Bruno,
I am using jonathan 2.0 (according to org.objectweb.jonathan.Release) and
StdStubFactory.
I found the definition of a method called hashcode() somewhere in the
jeremie.libs.stub_factory.stdRefImpl class
renaming this method to hashCode() seems to correct the problem.
Anne
>-----Message d'origine-----
>De: Bruno Dumant [mailto:[EMAIL PROTECTED]]
>Date: vendredi 4 mai 2001 12:20
>�: GERODOLLE Anne FTRD/DTL/GRE
>Cc: [EMAIL PROTECTED]
>Objet: Re: hashCode on stubs
>
>
>
>
>GERODOLLE Anne FTRD/DTL/GRE wrote:
>
>> Hello,
>>
>> I have a small question concerning Jeremie stubs.
>>
>> As far as I have experienced, two stubs corresponding to the
>same remote
>> object are considered "equals", but do not return the same hashCode.
>>
>> Note that two different java.rmi stubs to the same remote
>object return the
>> same hashcode.
>>
>> Is that a bug ?
>
>Apparently, yes. Two equal objects must have the same hashcode.
>
>Actually, the main difficulty is to check that two stubs
>correspond to the same
>remote object, because it requires comparing remote
>identifiers that are not
>necessarily comparable. In old versions of Jonathan, two stubs were not
>considered equal. But this was corrected quite a long time ago
>because Jonas
>needed it (to store objects in hashtables), through the
>implementation of an
>ad-hoc conservative mechanism, that ensures that if the
>equals() method returns
>yes, then the target objects are certainly equal. But, to my
>knowledge, the
>hashcode was always coherent with the result of the equals() method.
>
>If I remember well, the equals and hashcode methods are computed by the
>Reference object designated by the stub, not by the stub
>itself. Two different
>Reference implementations are available in Jeremie : one is
>present in the
>StdStubFactory (used by JIOP, thus in the classical
>client/server case), the
>other in the EventChannelFactory. In which case did you
>experience the problem
>?
>
>
>> What would be the best method if I want to build a hashtable
>> associating some values to remote objects ?
>
>Jonas already does something like that, and it works...
>
>> I can imagine some methods, like
>> using a key that is a "stub wrapper" :
>> class StubWrapper{
>> Object contents;
>> public boolean equals(Object other) {
>> if (! (other instanceof StubWrapper)) return false;
>> if (
>((StubWrapper)other).contents.equals(contents)) return
>> true;
>> return false;
>> }
>> public int hashCode(){ return 0;}
>> }
>>
>> but it does not look very nice...
>
>Indeed...
>
>I will have a look at your problem anyway. Which version of
>Jonathan are you
>using ?
>
>Bruno
>
To unsubscribe, send email to [EMAIL PROTECTED] and
include in the body of the message "unsubscribe jonathan".
For general help, send email to [EMAIL PROTECTED] and
include in the body of the message "help".
