Add action_sequence in testdriver

[chromium-wpt-export-bot]

After we expose test_driver.Actions API to web users, we add their
implementation in our testdriver file, and fix the wpt tests of
Actions API.

Bug: 893480
Change-Id: Ib02c0223208eeb2cc30c2ca35b98d5fc938baa2c
Reviewed-on: https://chromium-review.googlesource.com/c/1289119
Commit-Queue: Lan Wei <lanwei@chromium.org>
Reviewed-by: Navid Zolghadr <nzolghadr@chromium.org>
Reviewed-by: Dave Tapuska <dtapuska@chromium.org>
Cr-Commit-Position: refs/heads/master@{#606882}

[wpt-pr-bot]

Already reviewed downstream.

[foolip]

@Summerlw @NavidZ, this change has gotten stuck and not exported properly because of failing infrastructure/ tests. These aren't run in Chromium in precisely the same way as here, so that's unfortunately a possibility.

Are either of you able to take a look at the failures here and suggest a fix?

[foolip]

I've pushed a commit to update the expectations. That might allow this PR to land, but since the tests are still failing I presume there's more work to be done here anyway :)

[NavidZ]

I believe this has failed due to the pointerdown now getting the coordinates. w3c/webdriver#1348 is addressing the issue. I'm working on sending a pull request to webdriver spec as well.
@jgraham FYI.

[jgraham]

Yeah, this should never have landed. Having to update the metadata for infrastructure tests to expected: FAIL everywhere should be a pretty clear indication that there's something broken in the changeset (in this case that they're assuming an API that just isn't like the one that currently exists; that difference is indeed a missing piece of flexibility in the current API, but it's still not OK to just change tests to assume something totally nonstandard, even if the standard is lacking).

[foolip]

Sorry, I moved too fast here. The fact that the infrastructure tests were previously passing in Firefox wasn't apparent to me, I just treated it as "broken before, still broken" and needing some follow-up work.

What is the non-standard part here, though? I don't see any changes to the WebDriver layer, where is it hiding?

[foolip]

@lukebjerring yet another case where regression detection would have saved the day :)

[NavidZ]

Yeah. Now the changes are more hidden. Any changes to the serialized json coming out of send function of testdriver could break this. Basically we added a coordinate pair to the pointerdown action as well which we thought we needed to address some of the missing use cases. Then this get encoded in that json later.
As we discussed in w3c/webdriver#1348 we are going to instead add another flag to each command to suppress sending the command (and only let the driver update the state of input including coordinate and button state).
For now we leave those tests that need something like this as manual and essentially ignore those test and only migrate the tests that can work with the existing functionalities.

[foolip]

OK, so some of this code changed the JSON payload for the WebDriver commands. Then I understand how it could be non-standard.