Set width=height=1 of mouse like pointer events

third_party/WebKit/Source/core/events/PointerEventFactory.cpp

 // Copyright 2016 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "core/events/PointerEventFactory.h"
 
 #include "platform/geometry/FloatSize.h"
 
 namespace blink {
 
@@ skipping 19 matching lines @@
 
 const AtomicString& pointerEventNameForMouseEventName(
     const AtomicString& mouseEventName)
 {
 #define RETURN_CORRESPONDING_PE_NAME(eventSuffix) \
     if (mouseEventName == EventTypeNames::mouse##eventSuffix) {\
         return EventTypeNames::pointer##eventSuffix;\
     }
 
     RETURN_CORRESPONDING_PE_NAME(down);
-    RETURN_CORRESPONDING_PE_NAME(enter);

[mustaq, 2016/05/11 20:42:57]

Let's leave them as is. There's no harm in keeping this function complete, and
we don't want to close the way to easily create "pointerleave" from "mouseleave"
(e.g. for stylus-off-the-digitizer case).

[Navid Zolghadr, 2016/05/11 20:50:20]

Sure. I will add a DCHECK separately in the create function to only allow
move/up/down. Is that okay?

[Navid Zolghadr, 2016/05/12 20:13:21]

Done.

-    RETURN_CORRESPONDING_PE_NAME(leave);
     RETURN_CORRESPONDING_PE_NAME(move);
-    RETURN_CORRESPONDING_PE_NAME(out);
-    RETURN_CORRESPONDING_PE_NAME(over);
     RETURN_CORRESPONDING_PE_NAME(up);
 
 #undef RETURN_CORRESPONDING_PE_NAME
 
     ASSERT_NOT_REACHED();
     return emptyAtom;
 }
 
 } // namespace
 
@@ skipping 44 matching lines @@
     setBubblesAndCancelable(pointerEventInit, pointerEventName);
 
     pointerEventInit.setScreenX(mouseEvent.globalPosition().x());
     pointerEventInit.setScreenY(mouseEvent.globalPosition().y());
     pointerEventInit.setClientX(mouseEvent.position().x());
     pointerEventInit.setClientY(mouseEvent.position().y());
 
     if (pointerEventName == EventTypeNames::pointerdown
         || pointerEventName == EventTypeNames::pointerup) {
         pointerEventInit.setButton(mouseEvent.button());
-    } else {
-        // TODO(crbug.com/587955): We are setting NoButton for transition
-        // pointerevents should be resolved as part of this bug
+    } else { // Only when pointerEventName == EventTypeNames::pointermove

[mustaq, 2016/05/11 20:42:57]

The comment is not 100% correct: we are actually hacking
mosueEventType="mousedown" to ultimately create transition events, right?

Let's remove the comment.

[Navid Zolghadr, 2016/05/11 20:50:20]

No. The boundary events use another create function which clone the pointerevent
properties and only change the type and target.

[mustaq, 2016/05/12 14:03:09]

The create() call in sendMouseAndPossiblyPointerNodeTransitionEvents looks
confusing then. Looks like there is no easy way to avoid the create() call
there, right? If I am right, please rename the created event there to "dummy..."
at least.

A related thing: please update the doc for processPendingPointerCapture() to
somehow say that it doesn't care about the button state. It looks confusing from
spec perspective that the dummy events created here with button=none may not
release capture.

[Navid Zolghadr, 2016/05/12 14:36:43]

I'm not quite sure about the source of confusion here. Can you explain a little
bit more?

Sure. I will rename the variable to "dummyPointerEvent".

Regarding processPendingPointerCapture, it cares about neither button nor
buttons. The processPendingPointerCapture is defined in the spec here

https://w3c.github.io/pointerevents/#process-pending-pointer-capture

and this method was trying to implement exactly that. Whether a release will
happen or not depending on the explicit release call or buttons value that is
something else and when it happens it only sets the "pending pointer capture
target" variable. On the other hand, when processPendingPointerCapture is called
it is processing the assignment of "pointer capture target" from "pending
pointer capture target" and sends the got/lostpointercapture. Is this okay?
Maybe I can put a link to the spec section as comment before that function. Do
you want me to do that or will it be too tight to the spec?

[Navid Zolghadr, 2016/05/12 20:13:21]

Done.

         pointerEventInit.setButton(NoButton);
     }
     pointerEventInit.setPressure(getPointerEventPressure(
         mouseEvent.pointerProperties().force, pointerEventInit.buttons()));
 
+    // Set width/height to 1 because we don't get any information from lower

[mustaq, 2016/05/11 20:42:57]

"...because it matches Edge (and supported by the spec)."

[Navid Zolghadr, 2016/05/11 20:50:20]

Sure.

[Navid Zolghadr, 2016/05/12 20:13:21]

Done.

+    // lever APIs.
+    pointerEventInit.setWidth(1);
+    pointerEventInit.setHeight(1);
+
     UIEventWithKeyState::setFromPlatformModifiers(pointerEventInit, mouseEvent.getModifiers());
 
     // Make sure chorded buttons fire pointermove instead of pointerup/down.
     if ((pointerEventName == EventTypeNames::pointerdown
         && (buttons & ~MouseEvent::buttonToButtons(mouseEvent.button())) != 0)
         || (pointerEventName == EventTypeNames::pointerup && buttons != 0))
         pointerEventName = EventTypeNames::pointermove;
 
 
     pointerEventInit.setView(view);
@@ skipping 236 matching lines @@
     if (properties.pointerType
         == WebPointerProperties::PointerType::Mouse)
         return PointerEventFactory::s_mouseId;
     IncomingId id(properties.pointerType, properties.id);
     if (m_pointerIncomingIdMapping.contains(id))
         return m_pointerIncomingIdMapping.get(id);
     return PointerEventFactory::s_invalidId;
 }
 
 } // namespace blink