Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Zero divide is missing from normalize/normalize3 (glmath.h/glmath.cc) functions #177

Open
m-7761 opened this issue Mar 12, 2022 · 2 comments

Comments

@m-7761
Copy link

m-7761 commented Mar 12, 2022

https://github.com/zturtleman/mm3d/blob/master/src/libmm3d/glmath.cc#L875-L892

(Also found inline in the header and a Quaternion method.)

FWIW I was just testing the "Boolean" (model_bool.cc) code (adding texture coordinates to it) and it produces 0 length vectors. TBH I'm surprised this hasn't been an issue before now.

@m-7761
Copy link
Author

m-7761 commented Mar 12, 2022

Sorry, if you're interested here's some code for the Boolean function. I'm making an effort to fix the many editing functions that disregard texture mapping.

BTW, there second block of code prevents Model::booleanOperation from processing the split off triangles since it uses the same list for this. I get the same pattern in my test scenario (see update/note in next reply.) I think something that's missing is when a triangle is fully inset (like a stencil) inside a coplanar triangle. I suspected this wouldn't work because only two vertices are generated when cutting triangles, and this would need three. I think this kind of stenciling should be supported since it's very useful. Just another thought. (What happens is the inner triangle's normal/winding is reversed for some reason--some other part of the CSG logic I presume--but no cuts are made.)

Anyway, it's a wonder Kevin did all of this work but didn't bother to factor in texture mapping. He must've been quite a character going by a lot of the code and programmer's comments.

	//2022: This is new fix up code for texture mapping.
	//I couldn't find a way to make it work without the
	//normal vector not being normalized, so the 
	//https://www.scratchapixel.com/lessons/3d-basic-rendering/ray-tracing-rendering-a-triangle/barycentric-coordinates
	float st3[2][3];
	model->getTextureCoords(ut.tri,st3);
	float st1[2],st2[2];
	{
		auto &v0 = *(Vector*)ut.coord[0];
		auto &v1 = *(Vector*)ut.coord[1]; 
		auto &v2 = *(Vector*)ut.coord[2];

		Vector edge1 = v2-v1, edge2 = v0-v2;
				
		//auto &N = *(Vector*)ut.norm;
		double denorm = ut.denorm;
		double N[3]; //Un-normalize N?
		for(int i=3;i-->0;) N[i] = ut.norm[i]*denorm;
		//Revert saved length to N.dot(N) and prefer 
		//reciprocal for optimal division down below.
		denorm = 1/(denorm*denorm);

		for(int i=2;i-->0;)
		{	
			auto &P = *(Vector*)(i?nc2:nc1);
			
			//NOTE: normalize3 could be skipped if N wasn't
			//itself normalized... in which case u/v need to
			//be divided by N.dot3(N).
			double u = denorm*dot3(N,edge1.cross3(P-v1).getVector());
			double v = denorm*dot3(N,edge2.cross3(P-v2).getVector());
			double w = 1-u-v;

			float *st = i?st2:st1; for(int j=2;j-->0;)
			{
				st[j] = (float)(u*st3[j][0]+v*st3[j][1]+w*st3[j][2]);
			}			
		}
	}

	bool setUt = true;

	int group = model->getTriangleGroup(ut.tri);

	// now add the remaining triangles to the list
	for(auto&ea:ntl)	
	if(model_bool_isValidTriangle(model,ea.v[0],ea.v[1],ea.v[2]))
	{
		int tri; if(setUt)
		{
			tri = ut.tri; setUt = false;

			model->setTriangleVertices(tri,ea.v[0],ea.v[1],ea.v[2]);			
			model_bool_initUnionTriangle(model,ut,tri);
		}
		else
		{
			tri = model->addTriangle(ea.v[0],ea.v[1],ea.v[2]);
			if(group>=0) model->addTriangleToGroup(group,tri);

			model_bool_UnionTriangle nut;
			model_bool_initUnionTriangle(model,nut,tri);
			utl.push_back(nut);
		}

		float st[2][3]; for(int i=3;i-->0;) //2022
		{
			float &s = st[0][i], &t = st[1][i]; 
			int cmp = ea.v[i];
			if(cmp==v1){ s = st1[0], t = st1[1]; }
			if(cmp==v2){ s = st2[0], t = st2[1]; }
			if(cmp==v3[0]){ s = st3[0][0], t = st3[1][0]; }
			if(cmp==v3[1]){ s = st3[0][1], t = st3[1][1]; }
			if(cmp==v3[2]){ s = st3[0][2], t = st3[1][2]; }
		}
		model->setTextureCoords(tri,st);
	}

	if(setUt) log_error("No valid triangles in any splits\n");

@m-7761
Copy link
Author

m-7761 commented Mar 16, 2022

EDITED/UPDATE: Please disregard the (stupid) bit about the outer loop (I've removed that code and struck out that sentence from the previous comment/reply.) It's obvious that the split off triangles are still required to fully represent their source triangles. I'm sure the originals could be retained in a non-destructive way, but the splits themselves would still have to be applied in sequence. (As opposed to all at once.)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant