QBlitFramebuffer's sourceRect and destinationRect were defined as QRectF in the API.

Under the hood, GraphicsContext blindly passes left(), top(), left()+width() and top()+height() to glBlitFramebuffer. This is pretty bad since glBlitFramebuffer uses the GL coordinate system (Y == bottom left corner). QRect(F) is badly suited for representing such rectangles. So once a non-fullscreen blit is needed, applications will likely attempt to convert their QRectF to a GL-friendly rect using setY, but that will internally alter the QRectF's height.

Rather, QBlitFramebuffer should state that the rectangles are in Qt coordinates (Y == top left). This would not just allow sane rectangle manipulation on the client side, but would also be more future-proof (given that no other graphics APIs use the Y==bottom left convention)

Code like the following:

                    QRectF srcRect(data->layerPos, data->layerSize);
                    // blitframebuffer takes GL coords (Y==bottom left), convert since source is a partial area only
                    srcRect.setY(m_outputPixelSize.height() - (srcRect.y() + srcRect.height()));
                    srcRect.setHeight(data->layerSize.height()); // ugh, fix up the height since setY altered it; QRectF is not really suitable for GL rects
                    bgBlit->setSourceRect(srcRect);

                    QRectF dstRect(QPointF(0, 0), srcRect.size());
                    bgBlit->setDestinationRect(dstRect);

could then be simplified to a more sane:

                    QRectF srcRect(data->layerPos, data->layerSize);
                    bgBlit->setSourceRect(srcRect);

                    QRectF dstRect(QPointF(0, 0), srcRect.size());
                    bgBlit->setDestinationRect(dstRect);