hyc@symas.com writes:
I think you need to check what back-ldif is doing that triggers the false detection of a rename op.
No time now, but I expect it's because of the following. From memory, I hope I get the details right:
A back-ldif entry file only contain the entry's RDN, not the DN. The exact spelling of the suffix DN is not stored anywhere, except that of its RDN. And if you have stored the suffix entry "dc=NO" and add "dc=uio,dc=no", the lowercase in "dc=no" is lost.
Furthermore the backend does not read parent and grandparent entries to construct the right DN. Instead it uses the baseDN of the request, though when it reads an entry it does replace the RDN with that from the entry file. Thus a subtree search from the suffix ends up recursively constructing the right DNs, but a search based further down does not.
We could use the spelling of the suffix from slapd.conf's suffix. However that doesn't help operations with baseDN below the suffix, and unless we do read parent DNs it makes treatment of the suffix and other DNs inconsistent. Or we could cache all the DNs, but.. well, this is just back-ldif.