If a spec file has a failing test, it'll fail when ran from the command-line. But if it has an in-your-face flat-out wrong JavaScript syntax error, Rhino spits out an error to stderr but does not report a failure. The overall build reports success.
The problem here is that Rhino's "load()" function swallows JavaScript exceptions due to syntax errors. We'll have to patch Rhino or more likely write a different Java load function that correctly reports errors.
Can't duplicate this using Blue Ridge. (I'm seeing it quite a bit in the env.js qunit test suite, but for some reason Blue Ridge appears immune.) Marking as won't fix and closing.
If a spec file has a failing test, it'll fail when ran from the command-line. But if it has an in-your-face flat-out wrong JavaScript syntax error, Rhino spits out an error to stderr but does not report a failure. The overall build reports success.
The problem here is that Rhino's "load()" function swallows JavaScript exceptions due to syntax errors. We'll have to patch Rhino or more likely write a different Java load function that correctly reports errors.