On 2020-11-01 7:56 am, Jared Finder wrote:
> On 2020-11-01 5:39 am, Stefan Monnier wrote:
>> Hi Jared,
>> 
>> I really like this new feature but have just one comment/question?
>> 
>>> +  previous_help_echo_string = help_echo_string;
>>> +  help_echo_string = Qnil;
>>> +
>>> +  note_mouse_highlight(XFRAME(frame), XFIXNUM (mouse_x), XFIXNUM 
>>> (mouse_y));
>>> +
>>> +  if (!NILP (help_echo_string)
>>> +      || !NILP (previous_help_echo_string))
>>> +    {
>>> +      gen_help_event (help_echo_string, frame, help_echo_window,
>>> +                      help_echo_object, help_echo_pos);
>>> +    }
>> 
>> I see this exact same code in other C files.
>> Could we move it to a file where we can share it instead of having
>> N copies?
> 
> I completely agree, not just for this code but also for mouse handling
> in general.  I think there should be a shared mouse interface with
> individual C functions for each type of mouse event: mouse move, mouse
> click, etc.  Translating OS-specific events to this shared
> functionality would continue to be OS-specific, but the actual
> handling of these events, such as this logic, would be fully shared.
> For example, this would unify the different codepaths between TTY
> menus for GPM, xterm-mouse, and NT Emacs.
> 
> I would be happy to help with this next.  However I need some help.  I
> can only locally build and test for Linux terminal with xterm-mouse or
> GPM handling the mouse.  Is there someone who can help for other
> platforms?  And should the GUI platforms be included as well?  (I
> suspect yes is the right answer.)

Toward proving that the code could be shared, I refactored the GPM mouse 
logic so that it was clearly apparent how to share it with 
handle-lisp-mouse-motion.  That patch is attached.  I'd like to make 
sure this looks like an appropriate change to make.  If so, I will 
finish up the patch.

Two specific questions:

1. To enable sharing logic, I need to encode a handful of assumptions 
that I believe are true today (example: no need to handle quit-char in 
GPM handling).  Do these assumptions look reasonable?

2. In what file should such a shared function go?  My initial thought is 
a new file "mouse.c" as it would hold shared mouse logic.

   -- MJF