[m-rev.] diff: Delete unused imports.

[Peter Wang]

On Wed, 30 Mar 2022 14:09:03 +1100 "Zoltan Somogyi" <zoltan.somogyi at runbox.com> wrote:
> 
> 
> On Wed, 30 Mar 2022 13:26:36 +1100, Peter Wang <novalazy at gmail.com> wrote:
> > > The main reason why you have to import a module A in the interface
> > > of module B is that one of the arguments of a predicate or function
> > > exported by module B refers to a type defined in module A.
> > 
> > But does a module C that only calls the predicate/function exported by
> > module B also need to import module A?
> 
> No, it does not, but that is not the point I was trying to make. The sentence
> you quoted was not making an argument; it was only a prelude to the argument
> in the next paragraph.
> 
> > The type defined in module A is
> > never explicitly named in module C, nor constructed/deconstructed in
> > module C, so does it count as a "use"?
> 
> I am not talking about any third module C at all.
> 
> I am saying that module B needs to import module A if one of its variables
> has a type that contains a type constructor defined in module A.
> 

Ok, great.

My question was prompted by this:

> Actually, the reason why the import of vartypes was flagged as unused
> with --intermod-opt was a known limitation that we discussed in october:

If flagging vartypes as unused is due to a limitation, then would the
opposite behaviour, i.e. flagging vartypes as USED in call_gen.m,
be correct? Anyway, we both agree the answer to that is "no".

> > The reference manual doesn't
> > explain what it considers a "use".
> 
> Module A uses module B if it needs information from module B's interface
> to compile properly. While could and probably should give examples of
> what counts as as use, I think we should avoid any statement of the form
> "the above uses are the only possible uses", since such statements
> could be invalidated by future changes to the language.

Right. Still, we can always make changes to the language
as long as the reference manual is updated.

Peter