The attachment is a test case that Firefox currently fails. I just submitted it to the W3C test suite in https://github.com/w3c/csswg-test/pull/852 Custom property values is conceptually sequences of tokens, not strings. Firefox represents them as strings internally, and when substituting variable references with the corresponding value inserts `/**/` between adjacent tokens that would otherwise be re-parsed as a single token. It relies on the table  in the css-syntax spec, but that table has not been updated when the definition of identifier tokens was extended  to allow `--` prefixes. This was just fixed .  https://drafts.csswg.org/css-syntax/#serialization  http://hg.csswg.org/drafts/rev/ab8b75f1fa88  http://hg.csswg.org/drafts/rev/92d088e07ca2
This new eCSSTokenSerialization_Symbol_Minus case is the same as the earlier eCSSTokenSerialization_Symbol_Hash case, isn’t it? Also, nice, more tests I didn’t know about. I might steal them for Servo :)
Thanks, yes looks like I can join those two cases. If you want to convert my variables-related mochitests into Web Platform Tests at any point, please be my guest. :-)
Assignee: nobody → cam
Status: NEW → ASSIGNED
Attachment #8659635 - Flags: review?(simon.sapin)
Attachment #8659635 - Flags: review?(simon.sapin) → review+
You need to log in before you can comment on or make changes to this bug.