directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Karasulu (JIRA)" <j...@apache.org>
Subject [jira] Commented: (DIRSERVER-1104) Mixing Attribute value types results in write failures
Date Wed, 21 Nov 2007 04:39:43 GMT

    [ https://issues.apache.org/jira/browse/DIRSERVER-1104?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12544164
] 

Alex Karasulu commented on DIRSERVER-1104:
------------------------------------------

I went in to fix this thinking it would be simple and ran into all these problems:

(1) Why was the serialization structure of the partition changed?  We are now storing an extra
byte at the beginning of a value to indicate that it is a byte[] or a String.

The presumption that all values of an attribute are either a byte[] or a String is incorrect
so doing this is a problem unless we can convert the value to the proper format correctly
by considering the syntax of the value.  This however is not a simple matter.  There may be
no way this far into the persistence code to determine the syntax associated with a value
being persisted.  This alteration was a bad move IMHO and screws us a bit.

(2) Why has all this code been moved to ldap-shared when it deals specifically with the serialization
mechanism used by the JDBM partition implementation?

Too many things are being put into the ldap-shared module.  It's a jumbled up mess in there
and it's only getting bigger.  Only put things into shared if it really can be reused.  This
very specific serialization mechanism may not be.




> Mixing Attribute value types results in write failures
> ------------------------------------------------------
>
>                 Key: DIRSERVER-1104
>                 URL: https://issues.apache.org/jira/browse/DIRSERVER-1104
>             Project: Directory ApacheDS
>          Issue Type: Bug
>          Components: core
>    Affects Versions: bigbang, 1.5.1
>            Reporter: Alex Karasulu
>            Assignee: Alex Karasulu
>            Priority: Blocker
>             Fix For: bigbang
>
>         Attachments: error.log, offending.ldif
>
>
> This new bug is really serious and tricky. It has eluded us but comes out now thanks
to a crazy size effect it has on Windows.  
> First a characterization of the bug.  An entry (Attributes object) may be valid and still
contain an Attribute which has both byte[] and String array values.  Some may suggest that
this is invalid since attributes are, according to their syntax, either binary of human readable.
 However this is not the case.  The fact that an attributeType is human readable has no bearing
on how the user supplies the value.  Human readable data can be provided as binary information
so long as it still conforms to the syntax of the attribute.
> Here's an example entry which would cause such a failure:
>     dn: cn=person1,ou=system
>     objectClass: organizationalPerson
>     cn: person1
>     sn: sn_person1
>     seealso: cn=Good One,ou=people,o=sevenSeas
>     seealso:: Y249QmFkIEXDqWvDoCxvdT1wZW9wbGUsbz1zZXZlblNlYXM=
> This entry will cause the AttributeSerializerUtils.serialize() method to blow a ClassCastException.
 Note the log of the error can be found attached to this issue.  

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message