IllegalStateException in JSONRPC.init()

[almende]

IllegalStateException in JSONRPC.init()

[tracts-root]

Figured I would just confirm: simply omitting the throwable argument from the super call works fine. So that would read

public JSONRPCException(final CODE code, final String description, final Throwable t) {
    super(description);
    init(code, description, t);
}

[ludost]

Hi Erik,

Which release/version are you referring to? In the Master I've done a fix that probably fixes your mentioned issue, but it had not been ported back to 2.2.1 as it was contained in a larger unrelated commit. I'll patch 2.2.1 now, can you check if that helps?

[tracts-root]

Hi Ludo,

Thanks! I think the extra clause you added to init(), namely to check that getCause() == null before attempting to call initCause should work fine in most cases. However if we called the JSONRPCException constructor with the Throwable argument set to null (which according to the java docs is actually a reasonable thing to do) then it will still go wrong.

This latter cornercase might be a minor advantage of my suggested fix -- simply not passing the t argument to the super constructor -- compared to this extra test inside init.

[tracts-root]

@ludost Sorry didn't mean to close it just yet, at least until you read my comment :-)

[tracts-root]

@ludost Looking at the source for Throwable it looks like the test getCause()==null will not work... http://grepcode.com/file/repository.grepcode.com/java/root/jdk/openjdk/7u40-b43/java/lang/Throwable.java#Throwable
The cause field is initialized to this, not null, so I expect your test will not do what it should when one of the constructors without a throwable is used.

So I guess the way to get this working properly for all constructor variants is not to pass any throwable to the super constructor and remove the getCause() test from init?

[ludost]

The problem is that super(t) is different than setCause(t). And I like to have the behaviour of super(t) :)

But back to your statement, which is incorrect: the getCause()==null will work:

public synchronized Throwable getCause() {
     return (cause==this ? null : cause);
}

So, please test my latest commit in your code, I believe it will fix the issue.

PS: I fail to see why putting a null-cause would lead to a problem? Throwable seems to handle that case correctly in its constructors and hopefully JSONRPCException as well.

[tracts-root]

Ah, I had not noticed this quirk in getCause. I did test it, actually, and my immediate problem is resolved. I still wonder about calling a constructor with t==null but as far as I know we're not doing that.

Thanks!

Oh and just curious, what is the difference between super(t) and setCause(t) that you mentioned?

[ludost]

An indication about the safety of the constructors is this statement from one of these constructors:

 detailMessage = (cause==null ? null : cause.toString());

Which explicitly checks for null causes.