Re: value to function?

[Jim Blandy <jimb at redhat dot com>]

Andrew Cagney <ac131313@redhat.com> writes:
> > Andrew Cagney <ac131313@redhat.com> writes:
> >
> >> If the proposal to pass a function value to push_dummy_call is going
> >> to fly there's going to need to be a method that converts an arbitrary
> >> value into a function/method value.
> > A new gdbarch method to be called from where, specifically?  This
> > area
> > is a mess (just try describing the behavior of find_function_addr...),
> > so I'm not sure what you mean.
> 
> Now I think I'm missing something.  What new architecture method?
> What mess?

Never mind.  You said "there's going to need to be a method that
converts..."  Since you said "method", I assumed you didn't mean
"function", so I was trying to figure out what sort of method you
meant.

> > Kind of like find_function_addr() and the oposite of value_coerce_function.
> 
> 	pre (I guess that value isn't void or NULL);
> 	struct value *foo (struct value *);
> 	post (VALUE_TYPE (foo (value)) == TYPE_CODE_FUNC
> 		|| VALUE_TYPE (foo (value)) == TYPE_CODE_METHOD));
> 
> it would be used in relative proximity to push_dummy_call and
> return_value (ex, replace find_function_addr).

Sure, this is just a nicer packaging of find_function_addr, which
seems to be doing two separable jobs:

- coercing various non-function values into callable things, and 

- extracting that callable thing's entry point address and return
  type.

A function value carries both those data in a handy and obvious way.