Test cases are defined by .ccl files, one .ccl file is one testcase. 

If the test case filename contains the string "match", it is treated as a match rule, otherwise a disambiguation rule. Match rules require CCL input files as described below.

A test case loads a corpus from an .xml file thet is not an .out.xml file from the test case directory, or directories above if there are none.
Behavior is undefined if there is more than one .xml file in a directory. Only the first chunk is processed (all sentences from the chunk).
If the xml filename contains the string "ccl", it will be read as a CCL format xml as opposed to the default XCES.

A foo.ccl file should be accompanied by a foo.out.xml file defining the expected output. Output is compared intelligently, lexeme order / duplicates does not matter.
If the output filenale contains the string "ccl", it will be read as a CCL file as in the input file. Additionally, annotations will be compared.

A magic file with the extension .is-the-tagset defines the tagset for testcases in this directory and all subdirectories, unless overrden by another .is-the-tagset file.