db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Tiago Espinha <ti...@espinha.net>
Subject Re: emma nightly runs - are they insane only?
Date Wed, 21 Mar 2012 11:41:31 GMT
Hi Mike,

I only read this e-mail now, that I'm looking for Emma reports on the
mailing list. It's great that you want to bring Emma to the GSoC project! I
was planning on doing the same.

It should be possible to bring two students to work on this and have them
work in different packages, right? What do you think?

Looking at an important package like "org.apache.derby.client.net" I see
many classes with poor method coverage. There's more too:

- org.apache.derby.iapi.types - several classes not tested at all and
others with poor coverage.
- org.apache.derby.iapi.jdbc - could use extensive improvements.
- org.apache.derby.impl.io - another package with poor coverage.

Then again, like you said, we don't really know whether this is being ran
against sane or insane jars...

I'm trying out my own Emma run against sane jars and so far I've been
getting many failures and errors, not sure whether that has to do with the
Emma instrumentation or with my own setup... will try to figure that out
later.

If we're gonna have students using Emma as a reference for the target
coverage they should achieve, we should guide the students on how to get
Emma  and Derby working together.

Tiago

On Wed, Mar 14, 2012 at 6:36 PM, Mike Matrigali <mikem_app@sbcglobal.net>wrote:

> I am looking at generating a JIRA issue that is emma based new tests for
> the google summer of code.  In identifying what would be good modules
> to go after I started drilling down it looks like the public report is
> on insane code which is a reasonable first step, but led me to multiple
> cases saying something was untested which i believe would be tested
> if it run again under insane.
>
> Can emma be run on the sum of 2 tests runs?  For a clear report on
> coverage it would be nice to see an emma the sum of running our tests
> against sane and insane.  Then it would be very clear that the uncovered
> paths are really not tested.
>
> I am not sure how hard that would be for public reports.  Less useful but
> maybe easier might be to generate a sane report and an insane report.
>
> /mikem
>

Mime
View raw message