[m-rev.] for review: Fix abstract equivalence types.

[Zoltan Somogyi]

On Sun, 20 Aug 2017 11:21:00 +1000, Peter Wang <novalazy at gmail.com> wrote:
> The regression was introduced in commit 2856408ab0fbf74c272a999b9c5c7e269001e3fe
> "Separate the different kinds of sections in aug_comp_units."

Thanks for tracking that down.

> --- /dev/null
> +++ b/tests/invalid/abstract_eqv.m
> @@ -0,0 +1,19 @@
> +%---------------------------------------------------------------------------%
> +% vim: ts=4 sw=4 et ft=mercury
> +%---------------------------------------------------------------------------%
> +%
> +% Mantis bug 411. Abstract types should not be replaced by their equivalent
> +% types.

I would word that as "abstract_eqv.m should not be able to see that
abstract_eqv2.m has *privately* defined foo to be equivalent to bar".

Other than that minor quibble, the diff is fine.

Zoltan.