RFR: 8320965: Scrolling on a touch enabled display fails on Wayland [v3]

[Jose Pereda]

On Mon, 19 Feb 2024 00:35:48 GMT, Thiago Milczarek Sayao <tsayao at openjdk.org> wrote:

>> Jose Pereda has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Add compile-time checks to GdkSeat
>
> A shot in the dark since I don't own a touch enabled monitor:
> 
> Test 1:
> 
>  Add `GDK_SCROLL_MASK` on the original `gdk_pointer_grab` function;
> 
> Test 2:
> 
> This PR uses `GDK_SEAT_CAPABILITY_ALL_POINTING` which includes the touch masks.
> 
> So the equivalent would include `GDK_TOUCH_MASK` on `gdk_pointer_grab`.
> 
> 
> I would bet on option 2.

Thanks @tsayao, I can definitely test those suggestions.

If I get it right, you are proposing this change:


status = gdk_pointer_grab(gdkWindow, owner_events, (GdkEventMask)
                                            (GDK_POINTER_MOTION_MASK
                                                | GDK_POINTER_MOTION_HINT_MASK
+                                                | GDK_TOUCH_MASK // or GDK_SCROLL_MASK
                                                | GDK_BUTTON_MOTION_MASK
                                                | GDK_BUTTON1_MOTION_MASK
                                                | GDK_BUTTON2_MOTION_MASK
                                                | GDK_BUTTON3_MOTION_MASK
                                                | GDK_BUTTON_PRESS_MASK
                                                | GDK_BUTTON_RELEASE_MASK
                                            NULL, cursor, GDK_CURRENT_TIME);


But I don't see how this would fix the issue of the failing test on Linux that happens also without a touch enabled display. I tried both, the test still fails (two out of 5 times or so). Also, this is called only if there is no gdk_seat_grab function (which happens before GTK 3.20).

However, if I use `GDK_SEAT_CAPABILITY_NONE` instead of `GDK_SEAT_CAPABILITY_ALL_POINTING`, it passes 10 out of 10, but that would cancel pointer/touch events, wouldn't it?

Is there any chance that the robot implementation (via XTest, that uses `XTestGrabControl`) might not work fine with the new seat_grab approach after GTK 3.20+?

-------------

PR Comment: https://git.openjdk.org/jfx/pull/1305#issuecomment-1952237235