See bullet point two of: http://dev.w3.org/csswg/css-syntax/#changes-CR-20140220

See: http://dev.w3.org/csswg/css-syntax/#consume-a-declaration0
After consuming the colon token, it states the declaration's value is the consumption of every subsequent component value, until we reach an EOF token.

Identifiers may now begin with "--".

The semicolon token should not be respected.

See http://dev.w3.org/csswg/css-syntax-3/#consume-an-ident-like-token

Quoted URLs are now tokenized as functions

…05c' into incremental

The tokenizer is now fully incremental, rather than yielding a block/function at a time.

The convention is now to take `&mut Parser` as input for parsing, and only consume as necessary.

Let’s assume that str::len() calls are optimized away.

This would help taking input incrementally (rather than having a single `&str`) if we want to do that.

Conflicts:
	Cargo.toml
	src/color.rs
	src/parser.rs
	src/serializer.rs
	src/tests.rs
	src/tokenizer.rs

(Most values are still `Owned`, so we don’t take much advantage this yet.)

…s or NUL.

…es or NUL.

…pes or NUL.

Previous revisions of it were not because they consumed the delimiter
token.

"Nested" or "delimited" parsers can now only be used in a closure given
to the corresponding `Parser` method. This allows doing cleanup after
calling the closure, rather than in a `Drop` implementation.

This more local reasoning makes thing easier to understand
and less likely to be buggy.

…rser::parse_str.

A cloned parser is "detatched" from its parent and will progress independently,
but it keeps referencing the same input.

…std::cell::Cell.)

Its content is now merged into tokenizer.rs.

This simplifies the code by having less special cases.

It wasn’t really necessary, and confusigly had a different meaning than Parser::is_exhausted().

…efore.

The new convention is that any parsing function returning `Err(())`
may or may not have consumed any number of tokens from the input.

Therefore, any the caller wishes to recover and try parsing something else,
it should call `parser.reset(position)` where `position` is the result of
`parser.position()` that was saved beforehand.

The Parser::try helper does this for you by taking a closure returns Result,
and resetting the position if the result is Err.

…omments.

It’s cheap, and could be useful for tools such as pre-processors.