June 2021 (Vol 2) - New features, stability fixes, Interop fixes and more!

<td><td>WeakPtrs to the same object were dereferenced on different
threads.</td></td>

<td><td>Waiting for input to be processed on main</td></td>

<td><td>Waiting for a main thread hit test on the compositor</td></td>

<td><td>WidgetInputHandlerManager is now ref counted.</td></td>

<td><td>Safe to pass around the “this” pointer for callbacks that will run
on the compositor.</td></td>

<td><td>Use AsWeakPtr only for callbacks that will be run on the main
thread.</td></td>

<td><td>Fixes 5 crashing tests!</td></td>

<td><td>No longer need time-range for a scroll timeline; Times are converted
to percentages. (<a
href="https://github.com/w3c/csswg-drafts/pull/6410">patch</a>)</td></td>

<td><td>Support duration = ‘auto’ (<a
href="https://github.com/w3c/csswg-drafts/pull/6337/">patch</a>)</td></td>

<td><td>Normalization procedure to convert time-based animations to
progress-based.</td></td>

<td><td>Animations are scaled to 100% with iteration and delays adjusted
accordingly.</td></td>

<td><td>Smooth integration with web-animations API</td></td>

<td><td>mustaq@ <a
href="https://github.com/w3ctag/design-reviews/issues/655">kicked off</a> a
TAG review for the <a
href="https://wicg.github.io/capability-delegation/">explainer</a> and the
<a href="https://wicg.github.io/capability-delegation/spec.html">draft
spec</a>.</td></td>

<td><td>Also investigated possible shipping options and finalized <a
href="https://docs.google.com/document/d/1L66B1QtqHCzAKlLQXdtv-YCmlXrJhi2Je2Vo91XWMsQ/edit?usp=sharing">a
plan</a> working with smcgruer@ (Web Payments) and jyasskin@ (Web
Standards).</td></td>

Scroll Unification: fix autoscroll crash

kevers@ fixed an autoscroll crash. There are few problems including:

Here are the fixes:

Scroll Unification: untangling iframe issues

skobes@ landed the CL to handle the case when a scroll hits an iframe that isn’t scrollable for some reason. In this case, we need to look for scrollable container in the parent frame.

Composite BG-color animation

xidachen@ fixed two crash cases.

The first case is shown above where the frame is detached during animation. The fix landed here.

The second case is shown above, and the crash is due to antialiasing that changes the draw operation bounds. This is the fix.

Progress based timelines/Scroll Timelines

kevers@ made a lot of progress on specs.

Highlights:

Here’s an example of un-timed timelines:

Another example of empowering progress based animations

Capability Delegation is ready for TAG review

<td><td>infinity - infinity</td></td>

<td><td>0 / 0</td></td>

<td><td>infinity / infinity</td></td>

<td><td>Division by zero</td></td>

DCHECK in ConvertTimingForCompositor

This function did a bunch of math to check whether the result is finite or not, then either abort or carry on. But what if the intermediate result is indeterminate?

To solve that, we check intermediate results if there is any risk of the following in subsequent calculations:

FullscreenControllerInteractiveTest flaky fix

mustaq@ found an issue that might cause flakiness in a large suite of tests: an initialization problem with EmbeddedTestServer could be causing flakiness in all FullscreenControllerInteractiveTests! A fix has been landed, and we’re hopeful that flakiness will be greatly reduced moving forward.

<td><td>Through devrel outreach (<a
href="https://docs.google.com/document/d/1-pqljDgzzgRVve2oWUq237JPgtT12wcoctxbtBaHZdU/edit#heading=h.1xrzhx8chbuz">doc</a>)</td></td>

<td><td>Push a spec change (<a
href="https://github.com/w3c/csswg-drafts/issues/6406">proposal</a>)</td></td>

Fix interpolation of integers

The test case above demonstrates a bug: the 8 digit CSS number was serialized to 6 significant digits, and the out-of-bounds treatment turned it into a 0..

Below is the fix to the problem.

overscroll-behavior

Our ongoing effort to reduce mis-usage of overscroll-behavior (details here) moved forward with a fix landing on Google Search.

The first graph shows the usage on Google, and the second graph shows the usage of all page load, which is about 0.3%.

There are two ways to move this forward:

Position: sticky 100% pass rate on chrome!

flackr@ wrapped up one of our team’s key product excellence/interop OKRs by fixing all WPT tests for position:sticky. Thanks to all the folks on the DOM team that contributed. (link)