I was trying to preserve the external behavior that would occur
with the old code if you did not provide jmxtools.jar at runtime.
As implemented, both the new code and the old code would result in
a NoClassDefFoundError if it was deployed without jmxtools.jar.
However, the likelihood that some client code out there has
something like:
try {
new Agent().start();
} catch (NoClassDefFoundError ex) {
}
is vanishingly small, so switching to a RuntimeException is not a
big issue to me. I'd prefer not to revert the whole change, but
to change the Error constructors to RuntimeException constructors
which I'll do and see what you think.
Yes, thanks, although I _wish_ we could use this slightly different
pattern instead (as I'm sure you probably agree):
throw new RuntimeException("Failure starting... ", e);
except it needs 1.4 source compatibility. *sigh*. bummer.
One other note on the commit, I notice the addition of JavaDoc
comments on things like no-arg constructors. I'm -0 on adding
JavaDoc that adds no value at all ("create new instance"), that
really gets under my goat here at work it just seems pointless, a
waste of effort and makes the code actually harder to read if you ask
me. Simple clean code that is self-explanatory should be the goal IMHO.
Paul
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
