309067 - Implement :-moz-empty-except-children-with-localname(localname)

Status: RESOLVED FIXED

(Core :: CSS Parsing and Computation, defect, P2)

Description

[Boris Zbarsky [:bzbarsky]]

From bug 8131:

> According to the behaviour spec, we need to apply
> different styles to AREA elements if the IMG element is broken and the MAP
> element only contains AREA elements.

And similarly, the patch for bug 11011 has in it:

+/* Nonempty applets should just show their kids.
+   XXXbz do we need a selector that will ignore <param> elements? */

So perhaps we should have a selector like :-moz-empty-except(ns|localname) that
would match if the only kids are either ns:localname, or ignorable whitespace,
comments, etc?

Comment 1

[Hixie (not reading bugmail)]

Sounds like it would have the same potential problems as my more wide-reaching
:matches() suggestions, namely it would require a crawl of the element's
children which would potentially be slow. Other than that, seems fine to me.

Comment 2

[Boris Zbarsky [:bzbarsky]]

It only requires a crawl if we have a bunch of kids that have that fully
qualified name.  As soon as we hit any kid that doesn't, we know we don't match.

So it's really not any slower than :-moz-only-whitespace...

Comment 3

[Hixie (not reading bugmail)]

True.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Attachment: Something silly along these lines

I can try to think of a less short but still scary name

Comment 5

[Hixie (not reading bugmail)]

Does that support ns|localname, or just localname?

Comment 6

[Boris Zbarsky [:bzbarsky]]

That supports just localname and will only match if the ns of the kid is the
same as the ns of the node we're matching.

Given the (internal) uses for it we want, that restriction was acceptable, and I
wasn't quite prepared to refactor the ns|localname parsing code such that I
could reuse it here.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Oh, and this treats the localname as a literal, so * gets no special treatment
(and in fact won't parse, since that's not an ident).

Comment 8

[Hixie (not reading bugmail)]

aha, I missed the bit in the patch where you checked the parent's namespace.
Yeah,  that sounds perfect. Cunning.

Comment 9

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 197231 [details] [diff] [review]
Something silly along these lines

>Index: dom/locales/en-US/chrome/layout/css.properties
>+PEPseudoClassArgEOF=argument to pseudo-class selector
>+PEPseudClassArgNotIdent=Expected identifier for pseudo-class parameter but found '%1$S'.
>+PEPseudoClassNoClose=Missing closing ')' in pseudo-class '%1$S'.
>+PEPseudoClassNoArg=Missing argument in pseudo-class '%1$S'.

You lost the "o" on the second one of these.

Please test that the error messages still work to double-check that there
weren't other similar mistakes.

Comment 10

[Boris Zbarsky [:bzbarsky]]

Attachment: Testcase to exercise the error messages

Comment 11

[Boris Zbarsky [:bzbarsky]]

Attachment: Updated to comments

Comment 12

[Boris Zbarsky [:bzbarsky]]

Fixed.