consume_url does not record newlines

[haxney]

In consume_url(), a newline character is consumed without calling
tokenizer.new_line(). This can be fixed by replacing the first match block
of consume_url() with

match tokenizer.current_char() {
    '\t' | ' ' => tokenizer.position += 1,
    '\n' => { tokenizer.position += 1;
                tokenizer.new_line(); },
    "\"" => return consume_quoted_url(tokenizer, false),
    '\'' => return consume_quoted_url(tokenizer, true),
    ')' => { tokenizer.position += 1; break },
    _ => return consume_unquoted_url(tokenizer),
}

[SimonSapin]

So the parsed tree is correct, except for the line numbers? Could you give an example of CSS that is affected by this?

[haxney]

The problem is when there is a newline in the url() token, as in the following
file:

a {
  url(
  http://example.com
  );
  clear:left;
}

The following test fails with 'The tokenizer is missing a tokenizer.new_line() call somewhere.', tokenizer.rs:104

#[test]
fn url_newline() {
    let thing = tokenize("a {\n  url(\nhttp://example.com\n);\n  clear:left;\n}").to_owned_vec();
    println(list_to_json(&thing).to_str());
}