Re: [AD] timer_mutex again

[Eric Botcazou <ebotcazou@xxxxxxxxxx>]

> Thanks, I'll add that.  But this is not something that could happen with
> the existing code, is it?  If so, I'm stumped seeing how it could arise.

What existing code?  The timer code _with_ the mutex guarantees that this 
cannot happen.  The timer code _without_ the mutex doesn't.

For example for show_mouse(NULL):

void show_mouse(BITMAP *bmp)
{
   if (!mouse_driver)
      return;

   remove_int(mouse_move);

   if (_mouse_screen) {
      acquire_bitmap(_mouse_screen);

      if (gfx_capabilities & GFX_HW_CURSOR) {
	 gfx_driver->hide_mouse();
	 gfx_capabilities &= ~GFX_HW_CURSOR;
      }
      else
	 draw_mouse(TRUE, FALSE);

      release_bitmap(_mouse_screen);
   }

   _mouse_screen = bmp;


mouse_move draws on _mouse_screen so it could end up dereferencing the NULL 
pointer, were remove_int not guaranteed to return only after mouse_move has 
itself returned.  IIRC that's the very problem that prompted me to devise this 
stuff.

-- 
Eric Botcazou