Render to texture problems

Question

I'm tyring to get rendering to texture working, but am having a bit of trouble. I'm trying to render a simple quad to the screen to make sure everything is working.

    target->enable();

    glBegin(GL_QUADS);

    glColor4f(1.f, 0.f, 0.f, 1.f);
    glVertex3f(0.f, 0.f, 0.f);

    glColor4f(0.f, 1.f, 0.f, 1.f);
    glVertex3f(16.f, 0.f, 0.f);

    glColor4f(0.f, 0.f, 1.f, 1.f);
    glVertex3f(0.f, 16.f, 0.f);

    glColor4f(1.f, 1.f, 0.f, 1.f);
    glVertex3f(16.f, 16.f, 0.f);

    glEnd();

    target->disable();

When I draw the quad to the screen normally, the quad renders as expected. However, when I draw the quad with the render target enabled, the quad gets rendered at the bottom left corner of the screen and ends up not being rendered to the target.

RenderTarget::RenderTarget(GraphicsDevice *graphics, GLuint width, GLuint height) {

    mWidth = width;
    mHeight = height;

    // First create the depth buffer
    glGenRenderbuffers(1, &mDepth);
    glBindRenderbuffer(GL_RENDERBUFFER_EXT, mDepth);

    // Tell OpenGL that this renderbuffer is going to be a depth buffer
    glRenderbufferStorage(GL_RENDERBUFFER_EXT, GL_DEPTH_COMPONENT, mWidth, mHeight);
    glFramebufferRenderbuffer(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, mDepth);

    glBindRenderbuffer(GL_RENDERBUFFER_EXT, 0);

    // Create the frame buffer texture
    glGenTextures(1, &mTexture);
    glBindTexture(GL_TEXTURE_2D, mTexture);

    glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, mWidth, mHeight, 0, GL_RGBA, GL_UNSIGNED_BYTE, NULL);

    glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
    glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
    glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
    glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);

    glBindTexture(GL_TEXTURE_2D, 0);

    // Create the frame buffer
    glGenFramebuffers(1, &mFbo);
    glBindFramebuffer(GL_FRAMEBUFFER_EXT, mFbo);

    // Attach the texture
    glFramebufferTexture2D(GL_FRAMEBUFFER_EXT, GL_COLOR_ATTACHMENT0_EXT, GL_TEXTURE_2D, mTexture, 0);

    //Attach the depth buffer
    glFramebufferRenderbuffer(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, mFbo);

    glBindFramebuffer(GL_FRAMEBUFFER_EXT, 0);


}

RenderTarget::~RenderTarget() {
    glDeleteBuffers(1, &mFbo);
    glDeleteBuffers(1, &mDepth);
    glDeleteTextures(1, &mTexture);
    mFbo = 0;
    mDepth = 0;
    mTexture = 0;
}

void RenderTarget::enable() {
    // Bind the frame buffer
    glBindFramebuffer(GL_FRAMEBUFFER_EXT, mFbo);
    // store the glViewport and glEnable states
    //glPushAttrib(GL_VIEWPORT_BIT | GL_ENABLE_BIT);

    glClearColor(1.f, 0.f, 0.f, 1.f);
    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
    glClearColor(0.f, 0.f, 0.f, 0.f);

    glLoadIdentity();
}

void RenderTarget::disable() {
    // Restore glViewport and glEnable states
    glPopAttrib();
    glBindFramebuffer(GL_FRAMEBUFFER_EXT, 0);
}

Answer 1

Your enable is resetting the current matrix (most likely modelview) with glLoadIdentity. This explains why your square is being moved to the corner of the screen.

You've commented out the call to glPushAttrib but not the matching call to glPopAttrib. If you don't use glPushAttrib anywhere else this should set an error flag and leave GL state alone, but I don't know that's the case and it's asking for trouble to assume that the attribute stack is and always will be empty.

This documentation for glBindFrameBuffer suggests that you ought to be using GL_FRAMEBUFFER or GL_DRAW_FRAMEBUFFER as the first argument:

target must be either GL_DRAW_FRAMEBUFFER, GL_READ_FRAMEBUFFER or GL_FRAMEBUFFER. If a framebuffer object is bound to GL_DRAW_FRAMEBUFFER or GL_READ_FRAMEBUFFER, it becomes the target for rendering or readback operations, respectively, until it is deleted or another framebuffer is bound to the corresponding bind point. Calling glBindFramebuffer with target set to GL_FRAMEBUFFER binds framebuffer to both the read and draw framebuffer targets.

The enums with _EXT suffix in your code likely pertain to a version of OpenGL previous to the version documented there. Tutorials are prone to allowing details like this to fall out of date.

Check your environment & rely on the documentation that matches your version of OpenGL. If for some reason you can't easily determine your version of OpenGL, check for the GL_INVALID_ENUM error after calling glBindFrameBuffer.

Answer 2

This code seems wrong:

//Attach the depth buffer
glFramebufferRenderbuffer(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT, GL_RENDERBUFFER_EXT, mFbo);

What you are doing here is that you are trying to attach mFbo as a render buffer to mFbo. However, you already have your render buffer attached correctly (which is done in the first call to glFramebufferRenderbuffer()). I think removing that code should fix your problem.

Answer 3

As I suspected, my problem was setting glViewport, and glOrtho correctly when rendering the FBO. This code needs to be cleaned up, but here is how I fixed it.

void RenderTarget::enable() {
    // Bind the frame buffer

    glBindFramebuffer(GL_FRAMEBUFFER_EXT, mFbo);
    // store the glViewport and glEnable states
    //glPushAttrib(GL_VIEWPORT_BIT | GL_ENABLE_BIT);

    glViewport(0, 0, mWidth, mHeight);
    // projection matrix
    glMatrixMode(GL_PROJECTION);
    glLoadIdentity();

    // orthographic projection
    glOrtho(0, mWidth, 0, mHeight, -1, 1);

    glMatrixMode(GL_MODELVIEW);
    glLoadIdentity();

    glClearColor(1.f, 0.f, 0.f, 1.f);
    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
    glClearColor(0.f, 0.f, 0.f, 0.f);

    glLoadIdentity();
}

void RenderTarget::disable() {
    // Restore glViewport and glEnable states
    //glPopAttrib();
    glBindFramebuffer(GL_FRAMEBUFFER_EXT, 0);
    glViewport(0, 0, 1024, 768);
    glMatrixMode(GL_PROJECTION);
    glLoadIdentity();

    // orthographic projection
    glOrtho(0, 1024, 768, 0, -1, 1);

    glMatrixMode(GL_MODELVIEW);
    glLoadIdentity();
}