[m-rev.] for review: Define behaviour of string.to_char_list (and rev) on ill-formed sequences.

[Peter Wang]

On Mon, 21 Oct 2019 17:01:42 +1100 (AEDT), "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> 
> On Mon, 21 Oct 2019 16:53:07 +1100, Peter Wang <novalazy at gmail.com> wrote:
> > Having tried it, tainting the forward modes with `pragma obsolete'
> > creates a lot of busy work to convert predicate calls to function calls,
> > essentially for no gain. The reverse modes are hardly ever used in the
> > Mercury system, and probably the same is true in other projects.
> > 
> > I think simply deleting the reverse modes (with a suitable deprecation
> > period) will annoy fewer people.
> 
> The warning for calls to obsolete predicates is done by code in simplify_goal_call.m
> that has access to the mode of the call. There is no reason why the obsolete pragma
> can't be extended to allow the specification of a mode. At the moment, its first arg
> must have the form predname/arity, but the compiler already has code to parse
> terms of the form predname(modelist), which is used by e.g. pragma foreign_export.
> We could extend obsolete pragmas to allow this form as well, and if used, we would
> generate the obsolete warning only in the nominated mode.
> 
> If you want, I can implement this in a day or two.

Thanks, that would be useful for my proposed course of action.
Most likely the feature would see little use other than this instance,
but I think it's worth adding if the code is mostly there already.

Peter