opengl - Make Camera Look at point using a world transformation matrix? -

variants of question might have been asked on site, none of answers found worked in case.

i trying make camera @ point. camera has world transformation matrix, used calculate front vector, eye position when making view matrix finally. camera treated other object in scene , has world transform. want able modify world transform, , should automatically affect camera. want modify world transform of camera "lookat" point.

for this, modify rotation term of world transform (r in t * s * r) this:

void worldtransform::setlookat( float x, float y, float z ) {      vec3 lookpoint = vec3 ( x, y, z );      vec3 lookdir = normalize( lookpoint - m_position  );      float looklengthonxz = sqrtf( lookdir.z*lookdir.z + lookdir.x*lookdir.x );     m_rotationx = degrees(atan2f( lookdir.y, looklengthonxz ));     m_rotationy = degrees(atan2f( lookdir.x, lookdir.z ));      rotateabs( m_rotationx, m_rotationy, m_rotationz );      updatematlocal( ); } 

basically, have function can set rotation (rotateabs()), calculate pitch, yaw , roll angles , pass them there.

while calculating view matrix, do:

void camera::update( ) {     if (m_ptransform.m_pointer == nullptr) return;      vec3& position = m_ptransform->getposition( );     m_look         = m_ptransform->getforward( );     m_lookat       = position + m_look;      //calculate new matrix     m_view = glm::lookat( position, m_lookat, glm::vec3( 0.0f, 1.f, 0.0f ));   } 

this sort of works, camera starts deviating should looking @ after while. doing wrong?

create view "look at" matrix camera @ point "eye", looking "at", i.e. mine looks this:

matrix4x4<t> & lookat(vec3<t> const & eye, vec3<t> const & at, vec3<t> const & up) {     auto d = vec3<t>();     auto u = vec3<t>();     auto r = vec3<t>();      d = unit(at - eye);                          u = unit(up);                            r = unit(cross(d, u));           u = cross(r, d);      v[0] = r[0];    v[1] = u[0];    v[2] = -d[0];   v[3] = 0;     v[4] = r[1];    v[5] = u[1];    v[6] = -d[1];   v[7] = 0;     v[8] = r[2];    v[9] = u[2];    v[10] = -d[2];  v[11] = 0;     v[12] = 0;      v[13] = 0;      v[14] = 0;      v[15] = 1;      return operator *= (math::translate<t>(-eye[0], -eye[1], -eye[2])); } 

. won't need world matrix move eye. or did misunderstand question?