TypeError: Invalid value used as weak map key during destroy()

[Turbo87]

I've just been reported the following error running v2.10.0-beta.1:

TypeError: Invalid value used as weak map key
  at peekMeta(bower_components/ember/ember.prod.js:17979:0)
  at destroy(bower_components/ember/ember.prod.js:17632:0)
  at deleteMeta(bower_components/ember/ember.prod.js:18017:0)
  at destroy(bower_components/ember/ember.prod.js:21348:0)
  at apply(bower_components/ember/ember.prod.js:33980:0)
  at invoke(bower_components/ember/ember.prod.js:335:0)
  at flush(bower_components/ember/ember.prod.js:396:0)
  at flush(bower_components/ember/ember.prod.js:520:0)
  at end(bower_components/ember/ember.prod.js:590:0)
  at apply(bower_components/ember/ember.prod.js:704:0)
  at join(bower_components/ember/ember.prod.js:735:0)
  at arg1(bower_components/ember/ember.prod.js:10514:0)
  at _trigger(bower_components/ember/ember.prod.js:625:0)
  at end(bower_components/ember/ember.prod.js:601:0)
  at apply(bower_components/ember/ember.prod.js:704:0)
  at apply(bower_components/ember/ember.prod.js:735:0)
  at join(bower_components/ember/ember.prod.js:20303:0)
  at callback(bower_components/ember/ember.prod.js:38386:0)
  at flaggedInstrument(bower_components/ember/ember.prod.js:16560:0)
  at handleEvent(bower_components/ember/ember.prod.js:38385:0)
  at handleEvent(bower_components/ember/ember.prod.js:37281:0)
  at _bubbleEvent(bower_components/ember/ember.prod.js:37839:0)
  at apply(bower_components/ember/ember.prod.js:37761:0)
  at elem(bower_components/jquery/jquery.js:5095:0)
  at elem(bower_components/jquery/jquery.js:4766:0)

Stack trace without source maps

TypeError: Invalid value used as weak map key
  at e.peekMeta.O(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:9:2391)
  at s.destroy(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:9:287)
  at Object.p [as deleteMeta](/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:8:31574)
  at Object.u [as destroy](/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:9:24575)
  at p.PrototypeMixin.n.Mixin.create.s._scheduledDestroy(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:12:12474)
  at a.invokeWithOnError(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:12300)
  at Object.a.flush(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:12756)
  at Object.l.flush(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:13526)
  at Object.h.end(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:13871)
  at Object.h.run(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:14943)
  at Object.h.join(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:15067)
  at Array.p(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:6:23101)
  at Object.h._trigger(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:14130)
  at Object.h.end(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:14006)
  at Object.h.run(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:14943)
  at Object.h.join(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:4:15067)
  at Function.u.join(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:9:19837)
  at ? (/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:13:13972)
  at Object.e.flaggedInstrument.d [as flaggedInstrument](/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:8:24792)
  at Object.t.assign.handleEvent(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:13:13896)
  at e.default.n.Mixin.create.l.handleEvent(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:13:4511)
  at e.default.r.Object.extend._bubbleEvent(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:13:7102)
  at HTMLAnchorElement.<anonymous>(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:13:6336)
  at HTMLBodyElement.ce.event.dispatch(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:3:1695)
  at HTMLBodyElement.v.handle(/assets/vendor-6d4f9fce143a0a7254965d0f2c73d4b2.js:2:30415)

This was reported from a production build with source maps, so don't trust the stack trace too much... according to sentry.io the browser that hit this was Chrome Mobile 38 on Android 6.0.

@backspace reported a similar issue in #dev-ember (feel free to paste your call stack here as well)

[Serabe]

Is this with an empty app?

[backspace]

Yes, travis-web has a similar error in Chrome 53 and 56 (canary). Most things don’t render, presumably because of it?

Uncaught TypeError: Invalid value used as weak map key
    at WeakMap.set (native)
    at WeakMap_setMeta (http://localhost:4200/assets/vendor.js:34224:19)
    at Object.meta (http://localhost:4200/assets/vendor.js:34325:5)
    at Object.watchKey (http://localhost:4200/assets/vendor.js:37418:37)
    at new RootPropertyReference (http://localhost:4200/assets/vendor.js:29345:21)
    at RootReference.get (http://localhost:4200/assets/vendor.js:29251:44)
    at Object.referenceFromParts (http://localhost:4200/assets/vendor.js:59176:35)
    at CompiledPositionalArgs.evaluate (http://localhost:4200/assets/vendor.js:61120:43)
    at CompiledArgs.evaluate (http://localhost:4200/assets/vendor.js:60593:57)
    at VM.evaluateArgs (http://localhost:4200/assets/vendor.js:68278:54)

Let me know if I can provide debugging help somehow. I’m not sure where to look since I don’t see any application code in the stacktrace.

[Turbo87]

@Serabe nope, this was from one of my production apps

[Serabe]

Can any of you try to provide a reproduction? I'll try to take a look at travis-web later this week otherwise

[Turbo87]

unfortunately I haven't been able to reproduce this yet

[backspace]

While I get the error with every load on a local travis-web server, I can’t get it to happen in a deployed production build, sorry. I do get other mysterious errors that you can see here, but I don’t know if they’re related.

[rwjblue]

Seems related to #14499, but I'd need to dig in a bit more.

[backspace]

I added a conditional breakpoint to look more into it, but I’m too unfamiliar to have much insight. Here’s a screenshot of one of the levels of the stacktrace:

If a deeper entry would be more helpful, let me know!

[krisselden]

The common error message here does not make these all the same issue. We need stack traces in each case like @Turbo87 and @Serabe added, those are separate issues.

@rwjblue #14499 is about node 0.12 in which get() with a non object key fails. This does seem to match the stack that @Turbo87 gave and is fixed in #14499 but is not the same issue as @Serabe

@Turbo87 can you confirm if your error is when running in an old v8?

[Turbo87]

@krisselden I haven't been able to trigger it with the current Chrome versions that are installed on my MacBook and mobile devices. The only user that was able to trigger it was reportedly using Chrome 38 as stated above.

[backspace]

If I understand you correctly, @krisselden, are you suggesting that I open another issue? Is there more detail I can give apart from pasting the stacktraces and including the browser versions?

[krisselden]

@backspace yes, the issue that @Turbo87 reported is fixed with #14499 please don't open issues with just a TypeError. It should say TypeError during some operation, the error with old v8 and destroy() has been fixed. The error with ref get() bailing out is a different.

[backspace]

I’ll open a separate issue, sorry for my lack of understand. Can you see that it’s difficult for someone not familiar with the internals to know what portion of the stacktrace is important and should be extracted? You’re saying the destroy was the important part for this issue but that get is the important one from travis-web. Without this discussion here, I don’t see how I could know that 😞

[krisselden]

@backspace I'm sorry.