Hi Alex!
With "a diff" you mean a subversion diff output? (Probably not Unix diff
...)
That is, I just enter "svn diff" (found this in a svn tutorial) and you
are happy with the output (that is, I attach this output as a textfile)?
Sorry for the probably stupid question, but I'm not quite familiar to
subversion yet. Normally using CVS/Clearcase from Eclipse-plugins (ok,
lame).
Thanks in advance,
Stefan
Alex Karasulu (JIRA) wrote:
[ http://issues.apache.org/jira/browse/DIREVE-215?page=comments#action_12317931 ]
Alex Karasulu commented on DIREVE-215:
--------------------------------------
Thanks a bunch Stefan but next time please submit a diff instead of the file
itself. These are easier to review and apply but hey this is better than most
:).
I agree about the stack trace however that was for debuging purposes. I will
use the log level to not show this on the client as well as print the stack
trace out to the server log file. It's pretty helpful to get the trace on the
client side when you're working on separate machines. However for production
this should not be the case. Again using nlog4j should help us out
considerably in terms of controlling this behavoir.
Wrong error messages for some operations
----------------------------------------
Key: DIREVE-215
URL: http://issues.apache.org/jira/browse/DIREVE-215
Project: Directory Server
Type: Bug
Components: protocol
Versions: 0.9.1
Reporter: Stefan Zoerner
Assignee: Alex Karasulu
Priority: Minor
Fix For: 0.9.2
Attachments: DeleteHandler.java
If a client tries to remove an entry, and the deletion fails because the entry
is not a leaf (error code 66), the LDAP message is wrong.
Status code and Exception type within the server are correct (as detected by the
corresponding test case). But the message text itself contains the words "failed to
add entry". This information is sent to the client, which may display it, e.g.
* JNDI includes it during creation of appropriate Exception
(com.sun.jndi.ldap.LdapCtx.mapErrorCode(66, "failed to add entry ...").
* Command line tools like ldapmodify display it on the console.
Example: If the following LDIF is applied to ApacheDS
---
# create entry
dn: ou=noLeaf,ou=system
changetype: add
ou: noLeaf
objectclass: top
objectclass: organizationalUnit
# create entry below the first
dn: ou=leaf,ou=noLeaf,ou=system
changetype: add
ou: leaf
objectclass: top
objectclass: organizationalUnit
# delete subtree
dn: ou=noLeaf,ou=system
changetype: delete
---
it goes
$ ldapmodify -h magritte -p 10389 -D uid=admin,ou=system -w **** -f
deleteSubtree.ldif
adding new entry ou=noLeaf,ou=system
adding new entry ou=leaf,ou=noLeaf,ou=system
deleting entry ou=noLeaf,ou=system
ldap_delete: Operation not allowed on nonleaf
ldap_delete: additional info: failed to add entry ou=noLeaf,ou=system:
... (very long Stack trace deleted) ...
$
I modified the error message within class org.apache.ldap.server.protocol.DeleteHandler to
"failed to delete entry ..." (attached), although I would prefer a more detailed message
(like "Not Allowed On Non-leaf", as Sun Directory does). But this is not possible at that
general place without an ugly switch or so.
The same "defect" (probably a cut/copy/paste of error message from AddHandler)
is present in CompareHandler, ModifyDnHandler, and ModifyHandler.
I would also suggest not to send the complete stack trace back to the client. I
have removed it in the attached version of the DeleteHandler and had a better
command line experience with ApacheDS (the stack traces should go in a server
log only from my point of view). This applies to other handlers (e.g. Add) as
well.
