Re: (ITS#3546) Sync rep provider and server crash on SIGTERM

[m.d.t.evans@qmul.ac.uk]

Complete guesswork on my part... undoing a change made in 1.295 seems to
fix this behaviour.

bash-3.00$ diff -u connection.c.0 connection.c
--- connection.c.0      2005-01-20 17:01:06.000000000 +0000
+++ connection.c        2005-02-17 10:15:14.578895942 +0000
@@ -772,17 +772,10 @@
 {
        /* c_mutex must be locked by caller */

-       Operation *o, *next, op = {0};
-       SlapReply rs = {0};
+       Operation *o;

-       op.o_conn = c;
-       op.o_connid = c->c_connid;
-       op.o_tag = LDAP_REQ_ABANDON;
-       for ( o = LDAP_STAILQ_FIRST( &c->c_ops ); o; o=next ) {
-               next = LDAP_STAILQ_NEXT( o, o_next );
-               op.orn_msgid = o->o_msgid;
+       LDAP_STAILQ_FOREACH(o, &c->c_ops, o_next) {
                o->o_abandon = 1;
-               fe_op_abandon( &op, &rs );
        }

        /* remove pending operations */


I've also found that it seems to fix provider core dumps that happen
when the following search is performed:

openldap/bin/ldapsearch -H ldap://localhost:11389 -b"<base>" -E sync=rp

The dumps do not happen when a sync=ro search is done.

Cheers,
Martin.

-- 
-- Dr MDT Evans, Computing Services, Queen Mary, University of London