Bug 148408

Summary:	Some scroll snap overflow tests are flaky
Product:	WebKit	Reporter:	Alexey Proskuryakov <ap>
Component:	Tools / Tests	Assignee:	Nobody <webkit-unassigned>
Status:	NEW ---
Severity:	Normal	CC:	bdakin, bfulgham, simon.fraser, wenson_hsieh
Priority:	P2
Version:	Other
Hardware:	Unspecified
OS:	Unspecified
See Also:	https://bugs.webkit.org/show_bug.cgi?id=148407 https://bugs.webkit.org/show_bug.cgi?id=165196

Description Alexey Proskuryakov 2015-08-24 17:28:26 PDT

Some scroll snap tests have become flaky when asynchronous. This is reproducible locally when repeating them many times, e.g. "--repeat 500 -f".

tiled-drawing/scrolling/scroll-snap/scroll-snap-mandatory-borders.html
tiled-drawing/scrolling/scroll-snap/scroll-snap-mandatory-overflow.html
tiled-drawing/scrolling/scroll-snap/scroll-snap-mandatory-rotated.html
tiled-drawing/scrolling/scroll-snap/scroll-snap-mandatory-padding.html

Comment 1 Alexey Proskuryakov 2015-08-24 17:35:27 PDT

Also tiled-drawing/scrolling/scroll-snap/scroll-snap-iframe.html

Marked as flaky in r188899.

Comment 2 Alexey Proskuryakov 2015-08-27 10:18:03 PDT

There is a clear opportunity for a race between the end of regular scroll and the beginning of momentum scroll.

Some potential approaches:

- tell eventSender.callAfterScrollingCompletes whether it needs to wait for momentum scroll;

- this could be automagic, because all scrolls are started via eventSender functions, and we do post the momentum scrolling events before calling eventSender.callAfterScrollingCompletes;

- or it may be possible to have an eventSender function that both finishes regular scroll and starts a momentum one, or some other higher level description of a scroll that we'd use with eventSender. That way, we would keep the knowledge that there is a momentum scroll coming for longer.