In An+B notation, it should be invalid to have comment between "A" and "n"

[upsuper]

Things like :nth-child(-1/**/n+1) should be invalid.

It is not clear to me whether the spec matches that.

cc @SimonSapin

[SimonSapin]

What is "should" if not according to a spec?

[upsuper]

The spec says inserting whitespace there is invalid, so presumably adding comment is invalid as well? I suspect it indicates that An should be an unbreakable token?

[upsuper]

So, yeah, An is a dimension token according to the spec, so it should be unbreakable: https://drafts.csswg.org/css-syntax-3/#the-anb-type

[SimonSapin]

Without the comment, -1n+1 tokenizes as a dimention token with unit n, followed by a number token with an explicit plus sign. This matches <n-dimension> <signed-integer> in https://drafts.csswg.org/css-syntax/#the-anb-type.

With the comment it becomes an integer token, then an ident token, then an integer token with a plus sign. This does not match that grammar.

[SimonSapin]

If I add "-1/**/n+1", null, (expecting a parse error) to src/css-parsing-tests/An+B.json, unit tests still pass. So is there anything to do here?

[upsuper]

Oh, it seems I misunderstood the test result. It is Gecko who's wrong on this.

[SimonSapin]

(Gecko was correct according to the Selectors Level 3 definition: ['-'|'+']? INTEGER? {N} [ S* ['-'|'+'] S* INTEGER ]?, but that bit of grammar was weird of sort-of incompatible with the usual CSS tokenization.)