Hrm. That's surprising. https://w3c-test.org/html/browsers/origin/origin-of-data-document.html passes when I load it in canary. I wonder why content shell decides that function shouldn't throw...

I'll take a look, thanks!

Running ToT `content_shell` against https://w3c-test.org/html/browsers/origin/origin-of-data-document.html shows a passing test, and copy-pasting the test into a local layout tests seems successful as well (see https://codereview.chromium.org/2344833002). Strange! :)

Are we modifying the test in some way when importing it? Is our `testharness.js` up to date? Can I run the importer locally to see what's going on?

Maybe it's because the he tests are being imported into imported/wpt/, not http/?

Regardless, patching in the roll CL allows me to reproduce this locally: https://codereview.chromium.org/2331373005

Hrm. Are these tests not being run through the WPT server, but from a `file:` URL or something similar?

I don't understand what's going on here. I can also reproduce the error if I load that patch locally, but if I load the file itself (e.g. file:///ssd/blink/src/third_party/WebKit/LayoutTests/imported/wpt/html/browsers/origin/origin-of-data-document.html) then the test passes. I need to munge paths though to ensure that testharness loads correctly... I'm not sure how the test works at all if the file is being loaded directly. Do we parse the file at all when importing to make `/resources/` mean something different? There's probably a wiki page on this process somewhere...

Looks like it should be running wptserve: https://cs.chromium.org/chromium/src/third_party/WebKit/Tools/Scripts/webkitpy/layout_tests/servers/wptserve.py.

#6 is correct. All imported/wpt tests are being loaded as file:, unless the flag --enable-wptserve is given when running the tests. In the future we'd like to enable wptserve ( bug 360762 ) but in the meantime we've been blacklisting such tests by putting them in W3CImportExpectations.

We do parse the file to make file:///{common,images,media,resources}/ redirect to third_party/WebKit/LayoutTests/imported/wpt/* (https://cs.chromium.org/chromium/src/content/shell/renderer/layout_test/blink_test_runner.cc?l=224)

So the next step for this is just to add a line to W3CTestExpectations to skip that test for now (OR: when importing, add a Failure expectations in TestExpectations and a Pass expectation in WPTServeExpectations.)

> So the next step for this is just to add a line to W3CTestExpectations to skip that test for now

Looks like that has been done already: https://chromium.googlesource.com/chromium/src/+/ccabfd31fd88e080dfb577e6fcc3376ef1f56238

The timing looks close, perhaps this one build just didn't have that commit -- the commit just adds the [Skip] line; the test appears to have been imported previously.

Removing Sheriff label, but I'll leave it to mkwst / qyearsley to actually mark it as fixed, once everyone agrees that's true.

Thanks iclelland -- the file is now listed in W3CTestExpectations under " crbug.com/490940 : Needs to load extra HTML resources." -- but arguably it should be listed under "Requires http server."

This might matter because later if we enable wptserve then we'll try to enable all those tests in that category at once, and if it's under " crbug.com/490940 " then we might forget to try to enable it.

mkwst, what do you think?