Xidorn Quan [:xidorn] UTC+11 Assignee
	Description • 7 years ago

See the testcase.

There are three custom properties in the target element:
> --varA: var(--varB);
> --varB: var(--varA);
> --varC: var(--varB);

apparently, all of the three should be invalid, because --varA and --varB forms a dependency cycle, and --varC refers to this cycle.

However, stylo fails to mark --varC invalid in this case, but instead inherits it from the parent.

Both Gecko and Blink agree on the correct behavior (that --varC computes to its initial value rather than inheriting).

Web platform test css/css-variables-1/variable-substitution-variable-declaration.html checks this behavior.

	Xidorn Quan [:xidorn] UTC+11 Assignee
	Comment 1 • 7 years ago

I wouldn't say this is super important, but it is probably something we want to fix for shipping stylo.

	Xidorn Quan [:xidorn] UTC+11 Assignee
	Comment 2 • 7 years ago

The current algorithm is actually very unreliable.

For a testcase like this, the cycle removal can ignore certain items depending on the order of traversal. Moving declarations around within #outer rule can lead to different result.

	Xidorn Quan [:xidorn] UTC+11 Assignee
	Comment 3 • 7 years ago

Actually, comment 2 and comment 0 are different issues. There is no test for the issue of comment 2 at the moment.

I'll file a new bug for that one.

	Xidorn Quan [:xidorn] UTC+11 Assignee
	Comment 5 • 7 years ago

https://hg.mozilla.org/integration/autoland/rev/8dad665a3b1a
https://hg.mozilla.org/integration/autoland/rev/d2b23d922b73

	Bobby Holley (:bholley)
	Comment 6 • 7 years ago

Please either request uplift or set 57 status to wontfix.

	Xidorn Quan [:xidorn] UTC+11 Assignee
	Comment 7 • 7 years ago

I don't think it's important to uplift this. It is a behavior change, but mostly for handling error case.