On Thu, Jan 27, 2005 at 05:48:15PM +0100, Tels wrote: > If I understand the current test model correctly, the seperation also has > the advantage that if the test dies, the parser is largely unaffected. It > even tells you *where* it died (e.g. "after test #7"). If you have > test/parser in one process, a die() might tear it down completely, and > you get only "test process died" out of it.
It would have to be something beyond a normal die() or exit() because those can all be trapped. A segfault for example. But yes, it does mean we can test ourself with ourself no matter what. It also avoids mucking up the test environment so each test script knows exactly what's been loaded.
