save camera settings on selection to avoid different result when re-projecting the selection 2d screen points transformation (does not compile because of canvas.h cyclic dependency)

src/canvas_modes.cpp

@@ -493,7 +493,7 @@ void CanvasModeMaskFree::on_MouseEvent(MouseEvent* me, glm::vec2& loc)
         {
             if (!m_points.empty())
             {
-
+                m_selection_cam = canvas->get_camera();
                 //m_points2d = poly_intersect(poly_remove_duplicate(m_points2d), canvas->face_to_shape2D(0));
                 auto drawer = [this](const glm::mat4& camera, const glm::mat4& proj) {
                     //glEnable(GL_BLEND);
@@ -910,6 +910,9 @@ void CanvasModeTransform::enter()
     canvas->m_smask_active = false;
     auto points = m->m_points2d;
 
+    canvas->push_camera();
+    canvas->set_camera(m->m_selection_cam);
+
     glm::vec2 bb_min(FLT_MAX);
     glm::vec2 bb_max(-FLT_MAX);
     for (auto p2d : points)
@@ -1079,6 +1082,8 @@ void CanvasModeTransform::enter()
 
         m_source_image.destroy();
     }
+
+    canvas->pop_camera();
 }
 
 void CanvasModeTransform::leave()