https://bugs.openldap.org/show_bug.cgi?id=10091
--- Comment #2 from philip.schildkamp@uni-koeln.de --- As a further note I might add, I'm defaulting to Apache Directory Studio to visually manage the server. This results in the first request upon connection to be against the `cn=subschema`. I've thereby narrowed the occurring segfault down to the following `ldapsearch` query, run against a `slapd` with the aforementioned configuration:
ldapsearch -b cn=subschema -s base '(objectClass=subschema)' +
When requesting all attributes (`+`) the error occurs. When only, e.g., requesting all the `objectClasses` from the `cn=subschema` the error vanishes. Which led me on to try this:
ldapsearch -b cn=subschema -s base '(objectClass=subschema)' memberOf
Which, again, resulted in `slapd` segfaulting. At this point I've verified that reducing the `dynlist` configuration to exclude the optional `+<memberOf-ad>@<static-oc>` options also alleviates the error (it does). And I was able to narrow the occurrence of the segfault down to the following, simple `ldapsearch` query:
ldapsearch memberOf
From the stacktrace I've been able to figure out that the segfault is cause by a null pointer when accessing `o.o_bd->be_search(...)`, as the backend db (`o_db`) isn't set. But any of my further conclusions would more or less be guesswork at this point.