Save old/new indexmasks to allow processing to resume if slapd is
stopped and restarted. Save last entryID indexed to allow resume
after restart or pool pause.
- LDAP MODRDN handling
- support paged response
- add wt_extended
- add config emit
- wt_key_read() return WT_NOTFOUND if not found key.
- add ext_candidates()
- fix idlcache session name
- fix warning
- don't reuse idlcache cursor
- set correct pid when modrdn with newsuperior
- fix condition bug
- fix send_search_entry() error handling
- fix for referral handling
- fix for readonly mode
- fix sizelimit response
- support modrdn
- improve modify handling
- clear ancestor idlcache
- fix for multi-DIT
- IMPORTANT CHANGES: Compatibility is broken with previous database table, please restore database from LDIF.
- checking for scope=children
- sort dn2idl result
- fix cursor leak
- support db_open with readonly mode
- add wt_tool_entry_delete
- initialize comp variable
- support referrals
- implement wt_tool_dn2id_get() and wt_tool_entry_modify() for slapadd -w
- skip redundant scan, and more debug message
- fix OID conflict with back-passwd
- no need to close session, It may cause SEGV.
- fixed wt_dn2entry for empty DN
- support multiple database
- Construct wiredtiger's config parameter. It allow multi line wtconfig settings
- add idlcache
- fix concurrent modification to a entry with multi values
- prevent to add duplicate dn entry
- suppress error message "search_near failed: WT_NOTFOUND"
- update Debug statements
- back-wt does not support subtree rename
- fix for @ondra review
- update slapd-wt.5 and warning for mode option
- add back-wt test into test target
- add scope checking
Add a check to see if pwdChangedTime was actually present on the entry. If not, skip the expiry check.
Additionally change the debug log statement to TRACE instead of ANY, as the message is informational.
Some config changes need to be prepared at entry modification time and
then checked/committed at the end. Such a check can fail, so we need to
do this early enough and also commit the opposite change during a revert
step.
The assumption is that the original version is consistent and bconfig
can accurately reconstruct it, so the revert cleanups will not fail and
so all of them get to be run. We set ca->reply.err so that the cleanups
can be made aware and comply with this assumption.
