Parse issue when using Zombie

[its-danny]

Issue was mentioned here, but told to post here as well.

When trying to run any tests, it errors out with this:

/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/tokenizer.js:62
            throw(e);
^
Error
at Object.appendChild (/usr/local/lib/node/.npm/jsdom/0.1.22/package/lib/jsdom/level1/core.js:1312:13)
at Object.insert_html_element (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/parser/before_html_phase.js:41:21)
at Object.processStartTag (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/parser/before_html_phase.js:29:7)
at EventEmitter.do_token (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/parser.js:94:20)
at EventEmitter.<anonymous> (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/parser.js:112:30)
at EventEmitter.emit (events:31:17)
at EventEmitter.emitToken (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/tokenizer.js:84:7)
at EventEmitter.emit_current_token (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/tokenizer.js:813:7)
at EventEmitter.tag_name_state (/usr/local/lib/node/.npm/html5/0.2.5/package/lib/html5/tokenizer.js:358:8)
at EventEmitter.<anonymous> (/usr/local/lib/node/.npm/html5/0.2.5/package

[aredridel]

Indeed, looks like a bug.

(Also, the key bit is the data being parsed: http://beaconpush.com/)

[aredridel]

Which version of node are you using?

[its-danny]

v0.2.6

[aredridel]

Hrm. Okay. Just not sure what the error is; it's rethrown by the parser, since it's a real exception and not just the input stream draining, but it's not printing out what the error is (this is unusual -- the error is rethrown exactly)

Is there code that triggers it that I can look at?

[its-danny]

It seems like trying to run any Zombie tests triggers it, but here's a quick example:

var zombie = require('zombie')
  , assert = require('assert')

 zombie.visit('http://localhost:9292', { debug : true }, function(error, bro
  if (error) throw error

  assert.ok(browser.querySelector('#login'))
})

[aredridel]

That's cut off on the right columns there.

[its-danny]

D'oh, it's function(error, browser) {

[aredridel]

Filling it in with "ser) {", it runs for me if I disable the call to fixQueue in zombie. It looks to me like it's just not patching jsdom right -- but I don't get your original error at all.

[funkatron]

I'm pretty much stuck on this atm; can't run any Zombie tests. If someone has specific workaround instructions (I'm a meganewb), that would be very appreciated. thanks!

[clintecker]

We're also seeing parsing errors on http://arstechnica.com --- I think maybe because of relative script src URIs (our ad calls)?

[assaf]

@aredridel you must have two instances of JSDOM there, one loaded by HTML5, one loaded by Zombie. That's why Zombie can't find the fixQueue method that it adds to HTMLDocument. But running without those patches is running an incomplete test (scripts will load in wrong order, some events not handled, etc)

@clintecker if the problem is in resolving the src URL, it lies either in Zombie or JSDOM.

[rubys]

I'm seeing this problem too... without using Zombie. On a fresh install of node, npm, and html5 on Ubuntu 10.04-1.

/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/tokenizer.js:62
                throw(e);
    ^
Error
    at Object.appendChild (/home/rubys/lib/node/.npm/jsdom/0.1.23/package/lib/jsdom/level1/core.js:1312:13)
    at Object.insert_html_element (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/parser/before_html_phase.js:41:21)
    at Object.processStartTag (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/parser/before_html_phase.js:29:7)
    at EventEmitter.do_token (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/parser.js:94:20)
    at EventEmitter. (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/parser.js:112:30)
    at EventEmitter.emit (events:31:17)
    at EventEmitter.emitToken (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/tokenizer.js:84:7)
    at EventEmitter.emit_current_token (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/tokenizer.js:813:7)
    at EventEmitter.after_attribute_value_state (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/tokenizer.js:557:8)
    at EventEmitter. (/home/rubys/lib/node/.npm/html5/0.2.7/package/lib/html5/tokenizer.js:59:25)

Here's my script:

var http = require('http'),
    html5 = require('html5'),
    jsdom = require('jsdom').jsdom(),
   window = jsdom.createWindow(null, null, {parser: html5});

var rubix = http.createClient(80, 'intertwingly.net');
var request = rubix.request('GET', '/blog/',
  {'host': 'intertwingly.net'});
request.end();
request.on('response', function (response) {
  var length = 0;
  var data = new Buffer(parseInt(response.headers['content-length']));
  response.on('data', function (chunk) {
    chunk.copy(data, length, 0, chunk.length);
    length = length + chunk.length;
  });
  response.on('end', function () {
    // console.log(data.toString('utf-8'));
    var parser = new html5.Parser({document: window.document});
    parser.parse(data);
    console.log(window.document.innerHTML);
  });
});

[aredridel]

Looks like it's jsdom's non-textual error messages bite again.

It's trying to double-add a document element. I'll see about fixing that.

[aredridel]

Got it! The document passed in wasn't empty (thanks, jsdom!), I detect that now and handle it (using the existing document tag, clearing its children.)

Fixed in v0.2.9

[assaf]

With 0.2.9 I get "Cannot call method 'toLowerCase' of undefined" from lib/html5/treebuilder.js:176:36

Also before_html_phase calls console.debug, which is not defined in node 0.2.6. Assume this should be a call to HTML5.debug.

[aredridel]

I've fixed the debug statement.

Looking at the rest.