Apologies if this is in the wrong forum as perhaps it's a general problem, but I am specifically using the DirectX 11 runtime. I'm sure my problem is some basic matter that I just don't understand yet and I'm hopeful someone can help. I'm new to graphics, and sometimes I don't know what it is that I don't yet know.

Here's my situation. I have a dirt-simple DX11 renderer into which I'm loading mesh data that was parsed in from FBX (created in Maya and Blender). As long as the mesh is a simple primitive like a cube or a sphere everything works wonderfully. The mesh is recreated faithfully on my screen, and in fact I was getting to work on pixel shaders when I noticed something odd.

If the mesh is less simple -- say, if I grabbed a face in Maya and moved or extruded it a bit -- then that extruded geometry does not clip properly when I render the mesh. Exhibit A:

As you can see, the modified geometry doesn't cull correctly.

I didn't know what to make of that so I made a couple of geometry shaders to help me debug. That's what you see in the video. The lines are the calculated face normals and the vertex colors are coded R, G, B in the order received by the geometry shader. I had originally thought that the offending polys might be getting wound cw rather than ccw as a result of operations in Maya, but they all appear to be ccw. The calculated normals also appear to be pointing the right way.

I've looked at my depth/stencil states and they appear fine, but I expected that since I can draw spheres and cubes and such correctly all day. My near clip plane isn't set to zero; I thought of that.

Frankly I don't know what else to check. This has been driving me batty all day. Please send help.

** edit

Just wanted to add that I do have the debug layers going and I'm checking my HRESULTS and nothing is amiss as far as that goes. And I'm drawing with an opaque blend state.

This is a correct forum.

It looks like your algorithm is ignoring detph.

It would be a small bug in your shader.

Maybe you would paste your HLSL here.

That would be nice but if that were the case wouldn't everything be drawing incorrectly? In an effort to nail down the problem I've been running a very simplified shader on this:

cbuffer mat_camera : register( b0 )
{
	matrix gView;
	matrix gProjection;
	float3 gCamPosition;
	float3 gCamTarget;
};

cbuffer mat_object : register( b1 )
{
	matrix gWorld;
};



struct VS_IN
{
	float3 position	: POSITION;
	float3 normal	: NORMAL;
	float3 tangent	: TANGENT;
	float3 binormal	: BINORMAL;
	float2 uv		: TEXCOORD0;
};

struct VS_OUT
{
	float4 position : SV_POSITION;
	float3 normal	: NORMAL;
	float3 tangent	: TANGENT;
	float3 binormal	: BINORMAL;
	float2 uv		: TEXCOORD0;
};

VS_OUT main( VS_IN IN )
{
	VS_OUT OUT;

	OUT.position = float4( IN.position.xyz, 1.0 );
	OUT.position = mul( OUT.position, gWorld );
	OUT.position = mul( OUT.position, gView );
	OUT.position = mul( OUT.position, gProjection );

	OUT.normal = normalize( mul( IN.normal, ( float3x3 )gWorld ) );

	OUT.tangent = IN.tangent;
	OUT.binormal = IN.binormal;
	OUT.uv = IN.uv;

	return OUT;
}

Texture2D		gTex;
SamplerState	gSampler;


struct PS_IN
{
	float4 position : SV_POSITION;
	float3 normal	: NORMAL;
	float3 tangent	: TANGENT;
	float3 binormal	: BINORMAL;
	float2 uv		: TEXCOORD0;
};


float4 main( PS_IN IN ) : SV_TARGET
{
	float4 pixel = gTex.Sample( gSampler, IN.uv );
	return  pixel;
}

It looks like a depth stencil issue to me. Can you post your depth stencil related code and/or double check it with a frame debugger?

I've looked at my depth/stencil states and they appear fine, but I expected that since I can draw spheres and cubes and such correctly all day.

Your spheres and cubes render correctly because of back-face culling (i.e. you never have two front-faces "overlapping").

Solved. So after all that it turned out to be the order of operations in my handler for WM_SIZE. There's a very good reason why it was acting like there was no depth buffer: the call to OMSetRenderTargets was BEFORE I recreated the depth view. Oh, deary me. Obvious really when I saw it. And of course OMSetRenderTargets has to be a void-return; an HRESULT_STUPID_USER return might have saved me the time.

Mona, you were spot on. Thanks for pointing me in the right direction.

** edit

In retrospect, you know what else should have been a dead giveaway? The linestream coming out of my geometry shader, back when I thought it might have been poly winding. The lines weren't depth-clipped at all. Bloody hell I must have taken a stupid pill this morning.