Error response status: 13 after connect with ie6

[michaelmior]

I get the error below consistently. With ie9 I just get "socket disconnected". Both these messages appear after a sizeable delay. Chrome and Firefox seem to work fine.

$ SAUCE_USERNAME=XXX SAUCE_ACCESS_KEY=XXX repl ie6
… setting up tunnel
… booting up internet explorer (6) on Windows XP
… ready!
ie6 › window

Error: Error response status: 13. 
    at Webdriver._newError (/usr/lib/node_modules/browser-repl/node_modules/wd/lib/webdriver.js:80:13)
    at /usr/lib/node_modules/browser-repl/node_modules/wd/lib/webdriver.js:156:25
    at /usr/lib/node_modules/browser-repl/node_modules/wd/lib/webdriver.js:411:5
    at Request._callback (/usr/lib/node_modules/browser-repl/node_modules/wd/lib/webdriver.js:281:7)
    at Request.self.callback (/usr/lib/node_modules/browser-repl/node_modules/wd/node_modules/request/request.js:122:22)
    at Request.EventEmitter.emit (events.js:98:17)
    at Request.<anonymous> (/usr/lib/node_modules/browser-repl/node_modules/wd/node_modules/request/request.js:888:14)
    at Request.EventEmitter.emit (events.js:117:20)
    at IncomingMessage.<anonymous> (/usr/lib/node_modules/browser-repl/node_modules/wd/node_modules/request/request.js:839:12)
    at IncomingMessage.EventEmitter.emit (events.js:117:20)

[TooTallNate]

Dup of #13 really. The 13 response status appears to happen after Internet Explorer crashes (read: the entire browser quits) on the sauce vm.

[TooTallNate]

Any help on figuring out where/why the crash is occurring would be very helpful :)

[rauchg]

@admc seen this coming from sauce labs quite a bit. Any insight?

cc @defunctzombie

[defunctzombie]

Yea. There is more info in the error object. The value field I think. The
WD module is kinda annoying when it makes errors cause it doesnt put the
human text in the message field.
On Feb 19, 2014 1:16 AM, "Guillermo Rauch" [email protected] wrote:

@admc https://github.com/admc seen this coming from sauce labs quite a
bit. Any insight?

cc @defunctzombie https://github.com/defunctzombie

Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-35469607
.

[santiycr]

I bet you this is connected to the complexity of the window element. How do you serialize elements before pulling it from Sauce Labs?

[TooTallNate]

@santiycr We're using our util-inspect module, which is node.js's util.inspect() function ported to the browser for old ES3 browsers.

These host objects have already been giving us a wide range of issues with operations that would normally be fine with well-behaving JS objects. No doubt in my mind that some for (var i in window.document.foo.bar) ends up crashing the browser or something silly like that :p Hopefully we can narrow it down soon.

[defunctzombie]

@TooTallNate when you see that error 13 stuff, make sure to look at the other properties of the error object as they will have additional information about the error.

[TooTallNate]

@defunctzombie Cool, will do. Does the 13 error code have any generic meaning in the webdriver protocol?

[defunctzombie]

Not that I am aware of but I dont know anything about the protocol.
On Feb 19, 2014 5:49 PM, "Nathan Rajlich" [email protected] wrote:

@defunctzombie https://github.com/defunctzombie Cool, will do. Does the
13 error code have any generic meaning in the webdriver protocol?

Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-35559814
.

[TooTallNate]

Lol, lovely:

13  UnknownError     An unknown server-side error occurred while processing the command.

From https://code.google.com/p/selenium/wiki/JsonWireProtocol#Response_Status_Codes

[defunctzombie]

Rofl
On Feb 19, 2014 6:06 PM, "Nathan Rajlich" [email protected] wrote:

Lol, lovely:

13 UnknownError An unknown server-side error occurred while processing the command.

From
https://code.google.com/p/selenium/wiki/JsonWireProtocol#Response_Status_Codes

Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-35561407
.

[michaelmior]

I noticed that window.parent seems to be what's causing the issue. Given that it takes some time to blow up, I wonder if it's the circular reference causing the issue. That is, window.parent == window in this case.

[TooTallNate]

@michaelmior For IE6, at least, window.parent === window is false :( But actually, window.parent === window.parent.parent is true. Regardless though, the util-inspect function should be able to handle cyclic references fine, even in IE6:

㋛ ~/browser-repl (master) ❯ node repl.js ie6
… setting up tunnel
… booting up internet explorer (6) on Windows XP
… ready!
ie6 › a = {}
{}
ie6 › a.a = a
{ a: [Circular] }
ie6 › a
{ a: [Circular] }

[michaelmior]

Correct, I meant to say window.parent === window.parent.parent. Perhaps nothing to do with circular references then. But I did notice trying to log window.parent fails as well.

[santiycr]

Wouldn't be surprised if from the browser side this just manifests as one
of those good old "Script taking too long" popups of death.
I agree that is something that can get fixed (or hacked, I'd say) by
tweaking the inspect module you guys are using.

Let me know if you need me to help out from the Sauce side to show you guys
how to get more visibility into what's going on in the browser.

On Wed, Feb 19, 2014 at 3:51 PM, Michael Mior [email protected]:

Correct, I meant to say window.parent === window.parent.parent. Perhaps
nothing to do with circular references then. But I did notice trying to log
window.parent fails as well.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-35566718
.

[michaelmior]

FWIW, I tried running a loop that just spins.

for (var i=0; i<1000000000; i++) { Math.pow(2, 999); }

In this case, I just get socket disconnected instead of error 13. Subjectively, the delay is also quite a bit longer than in the case of trying to log window.

[defunctzombie]

I kinda feel like this should be handled by saucelabs / the web driver
server better. Since it can be reproduced (from my understanding) I would
imagine it should be possible for a sauce person to run it and see the
failure?
On Feb 20, 2014 8:31 AM, "Michael Mior" [email protected] wrote:

FWIW, I tried running a loop that just spins.

for (var i=0; i<1000000000; i++) { Math.pow(2, 999); }

In this case, I just get socket disconnected instead of error 13.
Subjectively, the delay is also quite a bit longer than in the case of
trying to log window.

Reply to this email directly or view it on GitHubhttps://github.com//issues/20#issuecomment-35621224
.