For running all my test classes automatically, I look for all class files inside a dedicated directory, convert the path to a package name and check if this class implements the given interface:
try {
Class<? > myTestClass = Class.forName( constructedClassName );
if( myTestClass.isInstance( MyTestInterface.class ) ) {
testCollection.add( myTestClass );
}
}
catch( Error e ) {
// ignore, no valid test class
}
Today I ran into an ugly bug (see this SO question) using this technique.
Question:
How can I collect all my test classes without having to ignore Errors that can occur with classes I'm not interested in?
You've kind of painted yourself into a corner here ...
What I'd do is one (or more) of the following:
fix the offending classes so that they do load
put the classes into different directories, create lists of names, or use pattern matching to discriminate between the classes that you do / don't want to add to
testCollection
And log the Errors of course!! Maybe log them somewhere different, but if you do that leave a loud message in the main log that tells someone where to look for details.
What good is a class that does not load? Why not just fix them first? Wouldn't you have to fix them at some point?