Checkins and test passing
Malcolm.Wallace at cs.york.ac.uk
Mon Oct 23 05:57:38 EDT 2006
Ian Lynagh <igloo at earth.li> wrote:
> I think it would be a nice requirement that for a test to be marked as
> expected-to-fail there must be an open bug about it.
Hmm. Isn't the meaning upside down here? I thought an "expected
failure" was a program that is supposed to fail by the rules of the
language. Therefore there is no bug. An "unexpected" failure would be
one representing a compiler bug. We may well know what the bug is, and
what is needed to fix it, so in that narrow sense the failure is
expected, but I would certainly have thought that, given the choice
between fixing an unexpected or expected failure, the unexpected one
is the worrisome one.
If the "expected/unexpected" terminology is confusing, then maybe it
would be better to rename them to "desired/undesired" failures?
More information about the Cvs-ghc