qml porting changes the opengl scene placement

Hello forum,

I had a simple opengl triangle application . Check the screen-shot:

http://imgur.com/ZNO7bcu

When I port the same application to qml following the documentation - "Scene Graph - OpenGL Under QML", I do get the triangle as an underlay, but the placement of the triangle changed as follows:

http://imgur.com/6NVKDnC


The basic OpenGL is exactly same as the original one . Thecode sippet that shapes the triangle is as follows:

Code:

---

//setup triangle geometry
	//setup triangle vertices
	vertices[0].color=glm::vec3(1,0,0);
	vertices[1].color=glm::vec3(0,1,0);
	vertices[2].color=glm::vec3(0,0,1);
 
	vertices[0].position=glm::vec3(-1,-1,0);
	vertices[1].position=glm::vec3(0,1,0);
	vertices[2].position=glm::vec3(1,-1,0);
 
	//setup triangle indices
	indices[0] = 0;
	indices[1] = 1;
	indices[2] = 2;

---

I am getting different placement of the triangle with qml where opengl is underlay.

Any thoughts?


Thanks

No reply so far. Here goes the link to the source which you can download and run at your end.

https://sajis997@bitbucket.org/sajis...anglescene.git

Then you can compare with the screen-shot i sent in my first post and provide me with some hint.


Thanks

Do you remember to initialize and reset the GL environment every frame as mentioned in the docs (QQuickWindow::​resetOpenGLState())?

I believe I am releasing the binding resources after every render call explicitly:

Code:

---

void TriangleScene::render()
{
..................................
.................................
    //draw the triangle
    glDrawElements(GL_TRIANGLES,3,GL_UNSIGNED_SHORT,0);
 
    glDisableVertexAttribArray(mTriangleShader->getAttribute("vVertex"));
    glDisableVertexAttribArray(mTriangleShader->getAttribute("vColor"));
    glBindVertexArray(0);
 
    mTriangleShader->UnUse();
 
}

---

Anyway , i tried as you suggested and it did not change the behavior. I hope that you skimmed through the source that I provided you with. I called the function in the following manner:

Code:

---

void TriangleSceneItem::handleWidowChanged(QQuickWindow *win)
{
    if(win)
    {
        QSurfaceFormat f = win->format();
        f.setMajorVersion(4);
        f.setMinorVersion(3);
        f.setSamples(4);
        f.setStencilBufferSize(8);
        f.setProfile(QSurfaceFormat::CompatibilityProfile);
 
        win->setFormat(f);
 
        win->setClearBeforeRendering(false);
 
        connect(win,SIGNAL(beforeSynchronizing()),this,SLOT(sync()),Qt::DirectConnection);
        connect(win,SIGNAL(sceneGraphInvalidated()),this,SLOT(cleanup()),Qt::DirectConnection);
        connect(win,SIGNAL(beforeRendering()),mScene,SLOT(paint()),Qt::DirectConnection);
 
        win->resetOpenGLState();
    }
}

---

You are not preparing the viewport before rendering thus you are using the one from QtQuick which is probably not what you wnat.

I am setting the viewport in the following function and it is called before the render call. I have checked it with the debugger.

Code:

---

void TriangleScene::paint()
{
    if(!mInitialized)
        initialise();
 
    resize(mWinWidth,mWinHeight);
    render();
}
 
 
void TriangleScene::resize(int w, int h)
{
    if(h < 1)
        h = 1;
 
    mWinWidth = w;
    mWinHeight = h;
 
    glViewport(0,0,(GLsizei)mWinWidth,(GLsizei)mWinHeight);
 
    mProjectionMatrix = glm::ortho(-1,1,-1,1);
}

---

So you are calling resize passing in the values you are setting in the function? Are you sure this is correct?

Quote:

---

Originally Posted by wysota

So you are calling resize passing in the values you are setting in the function? Are you sure this is correct?

---

Which function are you pointing at ? is that paint() ? The mWinWidth , mWinHeight are initialized before with the following snippet:

Code:

---

void TriangleSceneItem::sync()
{
    if(mScene)
    {
        QTimer *timer = new QTimer();
        connect(timer,SIGNAL(timeout()),this,SLOT(updateTime()),Qt::DirectConnection);
        timer->start();
 
        mScene->setViewportSize(window()->width() * window()->devicePixelRatio(),
                                window()->height() * window()->devicePixelRatio());
    }
}

---

You can see that the viewport setting(setViewportSize(....)) is called with every signal :

Code:

---

connect(win,SIGNAL(beforeSynchronizing()),this,SLOT(sync()),Qt::DirectConnection);

---

I am referring to resize().

Your sync() leaks memory.

Quote:

---

Originally Posted by wysota

So you are calling resize passing in the values you are setting in the function? Are you sure this is correct?

---

Could you point out what is wrong as long as the width and height are already set during sync().

