On Friday 11 January 2008 14:19:52 Ovid wrote: > So imagine you have a Web spider with 1000 tests and it fails at the > third test. If the behavior is in the harness, it stops processing > results but the test program keeps running unless you deliberately try > to kill the process. I think it's much cleaner that the thing which > generates the test output knows when to stop.
In this case yes, but imagine that you're debugging a test failure which you suspect is so catastrophic that subsequent assertions in the same file aren't helpful at all. In that case, the more important goal is to show only the first set of failure diagnostics so that you can fix them. In-suite test organization is very different from results display. -- s
