Separate stencil with FBO (NVidia) - opengl

Question

I'm trying to have a FBO with separate depth and stencil buffers.
I know that NVidia GPU's historically only supported packed depth/stencil.
However I stumbled on the ARB_texture_stencil8 extension and wonder how to use it against a FBO.
This code gives gl error 1159 on glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_STENCIL_ATTACHMENT_EXT, GL_TEXTURE_2D, fboStencilTexture, 0):
static GLuint fboId, rboDepth, rboStencil, TEXTURE_WIDTH, TEXTURE_HEIGHT;
if (!fboColorTexture) {
fboDepthTexture = fboId = rboDepth = rboStencil = TEXTURE_WIDTH = TEXTURE_HEIGHT = 0; // vid restart?
glGenTextures(1, &fboColorTexture);
glBindTexture(GL_TEXTURE_2D, fboColorTexture);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_REPEAT);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_REPEAT);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
}
if (!fboDepthTexture) {
glGenTextures(1, &fboDepthTexture);
glBindTexture(GL_TEXTURE_2D, fboDepthTexture);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
}
if (!fboStencilTexture) {
glGenTextures(1, &fboStencilTexture);
glBindTexture(GL_TEXTURE_2D, fboStencilTexture);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
qglTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
}
GLuint curWidth = r_virtualResolution.GetFloat() * glConfig.vidWidth, curHeight = r_virtualResolution.GetFloat() * glConfig.vidHeight;
if (curWidth != TEXTURE_WIDTH || curHeight != TEXTURE_HEIGHT) {
TEXTURE_WIDTH = curWidth;
TEXTURE_HEIGHT = curHeight;
glBindTexture(GL_TEXTURE_2D, fboColorTexture);
glTexImage2D(GL_TEXTURE_2D, 0, GL_RGB5_A1, TEXTURE_WIDTH, TEXTURE_HEIGHT, 0, GL_BGRA, GL_UNSIGNED_BYTE, NULL); //NULL means reserve texture memory, but texels are undefined
glBindTexture(GL_TEXTURE_2D, fboDepthTexture);
glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT16, TEXTURE_WIDTH, TEXTURE_HEIGHT, 0, GL_DEPTH_COMPONENT, GL_FLOAT, 0);
glBindTexture(GL_TEXTURE_2D, fboStencilTexture);
glTexImage2D(GL_TEXTURE_2D, 0, GL_STENCIL_INDEX8, TEXTURE_WIDTH, TEXTURE_HEIGHT, 0, GL_STENCIL_INDEX, GL_FLOAT, 0);
}
//-------------------------
if (!fboId) {
// create a framebuffer object, you need to delete them when program exits.
glGenFramebuffersEXT(1, &fboId);
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, fboId);
// attach a texture to FBO color attachement point
glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_2D, fboColorTexture, 0);
// attach a renderbuffer to depth attachment point
glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_TEXTURE_2D, fboDepthTexture, 0);
glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_STENCIL_ATTACHMENT_EXT, GL_TEXTURE_2D, fboStencilTexture, 0);
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
}
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, fboId);

Answer 1

The purpose of ARB_texture_stencil8 is to permit you to use stencil-only formats as textures. That is, reading from them in shaders.
This does not mean that you can separate your stencil and depth buffers. The idea behind stencil-8 textures is that you would generate their data by either copying the stencil portion of a depth/stencil texture, or you would render without a depth buffer entirely.
So you cannot use this extension to guarantee that you can render to separate depth and stencil images. That's still hardware dependent.
Also, if you're going to use new features like ARB_texture_stencil8, you shouldn't be combining them with old EXT_framebuffer_object APIs. So stop using glFramebufferTexture2DEXT and start using glFramebufferTexture.