Firefox 35: Passing arguments to executeScript isn't working

[lukeis]

Originally reported on Google Code with ID 8390

With Firefox 35, passing arguments to executeScript is not working like it used to:

Expected behaviour:

>> d.execute_script("return typeof arguments[0]", "foo")
=> "string"

Actual behaviour

>> d.execute_script("return typeof arguments[0]", "foo")
=> "undefined"

Reported by jari.bakken on 2015-01-15 13:57:18

[lukeis]

Console warning:

Security wrapper denied access to property length on privileged Javascript object.
Support for exposing privileged objects to untrusted content via __exposedProps__ is
being gradually removed - use WebIDL bindings or Components.utils.cloneInto instead.
Note that only the first denied property access from a given global object will be
reported.

Reported by jari.bakken on 2015-01-15 14:14:29

[lukeis]

I don't have time to dig into this further right now, but here's what I've found so
far:

The warning message suggests we use Components.utils.cloneInto with a Sandbox. Currently
we only use a sandbox if the document is in designMode, and then async script execution
is "not supported". 

Marionette appears to implement executeAsyncScript with a Sandbox, so it's likely possible.

https://dxr.mozilla.org/mozilla-central/source/testing/marionette/marionette-listener.js?from=marionette-listener.js&case=true#528

We could ignore executeAsyncScript for now and just force the sandbox code path for
Firefox 35, i.e.:

  https://gist.github.com/jarib/6a19bd42d2f90c9055b3

However this breaks a few tests: we're no longer able to call functions that are already
defined on the page through executeScript.

Reported by jari.bakken on 2015-01-15 15:13:11

[lukeis]

Reported by jari.bakken on 2015-01-15 15:13:50

[lukeis]

I can replicate this behaviour on OSX 10. 9.5 and Ubuntu using the selenium-webdriver
gem (2.44.0) and Firefox 35.0. Reverting firefox to 34.0 as downloaded from https://ftp.mozilla.org/pub/mozilla.org/firefox/releases/34.0/mac/en-US/
resolves the issue.

The issue is trigged with calls to Capybara::Selenium::Node#set on an <input> field

Reported by chris.atkins@lonelyplanet.com.au on 2015-01-16 02:46:12

[lukeis]

Issue 8389 has been merged into this issue.

Reported by jmleyba on 2015-01-16 03:26:00

[lukeis]

Issue 8398 has been merged into this issue.

Reported by jmleyba on 2015-01-16 20:16:16

[lukeis]

Issue 8387 has been merged into this issue.

Reported by jmleyba on 2015-01-16 20:17:06

[lukeis]

Issue 8393 has been merged into this issue.

Reported by barancev on 2015-01-17 21:02:48

[lukeis]

I also tried working around this by just JSON-serializing the args object, but quickly
realized that it would mean WebElement arguments would no longer work:

  https://gist.github.com/jarib/89406c5c5c019b72cd7c

Perhaps someone has better ideas?

Reported by jari.bakken on 2015-01-17 21:47:12

[lukeis]

Maybe parse the args and swap out webelement references for css selectors that you could
deference on the page side? That's basically what we do in Safari.

Reported by jmleyba on 2015-01-18 19:43:24

[lukeis]

I see this on Ubuntu 14.04 (x86_64) with selenium-webdriver gem 2.44.0 and Firefox 35.0,
and that Firefox 34.0.5 solves the problem.

Reported by derrelldurrett on 2015-01-18 20:37:49

[lukeis]

Thanks, but this is caused by a change in FF35 and will impact that browser on all platforms.

Reported by jmleyba on 2015-01-18 20:53:06

[lukeis]

Waiting for resolution, I confirm rolling back to a previous Firefox version (ex. 33.1)
makes selenium driver 2.44.0 usable again.
(OS : Ubuntu 14.04 x64)

Reported by dvdngrllo on 2015-01-19 15:44:02

[lukeis]

Issue 8325 has been merged into this issue.

Reported by barancev on 2015-01-20 06:47:54

[lukeis]

Issue 8416 has been merged into this issue.

Reported by james.h.evans.jr on 2015-01-21 15:40:24

[lukeis]

Issue 8418 has been merged into this issue.

Reported by jmleyba on 2015-01-21 19:47:07

[lukeis]

Hey, jari.bakken.  Thanks for looking into this.  I wanted to let you know that if you
fix it, my team and I would be happy to buy you some beers or coffees if that is your
style.

Reported by gerryster on 2015-01-22 19:12:54

[lukeis]

Workaround: Use the XPath locator for the web element and trigger a plain JavaScript
invocation.

See http://stackoverflow.com/questions/10596417

Reported by jan.mewes@epages.com on 2015-01-23 16:29:51

[lukeis]

jari.bakken, you need to change only 2 lines of code inside firefoxDriver.js:
https://gist.github.com/Kimosabe/9ad899c3ca630026e3b5

Tried to apply this changes to script inside wibdriver.xpi of selenium v44. Works fine
with firefox 35.

Reported by 4B.69.6D.6F on 2015-01-27 08:53:12

[lukeis]

Can you detail the change you made to the webdriver.xpi of sel v44. Need to get this
fixed asap. Thanks so much!

Reported by Chapman.Chris.H on 2015-01-27 17:29:36

[lukeis]

This issue was closed by revision 66d55c84463d.

