[css-cascade] When an import rule fails to load or has an unsatisfied condition, does the layer still count?

[xiaochengh]

Consider @import rules below:

@import url(url-not-found.css) layer(A);
@import url(sheet.css) layer(B) not all;
@import url(sheet.css) layer(C) supports(display: invalid-value);

While these are valid but ineffective imports, do the layer declarations still count (and hence affect the global layer ordering)? The current spec seems unclear about it.

I suppose for conditional imports whose conditions evaluate to false, the layer declarations should not count, because nested @layer rules inside false media/supports rules do not count.

I'm unsure about imports that fail to load, though. We may count the layer for a stable global layer ordering, or do not counter the layer to be consistent with media/supports conditions. Either as a resolution will be fine for me.

[mirisuzanne]

The way I understand the spec, we should already have implied answers to all of these, which we can clarify if we need to:

• layer A is established, but empty (stable layer order)
• layers B and C are not established since the global queries are false

[tabatkins]

The precise interleaving of layers and conditions in the import rule isn't specified, so we should be clearer about this, but I agree with @mirisuzanne on the details of how each case should work. The mental model should be that a layer is wrapped around the contents of the import (even if they're empty due to a load failure), and a condition is wrapped around the whole thing:

@import url(...) layer(foo) (condition: bar);

/* equivalent to */

@media (condition: bar) {
 @layer foo {
   ...import contents...
 }
}

[mirisuzanne]

Agenda+ to get this spec change approved by the group.

[css-meeting-bot]

The CSS Working Group just discussed [css-cascade] When an import rule fails to load or has an unsatisfied condition, does the layer still count?, and agreed to the following:

• RESOLVED: Accept proposal
• RESOLVED: Republish css-cascade-5

The full IRC log of that discussion

<fantasai> Topic: [css-cascade] When an import rule fails to load or has an unsatisfied condition, does the layer still count?
<fantasai> github: https://github.com//issues/6776
<fantasai> miriam: A few questions in here
<fantasai> miriam: All related to establishing layers inside of an @import statement.
<fantasai> miriam: One question is what to do if the import fails
<fantasai> miriam: other question is what to do if the conditions of the import don't match
<fantasai> miriam: consensus in the thread was that we should conceptualize this to "layer is wrapped around contents of the import, even if empty due to failure, and condition is wrapped around the whole thing"
<fantasai> miriam: We added language to spec to express it that way
<fantasai> miriam: layer is still added to layer order if import fails
<fantasai> miriam: but not added if the conditions don't match
<fantasai> miriam: that matches the behavior of if you had import inside of layer inside of condition
<fantasai> miriam: wanted to check with WG if this is acceptable
<fantasai> jensimmons: Makes a lot of sense from authoring point of view
<fantasai> +1 from me
<fantasai> RESOLVED: Accept proposal
<fantasai> miriam: This allows us to republish
<fantasai> RESOLVED: Republish css-cascade-5