Conflict with turbolinks 5.0.0.beta4

[szyablitsky]

Updated our application from version 2.3.0 of raven-js to 3.0.2.
Now I get the error every time click on any internal link

TypeError: a.match is not a function
  at r.Turbolinks.History.r.update (eval at <anonymous> (http://localhost:3500/vendor-bundle.js:3024:2), <anonymous>:8:11165)
  at r.Turbolinks.History.r.push (eval at <anonymous> (http://localhost:3500/vendor-bundle.js:3024:2), <anonymous>:8:10607)
  at e.Turbolinks.Controller.e.pushHistoryWithLocationAndRestorationIdentifier (eval at <anonymous> (http://localhost:3500/vendor-bundle.js:3024:2), <anonymous>:8:29305)
  at e.Turbolinks.Visit.e.changeHistory (eval at <anonymous> (http://localhost:3500/vendor-bundle.js:3024:2), <anonymous>:8:23290)
  at e.Turbolinks.BrowserAdapter.e.visitStarted (eval at <anonymous> (http://localhost:3500/vendor-bundle.js:3024:2), <anonymous>:8:8902)
...
(11 additional frame(s) were not displayed)

I think it is related to URL tracking (pushState) in new version of raven-js

[benvinegar]

Hi @szyablitsky – I'll have a fix for this shortly and we'll push a new release.

[benvinegar]

So, looking at the Turbolinks code, this is definitely being caused by a call to pushState – I'm guessing it's passing a non-string value for URL (since that's the only argument we inspect, and we do call .match on it). I've added a tighter check in #569.

[benvinegar]

Okay, better yet, we experimented a little more w/ pushState and saw that it coerces objects to strings. I think the integration test better captures the behavior now.