Description Philipp Hahn 2014-01-29 17:32:39 CET

Created attachment 5765 [details]
Fix JUnit output format

The generated JUnit XML Format is causing problems with Jenkins:

1. The testsuite/@timestamp attribute includes nano-seconds, which is not expected.

2. Instead of a class-name, currently the description is used to identify the test-case. This is a long test and quiet unhandy.

3. The class-name uses a file-name, where the section and test-case are separated by "/", while for Jenkins a dot (".") must be used. Because of that all tests are put in the "(root)" package and are not separated by sections.

4. The testsuite-element currently has a non-int @id attribute, which is not allowed.

5. The testsuite-element has a @id and @package attribute, which are not allowed for single-testsuite. Only when multiple suits are aggregated in a parent testsuites-element are those attributes allowed.

Impact:
1. The reports would be better readable, since the names are hard to read because of the long names.

2. Jenkins uses the name to calculate its trend; when the names change, the counter will probably restart.

3. The long description of each test would no longer be available in the Jenkins report; it could be printed to stdout/stderr instead if it should be still available through Jenkins.