On Thu, Sep 8, 2011 at 5:07 PM, Charles Oliver Nutter
<headius / headius.com> wrote:
> Matz: Is it official policy that all feature changes in Ruby must be
> accompanied by a test or spec? If not, why not? If so, why aren't they
> going into a non-MRI-specific test suite?
>
> Shouldn't ruby-core help maintain an implementation-agnostic Ruby
> behavioral test suite? If so, why isn't that happening, and why
> couldn't it be RubySpec? If not, why not?

I'm going to try to dial back the rhetoric a bit and lay out the
facts. Jump in where I go astray.

THE NEED:

* There must be a test suite all Ruby implementations can use to
confirm that they are compatible with "Ruby" the
language/core/libraries

* This suite must not be specific to any one implementation, and
should only include tests for features considered "standard" Ruby
(i.e. not implementation-specific).

* This test suite would ideally support running a subset of tests if
some are known to fail. This is necessary to maintain a watermark on
partial or in-progress implementations.

* All Ruby features should have a complete set of tests in this suite.

* All Ruby feature changes should be accompanied by a set of tests or
changes to tests in this suite.

How do we get there?

- Charlie