DND does not finish on desktop

[Hiroaki Yamamoto]

DND doesn't finish properly when the cursor button is released on desktop.

dnd-on-desktop.mp4

I confirmed removing the check for ctx.surface in cursor_button_release() (introduced in a252308) fixed the problem, though I don't know the purpose of this check.

diff --git a/src/input/cursor.c b/src/input/cursor.c
index eb359e7..4a9f6b9 100644
--- a/src/input/cursor.c
+++ b/src/input/cursor.c
@@ -1067,7 +1067,7 @@ cursor_button_release(struct seat *seat, uint32_t button,
        bool consumed_by_frame_context =
                handle_release_mousebinding(server, &ctx, button);
 
-       if (ctx.surface && !consumed_by_frame_context) {
+       if (!consumed_by_frame_context) {
                /* Notify client with pointer focus of button release */
                wlr_seat_pointer_notify_button(seat->seat, time_msec,
                        button, button_state);

[Consolatis]

Agree, that check doesn't appear to make sense for the cursor button release handler.

The commit linked is part of the original move to scene graph.
I assume what happened was that that condition is basically a copy from cursor_button_press() where the check likely makes sense.

CC @johanmalm just to ensure we don't miss any context here.

[Jens Peters]

Nice find. The same issue can also be seen in firefox when dragging a tab out of firefox. Would be cool to have this addressed.

[Johan Malm]

Agree, that check doesn't appear to make sense for the cursor button release handler.

The commit linked is part of the original move to scene graph. I assume what happened was that that condition is basically a copy from cursor_button_press() where the check likely makes sense.

CC @johanmalm just to ensure we don't miss any context here.

I'm fine with this patch. Can't think of why it's there either. Like you say - quite possibly cruft from pre-scene-graph.

I've tested with a few scenarios with scroll bars and selected text - but all seems fine.