Thanks for pointing it out. I thought that it is de-allocated automatically since the class is a QObject subclass. Anyway I made the necessay changes; made the timer member variable of the class and doing the connection there instead of initializing with each and every call of sync(); But it did not change the behavior either.

Where else should I look into ?

Quote:

---

Originally Posted by sajis997

Could you point out what is wrong as long as the width and height are already set during sync().

---

Just look at your code, you should spot it right away. Look at how mWinHeight and mWinWidth are used.

Quote:

---

Thanks for pointing it out. I thought that it is de-allocated automatically since the class is a QObject subclass.

---

It has no parent so no deallocation will take place.

I am not doing much of different that what is described in http://doc.qt.io/qt-5/qtquick-sceneg...l-example.html . Lets compare the sync() there and mine:

Code:

---

void Squircle::sync()
{
    if (!m_renderer) {
        m_renderer = new SquircleRenderer();
        connect(window(), SIGNAL(beforeRendering()), m_renderer, SLOT(paint()), Qt::DirectConnection);
    }
    m_renderer->setViewportSize(window()->size() * window()->devicePixelRatio());
    m_renderer->setT(m_t);
}

---

Whereas I am doing the following:

Code:

---

void TriangleSceneItem::sync()
{
    if(mScene)
    {
        connect(mTimer,SIGNAL(timeout()),this,SLOT(updateTime()),Qt::DirectConnection);
 
        mScene->setViewportSize(window()->width() * window()->devicePixelRatio(),
                                window()->height() * window()->devicePixelRatio());
    }
}

---

I am creating the timer for the animation and then setting the viewport size that is lated used during rendering. The renderer intialization is done inside the constructor.

As I said, look at the code of resize, not sync. By the way, what does initialize() do?

This what TriangleScene::initialize() do :

Code:

---

void TriangleScene::initialise()
{
    //return if already initialized
    if(mInitialized)
        return;
 
    if(!initializeOpenGLFunctions())
    {
        std::cerr << "Modern OpenGL Functions could not be initialized." << std::endl;
        std::exit(EXIT_FAILURE);
    }
 
    GL_CHECK_ERRORS;
 
    //make the projection and model-view matrix
    //identity matrices
    mProjectionMatrix = glm::mat4(1.0f);
    mModelViewMatrix = glm::mat4(1.0f);
 
    loadShaders();
 
    //setup the triangle geometry
    mVertices[0].color = glm::vec3(1.0,0.0,0.0);
    mVertices[1].color = glm::vec3(0.0,1.0,0.0);
    mVertices[2].color = glm::vec3(0.0,0.0,1.0);
 
    mVertices[0].position = glm::vec3(-1,-1,0);
    mVertices[1].position = glm::vec3(0,1,0);
    mVertices[2].position = glm::vec3(1,-1,0);
 
    //setup the triangle indices
    mIndices[0] = 0;
    mIndices[1] = 1;
    mIndices[2] = 2;
 
    //check for opengl errors
    GL_CHECK_ERRORS;
 
    //setup the triangle vao and vbo stuff
    glGenVertexArrays(1,&mVaoID);
 
    glGenBuffers(1,&mVerticesID);
    glGenBuffers(1,&mIndicesID);
 
    //get the size of the Vertex struct
    GLsizei stride = sizeof(Vertex);
 
    glBindVertexArray(mVaoID);
 
    glBindBuffer(GL_ARRAY_BUFFER,mVerticesID);
    //pass triangle vertices to the buffer object
    glBufferData(GL_ARRAY_BUFFER,  //specifies the target buffer object
                 sizeof(mVertices),//specifies the size in bytes
                 &mVertices[0], // pointer to the data
                 GL_STATIC_DRAW); //usage pattern of the data
 
    GL_CHECK_ERRORS;
 
    glEnableVertexAttribArray(mTriangleShader->getAttribute("vVertex"));
    glVertexAttribPointer(mTriangleShader->getAttribute("vVertex"),3,GL_FLOAT,GL_FALSE,stride,0);
 
    GL_CHECK_ERRORS;
 
    //enable vertex attribute array for the color
    glEnableVertexAttribArray(mTriangleShader->getAttribute("vColor"));
    glVertexAttribPointer(mTriangleShader->getAttribute("vColor"),3,GL_FLOAT,GL_FALSE,stride,(const GLvoid*)offsetof(Vertex,color));
 
    GL_CHECK_ERRORS;
 
    glBindBuffer(GL_ELEMENT_ARRAY_BUFFER,mIndicesID);
    glBufferData(GL_ELEMENT_ARRAY_BUFFER,sizeof(mIndices),&mIndices[0],GL_STATIC_DRAW);
 
    glDisableVertexAttribArray(mTriangleShader->getAttribute("vVertex"));
    glDisableVertexAttribArray(mTriangleShader->getAttribute("vColor"));
    glBindVertexArray(0);
 
    GL_CHECK_ERRORS;
 
    mInitialized = true;
 
}
 
I think the issue lies here and I am examining it right now. Any hint from your end within the function will accelerate the process.
 
 
Thanks

---