Rubber-banding in the DOM view of the WebInspector is really jumpy

[Beth Dakin]

* SUMMARY
If you try to rubber-band in the DOM view of the WebInspector, things get a bit jumpy.

* STEPS TO REPRODUCE
1. Open a web page, I used http://www.webkit.org.
2. Open the WebInspector and make sure the DOM tree view is visible.
3. Make sure the DOM tree view is scrollable by expanding as many of the nodes as you can (I option-clicked the body).
4. Scroll and rubber-band the view with your mouse cursor over the nodes (you will notice the blue hightlight moving from node to node staying under your cursor).

* RESULTS
As you rubber-band, the view will jump around a bit.

* REGRESSION
This is new with the changes to add rubber-banding in overflow areas.

* NOTES
If you rubber-band over the scrollbar track while it is visible or when the WebInspector is not the frontmost window, you will not see the jumping, but you will notice the blue hightlight also does not change.

rdar://problem/18166043

[Beth Dakin]

Created attachment 237821 [details]
Reduction

[Beth Dakin]

Created attachment 237822 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 237822 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=237822&action=review

> Source/WebCore/rendering/RenderLayer.cpp:1390
> +    if (!renderer().frame().settings().rubberBandingForOverflowScrollEnabled())
> +        return false;

Do you need to check the setting here, since presumably overhangAmount() will just return 0,0 otherwise?

[Beth Dakin]

(In reply to comment #3)
> (From update of attachment 237822 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=237822&action=review
> 
> > Source/WebCore/rendering/RenderLayer.cpp:1390
> > +    if (!renderer().frame().settings().rubberBandingForOverflowScrollEnabled())
> > +        return false;
> 
> Do you need to check the setting here, since presumably overhangAmount() will just return 0,0 otherwise?

True. Probably no need to belt-and-suspenders here.

[Build Bot]

Comment on attachment 237822 [details]
Patch

Attachment 237822 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/6264184811552768

New failing tests:
css3/flexbox/flexbox-overflow-auto.html
css3/flexbox/flexbox-baseline.html
css3/flexbox/child-overflow.html

[Build Bot]

Created attachment 237826 [details]
Archive of layout-test-results from webkit-ews-14 for mac-mountainlion-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: webkit-ews-14  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.5

[Beth Dakin]

Created attachment 237865 [details]
Patch

Here's a new patch that shouldn't break any tests or other ports. RenderLayer should call into ScrollAnimatorMac's implementation of isRubberBandInProgress() that will check to ensure there is a non-zero overhangAmount (like my first patch did) but it will also make sure you are in the middle of a momentum gesture.

[Darin Adler]

Comment on attachment 237865 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=237865&action=review

> Source/WebCore/rendering/RenderLayer.cpp:1399
> +    return false;
> +#else
> +    return false;
> +#endif

Could omit the else and share the "return false". Not sure if it would be better or not.

[Beth Dakin]

http://trac.webkit.org/changeset/173484

[Beth Dakin]

Thanks Darin!

[Simon Fraser (smfr)]

This caused bug 138525.