1428932 - Fix cascade-import-002 tests from CSS Cascading and Inheritance Level 3 test suite, due to its dependence on deprecated <meta http-equiv="Link"> tags

Status: RESOLVED INCOMPLETE

(Web Compatibility :: Site Reports, defect)

Description

[Tantek Çelik]

FF 59.0a1 fails four tests (g,h,q,r) in https://test.csswg.org/suites/css-cascade-3_dev/nightly-unstable/html/cascade-import-002.htm

Comment 1

[Daniel Holbert [:dholbert]]

We fail the same four tests regardless of whether stylo is enabled, FWIW.  So: not a stylo-specific issue.

Comment 2

[Daniel Holbert [:dholbert]]

The test and its resources can be downloaded at https://test.csswg.org/suites/css-cascade-3_dev/nightly-unstable.zip (maybe in some repo, too, but I couldn't find it inside of web-platform-tests)

The g,h failures seem to be because this markup fails to work:
> <meta http-equiv="Link" content='<support/cascade-import-002h.css>; rel="stylesheet"'>

And the q,r failure seems to be because this markup fails to work:
> <meta http-equiv="Link" content='<support/cascade-import-002r.css>; rel="stylesheet"'>

If I replace those with the more common <link rel="stylesheet"> markup, then the lines turn green. So, this seems to be us not honoring that markup.

Also worth noting: no other browser gets these lines right, AFAICT. In latest Chrome, Safari, and Edge: lines a-h are all black (rather than green) and lines q,r are red.  (Lines a-h are all black in Firefox, too, if I view the testcase locally -- extracted from the zip file linked above.)

Comment 3

[Daniel Holbert [:dholbert]]

Hmm - I don't see "Link" as one of the valid options for the http-equiv attribute here:
 https://html.spec.whatwg.org/multipage/semantics.html#attr-meta-http-equiv

The description of this attribute says:
> The http-equiv attribute is an enumerated attribute. The following table lists the keywords defined for this attribute.
...and "Link" is not listed in the table. So I think that means our behavior (disregarding this markup) is correct...?

It looks like this is a feature that we used to support, but it was deprecated/unshipped and AFAICT is intentionally unspecced (i.e. intentionally not a spec-compliant feature at this point):
 https://bugzilla.mozilla.org/show_bug.cgi?id=587928
 https://www.w3.org/Bugs/Public/show_bug.cgi?id=9797
 https://lists.w3.org/Archives/Public/public-html/2010Aug/0048.html

So we're actually correct, I think, and it's bogus that the test relies on this feature.  So this is INVALID in that there's no gecko bug here. I think we should file a testsuite bug, wherever these tests are maintained - tantek, would you mind doing that?

Comment 4

[Daniel Holbert [:dholbert]]

(feel free to reopen this bug and morph it to Tech Evang or something, if you prefer, too.)

Comment 5

[Tantek Çelik]

Thanks for tracking down this test dependence on a non-feature dholbert!

I finally tracked down how to file a testsuite bug on this test file (was pretty non-obvious, something I'm separately going to work on improving), and have now done so:

https://github.com/w3c/web-platform-tests/issues/9910

Morphed to Tech Evangelism as suggested.

Comment 6

[Mike Taylor [:miketaylr]]

Thanks Tantek, let's go ahead and close this as a dupe of the bug you filed.

(But awkwardly, we can't do that, so let's close it as INCOMPLETE).