1400275 - Assertion failure: givenProto == reflectorProto (How are we supposed to change the proto now?), at dom/bindings/BindingUtils.cpp:3740

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect, P1)

Description

[Bob Clary [:bc] (inactive)]

Attachment: Fedora Debug Log

1. https://app.wespeke.com/connect

2. Assertion failure: givenProto == reflectorProto (How are we supposed to change the proto now?), at /builds/worker/workspace/build/src/dom/bindings/BindingUtils.cpp:3740
#01: mozilla::dom::binding_detail::DoGetOrCreateDOMReflector<mozilla::dom::RTCPeerConnection, (mozilla::dom::binding_detail::GetOrCreateReflectorWrapBehavior)0u> [dom/bindings/BindingUtils.h:1100]
#02: mozilla::dom::GetOrCreateDOMReflectorHelper<RefPtr<mozilla::dom::RTCPeerConnection>, true>::GetOrCreate [dom/bindings/BindingUtils.h:1863]
#03: mozilla::dom::RTCPeerConnectionBinding::_constructor [dom/bindings/BindingUtils.h:1885]
#04: js::CallJSNative [js/src/jscntxtinlines.h:294]
#05: js::CallJSNativeConstructor [js/src/jscntxtinlines.h:326]
#06: InternalConstruct [js/src/vm/Interpreter.cpp:599]
#07: js::Construct [js/src/vm/Interpreter.cpp:637]
#08: js::SpreadCallOperation [js/src/vm/Interpreter.cpp:4904]
#09: Interpret [js/src/vm/Interpreter.cpp:3036]

Windows, Linux; Beta/56, Nightly/57

Comment 1

[Andrew Overholt [:overholt]]

This feels bad (so P1) and I'm guessing you're the best person to fix, Boris, but please correct me on both accounts if needed.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Oh, this is exciting.  The page is subclassing RTCPeerConnection.  Specifically, in https://static.opentok.com/v2/js/opentok.min.js we have this bit (somewhat de-minified):

  canUseES6Class = function(e) {
    var t = e.RTCPeerConnection||e.webkitRTCPeerConnection||e.mozRTCPeerConnection,
        n = extendParentClass(t);
    if (n)
      try {
        if (n.prototype.foo="bar","bar"=== new n(null).foo)
          return !0
      } catch(e) {}
    return!1
  }

where extendParentClass is:

  extendParentClass = function extendParentClass(ParentClass){try{var ChildClass=eval('\n      "use strict"; // Chrome<49 requires strict mode\n      class ChildClass extends ParentClass {\n        constructor(...args) {\n          super(...args);\n          try {\n            this.__proto__ = ChildClass.prototype;\n          } catch (e) {}\n        }\n      };\n      ChildClass; // Ensure the final statement is returned in FF\n    ');if(ChildClass&&ChildClass.prototype instanceof ParentClass)return ChildClass}catch(e){}return null}

(not going to even try to prettyprint that).

Anyway, the upshot is that they have an RTCPeerConnection subclass.  This interface is JS-implemented, and for JS-implemented interfaces we create the reflector in the constructor, like so: http://searchfox.org/mozilla-central/rev/184f0c7888dd6abb32235377693b7d1fc0b75ac1/dom/bindings/Codegen.py#15359-15363

The good news is that we control all this code, so it's easy to fix.  ;)

Comment 3

[Boris Zbarsky [:bzbarsky]]

Attachment: Ensure we create reflectors for JS-implemented webidl interfaces with the right prototype when those interfaces are subclassed

MozReview-Commit-ID: 5yYD92NVVFQ

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/06ee47ac7e4d
Ensure we create reflectors for JS-implemented webidl interfaces with the right prototype when those interfaces are subclassed.  r=qdot

Comment 5

[Sebastian Hengst [:aryx] (needinfo me if it's about an intermittent or backout)]

https://hg.mozilla.org/mozilla-central/rev/06ee47ac7e4d