Re: [Asterisk-java-users] Possible bug in actionid generation?

[Stefan R. <sr...@re...>]

On Thu, 2005-04-21 at 18:05 -0500, Chad Kitching wrote:
> I'm sure I'll get harassed for this, but

no not at all, we all get enough harassment from the pure c folks
implementing asterisk :)

> I converted some of asterisk-java to C#, and in the process, 
> I may have run across a possible bug.

do you plan to release your port as open source?

> DefaultManagerConnection uses the hashcode of the object in 
> createInternalActionId, however, the Java specs do not specify that 
> hashcodes must be positive, so if a VM provides a negative value for 
> this function, Util.{get,strip}InternalActionId will fail to work, and 
> will return null.

Yes you are right. Though I didn't encounter any negative hash codes
yet...

> Either a different separator between the internal 
> actionid and the object actionid needs to be used, or the util functions 
> should perhaps start scanning at the second character instead of the first.

I fixed in in CVS and replaced the delimiter with a '#'.

> Now, I have no idea if any JVM is actually affected by this, and my 
> conversion job worked fine on Microsoft's CLR, but failed on Mono 
> because of the hashing implementation differences.

Yep - as I said i never encountered in on Sun's JVM or JRockit but the
spec clearly allows any int.

Thanks for your hint.

=Stefan