My program renders a scene from different viewpoints, saves results to textures, and then renders meshes using those textures into the output buffer. Ideally it should simulate a wide angle camera.
The scene itself is a box, with 6 square faces, illuminated by a point light source. If I render the faces using GL_POLYGON, the result (composed from different viewpoints) looks seamless. But if I use GL_QUADS I can see the edges between meshes for different channels (polygons are illuminated slightly differently).
This problem only happens when I use a point light, it looks fine with directional light.
Does GL_QUADS triangulate differently depending on the frustum or something?