quanah@zimbra.com wrote:

--On Wednesday, August 13, 2014 8:27 AM +0000 dieter@dkluenter.de wrote:

...

Sorry, it is back-ldap with pcache and it seems to be a pcache problem. When disabling pcache, slapd runs fine.

a) Confirm you still see this issue with current RE24 if you want it fixed in the 2.4 series, as in general we're planning on 2.4.40 to be the final release for 2.4

b) If you are able to reproduce this with current RE24, please provide an example configuration that causes the issue. test020 passes w/o issue.

The example config is invalid.

### include schema/core.schema include schema/cosine.schema include schema/inetorgperson.schema # modulepath ../servers/slapd/overlays moduleload pcache.la

database ldap uri ldap://example.com suffix dc=example,dc=com rootdn dc=example,dc=com rootpw xxx

overlay pcache directory /tmp/testr/db.1.a pcache mdb 50000 2 500 3600 pcacheAttrset 0 mail uid pcacheAttrset 1 cn telephoneNumber pcachePersist TRUE pcacheTemplate (mail=) 0 10800 7200 pcacheTemplate (uid=) 1 10800 7200 index objectclass pres index mail,uid,telephoneNumber pres,eq

database monitor ###

"directory" is not a valid config keyword for the pcache overlay. It is a valid keyword for back-mdb, but here it was given before the "pcache mdb..." line so slapd doesn't yet know that back-mdb will be used. The slapo-pcache(5) manpage clearly states what directives are valid.

Personally I'd be fine with leaving this to SEGV as it does. Catching user errors doesn't help much in the long run; either way slapd will fail to start so there's no difference either way.

But in this case, the SEGV is now prevented in git master.