Selenium Error When Raven is Initialized

[bluecollarcoder]

All my Selenium tests pass before I integrated Raven. Tests still pass if I load the Raven JS file but not initialize Raven. Tests fail in Firefox ONLY with:

Error: [frame("even-iframe-embed1")] Error response status: 13, UnknownError - An unknown server-side error occurred while processing the command. Selenium error: Permission denied to access property "__raven__"

if Raven is loaded and initialized. Here is the initialization code:

window.Raven.config('<%=loggerKey%>', {
  release: '<%=version%>'
}).install();

Any help with this issue will be appreciated.

[darkkn8t]

+1 Facing the same issue when running on Firefox using Selenium

[benvinegar]

Is there any possibility of either of you putting together a repository with a reproducible example?

Questions I'd have:

• Does the error occur when install() is called? Or later during test execution?
• Does the error occur on a particular type of event, e.g. an event listener or XHR handler?

[bluecollarcoder]

Error seems to happen during install(). If I remove that call, tests work; if I add it back in, tests fail. I ran across this problem on an internal project. I will try to put together an example.

@srshah19 If you have something that you could share, I would appreciate it.

[darkkn8t]

Unfortunately I will not be able to share the repo. But I can try and explain what is happening:

Our app is a Django Project.

1. Using Raven JS directly from the CDN in the base template.
2. Calling the Raven.captureException(error); on a function().then().catch() block.

Selenium tests only fail when using Firefox. Here is the exception: Unexpected Exception: Error: Permission denied to access property "__raven__".
Does this help?

[bluecollarcoder]

After some more digging, it seems that the problem is caused by line 249:

// We don't wanna wrap it twice!
if (func.__raven__) {
    return func;
}

A console log right before this line shows that func is a native function (in this case it is the window.addEventListener function). So it would seem that a combination of WebDriver and FF throws exceptions when trying to modify a native function. Adding a try..catch before this line to handle the error case fixes the problem:

try {
  func.__raven__;
} catch (e) {
  return func;
}

@benvinegar However, I am not certain that this is the right long-term fix. Please advise.

[bluecollarcoder]

Also for completeness, I should mention that the caller function that is triggering this exception is line 683:

return orig.call(this, evt, self.wrap(fn), capture, secure);

in the function _wrapBuiltIns. Removing self.wrap() or moving it into the try block also fixes the problem.

[benvinegar]

We should probably do what you've done, which is try/catch the __raven__ access.

I notice that the folks at Bugsnag put try/catch around their entire wrap function. Looks like it was to solve the same problem.

So I'm in favor of going ahead with that solution. I can put a fix together w/ a test shortly and cut it w/ the same release. Or if you're so inclined, I'll accept a PR too :)

[bluecollarcoder]

@benvinegar I don't quite understand the reasoning and logic behind wrapping addEventListener and removeEventListener. It's going to take me some time to grok it and put in a fix. I can certainly do it, but might take a little while.

[benvinegar]

@bluecollarcoder – to be clear, I prefer the solution of putting the try/catch in Raven.wrap when accessing the __raven__ property.

[bluecollarcoder]

In that case, I will work on it later this week.

Just so I understand the fix, what is the correct behavior if func is a native function (thus throwing the exception)? Should it be returned as is (like in my workaround), or should it be wrapped anyway (like in the original logic)?

If the correct behavior is to wrap it, then we would still need to mark a function as wrapped without causing the permission exception. Maybe instead of attaching a flag to the function itself, we would need to use a collection that holds the wrapped functions.

Let me know your thoughts.

[bluecollarcoder]

@benvinegar Ha. Just saw that you posted a blog post about function wrapping. Very timely, thanks.

[dimeloper]

What's the status with this issue? Will it be fixed in an upcoming release?

[benvinegar]

@dkiriakakis I'll get a fix together today and we'll cut a release ASAP.

[darkkn8t]

Thanks @benvinegar !