From: "David Simmons" <david.simmons@smallscript.com>
References: <usmy1m39d.fsf@double.co.nz>
Subject: Re: OpenGL Example and float problem
Date: Sat, 16 Nov 2002 17:09:03 -0800
Lines: 66
Organization: SmallScript Corp
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2800.1106
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2800.1106
Message-ID: <yvexrXdjCHA.2252@minos.QKS.COM>
Newsgroups: comp.lang.smallscript
Path: minos.QKS.COM
Xref: minos.QKS.COM comp.lang.smallscript:2156
NNTP-Posting-Host: mail.smallscript.llc 207.213.215.173

"Chris Double" <chris@double.co.nz> wrote in message
news:usmy1m39d.fsf@double.co.nz...
> I threw together an example of using OpenGL in SmallScript. It's
> available at:
>
> http://www.double.co.nz/smallscript/OpenGLExample.zip
>
> For some reason passing floats to an FFI function don't seem to
> work. Is there a problem there? For example, in the OpenGL example I
> have code like:
>
>     glColor3ub(255, 0, 0).
>     glVertex2i(0, 1).
>
> If I replace 'glColor3ub' with 'glColor3f' or 'glColor3d' then the
> example doesn't work. eg:
>
>     glColor3f(1.0, 0, 0).
>
> (When using floats the range is from -1 to 1.0. When using unsigned
> byte (ub) the range is 0 to 255).
>
> No error occurs, it's just the function seems to silently fail.

The "auto-marshalling" cannot determine how to pass floats without a hint.
So it is not passing the floats correctly. I.e., it does not know if they
should be passed in a register, as a pointer, etc. By default it will pass
them by reference as <double*>, but you really want them passed by-value as
<double>. So you have to explicitly override the auto-marshaller's default.

You need to provide a hint, as follows:

"" Presuming you have the OpenGL dll wrappered as follows:
Namespace name: OpenGL dll: OpenGL
{
    "" Declare an extern prototype to hint how values
    "" should be marshalled.
    Function [<$extern>
    glColor3f(<~double>a, <~double> b, <~double> c)
    ]
}

Where "~" says ignore the type information for the purposes of multi-method
binding (this just makes it a little faster), and the "double" says how the
value should be marshalled. Where <double> is a pseudonym for <Float64>
which is a pseudonym for <Float>. You can see the actual marshaller code in
the <Float> methods for #ffiMarshallToFFV, and #ffiMarshallFromFFV.

For reference to additional details on float operations, see the c.l.s#
discussions "Re: Float Arrays: FFI and memory layout" from "Feb-2002" which
I had with Jaco van der Merwe <jvdm@crusader.co.za>.

-- Dave S. [www.smallscript.org]

>
> I've also done an example using the WildTangent game library [1]. This
> tests my IDL generator code. But this library uses floats and those
> functions also silently fail.
>
> [1] http://www.wildtangent.com/
>
> Chris.
> --
> http://www.double.co.nz/smallscript