The situation today is any test case covering a negative syntax error must be wrapped within an eval call to accurately detect the exception and it's type. Most JavaScript implementations will follow a different code path for eval'ed versus literal program syntax meaning we're missing some coverage here.
To fix this, one solution would be to move the metadata from existing test case/test objects into a separate file of some format. In this manner, the separate file would indicate whether an exception is expected, it's type, etc., and the test case could have the literal syntax expected to fail.
This enhancement also seems critical for test cases specifically covering global scoping.
I'm guesstimating around ten days will be needed to implement this. Should we go the easy route of simply extending the test harness to support a new type of test cases, this could take far less time. On the other hand, modifying all test cases to follow this paradigm could take considerably longer.