If there are multiple dependent projects in an eclipse workspace, it can be tedious to run the unit test of all these projects manually. Often, a failed unit test caused by a change in a different project than where the unit test is located is only discovered with a significant time gap. From there, it becomes difficult to trace back to the initial change causing the unit test to fail.
The Infinitest eclipse plug in constantly executes all relevant unit test defined in the workspace in the background. If a source file is saved, Inifitest will search for affected unit tests and execute these. If they fail, a Java problem is reported.
Below some simple steps to install and configure Inifitest:
Download
Infinitest can be downloaded using their update site: http://update.improvingworks.com
Just follow the instructions on their site to install the plugin.
Activate
Once the Infinitest plug in is installed and the eclipse IDE restated, it should start searching for test cases.
If code is changed (and saved), it should start reporting any failed test cases as Java problem.
Possible Problems
Infinitest can be disabled in the eclipse preference window. It adds a top level item “Infinitest” to the eclipse preference tree.
I encountered some problems with GWTTestCases. For every GWTTestCase in my projects, Infinitest reported:
Description Resource Path Location Type
NoClassDefFoundError (com/google/gwt/dev/cfg/Condition) in GwtTestGwtCollaborationScenario. GwtTestGwtCollaborationScenario.java /nxServerGwt/src/test/java/nx/servergwt/tests/hosted line 0 Infinitest Test Failure
The only solution I found was to remove these test cases from the eclipse projects build path (right click the .java file and select Build Path / Exclude).
The Full Stack Blog 
“a failed unit test caused by a change in a different project”.. Then you are not UNIT testing, mocks or friends, API’s should not change is such a way that unit tests fail elsewhere.
Thank you for your comment! I see your point … however, I disagree: I think it often happens that the UNIT tests for a particular API do not test ALL possible test cases, in this case, test cases, which use the API indirectly can still fail even though all test cases for the API pass. In this case, apart from fixing the bug a new test case should be added for the API, of course!