OpenGL is not culling faces properly when drawing OBJ model

Question

I'm trying to render a Teapot model from an OBJ file. I'm using the Fixed Function rendering pipeline, and I cannot change to the Programmable Pipeline. I would like to have some basic lighting and materials applied to the scene as well, so my teapot has a green shiny material applied to it. However, when I rotate the teapot around the Y-Axis, I can clearly see through to the back side of the teapot.

Here's what I've tried so far:

• Changing the way OpenGL culls the faces (GL_CCW, GL_CW, GL_FRONT, GL_BACK) and none produce the correct results.

• Changing which way OpenGL calculates the front of the faces (GL_FRONT, GL_CCW, GL_BACK, GL_CW) and none produce the correct results.

• Testing the OBJ file to ensure that it orders its vertices correctly. When I drag the file into https://3dviewer.net/ it shows the correct Teapot that is not see-through.

• Changing the lighting to see if that does anything at all. Changing the lighting does not stop the teapot from being see-through in some cases.

• Disabling GL_BLEND. This did nothing

Here is what I currently have enabled:

    glLightfv(GL_LIGHT0, GL_AMBIENT, light0Color);
    glLightfv(GL_LIGHT0, GL_DIFFUSE, light0DiffColor);
    glLightfv(GL_LIGHT0, GL_SPECULAR, light0SpecColor);
    glLightfv(GL_LIGHT0, GL_POSITION, position);
    glLightModelfv(GL_LIGHT_MODEL_AMBIENT, ambientIntensity);
    glEnable(GL_LIGHTING);
    glEnable(GL_LIGHT0);
    glEnable(GL_NORMALIZE);
    glEnable(GL_DEPTH_TEST);
    glEnable(GL_CULL_FACE);
    glCullFace(GL_CCW);
    glFrontFace(GL_CCW);

Here are the material properties:

    float amb[4] = {0.0215, 0.1745, 0.0215, 1.0};
    float diff[4] = {0.07568, 0.61424, 0.07568, 1.0};
    float spec[4] = {0.633, 0.727811, 0.633, 1.0};
    float shininess = 0.6 * 128;
    glMaterialfv(GL_FRONT, GL_AMBIENT, amb);
    glMaterialfv(GL_FRONT, GL_DIFFUSE, diff);
    glMaterialfv(GL_FRONT, GL_SPECULAR, spec);
    glMaterialf(GL_FRONT, GL_SHININESS, shininess);

Here is the rendering code:

    glClearColor(0.0, 0.0, 0.0, 1.0);
    glClearDepth(1.0);
    glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
    glLoadIdentity();
    glTranslatef(0, 0, -150);
    glRotatef(r, 0.0, 1.0, 0.0);
    glScalef(0.5, 0.5, 0.5);
    r += 0.5;
    m.draw(0, 0, 0);

I'm not sure if it's the cause of the problem, but I've included the model loading code below just in case it's relevant:

            while(std::getline(stream, line))
            {
                if (line[0] == 'v' && line[1] == 'n') // If we see a vertex normal in the OBJ file
                {
                    line = line.substr(3, line.size() - 3); // Removes the 'vn ' from the line
                    std::stringstream ss(line);
                    glm::vec3 normal;
                    ss >> normal.x >> normal.y >> normal.z;
                    tempNormalData.push_back(normal);
                }

                if (line[0] == 'v') // If we see a vertex on this line of the OBJ file
                {
                    line = line.substr(2, line.size() - 2); // Removes the 'v ' from the line
                    std::stringstream ss(line);
                    glm::vec3 position;
                    ss >> position.x >> position.y >> position.z;
                    tempVertData.push_back(position);
                }

                if (line[0] == 'f') // If we see a face in the OBJ file
                {
                    line = line.substr(2, line.size() - 2); // Removes the 'f ' from the line
                    std::stringstream ss(line);
                    glm::vec3 faceData;
                    ss >> faceData.x >> faceData.y >> faceData.z;
                    tempFaceData.push_back(faceData);
                }
            }

