Messed Up OpenGL Depth Buffer? - c++

Question

I have something rather strange going on at the moment with my code. I am running this on a BlackBerry Playbook and it is OpenGL ES 1.1
EDIT 4: I deleted everything I have posted to simplify my question.
I took the code and simplified it to drawing two overlapping triangles. Here is the array containing the coordinates as well as an array containing colours:
GLfloat vertices[] =
{
// front
175.0f, 200.0f, -24.0f,
225.0f, 200.0f, -24.0f,
225.0f, 250.0f, -24.0f,
// back
200.0f, 200.0f, -25.0f,
250.0f, 200.0f, -25.0f,
250.0f, 250.0f, -25.0f
};
static const GLfloat colors[] =
{
/* front */ 1.0f,0.0f,0.0f,1.0f,1.0f,0.0f,0.0f,1.0f,1.0f,0.0f,0.0f,1.0f, //Red
/* back */ 0.0f,1.0f,0.0f,1.0f,0.0f,1.0f,0.0f,1.0f,0.0f,1.0f,0.0f,1.0f //Green
};
Please note that my coordinates are 0 to 1024 in the x direction and 0 to 600 in the y direction as well as 0 to -10000 in the z direction.
Here is my setup code which reflects this:
glClearDepthf(1.0f);
glClearColor(1.0f,1.0f,1.0f,1.0f);
glEnable(GL_DEPTH_TEST);
glDepthMask(GL_TRUE);
glShadeModel(GL_SMOOTH);
glViewport(0, 0, surface_width, surface_height);
glMatrixMode(GL_PROJECTION);
glLoadIdentity();
glOrthof(0, surface_width, 0, surface_height, 0, 10000);
glMatrixMode(GL_MODELVIEW);
glLoadIdentity();
glEnable(GL_DEPTH_TEST);
glDepthMask(GL_TRUE);
I have depth enabling in two places as I was trying to rule out the possibility that it was supposed to be used while a certain matrix mode was chosen.
Lastly here is my render code:
void render()
{
//Typical render pass
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
glEnableClientState(GL_VERTEX_ARRAY);
glVertexPointer(3, GL_FLOAT, 0, vertices);
glEnableClientState(GL_COLOR_ARRAY);
glColorPointer(4, GL_FLOAT, 0, colors);
glDrawArrays(GL_TRIANGLES, 0 , 6);
glDisableClientState(GL_VERTEX_ARRAY);
glDisableClientState(GL_COLOR_ARRAY);
//Use utility code to update the screen
bbutil_swap();
}
The issue is that no matter what I do the green triangle is always overlayed over the red one. Changing z values either way has no effect on the finished image. I cannot figure this out.

Answer 1

By default, depth testing is disabled. You have to enable it with glEnable(GL_DEPTH_TEST). The reason why it is working when you enable culling is because the back facing triangles are not drawn, and since a cube is a convex polyhedron, no front-facing quad will ever overlap another front-facing quad. If you try to render a second cube, however, you will see depth problems as well, unless you enable depth testing.

Answer 2

I finally got it to work. The issue was with EGL setup code that I used that was provided. In bbutil.c (in my case .cpp) there is some code:
if(!eglChooseConfig(egl_disp, attrib_list, &egl_conf, 1, &num_configs)) {
bbutil_terminate();
return EXIT_FAILURE;
}
(that is not all the code in the file but its the important bit)
This basically freaks if the given attribute list is nor supported. Up higher in the file attrib_list is set as follows:
EGLint attrib_list[]= { EGL_RED_SIZE, 8,
EGL_GREEN_SIZE, 8,
EGL_BLUE_SIZE, 8,
EGL_SURFACE_TYPE, EGL_WINDOW_BIT,
EGL_RENDERABLE_TYPE, 0,
EGL_NONE};
There is no depth buffer specified. Now if you look in the EGL spec it says no depth is the default. BINGO, that's the problem. So I just modified it to look like this:
EGLint attrib_list[]= { EGL_RED_SIZE, 8,
EGL_GREEN_SIZE, 8,
EGL_BLUE_SIZE, 8,
EGL_SURFACE_TYPE, EGL_WINDOW_BIT,
EGL_RENDERABLE_TYPE, 0,
EGL_DEPTH_SIZE, 24,
EGL_NONE};
Note the EGL_DEPTH_SIZE and the 24. This sets the depth buffer to 24 bits. On the PlayBook 32 throws a segmentation fault although usually 32 is not supported anyways. Perhaps this will help someone out there trying to figure out why the provided include is causing this funny result I described as my problem.