2014/1/4 Eliot Miranda <eliot.miranda@gmail.com>
Hi Nicolas,

On Wed, Jan 1, 2014 at 1:16 AM, Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com> wrote:

2013/12/31 Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com>
I forgot to tell this was in COG branch.
This file is generated, and the type is inferred from InterpreterProxy>>positive64BitIntegerFor:
I guess that a correct <returnTypeC:#sqInt> annotation is missing here.

The question is why using InterpreterProxy from the plugin code generator?
I find it strange to scatter (and duplicate) C-generation hints in a code that is not intended to be C-generated...

The problem is that the InterpreterProxy interface wasn't well-documented in Smalltalk.  Instead the definitive code is in platforms/Cross/vm/sqVirtualMachine.[ch].  I think it is better to have InterpreterProxy define the return types (via the pragmas) and to use these definitions for tests.  See VMInterfaceConsistencyTests>>#testInterpreterProxyInterface.

Since the default return type is sqInt there's no pragma to specify it.  But I agree it would be preferrable to be explicit.

Currently, it's not preferable, it's mandatory because the return type is inferred as sqLong which is incorrect.

2013/12/31 Nicolas Cellier <nicolas.cellier.aka.nice@gmail.com>
positive64BitIntegerFor return type is declared sqLong, but should be sqInt (it returns an oop).