7 replies to this topic

[NimrodPSI]

Hello all, I have recently begun the switch from old OpenGL to the newer specification that uses VAO's and VBO's for like... everything.
I am having alot of trouble getting texturing to work with GLSL and my VAO's, I managed to get lighting and color with GLSL working earlier(using an individuale VBO for ther verts, color, and normals) so i know my view/projection/model matrixies are all okay. I manually generate all the verticies of a cube into v[108] (im not using TRIAGLE_STRIPS right now so its 2 tri's a side) and then the tex coords for each vertex into t[72]. Since each vertex of 3 floats only requires two tex coords right? Might be useful to know im going off Swiftless's OpenGL 4 GLSL tutorials so I'm using glm and glew.

Also I use glBindAttribute to always pass the vertices position as a vec3 and to pass the tex coords as a vec2 into my Shader program(the same way i passed verts, color and normals earlier)

Any way heres some code:

This generates my VAO from 2 VBO's, one fro my verts and the other tex coords.

glGenVertexArrays(1, &vaoID);
glBindVertexArray(vaoID);

glGenBuffers(2, &vboID[0]);
//Verticies
glBindBuffer(GL_ARRAY_BUFFER, vboID[0]);
glBufferData(GL_ARRAY_BUFFER, 108 * sizeof(GLfloat), v, GL_STATIC_DRAW);
glVertexAttribPointer((GLuint)0, 3, GL_FLOAT, GL_FALSE, 0, 0);
glEnableVertexAttribArray(0);
//Tex Coords
glBindBuffer(GL_ARRAY_BUFFER, vboID[1]);
glBufferData(GL_ARRAY_BUFFER, 72 * sizeof(GLfloat), t, GL_STATIC_DRAW);
glVertexAttribPointer((GLuint)1, 2, GL_FLOAT, GL_FALSE, 0, 0);
glEnableVertexAttribArray(1);

glBindVertexArray(0);

Then I draw the cube like so:

shader->bind();
	 int projectionMatrixLocation = glGetUniformLocation(shader->id(), "projectionMatrix");
	 int viewMatrixLocation = glGetUniformLocation(shader->id(), "viewMatrix");
	 int modelMatrixLocation = glGetUniformLocation(shader->id(), "modelMatrix");
	 int texture_location = glGetUniformLocation(shader->id(), "texture_color");

	 glUniformMatrix4fv(projectionMatrixLocation, 1, GL_FALSE, &projectionMatrix[0][0]);
	 glUniformMatrix4fv(viewMatrixLocation, 1, GL_FALSE, &viewMatrix[0][0]);
	 glUniformMatrix4fv(modelMatrixLocation, 1, GL_FALSE, &modelMatrix[0][0]);
	  
	 glActiveTexture(GL_TEXTURE0);
	 glBindTexture(GL_TEXTURE_2D, texture[0]);
	 glUniform1i(texture_location, 0);

	
	 glBindVertexArray(vaoID);
	 glDrawArrays(GL_TRIANGLES, 0, 36);
	 glBindVertexArray(0);


shader->unbind();

...and my Vertex and Pixel(fragment) shaders respectivley:

#version 150 core

uniform mat4 projectionMatrix;  
uniform mat4 viewMatrix;  
uniform mat4 modelMatrix;

in vec3 in_Position;
in vec2 in_TexCoord;

void main(void){
	gl_TexCoord[0] = in_TexCoord;
		
	gl_Position = projectionMatrix * viewMatrix * modelMatrix * vec4(in_Position, 1.0);  
}

#version 150 core  

uniform sampler2D texture_color;

void main(void){
	gl_FragColor = texture2D(texture_color, gl_TexCoord[0].st);
		
}

The result is a white cube that seems to lose its scaleing compared to when I specify a solid gl_FragColor

My assumtion is something to do with gl_TexCoord[0] not being a vec2 like in_TexCoord but I can't seem to find much/any documentaion on GLSL when using custom arrays for everything. All the examples I've seen use gl_MultiTexCoord[0] to pull the tex coords from the now depriciated glMultiTexCoordxx()... I think .
I can post the vertex and tex coords initialization routine if that would help but its alot of crap lol.
I apologize if I'm not giving enough info or too much lol. I greatly appreciate any help given!!

