OpenJDK adoption by Java User Groups
Alan Bateman
Alan.Bateman at oracle.com
Sun Dec 11 13:27:18 PST 2011
On 09/12/2011 17:14, Jonathan Gibbons wrote:
> Stephen,
>
> I've browsed around the links for Clover. It looks very interesting,
> but it's not clear to me the constraints under which it operates.
>
> Note that OpenJDK uses jtreg for the publicly available regression
> tests, and that many of the tests rely on being able to start up one
> or more new JVMs for the duration of the test. In my experience this
> has been too much of a stretch for commercially available coverage
> tools to tolerate, and they typically can't cope and/or fall over. A
> common assumption for such tools is that all API tests can run
> sequentially in the same JVM.
>
> Do you know if there are any such restrictions with Clover?
>
> -- Jon
It would be interesting to know if this is possible. Another thing that
often gets us with off-the-shelf tools is that they don't recognize the
newest class file version or they only work with released JDK versions.
I don't know if this tool uses static or load-time instrumentation but
it reminds me that another useful thing that could be done is audit the
shell tests, and tests that use Runtime.exec to create additional VMs,
to make sure that they pass the VM options through to the child
processes. It's unlikely to change the code coverage numbers too much,
except perhaps tool code, but important to get the right results.
-Alan
More information about the jdk8-dev
mailing list