ffi_backend: convert numeric function args to pointers

[danini-the-panini]

This allows for passing integers as pointer arguments to functions when using the FFI backend. This is a workaround until we can get JRuby's FFI implementation to allow for it directly (see also jruby/jruby#8423)

[kou]

Could you add a test for this case?

[headius]

Passing integers as pointers might be a good feature to add, but patching it into JRuby led to failures in FFI specs that confirm integers are prevented from being passed as pointers. I've asked @danini-the-panini to propose it to ffi/ffi and we can discuss from there.

The workaround provided here will be needed until that discussion and feature can happen.

[danini-the-panini]

I've also allowed any integer coercible to passed in as a pointer address, to align with native Fiddle

[larskanis]

This looks good to me now.

[larskanis]

The idea of accepting integers where a pointer is expected in a function call is more or less rejected in ffi/ffi#1130 . So this PR seems to be the way to proceed.

[headius]

LGTM

[kou]

Suggested change

	def test_to_ptr_with_num
	def test_to_ptr_with_float

[danini-the-panini]

this isn't really testing float specific (see my comment below) but rather testing integer coercion. Perhaps this is a more apt name?

Suggested change

	def test_to_ptr_with_num
	def test_to_ptr_with_int_coercion

[kou]

Ah, it may be better that we reject Float for address. Because Float address is invalid.

@tenderlove What do you think about this?

[danini-the-panini]

The only reason I added this test is to make sure the functionality of the FFI backend matches the C implementation. I just picked an arbitrary builtin that can be coerced into an integer 🤷🏻‍♀️ There might be other types that make more sense, or I could just create a dummy int-wrapper and use that instead, e.g.

IntWrapper = Struct.new(:value) do
  def to_int
    value
  end
end

and then

Suggested change

	assert_equal ptr, Pointer[0.0]
	assert_equal ptr, Pointer[IntWrapper.new(0)]

[tenderlove]

Ah, this is just a test for coercion, I guess. @danini-the-panini maybe adding the wrapper will show the intention of the test better? I don't really have a strong opinion about float vs wrapper object as long as the intention of the test is clear. Either a comment or the wrapper class seems sufficient to me.

I assume this is calling to_int on things because it wants to accept other pointer instances as a parameter.

[danini-the-panini]

It's calling rb_Integer in C, and Kernel#Integer in the FFI backend

[danini-the-panini]

@kou @tenderlove I've updated it to use a wrapper and renamed the test method

[tenderlove]

It's calling rb_Integer in C, and Kernel#Integer in the FFI backend

ah, gotcha.

@kou @tenderlove I've updated it to use a wrapper and renamed the test method

Thank you!