[m-rev.] for review: extend modes of operation

[Julien Fischer]

Hi Zoltan,

On Mon, 5 Oct 2015, Zoltan Somogyi wrote:

> For review by Julien, since this extends his work.
> 
> The two parts I am most interested in feedback on are
> the exceptions in decide_op_mode (are they ok?

They look fine.

> do we need more exceptions?),

I'm not aware of any others.

> and whether it is ok that we now assume that rebuilding the option set
> should NOT change the mode of operation.

Again, I'm not aware of any reasons that it shouldn't be ok; in this
case it would be easiest to try it and see.

> Also, should the op_mode type itself be in globals.m, with
> decide_op_mode in handle_options.m?  The op_mode is used only in
> mercury_compile.m (now), but the latter would allow the deletion of
> the (slightly incorrect) redundant computation of Link, and it
> requires the former as a prerequisite.

Yes, it's a desirable change in any case since it reduces the amount
of clutter in mercury_compile.m.

The diff itself looks fine

Julien.