[m-rev.] for post-commi review: unify erlang implementation of mutables

[Julien Fischer]

On Fri, 30 Jan 2015, Peter Wang wrote:

> On Fri, 30 Jan 2015 00:06:39 +1100 (EST), "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
>> For review by anyone. However, what is worth reviewing is not
>> so much this diff, as the contents of add_mutable_aux_preds.m
>> after its installation.
>>
>> I also noticed that the code implementing mutables for the Erlang
>> backend assumed that trailing was not enabled. I checked, and I found
>> comments in the compiler that say that the Erlang backend does not
>> implement trailing, but I can find nothing in handle_options.m to generate
>> an error if users try to compile something with --grade erlang --use-trail.
>> Is this intentional? Peter, do you know?
>
> I'd never have thought to try that.  We also don't warn about lack of
> trailing support for Java and C#.  Similarly for profiling, and perhaps
> tabling.

It's essentially a manifestation of the same problem we discussed 
in this thread from last September:
<http://www.mercurylang.org/list-archives/reviews/2014-September/017300.html>.

The compiler does not do much *compile-time* checking of grade component
compatibility whether they are specified by via the --grade option,
or via individual grade component options.

Cheers,
Julien.