[css-properties-values-api] `initial-value` for `*` with arbitrary substitution values

[cdoublev]

If I understand the spec correctly, initial-value does not require to be validated when syntax is *.

But in Chrome, given the CSS rules below:

• the first rule is valid: initial-value: var(1) is discarded
• the second rule is invalid

@property --p1 {
  inherits: false;
  syntax: "*";
  initial-value: green;
  initial-value: var(1);
}
@property --p2 {
  inherits: false;
  syntax: "*";
  initial-value: green;
  initial-value: var(--initial);
}

I am completely confused by the parse result in Chrome. Anyway, should initial-value be invalid when it includes an arbitrary substitution value?

[tabatkins]

Hm, the spec is indeed unclear here. I prefer Chrome's behavior - validating the initial-value as normal for an untyped custom property. The spec just needs its conditions to be slightly rephrased to match.

[cdoublev]

To clarify my understanding:

• in the first rule, initial-value: var(1) makes the descriptor invalid and ignored
• in the second rule, initial-value: var(--initial) makes the rule invalid because it is computationally dependent

That is, var(1) is invalid at the syntax level of initial-value, as specified:

If a property contains one or more var() functions, and those functions are syntactically valid, the entire property’s grammar must be assumed to be valid at parse time.

And initial-value: var(--initial) is invalid at the context level of @property, as specified:

• The initial-value must be computationally independent.

If the above conditions are not met, the @property rule is invalid.

Computationally dependent values could have been specified as invalid at the context level of initial-value, but this does not seem important since authors will probably not make multiple declarations of initial-value.

So the specifications are already clear and I am closing this issue. I hope I am not missing something.

[cdoublev]

Indeed, I missed something. initial-value must be computationally independent only when syntax is *.

[cdoublev]

initial-value is technically a descriptor so var() parsing should not be triggered as it is only for properties, and initial-value: var(1) should be valid since var(1) matches <declaration-value>. This should also apply in CSS.registerProperty() (WPT), unless:

• initial-value should be parsed as a property
• var() (and other arbitrary substitutions) should always have a valid syntax, even when used in another context than a property value

This should then be explicitly specified in both cases.

But this is a different problem than a valid var(). Please let me know if I should open a separate issue.

---

There are these lines on WPT (it is about the first and the third):

// initial-value: not computationally independent
test_descriptor('initial-value', '3em', null, {'syntax': '"<length>"'});
test_descriptor('initial-value', 'var(--x)', null);

The spec requires initial-value to be computationally independent for @property to be valid, but only when syntax is not *.

An equivalent test exists for CSS.registerProperty(). So I guess there is a good reason.

I wonder which one it is, and whether it also applies to other substitution functions like attr() or first-valid().