Skip to content

Latest commit

 

History

History
123 lines (76 loc) · 6.07 KB

2024-11-05.md

File metadata and controls

123 lines (76 loc) · 6.07 KB

wpt RFCs/infra sync

Agenda Items for Nov 5, 2024

Cookie State

web-platform-tests/wpt#48106

Notes:

  • @jgraham: A change landed in the WPT runner chromium implementation that cookies are cleared between tests. But not implemented in other browsers. Historically, we let the tests clean up the cookies. Should have been a RFC.
  • @jcscottiii: Ball park figure
  • @jgraham: Been a few failures. This landed a few weeks ago.
  • @jonathan-j-lee: Should be fine to revert for now
  • @gsnedders: What we don't have is a way to deal with the rest of storage in a determinsitic way. Wish wptrunner had a way to restart, reset the browser between every test
  • @jgraham: Should revert. And then have the RFC to discuss this since it is reasonable. In WebDriver Classic, there is not a delete all cookies command. Maybe BiDi changes that? And I agree there is no current way to do that in storage. There is a restart before test flag that you could apply. It would just take a very long time to run. We do have verify mode that helps with groups of tests (runs with restarts between tests, and rerunning the test many times without a browser restart)

Action Items:

  • @jonathan - Revert 48106

RFCs

208 web-platform-tests/rfcs#208 - adding third party Ed25519, HPKE, and CBOR libs

Recap:

  • @gsnedders: Need to track down the newly unresolved comments with Apple legal.

Notes:

211 web-platform-tests/rfcs#211 - Support testdriver.js in other test types

Notes:

  • @jonathan-j-lee: Useful for customizable select. Made a prototype implementation. Works as you would use it in testharness tests
  • @jgraham: In favor of this. Will need to check if this is more effort for us. Would need a way to fall back to the other implementation.
  • @gsnedders: Things like select open what is technically a new window. And the web driver screenshot command takes a screenshot of a window. Which means it wouldn't be captured. And other things with user interaction where this becomes more apparent.
  • @jgraham: Not sure if we have the same problem with screenshot but select in webdriver can be a problem. But do not know enough details. We should figure out how to make this change, even if only for other use cases.
  • @gsnedders: I think it is worthwhile but there will be cases with more differences

Action items:

  • @jgraham and @gsnedders: in favor in general but will comment on RFC on some issues.

207 web-platform-tests/rfcs#207 - step_wait_promise method

Notes:

  • @sadym-chromium: No strong opinions. Okay with the existing one and keep with overhead
  • @gsnedders: Same here. Maybe we move ahead with this.

Action items:

  • Leave alone for now

212 web-platform-tests/rfcs#212 - Introduce require_webdriver_bidi metadata tag

Notes:

  • @sadym-chromium: Addressed @gsnedders initial comments. This is required for us to implement BiDi methods and avoid enabling for all tests.
  • @jgraham: Agree that is the right way to do a gradual opt-in to certain tests. Improvement from the previous RFC.
  • @gsnedders: There are plenty of other webdriver commands not supported cross browser
  • @jgraham: True but most of those commands are not supported cross browser.
  • @sadym-chromium: Our intenttion is to add tests for things not tested at the moment.
  • @jgraham: Hope this is just a temporary thing for the gradual migration.
  • @sadym-chromium: This RFC does not require all the browsers to implement BiDi and migrate to it.
  • @past: Are we ready to accept this RFC?
  • @jgraham: I need to read over it. But I don't have major concerns.
  • @gsnedders: I need to read over it again.

Action items:

  • Reviewers to read. Feel free to ping.

213 web-platform-tests/rfcs#213 - Wait for document.fonts.ready before running checkLayout() tests

Notes:

  • @jonathan-j-lee: Some flakiness. The web font is not loaded. Previously using content shell. That sideloaded the fonts. Probably more desireable for checkLayout to do that. Then people don't have to remember to add that.
  • @jgraham: This seems fine.
  • @gsnedders: My question: Whether we should wait until load before running things. We should land this with document.font.read and move from there. If we have more problems, we can come back.
  • @jgraham: One other thing to note: Removing the installation of fonts should require a RFC. But marked this RFC as approved.

Action items:

  • gsnedders to write a comment

205 web-platform-tests/rfcs#205 - Layout tests for non-browsers

Notes:

  • @jgraham: Skeptical about adding the complexity for this. But we can revisit if someone wants it.

Azure Pipelines migration

Notes:

  • @gsnedders all the full runs are on GitHub actions. PR runs still run on Azure pipelines. One thing that is wrong: Not running on the aligned epoch. So sometimes we missed the aligned runs or run a single commit twice.
  • @jgraham: Is there a GH action to make sure it runs on the aligned commit
  • @gnsedders: No. Depends on how smart we want to be. Can go with the good enough option.

Navigating the test web-platform-tests/wpt#48825

  • Should we relax this restriction? We have both testharness and reftests which try to do this.

Notes:

  • @gsnedders: If continue testing in frames, not sure how this will work with View Transitions
  • @jgraham: Defintely a pain to work with. Don't have a proposal without moving away from WebDriver. BiDi would work. Maybe we could hack something but it would be fragile. This is a strong motivation to move away from WebDriver classic to BiDi. That transition is not on the horizon.

Firefox android infra tests

Mobile results update

  • @gsnedders: Have some pieces working with Safari. Big risk: Figuring out the regressions with webdriver on iOS.