Testsuite policy - test case now passes but bug remains

[Martin v. Loewis]

> I'm pretty certain that the original bug is still there, as the failure is
> the same.  It's just that it is no longer triggered in the original reduced
> testcase.  Trivial and apparently unrelated changes to the testcase
> retrigger the problem - essentially declaring another unused array in the
> subroutine.

One thing that hasn't been mentioned: The testsuite is *not* a place
to keep track of bug reports - GNATS is the place for that.

So I think new test cases should *only* be added when a bug had been
fixed (or, at the discretion of maintainers, for any other reason they
please), to make sure that the same bug does not re-occur. Adding a
test case on its own (i.e. without knowing the fix for the testcase)
is pointless, IMO.

Regards,
Martin