Submission for slug-info

[daemon@potomac]

Path: potomac!daemon
From: garza@ghostwheel.DB.MCC.COM
Newsgroups: slug-info
Subject: Symbolics (Genera 7) BUG
Message-ID: <155@potomac.dc.ads.com>
Date: 21 Jan 87 04:12:00 GMT
Sender: daemon@potomac.dc.ads.com
Lines: 44


  There is a bug in Symbolics Common Lisp (Genera 7.0) when one calls

remhash in the body of the maphash function.  Guy Steele (page 285) points

out that if entries are added/deleted from a hash table while maphash is in

progress the results are unpredictable unless the function calls remhash to
                                       ------

remove the entry currently being processed.

Test::

   (setq x (make-hash-table))
   (dotimes (i 100)
      (setf (gethash i x) i))
   (maphash #'(lambda (k v) (format t "~% k = ~d v = ~d" k v)) x)
   (maphash #'(lambda (k v) (remhash k x)) x)
   (maphash #'(lambda (k v) (format t "~% k = ~d v = ~d" k v)) x)

The hash table should now become empty, but it is found to be not so.

However, if one indicates that the '= comparator is desired during

the allocation of the hash table, it works fine.

   (setq x (make-hash-table :test #'=))
   (dotimes (i 100)
      (setf (gethash i x) i))
   (maphash #'(lambda (k v) (format t "~% k = ~d v = ~d" k v)) x)
   (maphash #'(lambda (k v) (remhash k x)) x)
   (maphash #'(lambda (k v) (format t "~% k = ~d v = ~d" k v)) x)


Now the table is empty, as it should be.

PS: The above bug does not happen under Rel 6.
Jorge F. Garza