[karwosts]

I'm not sure about the gl_TexCoord[0] issue, but I would suggest just to define your own varying texcoord variable and not use the builtin one. Are you checking shaders and programs for compile errors?

You said you're using glBindAttribute, do you mean glBindAttribLocation?

[Jesse7]

If you read Swiftless, he says:

In GLSL 1.50, we no longer have gl_ModelViewMatrix, gl_ProjectionMatrix, gl_Vertex, etc, etc

So it appears you cannot use the built-in texture coordinate array. Declare your own varying vec2 variable, set it to the attribute texture coordinate in the vertex shader, and feed it to the fragment shader. Also, don't use the built-in fragment color variable. Declare your own vec4. Something like this:

#version 150 core

uniform mat4 projectionMatrix;  
uniform mat4 viewMatrix;  
uniform mat4 modelMatrix;

in vec3 in_Position;
in vec2 in_TexCoord;


out vec2 out_TexCoord;


void main(void){
		out_TexCoord = in_TexCoord;
				
		gl_Position = projectionMatrix * viewMatrix * modelMatrix * vec4(in_Position, 1.0);  
}

#version 150 core  

in vec2 out_TexCoord;
uniform sampler2D texture_color;
out vec4 FragColor;

void main(void){
		FragColor = texture2D(texture_color, out_TexCoord.st);
				
}

If that doesn't work, check the stuff that Karwosts mentioned above.

[NimrodPSI]

@karwosts - Yes I meant glBindAttribLocation, sorry I had literally been up all night working with the new GL specification and It was starting to hurt my brain lol.

Okay I dropped the usage of gl_TexCoord[0] and did as Jesse7 suggested but that still didn't work, although not because it's incorrect. The issue seems to be the texture loader I have been using all the time from all my intermediate mode stuff. Not sure what was wrong with it but replaceing it with Swiftless's RAW texture loader code and doing the things Jesse7 suggested worked like a charm. It may be worth noteing that in the fragment shader the line:

FragColor = texture2D(texture_color, out_TexCoord.st);

You don't need the .st at the end of our varying vec2 out_TexCoord, or at least it works without it. What is the purpose of the .st at the end of a vec2 varialbe?
Anyways I hate the RAW format so I guess it's time to write my own Targa loader.
Thank you all for your help!

[Jesse7]

What is the purpose of the .st at the end of a vec2 varialbe?
Anyways I hate the RAW format so I guess it's time to write my own Targa loader.

I guess .st is redundant since we have a vec2. For loading Targas and a slew of other formats, you could use FreeImage or DevIL. SOIL is another one but it doesn't seem to work with a core profile.

[NimrodPSI]

Well I had been using a pretty good Targa loader for a while and had even adapted the code to load my own height maps. Although I now have figured out why my texture loader wasn't working In the first place. I tried replaceing the filtering on the RAW texture loader to use mipmaps and used gluBuild2DMipmaps(); but as soon as I ran it resulted in the same black cube. So my old texture loader that also used mipmaps was wrong due to this. I've looked around a bit about GLSL mipmaping and found the command texture2dLod(); So I'm going to persue that route at some point.

[Jesse7]

I tried replaceing the filtering on the RAW texture loader to use mipmaps and used gluBuild2DMipmaps(); but as soon as I ran it resulted in the same black cube. So my old texture loader that also used mipmaps was wrong due to this. I've looked around a bit about GLSL mipmaping and found the command texture2dLod(); So I'm going to persue that route at some point.

Yeah, if you use deprecated code with a core profile you sometimes get weird stuff like black textures . I don't much about that GLSL command, but another way of generating mipmaps is by calling glGenerateMipmap () right after you load the texture:

// use trilinear filtering
glBindTexture(GL_TEXTURE_2D, textureId);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR_MIPMAP_LINEAR);
glTexParameteri( GL_TEXTURE_2D,GL_TEXTURE_MAG_FILTER, GL_LINEAR );
glTexImage2D(GL_TEXTURE_2D,0,GL_RGBA, w, h, 0, GL_RGBA,GL_UNSIGNED_BYTE,pixels);

// generate mipmaps
glGenerateMipmap(GL_TEXTURE_2D);

[NimrodPSI]

Confirmed and working! Thank you, that was much easier than all the hassle of generating my own and passing them to GLSL or something. Now I just need to brush up on some matrix maths and I'm good to go lol.