Following the same argument for removing policy from modify_window() that we applied in revision 347 I think it is modify_window() that shouldn't be interpreting state changes to imply resizing.
Moving that logic to before the call to handle_modify_window() in order to provide suggestions for position and size in the modifications argument will allow shells to customise the result. (We do that for surface creation already.)
Following the same argument for removing policy from modify_window() that we applied in revision 347 I think it is modify_window() that shouldn't be interpreting state changes to imply resizing.
Moving that logic to before the call to handle_ modify_ window( ) in order to provide suggestions for position and size in the modifications argument will allow shells to customise the result. (We do that for surface creation already.)