GLSL 4 provides the ability to configure how the depth test is performed. This gives us additional control over how and when fragments are tested against the depth buffer.
Many OpenGL implementations automatically provide an optimization known as the early depth test or early fragment test. With this optimization, the depth test is performed before the fragment shader is executed. Since fragments that fail the depth test will not appear on the screen (or the framebuffer), there is no point in executing the fragment shader at all for those fragments and we can save some time by avoiding the execution.
The OpenGL specification, however, states that the depth test is performed after the fragment shader. This means that if an implementation wishes to use the early depth test optimization, it must be careful. The implementation must make sure that if anything within the fragment shader might change the results of the depth test, then it should avoid using the early depth...