Thanks for looking at that, Ian. The extensive test suite which the original authors started and Ian and I have improved gives us a lot of leverage here; if the change keeps us green on all tests then we are probably safe.
On Sat, 2007-09-08 at 15:49 -0400, Ian Eslick wrote: > Let me look at the history for that file. I can't remember if I > fixed some of this on the working branch (darcs) or back during the > 0.9.0 release. There are a bunch of subtle interactions that can > disturb each other, but that does look like a problem doesn't it? > > Ian > > > On Sep 8, 2007, at 3:28 PM, Chris Laux wrote: > > > So I had a look at it myself and it turned out to be much easier > > than I > > had expected, which is what kept me from trying in the first place. > > > > The following method in src/db-bdb/bdb-collections.lisp is the > > culprit: > > > > (defmethod cursor-prev ((cursor bdb-cursor)) > > (if (cursor-initialized-p cursor) > > (let ((sc (get-con (cursor-btree cursor)))) > > (with-buffer-streams (key-buf value-buf) > > (multiple-value-bind (key val) > > (db-cursor-move-buffered (cursor-handle cursor) > > key-buf value-buf :prev t) > > (if (and key (= (buffer-read-oid key) (cursor-oid cursor))) > > (values t (deserialize key sc) > > (deserialize val sc)) > > (setf (cursor-initialized-p cursor) nil)))) > > (cursor-last cursor)))) > > > > That "cursor-last" call at the end looks very suspicious, and indeed > > removing it solves the problem. But I don't know where it belongs > > or if > > it can be safely removed. Any ideas? > > > > Cheers, > > > > Chris Laux > > > > > > Chris Laux wrote: > >> Hi again, > >> > >> I seem to have stumbled across another bug: the from-end argument > >> doesn't seem to be working, map-btree always calls with the same > >> (presumably last) entry when it is set. > >> > >> (defun test-btree-from-end (btree from-end) > >> (let (acc > >> (n 0)) > >> (map-btree #'(lambda (k v) > >> (push (list k v) acc) > >> (when (= (incf n) 4) > >> (return-from test-btree-from-end acc))) > >> btree :from-end from-end))) > >> > >>> (test-btree-from-end ** nil) > >> ((3398012969 #<MESSAGE oid:227324>) (3398012898 #<MESSAGE oid: > >> 226751>) > >> (3398012815 #<MESSAGE oid:226044>) (3398012772 #<MESSAGE oid: > >> 225621>)) > >> > >>> (test-btree-from-end *** t) > >> ((3398231130 #<MESSAGE oid:552110>) (3398231130 #<MESSAGE oid: > >> 552110>) > >> (3398231130 #<MESSAGE oid:552110>) (3398231130 #<MESSAGE oid: > >> 552110>)) > >> > >> My setup is BerkeleyDB 4.5.20 with Elephant 0.9.0 with SBCL 1.0.4 on > >> Linux/x86-64. > >> > >> Thanks, > >> > >> Chris Laux > >> > > _______________________________________________ > > elephant-devel site list > > elephant-devel@common-lisp.net > > http://common-lisp.net/mailman/listinfo/elephant-devel > > _______________________________________________ > elephant-devel site list > elephant-devel@common-lisp.net > http://common-lisp.net/mailman/listinfo/elephant-devel _______________________________________________ elephant-devel site list elephant-devel@common-lisp.net http://common-lisp.net/mailman/listinfo/elephant-devel
