Our UI is written from scratch so TC basically sees the whole app as one big object. It cannot differentiate the buttons and controls in the GUI. But TC does record by coordinates too so it's fine.
The big problem is, our UI has pop-up windows. When a pop up shows, the original screen in the back becomes inactive. Now, since TC cannot differentiate the original screen from the pop-up screen (again, all it sees is one big object), when I click on the pop-up, TC thinks I'm clicking on the original screen, which is inactive. Then it throws an error.
For example, this is what the keyword test records:
-- clicks at point (x1, y1) of object "dlg" //here I clicked on the button in UI to launch a pop-up window
-- clicks at point (x2, y2) of object "dlg" //here I clicked on the button on the pop up. The UI in the back is inactive now.
And the script would not execute the last step. The error is:
The window is inactive. The action cannot be executed.
If I can record simply the coordinates without the object, I will be able to move forward and utilize other features like image comparison...etc. I also prefer not to use image-based tools that utilizes VNC.