[Isolate-interest] Re: JSR 121 TCK available

Guillaume Desnoix guillaume at desnoix.com
Mon Oct 10 12:14:51 EDT 2005


> "Not what I call an open process."
> Better than some, worse than others.  Sun appears to be slowly
> migrating from JSRs being closed by default to them being open by
> default.

Yes, and this is a good move. Being open from the start can only 
increase the quality.

> The tests are just JUnit tests.  In order to compile or run them, you
> need the classes that they test.  The TCK download could simply
> include source for a dummy implementation that runs on Tiger, but
> fails all the tests.  My guess is that they didn't want to include a
> broken implementation.  Perhaps there is some JCP reason that a broken
> implementation can't be included.

And this is understandable. The javadoc and the test suite should be enough.

> Does the JDistro JSR-121 implementation depend on JDistro?

Yes. This is just a thin layer over the JDistro core. Basicaly, we just 
delegate the calls. Right now, we have PD1 and PD2. It has not yet been 
updated to the latest spec.

> Perhaps
> the official TCK could be improved with some of your tests.  More
> tests are almost always better than fewer.

I have only a few tests and they are certainly not good. OTOH, the 
results of the TCK could be interesting. The fact that the 
implementation is written in Java could bring some issues. Or at least 
it could show my misunderstanding of the API.

> I've toyed with the idea of providing an implementation of JSR-121
> based on Runtime.exec or ProcessBuilder that could be distributed as a
> standard extension.  It doesn't look like more than a week worth of
> work, but it's a week I won't have any time soon.

Same here. There is plan for multiple implementations of this module in 
JDistro (the current, one as you described, one based on the MVM and one 
for jNode). But yet done due to lack of time.

> In order to run the tests, I used Eclipse to quickly produce an
> utterly broken JSR-121 "implementation".  I thought about publishing
> it to this list just to save the curious a small bit of effort, but I
> didn't want to violate any of those agreements that I should have read
> closely.  The whole situation is a bit silly.

I agree.

Regards, Guillaume


More information about the Isolate-interest mailing list