- From: Ian Hickson <ian@hixie.ch>
- Date: Sat, 13 Dec 2008 01:15:16 +0000 (UTC)
On Fri, 12 Dec 2008, Jonas Sicking wrote:
>
> Currently tokenizing the following string (starting at Data state)
> "<!--foo" results in a parse error when hitting the 'f'.
As far as I can tell, it does not.
Assuming we're in a normal state of affairs, here are the states we visit
and the parse errors we emit as we are consuming the characters:
1. Data state: consume "<" and switch to tag open state.
2. Tag open state: consume "!" and switch to markup declaration open
state.
3. Markup declaration open state: see "--" next, consume those characters,
and switch to comment start state.
4. Comment start state: consume anything else ("f"), switch to comment
state.
5. Comment state: consume anything else ("o"), stay in comment state.
6. Comment state: consume anything else ("o"), stay in comment state.
7. Comment state: parse error, consume EOF, return to data state.
8. Data state: end.
No errors are emitted until the end of the string, as expected.
Am I missing something?
> It seems like the error is in the "Comment start dash state" (section
> 8.2.4.19). It should switch to 'comment state' when a '-' is consumed,
> which is not what it currently does.
Given "<!--foo" you should never hit the comment start dash state.
> One more thing I forgot to mention. Several of the states regarding
> comments refer to outputting 'the comment token' and 'the comment
> tokens data'. However there is no mention that I could find for when
> the comment token is created. Maybe this isn't an error but a general
> pattern?
The comment token is created in the first paragraph of the markup
declaration open state.
HTH,
--
Ian Hickson U+1047E )\._.,--....,'``. fL
http://ln.hixie.ch/ U+263A /, _.. \ _\ ;`._ ,.
Things that are impossible just take longer. `._.-(,_..'--(,_..'`-.;.'
Received on Friday, 12 December 2008 17:15:16 UTC