[m-rev.] diff: fix mantis bug 368

[Julien Fischer]

Hi Zoltan,

On Thu, 5 Feb 2015, Zoltan Somogyi wrote:

> I tried to make the test case invoke the compiler with --erlang even when
> the bootcheck is in a different grade, but that doesn't fit into our current setup.
>
> I tested it manually, and the new test case gets a compiler error without
> this fix, but not with this fix. However, since I don't have Erlang installed,
> I cannot test the fix fully. Peter, will you try it out?

I have tried it out with rotd-2015-02-06 and it still fails to compile
in the Erlang grade:

Making Mercury/erls/transform_hlds__ctgc__selector.erl
ctgc.selector.m:396: In
ctgc.selector.m:396:
`table_reset_for_normalize_selector_with_type_information_4'(di,
ctgc.selector.m:396:   uo):
ctgc.selector.m:396:   error: determinism declaration not satisfied.
ctgc.selector.m:396:   Declared `det', inferred `multi'.
ctgc.selector.m:396:   Disjunction has multiple clauses with solutions.
ctgc.selector.m:528: In clause for
ctgc.selector.m:528:
`table_reset_for_normalize_selector_with_type_information_4(di,
ctgc.selector.m:528:   uo)':
ctgc.selector.m:528:   mode error: argument 2 did not get sufficiently
ctgc.selector.m:528:   instantiated.
ctgc.selector.m:528:   Final instantiatedness of `HeadVar__2' was
ctgc.selector.m:528:   `mostly_unique',
ctgc.selector.m:528:   expected final instantiatedness was `unique'.
** Error making `Mercury/erls/transform_hlds__ctgc__selector.erl'.

Cheers,
Julien.