18 Jul, 2011

1 commit

  • Many thanks to Nemanja Lukic and his colleagues at Profiq S.R.O for this kind contribution.
    
    The contribution has been modified as follows:
    
    *) integrate as core feature of OpenDJ, not extension
    *) rename package names accordingly
    *) remove dependency on BouncyCastle.org MD4 implementation, by porting implementation to OpenDJ
    *) re-implement setOddParity method in order to avoid potential copyright issues with existing implementations.
    *) simplify configuration model and validation.
    
    TODO:
    
    *) support u: and dn: authid in password modify extended operation.
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@7116 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     

15 Jul, 2011

1 commit


08 Jul, 2011

1 commit


05 Jul, 2011

3 commits


04 Jul, 2011

1 commit

  • The conflict resolution code didn't consider the case where a single modify operation could contain multiple modifications on an attribute, especially a replace after adds or delete. All changeNumbers would then be equals. Replace always wins over all precedent operations on the attribute, as Modifications are ordered (Sequence of).
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@7090 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     

30 Jun, 2011

1 commit


29 Jun, 2011

2 commits


28 Jun, 2011

1 commit


24 Jun, 2011

2 commits


23 Jun, 2011

3 commits


22 Jun, 2011

1 commit

  • The value part of control specification is optional. And therefore there should be no : to finish the specification.
    Now if the specification ends with :, then it's the same as no value.
    Test added with the ManageDSAIt control.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@7049 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     

15 Jun, 2011

1 commit


13 Jun, 2011

2 commits


10 Jun, 2011

2 commits


09 Jun, 2011

3 commits

  • …ns servers which have been subjected to different purging policies.
    
    The issue of ECL change number inconsistency across servers is not tied to purging policies but to the fact that the CN is based on a lazy built index. As a result, the last change number in the index database depends on when Searches are done against the server and not what are really the changes available. The lastChangeNumber virtual attribute uses the index last CN and adds the changes received and eligible.
    
    The changes here are of fixing the following :
    After purging of the DraftCNDB, the next value will be incremented from last known value. As a result, changeNumbers are always increasing. Purging the DraftCNDB is now stopping at the first index that still points to a valid change in the changelog. The counter of changes has been fixed to take into account the fact that the changelog keeps at least one change older than the trim age in order to preserve the validity of the Cookie for at least "purge delay".
    Also contains some optimization of the DraftCNDBCursor to avoid reading the entries pointed by the cursor several times.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@6994 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     
  • Publish original attributes from the target entry as a single value LDIF blob in "includedAttributes" in a similar manner to the "changes" attribute. 
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@6993 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     
  • Remove targetUniqueID attribute and instead declare it as an alias of targetEntryUUID.
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@6991 41b1ffd8-f28e-4786-ab96-9950f0a78031
    matthew
     

08 Jun, 2011

2 commits


07 Jun, 2011

1 commit


06 Jun, 2011

2 commits


01 Jun, 2011

2 commits


31 May, 2011

2 commits


30 May, 2011

3 commits

  • …hanges with updates and purging.
    More specifically these changes are resolving the following issues :
    OPENDJ-57 - ECL: lastChangeNumber and firstChangeNumber reset to zero when the changelog is purged to empty
    OPENDJ-172 - External ChangeLog Cookie varies when searching with an empty cookie. Cookie should be reproducible.
    OPENDJ-173 - External ChangeLog cookies content is altered by Change purging and prevents from continuing search with a previous returned cookie.
    
    The changes are multiples and interdependant, hence a single commit.
    But : 
    - The DraftCNDB is now purged but always keep the last record. The firstKey value is updated accordingly.
    - The ReplicationDB is purged but always keep the last record before the trimdate.
    - If no cookie is specified, the initial cookie is computed from the ReplicationDomain StartState, but keeping only change numbers older than the trimdate. 
    - Fix a possible issue when searching a ChangeNumber, and a Count record is found, the changeNumber value was not correct.
    
    These changes mostly revert revision  6406 which meant to fix the same issue (cookies would become obsolete and a search using them returns an Unwilling To Perform), but only managed to hide the root cause of the issue.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@6943 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     
  • No brainer : If there's no value, create an empty MultiDomainServerState. The server already knows how to deal with them. 
    
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@6938 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     
  • Fix now check if the control is present in the search and returns it only if so.
    Tests have been added in unit-tests to verify this.
    
    git-svn-id: https://svn.forgerock.org/opendj/trunk@6936 41b1ffd8-f28e-4786-ab96-9950f0a78031
    ludo
     

19 May, 2011

2 commits


16 May, 2011

1 commit