Change wpt tests for <meta name=color-scheme> to not depend on the color-scheme computed value (before whatwg/html#7226)

[emilio]

This makes the tests valid both with and without the spec change
proposed above. Will send a follow-up PR to test the spec change.

This makes the assumption that the initial browser's color-scheme is
light, which I think is a reasonable assumption to run these tests
under.

An alternative would have to allow both "light" and "dark" wherever we
check for "normal" now (which loses test coverage in practice), or I
could complicate the test to compute the initial color-scheme using an
iframe or what not (but that complicates everything more).

[lilles]

These tests are valid even without the spec pull request, right?

[emilio]

These tests are valid even without the spec pull request, right?

No, because they assert that getComputedStyle(document.documentElement).colorScheme is normal. Otherwise, yes.

[emilio]

@lilles I guess I could remove that assert and land it, if you want?

[emilio]

(then add that assert in another PR)

[lilles]

@lilles I guess I could remove that assert and land it, if you want?

sgtm, what suits you best.

[emilio]

Alright, let's do that.

[emilio]

Sent the follow-up in #31268.

[emilio]

Err, wrong tab-complete :) #31271

[tabatkins]

This makes the assumption that the initial browser's color-scheme is
light, which I think is a reasonable assumption to run these tests
under.

This is reasonable, but it should be documented in the test as an assumption, so anyone producing a default-dark browser can tell why they're failing the test and that the test should be fixed, rather than their browser. ^_^

[emilio]

This is reasonable, but it should be documented in the test as an assumption, so anyone producing a default-dark browser can tell why they're failing the test and that the test should be fixed, rather than their browser. ^_^

#31274