815662 - querySelectorAll fails to throw an error on invalid selector syntax

Status: RESOLVED INVALID

(Core :: CSS Parsing and Computation, defect)

Description

[Dave Methvin]

User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.11 (KHTML, like Gecko) Chrome/23.0.1271.64 Safari/537.11

Steps to reproduce:

document.querySelectorAll(":not(.red") //note missing closing paren

http://jsfiddle.net/JhvqV/1/


Actual results:

This selector works and returns all elements without class "red".


Expected results:

An invalid selector should throw a syntax error exception.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Our behavior is an immediate consequence of applying the CSS end-of-file parsing rules, which include implying any ')' that's needed.  Note that this happens on the tokenizer level, so after tokenization the string provided matches the selectors grammar.

Last I checked, WebKit doesn't implement those rules correctly, which is why you're seeing a different behavior there...

Comment 2

[Dave Methvin]

I agree it looks *very* explicit in the CSS2 spec, http://www.w3.org/TR/CSS21/syndata.html#parsing-errors says "User agents must close all open constructs (for example: blocks, parentheses, brackets, rules, strings, and comments) at the end of the style sheet." By that measure Firefox is doing the right thing, as are Opera and Internet Explorer. Safari and Chrome throw an error.

The CSS3 syntax WD is a mess and I can't tell what it is trying to say, other than it's attempting to explicitly allow future extensions. In particular, http://www.w3.org/TR/css3-syntax/#principles says "...if a brace closing a declaration block is not present, the declaration block must be ignored since the closing brace is required to satisfy this grammar." This seems to directly contradict the CSS2 rule above about closing blocks.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Yeah, the CSS3 syntax draft you linked to is 9 years old and very much incomplete.  The current work on that stuff is http://dev.w3.org/csswg/css3-syntax/ which is still unclear to the point of uselessness on this.  Tab is going to define a separate selector parser at some point that querySelector can use, but it hasn't happened yet.

Comment 4

[Simon Sapin (:SimonSapin)]

The 2003 draft of CSS3 Syntax should definitely be ignored. Look either at CSS 2.1 or the new Syntax draft: http://dev.w3.org/csswg/css-syntax/ In both cases, the EOF handling rules make this selector valid, as Boris said. Selectors 4 is expected to be defined on top of the CSS tokenizer.