Re: Bug in processing surefire/failsafe groups

Previous Topic Next Topic
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

Re: Bug in processing surefire/failsafe groups

Tibor Digana

Did you receive my previous answer?
There is some problem with email delivery.

On Fri, Mar 30, 2018 at 8:58 PM, Christopher <[hidden email]> wrote:

> Found a bug in maven-failsafe-plugin (
> when using groups (JUnit
> categories):
> plugin/integration-test-mojo.html#groups
> plugin/integration-test-mojo.html#excludedGroups
> The problem is when the class corresponding to a configured group is not on
> the class path. This is particularly problematic in multi-module builds, if
> a group is specified which exists in one module but not another one.
> It appears the bug occurs because the configured group/category is loaded
> from the class path, and then compared to the classes to find matching
> tests. However, this seems backwards. If the tests were examined first,
> then the name of the groups could be matched against the configuration
> without loading to load the configured groups/category. Even if this is not
> possible due to some limitation on how these groups are passed to JUnit or
> some other restriction, a ClassNotFoundException (CNFE) should be handled
> more sanely: For example, in the case of a CNFE for the "groups" parameter,
> then it's clearly not possible for any tests to be included, so no tests
> should run. In the case of a CNFE for "excludedGroups", then it's clearly
> not possible for any tests to match, so none should be excluded.
> This bug has got me thinking about a related feature I'd like to see in
> maven-surefire-plugin and maven-failsafe-plugin: I'd like to see groups /
> excludedGroups specified by patterns, similar to `-Dtest=MyTest` instead of
> `-Dtest=org.apache.project.package.subpackage.testing.MyTest`, it would be
> nice to set `-Dgroups=MyGroup` rather than
> `-Dgroups=org.apache.project.package.subpackage.testing.
> categories.MyGroup`.
> If groups could be specified as patterns, it would be more obvious that a
> CNFE could not occur... if a pattern didn't match any classes, then it
> wouldn't result in a CNFE... it would simply return an empty set.
> Thanks,
> Christopher