> ad 2) "The guitests target is not part of the build sequence for good > reasons" - I know but how you find any new GUI bugs if the guitests are > considered broken and not executed? Maybe the next time we get 18 errors > instead of 17 but when the tests are ignored that one bug could cause > frustration within the JSPWiki user community when it escapes into the real > world - I had my five minutes of fame when a late change caused a NPE in my > commons-exec release - I think I got more than 20 mails with "btw, there is a > stupid NPE in this method". IMHO it is acceptable to state that 17 tests are > indeed broken but those 17 tests should be commented out to get overall > guitests working - for the remaining 17 tests we can create a JIRA and hope > for better times.
guitests runs the same tests as "tests", but it uses a different JUnit test runner which shows up a pretty gui :-). Unfortunately the classloader seems different between these two and I don't recall it ever being anything else but a source of grief... My vote would be to remove the entire guitests target and just concentrate on a single testrunner getting the tests right... /Janne
