On 19 Jan 2012, at 14:05, Chris Ridd wrote: > > On 19 Jan 2012, at 14:00, John Devitofranceschi wrote: > >> >> >> On Jan 19, 2012, at 8:21, Chris Ridd <chrisr...@mac.com> wrote: >> >>> >>> On 19 Jan 2012, at 12:39, John Devitofranceschi wrote: >>> >>>> Two bugs! Such a deal! >>> >>> No extra charge :-) >>> >>>> Yes, the cookie setting code is in the while (1) loop and the callback >>>> merely prints out the dn of the returned entries. >>> >>> You mentioned that the OpenLDAP ldapsearch command-line tool seemed to >>> work. Can you double-check it is getting multiple pages back (try using >>> pr=2/prompt), or whether it is stopping after the first page because of the >>> misplaced result cookie? >> >> OpenLDAP's ldapsearch works as expected. Multiple pages get returned. > > That's puzzling, and suggests my analysis is wrong. Can you get some > (snoop/tcpdump) packet traces from ldapsearch and perl up to and including > the first page search done? It'll have passwords and your data in so if you > want to send them off-list that's fine.
Just to follow up on-list - the snoops that John sent me both showed that Oracle consistently sends the paged results control in the wrong place. Wireshark also complained. Unless Oracle's shipping a version of ldapsearch that reads the control from the wrong place and counteracts their server bug, I'm puzzled how John's ldapsearch works. Chris