Bug 208566 – REGRESSION(r255957): Element with scroll-behavior:smooth isn't draggable after r255957

• Component: Scrolling
• Status: RESOLVED FIXED

[cathiechen]

The autoscroll uses m_autoscrollTimer to perform drag scroll animation, then uses RenderLayer::scrollRectToVisible to perform scroll.
In this scenario RenderLayer::scrollRectToVisible shouldn't consider scroll-behavior and scroll immediately.

[cathiechen]

Created attachment 392397 [details]
Patch

[cathiechen]

Created attachment 392398 [details]
Patch

[cathiechen]

Created attachment 392685 [details]
Patch

[cathiechen]

About the test for this patch, we need to create a drag action.
I tried to use eventSender, but it doesn't work. Maybe we can add a manual test?

[cathiechen]

Created attachment 392686 [details]
Patch

[Frédéric Wang (:fredw)]

Comment on attachment 392686 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=392686&action=review

LGTM, I wonder if we already have tests for this feature and whether we can adapt them for scroll behavior?

> Source/WebCore/rendering/RenderLayer.cpp:2797
> +void RenderLayer::scrollRectToVisible(const LayoutRect& absoluteRect, bool insideFixed, const ScrollRectToVisibleOptions& options, bool isAutoscrollInProgress)

I guess it should be AutoScrollInProgress?

[cathiechen]

(In reply to Frédéric Wang (:fredw) from comment #6)
> Comment on attachment 392686 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=392686&action=review
> 
> LGTM, I wonder if we already have tests for this feature and whether we can
> adapt them for scroll behavior?
> 

Good idea! I'll have a try!

> > Source/WebCore/rendering/RenderLayer.cpp:2797
> > +void RenderLayer::scrollRectToVisible(const LayoutRect& absoluteRect, bool insideFixed, const ScrollRectToVisibleOptions& options, bool isAutoscrollInProgress)
> 
> I guess it should be AutoScrollInProgress?

Done, changed it to autoscrollInProgress.

[cathiechen]

Created attachment 392708 [details]
Patch

[cathiechen]

Created attachment 392710 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 392710 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=392710&action=review

> Source/WebCore/rendering/RenderLayer.cpp:2802
> +    if (autoscrollInProgress)
> +        ASSERT(false);

hmm?

> Source/WebCore/rendering/RenderLayer.cpp:3040
> +    scrollRectToVisible(LayoutRect(currentDocumentPosition, LayoutSize(1, 1)), false, { SelectionRevealMode::Reveal, ScrollAlignment::alignToEdgeIfNeeded, ScrollAlignment::alignToEdgeIfNeeded, ShouldAllowCrossOriginScrolling::Yes }, true);

Boolean trap

[cathiechen]

Created attachment 392738 [details]
drag-smooth-scroll-element

Add test case.

[cathiechen]

Comment on attachment 392710 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=392710&action=review

Hi Simon,

Thanks for the review:)

>> Source/WebCore/rendering/RenderLayer.cpp:2802
>> +        ASSERT(false);
> 
> hmm?

Ah, sorry, this is used to find out the autoscroll tests. It'll remove in the new patch.

>> Source/WebCore/rendering/RenderLayer.cpp:3040
>> +    scrollRectToVisible(LayoutRect(currentDocumentPosition, LayoutSize(1, 1)), false, { SelectionRevealMode::Reveal, ScrollAlignment::alignToEdgeIfNeeded, ScrollAlignment::alignToEdgeIfNeeded, ShouldAllowCrossOriginScrolling::Yes }, true);
> 
> Boolean trap

Thanks, how about enum AutoscrollStatus { NotInProgress, InProgress } ?

[cathiechen]

Created attachment 392747 [details]
Patch

[cathiechen]

Hi Fred and Simon,

The new patch add a test using eventSender to simulate the drag action.
Please take a look, thanks:)

[cathiechen]

Created attachment 393121 [details]
Patch

[Frédéric Wang (:fredw)]

Comment on attachment 393121 [details]
Patch

This LGTM, but I think it would be better to have a test for WK2. Can't you use UIHelper (e.g. dragFromPointToPoint?). See how it is done for other tests in fast/scrolling.

[cathiechen]

(In reply to Frédéric Wang (:fredw) from comment #16)
> Comment on attachment 393121 [details]
> Patch
> 
> This LGTM, but I think it would be better to have a test for WK2. Can't you
> use UIHelper (e.g. dragFromPointToPoint?). See how it is done for other
> tests in fast/scrolling.

Hi Fred,
Thanks for the review!

It seems UIScriptControllerMac::dragFromPointToPoint for WebKitTestRunner is not implemented yet. I found UIScriptControllerMac::activateAtPoint which is related to mouse event and has implemented, is also using eventSender.
I also looked into some examples of autoscroll tests, they are also using eventSender.
So I guess eventSender is the common way to trigger an autoscroll in test.

Bug:42194 is tracing the incomplete implementation for eventSender. So the test would cover WK2 if Bug:42194 fixed.

Sorry, it seems my comments in changeLogs is confusing. I'll reword them.

[cathiechen]

Created attachment 393253 [details]
Patch

[cathiechen]

Created attachment 393378 [details]
Patch

[cathiechen]

Created attachment 393456 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 393456 [details]
Patch

Clearing flags on attachment: 393456

Committed r258382: <https://trac.webkit.org/changeset/258382>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Radar WebKit Bug Importer]

<rdar://problem/60404576>