[ApacheDS] New issues with name parser

[Alex Karasulu <ao...@bellsouth.net>]

Emmanuel,

I'm thinking some of the recent changes on the name parser may have 
introduced new problems.

These are really weird problems.  They happen only when running 
regression tests in maven within the server-unit module.  Yeah crazy.  
Basically to reproduce run "mvn -Dregression test" inside the 
apacheds/server-main module.  These are the tests that fail:

$ mvn -Dregressions test
[INFO] 
----------------------------------------------------------------------------
[INFO] Building ApacheDS Server Unit
....
T E S T S
-------------------------------------------------------
....
[surefire] Running org.apache.directory.server.AddObjectClassesToEntryTest
[surefire] Tests run: 3, Failures: 0, Errors: 3, Time elapsed: 0.113 sec 
<<<<<<<< FAILURE !!
[surefire] Running org.apache.directory.server.MiscTest
[surefire] Tests run: 7, Failures: 0, Errors: 7, Time elapsed: 0.177 sec 
<<<<<<<< FAILURE !!
[surefire] Running org.apache.directory.server.MatchingRuleCompareTest
[surefire] Tests run: 3, Failures: 0, Errors: 3, Time elapsed: 0.08 sec 
<<<<<<<< FAILURE !!

Now when these tests run inside eclipse there is no problem.  Here's 
what teh surefire report
looks like for these failures.  They are all pretty much the same trace:
========================================================================================================
o.a.d.shared.ldap.exception.LdapConfigurationException: Failed to 
normalize the suffix: ou=system
[Root exception is o.a.d.shared.ldap.exception.LdapInvalidNameException: 
Parser failure on name:
        ou=system
Antlr exception trace:
line 1:1: unexpected token: ou
        at 
o.a.d.shared.ldap.name.antlrNameParser.attributeTypeAndValue(antlrNameParser.java:192)
        at 
o.a.d.shared.ldap.name.antlrNameParser.nameComponent(antlrNameParser.java:120)
        at 
o.a.d.shared.ldap.name.antlrNameParser.name(antlrNameParser.java:69)
        at o.a.d.shared.ldap.name.DnParser.parse(DnParser.java:208)
        at o.a.d.shared.ldap.name.LdapName.<init>(LdapName.java:115)
        at 
o.a.d.server.core.configuration.DirectoryPartitionConfiguration.setSuffix(DirectoryPartitionConfiguration.java:203)
========================================================================================================

The suspect code is at 
o.a.d.server.core.configuration.DirectoryPartitionConfiguration.setSuffix(DirectoryPartitionConfiguration.java:203:

    /**
     * Sets the suffix of the {@link DirectoryPartition}.
     */
    protected void setSuffix( String suffix ) throws NamingException
    {
        suffix = suffix.trim();
        try
202:  {
203:        new LdapName( suffix );       <<<<<<<<================== HERE!!!
204:  }
        catch ( NamingException e )
        {
            throw new LdapConfigurationException( "Failed to normalize 
the suffix: " + suffix, e );
        }
        this.suffix = suffix;
    }

This is crazy though because how the heck is the parser failing on 
"ou=system" and succeeding on all test cases I cannot understand.  I 
think I'm loosing my mind on this one this is why I need a second 
opinion.  Please advise.

Thanks,
Alex

Re: [ApacheDS] New issues with name parser

[Alex Karasulu <ao...@bellsouth.net>]

Alex Karasulu wrote:

> Emmanuel,


Ok this is part II:

I have found a way to reproduce the error in eclipse.  Looks like 
running org.apache.directory.server.jndi.ServerContextFactoryTest before 
running org.apache.directory.server.AddObjectClassesToEntryTest makes 
the parser fail.  This parser is a reused static parser by the LdapName 
class.  I think something is going stale in the parser after the first 
tests are run. 

Try deleting the other tests minus these two in server-main module. Then 
run these two tests together in elcipse using the run option for all 
tests in the project.  Then you can reproduce the error within eclipse.

I am debugging now and for some reason the parser starts to see the 
first token "ou" in the case of the "ou=system" string as a WS token of 
type = 16.  This is a whitespace token.  I think deep down the lexer is 
messed up somewhere.  I'll try to isolate what is making the lexer freak.

Alex