db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kristian Waagan <kristian.waa...@oracle.com>
Subject Re: Alternative code coverage report using JaCoCo
Date Tue, 12 Jun 2012 09:04:25 GMT
On 12.06.2012 02:36, Bryan Pendleton wrote:
>> I've been looking into adding support for JaCoCo [1] for Derby
> Thanks for investigating this, Kristian.
> I looked at the report you published, and I found it clear and
> easy to read.
> Why would we choose one tool over another? Is there a reason to
> trust the reports produced by one tool more than those produced
> by another? Is it easier to configure or lower-overhead to operate
> one tool versus the other?

Hi Bryan,

Note that we don't have to choose one of the tools - we can use both if 
we want to.
When I started this effort we were having problems getting EMMA to 
produce results reliably. I think that issue has been resolved now (Knut 
made changes to part of our test framework).

I'd say both tools are just about as easy to configure, but JaCoCo may 
have a stronger focus on integration (currently ant, maven, Java agent).
I don't know exactly about overhead, but my gut feeling is they're in 
the same league. I've tried other tools that affected performance a lot 
(i.e. it took ages to run the tests) and that required a lot of 
resources (either during data recording or during report generation). 
Note that JaCoCo doesn't have a separate instrumentation/compilation step.
Based on a quick chat with Knut Anders offline the raw result files from 
JaCoCo may be somewhat smaller, but the files are so small anyway that 
it doesn't matter.
Report generation with JaCoCo is very fast.
I believe both tools are pretty simplistic in what data they record or 
derive, but it's enough for our current usage.

> Or are you just anticipating that Emma will fall into disrepair
> over time?

Yes, this is the main reason to look at another tool in my opinion.
With time, EMMA may get into trouble as new versions of Java are released.

There are also alternative ways to implement code coverage tools with 
Java 7, so in the future there may be a better tool available. However, 
the current level of overhead seems to be acceptable to me (when used 
with test runs).


> thanks,
> bryan

View raw message