   (etypecase stream
     ;; Don't use an OR type here
     (file-stream (stream-length stream))
     (synonym-stream (file-length
                      (symbol-value (synonym-stream-symbol stream))))
     (broadcast-stream (let* ((last (last-broadcast-stream stream)))
                         (if last
                             (file-length last)
                           0)))))

and the justification:

the recursion is there because i think it's valid for a
synonym-stream to point to a broadcast-stream and a
broadcast-stream to contain synonym-streams.

i didn't want to add any type-checking to stream-length
because it seems fair that it could be useful without the
type restriction.

this also means that stream-length specialized on synonym-
and broadcast-streams doesn't get called any more.

and now the catch:

the ansi-tests file-length.error.7 and file-length.error.8
contradict make-broadcast-stream.5.  the ansi spec seems to
contradict itself on this issue also.. i suspect that the
right thing is that make-broadcast-stream.5 is correct and
the result should be 0 and not a type-error.