🚀 go-pugleaf

I using the Berkeley DB hash routines (recently picked up
from gatekeeper.dec.com - all sccs id 5.?, updated as recently as
6/17/91), primarily using the supplied ndbm interface. I'm running them
on a Sun Sparc 1+ under SunOS 4.1.1.

I had a database corruption problem so I wrote a simple test program
and discovered a problem when doing the following:

A new database is created and opened.
Records are added to it (small key and data (less than 50 characters total)).
All the records are retrieved (they're there).
All the records are successfully deleted (no non-zero return codes).
Then, a call to dbm_firstkey returns a non-NULL key pointer (ack!!)
which turns out to be some garbled part of what was previously in
a data record.

This problem does NOT happen when doing this with 100 or 1000 records,
but does for a 10000 record database.

Questions:
Has this problem or a similar one been noted before?
Is there a fix?

Where is the best place to get the latest bug fixes and versions?

An other ideas?

Thank you.

-- 
Ron McEwen     | email : chucker@bnr.ca | phone: (613) 765-4591
Bell-Northern Research Ltd., Ottawa, Ontario, Canada