Non-conformant VertexAttribPointer implementation for OpenGL >= 3.0

[jonvaldes]

According to the OpenGL API docs ( http://www.opengl.org/sdk/docs/man3/xhtml/glVertexAttribPointer.xml ), the "pointer" element in glVertexAttribPointer has been repurposed in OpenGL 3.x to indicate an offset into the currently bound buffer, effectively treating that GLvoid* parameter as a 32/64 bit uint.

In simple cases, a "nil" value is enough to indicate a zero offset, but more complex cases require specific offsets, which because of the go-gl API design are not easy to convey (how to make an interface{} have a 0x00000005 address?)

This is why I believe a modification to the current API is required for this way of using this function.

I propose adding another function with a uintptr parameter, so integer values can be passed directly, just like they are in the C API. The function signature would look something like this:

func (indx AttribLocation) AttribPointerOffset(size uint, typ GLenum, normalized bool, stride int, offset uintptr)

Here's a diff for a possible implementation of this API change:

diff --git a/attriblocation.go b/attriblocation.go
index a9301d9..a4ce8fd 100644
--- a/attriblocation.go
+++ b/attriblocation.go
@@ -6,6 +6,7 @@ package gl

 // #include "gl.h"
 import "C"
+import "unsafe"

 // AttribLocation

@@ -48,6 +49,11 @@ func (indx AttribLocation) AttribPointer(size uint, typ GLenum, normalized bool,
                glBool(normalized), C.GLsizei(stride), ptr(pointer))
 }

+func (indx AttribLocation) AttribPointerOffset(size uint, typ GLenum, normalized bool, stride int, offset uintptr) {
+       C.glVertexAttribPointer(C.GLuint(indx), C.GLint(size), C.GLenum(typ),
+               glBool(normalized), C.GLsizei(stride), unsafe.Pointer(offset))
+}
+
 func (indx AttribLocation) EnableArray() {
        C.glEnableVertexAttribArray(C.GLuint(indx))
 }

[polyfloyd]

The existing function (AttribLocation.AttribPointer) is already capable of accepting and converting a uintptr to an offset. Or do you want that extra function to bypass the runtime reflection?

[jonvaldes]

Well, I totally missed that feature. Thanks for telling me about it!

May I then suggest adding some documentation to that function, as the library source code and examples don't say anything about that capability?

[pwaller]

Pull requests welcome!

[pwaller]

Closing this one since the original problem is resolved.