I'm not sure. It can be a "false" scrolling when you accidentally touched a Magic Mouse. I think we should use threshold on the phase end, to ignore accumulatedDelta less than 0.1
On Fri, Sep 30, 2016 at 6:18 PM, Sergey Bylokhov <sergey.bylok...@oracle.com> wrote: > What will be the difference if we will scroll by one line at the "begin > phase"? probably it will be better if the the scroll will start immediately > on first touch? > > > On 30.09.16 17:59, Alexander Scherbatiy wrote: >> >> >> Hello, >> >> Could you review the updated fix: >> http://cr.openjdk.java.net/~alexsch/8166591/webrev.06 >> >> - The CPlatformResponder.handleScrollEvent(...) is updated to dispatch >> a scroll event when delta or round delta is not equal to zero >> - The native scrollStateWithPhase: method is updated to have NSEvent as >> an argument >> >> Thanks, >> Alexandr. >> >> On 30/09/16 16:58, Sergey Malenkov wrote: >>> >>> In the CPlatformResponder: >>> >>> phase3 0 ~ 0.0 // mayBegan >>> phase2 0 ~ 0.0 // began >>> phase3 0 ~ 0.0222015380859375 >>> phase3 0 ~ 0.0234222412109375 >>> phase3 0 ~ 0.023956298828125 >>> phase3 0 ~ 0.0242919921875 >>> phase3 0 ~ 0.02447509765625 >>> phase3 0 ~ 0.0246124267578125 >>> phase3 0 ~ 0.024658203125 >>> phase3 0 ~ 0.0222015380859375 >>> phase3 0 ~ 0.0233306884765625 >>> phase5 1 ~ 0.0 // end >>> >>> In Java: >>> >>> wheelRotation=0,preciseWheelRotation=-0.0222015380859375 >>> wheelRotation=0,preciseWheelRotation=-0.0234222412109375 >>> wheelRotation=0,preciseWheelRotation=-0.023956298828125 >>> wheelRotation=0,preciseWheelRotation=-0.0242919921875 >>> wheelRotation=0,preciseWheelRotation=-0.02447509765625 >>> wheelRotation=0,preciseWheelRotation=-0.0246124267578125 >>> wheelRotation=0,preciseWheelRotation=-0.024658203125 >>> wheelRotation=0,preciseWheelRotation=-0.0222015380859375 >>> wheelRotation=0,preciseWheelRotation=-0.0233306884765625 >>> >>> We ignored first two events, because of 0 & 0.0 >>> We should not ignore last event, where 1 & 0.0 >>> Seems we need to fix DeltaAccumulator. >>> >>> >>> On Fri, Sep 30, 2016 at 2:25 PM, Alexander Scherbatiy >>> <alexandr.scherba...@oracle.com> wrote: >>>> >>>> Hello, >>>> >>>> Could you review the updated fix: >>>> http://cr.openjdk.java.net/~alexsch/8166591/webrev.05 >>>> >>>> The momentumPhase is used to detect the trackpad events. >>>> >>>> Thanks, >>>> Alexandr. >>>> >>>> >>>> On 30/09/16 14:58, Sergey Malenkov wrote: >>>>>> >>>>>> # C [AppKit+0x3a528e] -[NSApplication _crashOnException:]+0x6d >>>>>> >>>>>> The app is crashed as soon as I start scrolling. Investigating... >>>>>> May be it is my fault during backporting. >>>>> >>>>> Sorry, It was may fault. >>>>> >>>>> >>>>>> LWCToolkit.m: >>>>>> +// SCROLL EVENT MASK >>>>>> +#define SCROLL_PHASE_UNSUPPORTED 1 >>>>>> ... >>>>>> >>>>>> replace the comment with the following one: >>>>>> >>>>>> +// TRACKPAD SCROLL EVENT PHASE >>>>> >>>>> I discovered how we should detect mouse event properly: use both >>>>> properties phase and momentumPhase. >>>>> >>>>> >>>>> https://developer.apple.com/library/prerelease/content/documentation/Cocoa/Conceptual/EventOverview/HandlingTouchEvents/HandlingTouchEvents.html >>>>> >>>>> "The momentumPhase property helps you detect momentum scrolling, in >>>>> which the hardware continues to issue scroll wheel events even though >>>>> the user is no longer physically scrolling." >>>>> >>>>> if (phase==NULL) && (momentumPhase==NULL) -> this is a mouse event. >>>>> >>>>> scrolling by trackpad generates the following events: >>>>> >>>>> phase=mayBegan momentumPhase=null >>>>> phase=began momentumPhase=null >>>>> phase=continued momentumPhase=null >>>>> ... >>>>> phase=continued momentumPhase=null >>>>> phase=ended momentumPhase=null >>>>> phase=null momentumPhase=began >>>>> phase=null momentumPhase=continued >>>>> ... >>>>> phase=null momentumPhase=continued >>>>> phase=null momentumPhase=ended >>>>> >>>>> So, we should generate PHASE_UNSUPPORTED only >>>>> if ((phase == NULL) && (momentumPhase == NULL)) >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>>>> >>> >>> >> > > > -- > Best regards, Sergey. -- Best regards, Sergey A. Malenkov
