Tim Horton

> Do you expect zero impact on benchmarks, no chance of 1% regression somewhere? I am going to kick off some benchmarks; first I wanted to see what the tests...

This is a 3-5% motionmark regression, so I will find another way :(

> Let's extend this quirk to macOS? :) Definitely not going to regress correctness on macOS for an optimization that should never have landed :) I am going to fix...

I think that you should remove the cursor-determining code for all cases covered here from EventHandler in this patch, so that tests will reveal if you've done it right (right...

Also, your changelog formatting is a bit off. There should be a bugzilla link below the title, possibly a radar below that, then a "Reviewed by NOBODY (OOPS!)." and then...

Regardless of the answers above, please add a test to the interaction-region subdirectory of layout tests

This is just a rough first cut. I also generally followed existing patterns, even if they seemed questionable (especially re: naming oddities around "Process" vs. "Service" and similar).

> I have a few high-level questions: > > 1. When you say "loading/parsing s", is that intended to also include loading subresources? Or will that be something that is...

I stripped area codes for consistency, but I suppose taking the longest number we ever see for a contact isn't unreasonable.

Right, right. Right now we get numbers as they stand, and sometimes they're 10 digit, and sometimes they're 7 digit, and sometimes they're prefixed by a +, so we just...