possible hdb cache/modrdn issue

[Aaron Richton <richton@nbcs.rutgers.edu>]

One of my colleagues is writing an application and started to add some 
modrdn ops to the code today. He got confused (rightly so) at some entries 
disappearing, when modrdn from

cn=e,cn=d,dc=c,dc=b,dc=a

to

cn=e,cn=z,dc=c,dc=b,dc=a

wherein the changed entries would appear in slapcat output, but not in 
ldapsearch output. He also noted that modrdn to cn=e,dc=c,dc=b,dc=a was 
successful.

After playing around with test005 to make sure that the syntax was right, 
I started to suspect some sort of hdb caching issue. So we kicked the 
slapd in question, and sure enough, ldapsearch and slapcat matched 
following the restart.

I realize this isn't much to go on. For now, I'm just giving a heads up 
and requesting a general call of if anybody else has seen something like 
this. We'll try to base a repro off of test005 next week--I'm hopeful 
because this test database is on one of our simplest servers (no syncrepl, 
only scores of entries, etc). This is with 2.3.24.