What mutation-testing frameworks exist?

Question

In another question I asked if mutation-testing is useful in practice. As I didn't get any answers that satisfy me, I want to check the current tools myself. So I need an overview about current existing mutation-test-frameworks. Which of them are most usable and why?

• I program in Java, so I would prefer Java-tools, but I would risk a look at interesting frameworks for different languages.
• I want to integrate in an automatic build-process, so I would prefer tools that can be executed through command-line.

Answer 1

I only know of two frameworks, but they're both for Java :)

• Jester
• Jumble

I haven't used either of them, I'm afraid.

Answer 2

For Ruby there is Heckle, and a newcomer called Boo_hiss.

Answer 3

For the .Net community, there is NesTer, but it has some serious limitations. E.g. only supports C# and NUnit.

Does not appear to be actively maintained either, but it might be a starting point.

Answer 4

I took a look at Jester (the actual source code) and it seems to me that it does not support too many mutations. There is a file in there where these mutations are specified. I might be wrong about the above but what I definitely did not like was the mix between launching the tool from command line and the little GUI feedback interface. Why not give feedback in the command line like JUnit does when run outside an IDE?

Jumble is another thing :). It has a simple command line interface and comes with an Eclipse plugin too. The feedback is all text in the console. I am happy with this tool and I plan to write some ANT target to add it in my project continuous integration.

I am also looking at Javalanche but did not try it yet.

I'll have news in a few weeks.

Answer 5

Jester does provide a file for the mutations and they are limited. To some degree, you can add your own mutations to the file.

I've experimented with Jumble and Jester and I found that Jumble provides more mutations and better documentation. Additionally, I've had quick responses from the project owners when I've emailed them. One drawback to Jumble is that it operates on the bytecode using BCEL. That presents something of a learning curve for many developers.

My company, State Farm, wrote an Ant task that we may contribute back to the Jumble project. Based on what I've read in their mailing lists, others are working on an Ant task for Jumble too.

I'm looking at Javalanche as well. I’ll be glad to share what I know when I’m done.