[m-rev.] diff: improve error checking for instance decls

[Mark Brown]

On 26-Oct-2006, Julien Fischer <juliensf at csse.unimelb.edu.au> wrote:
> 
> On Wed, 25 Oct 2006, Mark Brown wrote:
> 
> >On 25-Oct-2006, Julien Fischer <juliensf at csse.unimelb.edu.au> wrote:
> >>Emit an error if abstract instances in the implementation of a module do
> >>not have a corresponding concrete instance.
> >
> >Why do we even allow abstract instances in the implementation?  (Likewise
> >for abstract types.)
> 
> 
> For the former I'm not sure - I was planning to disallow it when I
> implemented the concrete instance check which is why the original version
> never checked for it.  (The lack of such a check was the cause of
> Sebastian's problem this afternoon.)  I'm planning to implement the
> '--allow-instance-stubs' option at some point so abstract instances in
> the implementation will then have a use.
> 
> For the latter, if you have a foreign type you need an abstract type
> to declare it, i.e. the foreign_type pragma doesn't act as a type
> declaration.

Fair enough, on both points.

Cheers,
Mark.

--------------------------------------------------------------------------
mercury-reviews mailing list
Post messages to:       mercury-reviews at csse.unimelb.edu.au
Administrative Queries: owner-mercury-reviews at csse.unimelb.edu.au
Subscriptions:          mercury-reviews-request at csse.unimelb.edu.au
--------------------------------------------------------------------------