directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ole Ersoy <ole.er...@gmail.com>
Subject Re: [jira] Closed: (DIRSERVER-950) Server Left in Unstable State
Date Sat, 02 Jun 2007 20:28:27 GMT


Emmanuel Lecharny wrote:
> Ole Ersoy a écrit :
> 
>> Well,
>>
>> If you consider the server being left in an unrecoverable state due to 
>> basic JNDI operations then that's a good move.
>>
>> - Ole
> 
> Didn't you found that the cause was that there were some broken logic in 
> the test ? Here is your last comment :

Yes - I should have probably put a "Caution" around that statement.  There was broken logic
in the code the test was running that caused ObjectClass A not to be deleted before ObjectClass
B.  This then leads to the server being left in a state where JNDI operations on the context
containing ObjectClass A and ObjectClass B will fail.  For instance children of the context
cannot be listed, etc.  So the only way to recover is to reinstall the server.  

> 
> "When the test runs it is supposed to delete ObjectClass A first, but 
> the logic was broken"
> 
> If you have a clear test case to demonstrate a problem in the server, 
> and not in your test, then repoen the issue, and attach the test case.

The test is referenced in this comment:

https://issues.apache.org/jira/browse/DIRSERVER-950#action_12500469 

Sorry there were so many comments.  Especially the first one :-).  In the beginning I thought
this might be a tricky thing to track down, so that wanted to make sure I captured as much
information as possible.  I think if the server simply throws an exception when an attempt
to delete a schema entry that has other schema entries dependent on it is made, it will solve
the problem.  
> 
> I may miss something, but consider that the initial JIRA and the added 
> comments didn't help at all to understand where the problem is.


Sure - Sorry - I should have documented more clearly in the end.  Initially I just noticed
that I had to reinstall the server in order to recover, so I wanted to capture all the symptoms
around the process, before I attempted to isolate.  I'm changing things on my system so fast
that occasionally recovering the original state is very tricky.  This way at least there was
something.  I'll try to do it more cleanly though.

Thanks,
- Ole

Mime
View raw message