            if (tempVertData.size() != tempNormalData.size() && tempNormalData.size() > 0)
            {
                std::cout << "Not the same number of normals as vertices" << std::endl;
            }
            else
            {
                for (int i = 0; i < (int)tempVertData.size(); i++)
                {
                    Vertex v;
                    v.setPosition(tempVertData[i]);
                    v.setNormal(tempNormalData[i]);
                    vertices.push_back(v);
                }

                for (int i = 0; i < tempFaceData.size(); i++)
                {
                    Vertex v1 = vertices[tempFaceData[i].x - 1];
                    Vertex v2 = vertices[tempFaceData[i].y - 1];
                    Vertex v3 = vertices[tempFaceData[i].z - 1];
                    Face face(v1, v2, v3);
                    faces.push_back(face);
                }
            }
        }

Lastly, when I draw the faces I just loop through the faces list and call the draw function on the face object. The face draw function just wraps a glBegin(GL_TRIANGLES) and a glEnd() call:

    for (int i = 0; i < (int)faces.size(); i++)
    {
        auto& f = faces[i];
        f.draw(position);
    }

Face draw function:

glBegin(GL_TRIANGLES);
                glVertex3f(position.x + v1.getPosition().x, position.y + v1.getPosition().y, position.z + v1.getPosition().z);
                glNormal3f(v1.getNormal().x, v1.getNormal().y, v1.getNormal().z);

                glVertex3f(position.x + v2.getPosition().x, position.y + v2.getPosition().y, position.z + v2.getPosition().z);
                glNormal3f(v2.getNormal().x, v2.getNormal().y, v2.getNormal().z);

                glVertex3f(position.x + v3.getPosition().x, position.y + v3.getPosition().y, position.z + v3.getPosition().z);
                glNormal3f(v3.getNormal().x, v3.getNormal().y, v3.getNormal().z);
        glEnd();

I don't really want to implement my own Z-Buffer culling algorithm, and I'm hoping that there is a really easy fix to my problem that I'm just missing.

SOLUTION (thanks to Genpfault)

I had not requested a depth buffer from OpenGL. I'm using Qt as my windowing API, so I had to request it from my format object as follows:

format.setDepthBufferSize(32);

This requests a depth buffer of 32 bits, which fixed the issue.

Answer 1

in order to make face culling working you need to:

1. define winding rule

   glFrontFace(GL_CCW); // or GL_CW depends on your model and coordinate systems
   

2. set which faces to skip

   glEnable(GL_CULL_FACE);
   glCullFace(GL_BACK); // or GL_FRONT depends on what you want to achieve
   

   As you can see this is where you have a bug in your code as you are calling this with wrong parameter most likely causing new glError entries.

3. in case of concave mesh you need also depth buffer

   glEnable(GL_DEPTH_TEST);
   

   However your OpenGL context must have allocated depth buffer bits in its pixelformat during context creation. The most safe values are 16 and 24 bits however any decent nowadays gfx card can handle 32bit too. If you need more then you need to use FBO.

4. mesh with consistent polygon winding

   wavefront obj files are notorious for having inconsistent winding so in case you see some triangles flipped its most likely bug in the mesh file itself.

   This can be remedied either by using some 3D tool or by detecting wrong triangles and reverse their vertexes and flipping normal.

Also your rendering code glBegin/glEnd is written in very inefficient way:

glVertex3f(position.x + v1.getPosition().x, position.y + v1.getPosition().y, position.z + v1.getPosition().z);
glNormal3f(v1.getNormal().x, v1.getNormal().y, v1.getNormal().z);

for each of the component/operand you call some class member function and even making arithmetics ... The position can be done with simple glTranslate in actual GL_MODELVIEW matrix and if you got some 3D vector class try to access its components as pointer and use glVertex3fv and glNormal3fv instead that would be much much faster.