JCov and coverage numbers on OpenJDK

Martijn Verburg martijnverburg at gmail.com
Sat Mar 28 15:03:17 UTC 2015


Hi all,

Oracle's internal QA team were able to confirm that the numbers that the
Adoption Group were producing are very close (not a statistical significant
difference) to their numbers.  With validation that the numbers are
accurate, it would be good to start publishing these for the purpose of
guiding OpenJDK developers to areas that need more test coverage!

What steps would people like to take next?

I think the right home for these reports is in the quality group.  They
could host the code coverage reports and pro-actively release test coverage
numbers alongside the # tests passing/failing (as they do currently).

@Rory, is that feasible in the short term? I understand that there's
potentially some technical work to do and other hoops to jump through. If
it's not possible in the short term then perhaps the quality group could
reference the reports that the Adoption Group are hosting (with a caveat)
in the short term until that work can be completed.

Special thanks to John Oliver and Alexandre Iline for digging into this!


Cheers,
Martijn

On 4 March 2015 at 13:25, Ben Evans <benjamin.john.evans at gmail.com> wrote:

> Depending on timings, I can probably be free on Tuesday (I'm on GMT too).
>
> Thanks,
>
> Ben
>
> On Tue, Mar 3, 2015 at 10:32 AM, Martijn Verburg
> <martijnverburg at gmail.com> wrote:
> > Hi All,
> >
> > As some of you know we've been running some experiments in the Cloudbees
> > incubator to see if we can get accurate code coverage numbers using JCov
> on
> > the jdk9 forest in particular.
> >
> > John Oliver has gone back and reviewed the process and the numbers and we
> > *think* we've gone about it the right way.
> >
> > Before we even think about taking the next step to start producing these
> > numbers regularly in the incubator, we need to make sure that we've used
> > JCov correctly and that the numbers are not misleading.
> >
> > It would be great to have a technical call with John Oliver, Mani,
> someone
> > from Rory's team (the person who does the internal OpenJDK numbers?) and
> > probably Jonathan Gibbons.
> >
> > Does next Tuesday suit folks?  It all depends on timezones (John Oliver,
> > Mani and myself are GMT)
> >
> > Cheers,
> > Martijn
>


More information about the code-tools-dev mailing list