RE: MapImpl

[Fred Wan]

> Hmm.  I get that same output on Linux with Swarm-2.1.72.  With a new
> Swarm on Solaris, are you saying you don't get the same?

The date of the Solaris build is actually in my SWARMHOME path, it is April
2nd 2001. I'm dependent on system management for updates here (maybe luckily
so this time). On my NT I reinstalled versions (javaswarm.dll etc.) that
date back to March 21nd and the errors don't appear (but probably others
will...).

>
> In the short term, I would say you should use one of the java
> collections instead of swarm's map anyway.  I don't  know if there is
> any official word about that claim, but I've sensed it strongly from
> several of the experts.

Do you have any suggestions about java alternatives?

>
> I do agree the behavior you found is surprising/bugish.  Because I was
> bored I dug into your example more and I think this is what is
> happening. The java runtime system sees the output of two calls to new
> Integer(1) and new Integer(1) as different things. Note below I edit

Your explanation sounds plausible in that it suggests that MapImpl probably
compares the addresses of the objects. However, why this should happen in
the most recent version and not in the older ones is quite puzzling.


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.