Tablet pen map_to_output not applied correctly at startup

[Xyene]

I'm running sway master, and have two displays, arranged in the below configuration:

(One is 2560x1440, the other 1600x900.)

My sway config contains directives for mapping my pen to DP-1.

input 9580:110:HUION_420_Pen {
  map_to_output DP-1
  map_from_region 0.22x0.22 0.78x0.78
  left_handed enabled
}

However, on startup, my pen is not confined, and actually appears at the bottom-right corner of HDMI-A-2. Moving it over to DP-1 myself doesn't make it confined. Triggering a reload after startup via swaymsg reload correctly confines the pen, though.

Edit: this appears to me to be a race of sorts. My logs contain:

00:00:00.439 [backend/libinput/events.c:94] Added HDA ATI HDMI HDMI/DP,pcm=3 [0:0]                                      
00:00:00.439 [backend/libinput/events.c:94] Added HDA ATI HDMI HDMI/DP,pcm=7 [0:0]                                      
00:00:00.439 [backend/libinput/events.c:94] Added HDA ATI HDMI HDMI/DP,pcm=8 [0:0]                                      
00:00:00.439 [backend/libinput/events.c:94] Added HDA ATI HDMI HDMI/DP,pcm=9 [0:0]                                      
00:00:00.439 [backend/libinput/events.c:94] Added HDA ATI HDMI HDMI/DP,pcm=10 [0:0]                                     
00:00:00.439 [backend/libinput/events.c:94] Added HDA ATI HDMI HDMI/DP,pcm=11 [0:0]                                     
00:00:00.439 [backend/libinput/events.c:94] Added HUION 420 Pen [9580:110]                                              
00:00:00.439 [sway/input/input-manager.c:231] adding device: '9580:110:HUION_420_Pen'                                   
00:00:00.439 [sway/input/libinput.c:191] config_libinput_pointer('9580:110:HUION_420_Pen' on  '9580:110:HUION_420_Pen') 
00:00:00.439 [sway/input/libinput.c:106] left_handed_set(1)                                                             
00:00:00.439 [sway/input/seat.c:824] adding device 9580:110:HUION_420_Pen to seat seat0                                 
00:00:00.439 [sway/input/seat.c:630] Applying input config to 9580:110:HUION_420_Pen                                    
00:00:00.439 [sway/input/seat.c:644] Mapping input device 9580:110:HUION_420_Pen to output DP-1                         
00:00:00.439 [backend/libinput/events.c:94] Added HUION 420 Pad [9580:110]                                              
...
00:00:00.456 [backend/libinput/backend.c:101] libinput successfully initialized                                         
00:00:00.456 [backend/drm/drm.c:1320] Scanning DRM connectors                                                           
00:00:00.457 [backend/drm/drm.c:1383] Found connector 'DP-1'                                                            
00:00:00.457 [backend/drm/drm.c:1419] 'DP-1' connected                                                                  
00:00:00.457 [backend/drm/drm.c:1421] Current CRTC: 47                                                                  
00:00:00.457 [backend/drm/drm.c:1426] Physical size: 600x340                                                            
00:00:00.457 [backend/drm/drm.c:1443] Detected modes:                                                                   
00:00:00.457 [backend/drm/drm.c:1468]   2560x1440@59951 (preferred)                                                     
00:00:00.457 [backend/drm/drm.c:1468]   2560x1440@61029

So, it would seem that if the display isn't initialized before the output map is set, it won't take effect. Seemingly supporting this hypothesis, unplugging and replugging the tablet (rather than swaymsg reload) also works.

The code responsible is:

sway/sway/input/seat.c

Lines 653 to 658 in 58eb0c4

	struct sway_output *output = output_by_name_or_id(mapped_to_output);
	if (!output) {
	return;
	}
	wlr_cursor_map_input_to_output(seat->cursor->cursor,
	sway_device->input_device->wlr_device, output->wlr_output);

(output is NULL at that point in time, and the map is never set.)