Rewrite /wpt/cookies/http-state/attribute.html tests

[chromium-wpt-export-bot]

This is the first of a handful of CLs rewriting the legacy
tests in the ported wpt http-state tests. The plan is to continue
to refine and refactor and eventually rewrite (or delete) them all.

Bug: 1145300
Change-Id: I8a50939f4e2c95c1293ba5423577693ab2a10d9a
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2550362
Commit-Queue: Mike Taylor <miketaylr@chromium.org>
Auto-Submit: Mike Taylor <miketaylr@chromium.org>
Reviewed-by: Lily Chen <chlily@chromium.org>
Cr-Commit-Position: refs/heads/master@{#830803}

[wpt-pr-bot]

The review process for this patch is being conducted in the Chromium project.

[miketaylr]

Relates to #26190

[miketaylr]

@jgraham could you please re-trigger/re-run the wpt-firefox-nightly-stability check on TC?

Oh, false alarm. It's re-running now (I guess cause CQ is trying the final patchset).

Client ID github/67283|miketaylr does not have sufficient scopes

[miketaylr]

OK, I guess it's still flaky. Will investigate.

[miketaylr]

OK, I can reproduce locally with the following:

git fetch origin && git checkout -b chromium-export-cl-2550362 origin/chromium-export-cl-2550362
./wpt install --channel nightly firefox browser
./wpt install --channel nightly firefox webdriver

python3 ./wpt run --channel=nightly --verify --verify-no-chaos-mode --verify-repeat-loop=0 --verify-repeat-restart=10 --log-mach-level=info --log-mach=- -y --no-pause --no-restart-on-unexpected --no-headless --verify-log-full --binary="/Users/miketaylr/dev/wpt/_venv2/browsers/nightly/Firefox Nightly.app/Contents/MacOS/firefox" firefox cookies/attributes

[miketaylr]

To the best of my knowledge, the issue is related to trying to run promise tests that set and get cookies while opening a new window (via window.open()) which runs its own cookie tests that set and get cookies.

In Firefox, this works most of the time, but as the stability run shows, it can be flaky. Locally, I could get it to reproduce once every 3 or 4 refreshes. Either the cookies don't get set (unlikely, probably?), or there's a race and the get happens before the set, so document.cookie returns an empty string.

Pulling the child window tests into its own file resolves the issue for me (and is probably a better idea, in retrospect).

I'm not sure if this is related to what Ehsan is describing here, https://bugzilla.mozilla.org/show_bug.cgi?id=1331680#c2, or just bad test design on my side of things. But it's something related to the architecture of Firefox (or it's tickling a bug), as Chromium seems fine with it either way.

0:42.65 INFO ### /cookies/attributes/secure.https.html ###
 0:42.65 INFO |                Subtest                 | Results | Messages |
 0:42.65 INFO |----------------------------------------|---------|----------|
 0:42.65 INFO |                                        | OK      |          |
 0:42.65 INFO | `Set cookie for Secure attribute`      | PASS    |          |
 0:42.65 INFO | `Set cookie for seCURe attribute`      | PASS    |          |
 0:42.65 INFO | `Set cookie for for Secure= attribute` | PASS    |          |
 0:42.65 INFO | `Set cookie for Secure=aaaa`           | PASS    |          |
 0:42.65 INFO | `Set cookie for Secure space equals`   | PASS    |          |
 0:42.65 INFO | `Set cookie for Secure equals space`   | PASS    |          |
 0:42.65 INFO | `Set cookie for spaced Secure`         | PASS    |          |
 0:42.65 INFO | `Set cookie for space Secure with ;`   | PASS    |          |
 0:42.65 INFO 
 0:42.65 INFO ::: Running tests in a loop with restarts 10 times : PASS
 0:42.65 INFO :::
 0:42.65 INFO ::: Test verification PASS
 0:42.65 INFO :::

[miketaylr]

Yep, that fixed it.