Thanks David,
Since I’m not at all familiar with the intricacies of OpenGL/DirectX/Panda3D, here is a sketch (stuttering…) of what I’ve understood:
(1) in glGraphicsBuffer_src.h, make public:
GLuint _fbo;
PT(Texture) _tex[RTP_COUNT];
(2) then the code could look like
#include "glgsg.h"
#include "glGraphicsBuffer_src.h"
// assuming directx already on with d3ddev ptr
if( NULL == (d3d = Direct3DCreate9(D3D_SDK_VERSION))) exit(0); // create the Direct3D interface
....
HRESULT hr = d3d->CreateDevice(D3DADAPTER_DEFAULT, D3DDEVTYPE_HAL, hWnd,
D3DCREATE_SOFTWARE_VERTEXPROCESSING,
&d3dpp,&d3ddev);
...
// panda actions:
// set texture buffer (ABuffer) that will be written as a result of CameraA seiing the scene
PT(GraphicsBuffer) ABuffer =
DCAST(GraphicsBuffer,window->get_graphics_output()->make_texture_buffer("ABuffer", wtexture, htexture));
PT(Texture) ABuffertexture = ABuffer->get_texture();
camA = new Camera("camA");
camA_NP = camera.attach_new_node(camA);
ABuffer->set_sort(-200);
camA->set_scene(render);
PT(DisplayRegion) ADr = ABuffer->make_display_region(0,1,0,1);
ADr->set_camera(camA_NP);
// will do the same for additional cameras (B,C,D)
...
// If GraphicsBuffer is wglGraphicsBuffer get gl descriptors
if (ABuffer->get_type().get_name() == "GLGraphicsBuffer") {
PT(GLGraphicsBuffer) gl_ABuffer = DCAST(GLGraphicsBuffer,ABuffer);
// Register directx device for interop with opengl
gl_handleD3D = wglDXOpenDeviceNV(d3ddev); // interop device handle
// Create d3d surface (gImage)
d3ddev->CreateOffscreenPlainSurface(
ImageWidth,
ImageHeight,
D3DFMT_A8R8G8B8, D3DPOOL_DEFAULT, // Surface is in video memory
&gImage, NULL);
// Register this DirectX buffer as GL texture object
GLuint gl_name_image;
glGenTextures(1, &gl_name_image); // reserver le nom d'une texture
// prepares this DirectX object for use by the GL
HANDLE handle_image = wglDXRegisterObjectNV(gl_handleD3D, gImage,
gl_name_image, GL_TEXTURE_2D_MULTISAMPLE, WGL_ACCESS_READ_WRITE_NV);
// Lock the render target for GL access
wglDXLockObjectsNV(gl_handleD3D, 1, &handle_image);
// Now transfer openGL ABuffertexture into part of Directx gImage
BindFramebufferEXT(READ_FRAMEBUFFER_EXT, gl_ABuffer->fbo);
BindFramebufferEXT(DRAW_FRAMEBUFFER_EXT, gl_name_image );
BlitFramebufferEXT(0, 0, 1024, 768, 80, 80, 1024-80, 768-80,GL_COLOR_BUFFER_BIT,GL_NEAREST);
// Do the same for cameras B,C & D
// Unlock gImage from openGL
wglDXUnlockObjectsNV(gl_handleD3D, 1, &handle_image);
// Do some additional processing in directx on gImage
....
// Get the Backbuffer then Stretch the Surface on it.
RECT srcRect= { 0, 0, gImageWidth/2, gImageHeight/2};
RECT dstRect= { 0, 0, gImageWidth, gImageHeight};
d3ddev->GetBackBuffer(0, 0, D3DBACKBUFFER_TYPE_MONO, &gBackBuf);
d3ddev->StretchRect(gImage, &srcRect, gBackBuf, &dstRect, D3DTEXF_NONE);
// then Direct3D presents the results on the screen
d3ddev->Present(NULL, NULL, NULL, NULL); // displays the created frame
}
By far I’m not sure that this is the correct approach, and obviously this rough code is not likely to work as such 
How to make it work?? 
But there are many, many OpenGL extensions. I’m guessing that the trailing X adds a new revision to the original extension. Are you sure that your graphics driver actually provides the extensions that you’re trying to call? If you call an extension that isn’t provided, you will certainly crash miserably.