Reported by jmleyba on 2015-01-28 00:24:26

• Status changed: Fixed

[lukeis]

Any plans on rolling out a new Selenium release, available in Maven Central?

Reported by yamert on 2015-01-28 16:30:53

[lukeis]

Release is still blocked on fixing native events for FF35.

Reported by jmleyba on 2015-01-29 05:14:59

[lukeis]

Issue 8443 has been merged into this issue.

Reported by jmleyba on 2015-01-31 00:13:41

[lukeis]

I'm really waiting for this to be made part of a (Rubygems) release, because it is a
huge blocker on a project right now. Anyhow, thanks for fixing this in the first place.

Reported by pfrasa on 2015-02-03 11:18:53

[lukeis]

@pfrasa: I've pushed a pre-release gem, 2.45.0.dev3, that includes the fix for this
issue.

Reported by jari.bakken on 2015-02-03 13:15:41

[lukeis]

Is there also a pre-release or a snapshot version of selenium-java artifact available
in maven repo?

Reported by bernard.i.lago on 2015-02-04 02:25:51

[lukeis]

wow, that was fast. Many thanks, everything works perfectly now. :)

Reported by pfrasa on 2015-02-04 09:37:35

[lukeis]

Am also very interested in a pre-release or a snapshot version of selenium-java artifact!

Reported by christian.zangerl on 2015-02-04 10:14:21

[lukeis]

This is a blocker for us as well. +1 for Java maven release.

Reported by AdamJasonDev on 2015-02-04 15:21:44

[lukeis]

When will a stable release come out with this fix?

Reported by akbardhedhi on 2015-02-04 15:46:49

[lukeis]

Great work, and excellent response time here guys. Was wondering when the pre-release
for this will be out on nuget?

Reported by rankin.dc on 2015-02-04 15:57:33

[lukeis]

As a user of protractor, I'd like to try this fix out. How can I install the latest
version of selenium-webdriver?

Thanks for the quick turnaround on this.

Reported by andrew.eisenberg on 2015-02-05 22:43:10

[lukeis]

I am using the python version of Selenium. is there any quick fix for it?

Reported by yizhongji on 2015-02-06 20:16:38

[lukeis]

FYI I have built my own version for dot net - see https://code.google.com/p/selenium/wiki/ReleasingSelenium

Reported by david.rankin@norlandtech.com on 2015-02-09 14:56:06

[lukeis]

Issue 8477 has been merged into this issue.

Reported by jmleyba on 2015-02-10 17:04:40

[lukeis]

@David, Could you please provide me the DLLs, where this issue is fixed. It will be
very helpful.

Reported by Gurudatt.gd04 on 2015-02-16 12:17:49

[lukeis]

Have anyone resolved this issue without downgrading firefox? If so please share your
insight on the same.

Reported by nirmalspecial on 2015-02-17 06:03:01

[lukeis]

"Release is still blocked on fixing native events for FF35."

Any update on this? FF36 will be out soon. Shouldn't these sort of things be resolved
in advance of stable releases of Firefox? Isn't that the whole purpose of Mozilla having
aurora/beta releases available months in advance?

Reported by treguard0439 on 2015-02-17 10:04:56

[lukeis]

Do we have a timeline on a release, yet?

Reported by qtrolazyg on 2015-02-19 16:40:17

[lukeis]

@GURUDATT
Attached are the files. Hopefully they help. if you have problems contact me directly,
I only saw this message recently.

David

Reported by rankin.dc on 2015-02-23 12:48:50

---

- _Attachment: [SELENIUM_DLLS.zip](https://storage.googleapis.com/google-code-attachments/selenium/issue-8390/comment-41/SELENIUM_DLLS.zip)_

[lukeis]

Is there anyway to patch this issue in the current build? Our manual test team just
found a regression in Firefox 35 that would have been caught by our automated tests
weeks ago if we could update. It seems like this is a major blocker for some folks.

Reported by AdamJasonDev on 2015-02-23 21:08:49

[lukeis]

 @AdamJaso...@gmail.com  There is another conversation about similar issues (FF 36),
issue 8399.  There they talk about a dev version that is available.

Reported by qtrolazyg on 2015-02-25 20:10:06

[lukeis]

Is this Javascript issue in firefox 35 fixed in selenium latest version or any patches
released for this fix. Please let me know.

Reported by rameshnss on 2015-02-26 16:30:07

[lukeis]

selenium 2.45 is out and works well with FF release (36)

Reported by qtrolazyg on 2015-03-01 00:23:03

[lukeis]

I upgraded FF to 36 and Selenium to 2.45. In command with Javascript I got error  <Exception
Found: org.openqa.selenium.WebDriverException: Error: Syntax error, unrecognized expression:
By.cssSelector: .email_addresses_tipsy

for FF 35 and Selenium to 2.44 - this error is not appear in my tests

Reported by Antinua888 on 2015-03-05 23:26:47

[lukeis]

> selenium 2.45 is out and works well with FF release (36)

Not really... https://code.google.com/p/selenium/issues/detail?id=8551

Reported by steven.mcphillips@lonelyplanet.com.au on 2015-03-05 23:40:16

[lukeis]

Reported by luke.semerau on 2015-09-17 18:25:06

• Labels added: Restrict-AddIssueComment-Commit