Correction: I've just received the same error for an Android device. After checking again I found two other earlier reports of this error from Android devices. So, it is not just iOS as I had incorrectly stated earlier, but no PCs or Macs have reported this error.

Is this the line you are saying the error is firing on?

self._scratchBuffer = self.ctx.createBuffer(1, 1, 22050);

Yes. Apparently, in rare situations on iOS and (even rarer) on Android, the value of ctx is null on that line.

Are you 100% it is referencing that line? A few lines above it already checks if self.ctx is set, so I'm not sure how it could possibly be null at that line.

Yes, I'm sure. That is the only line in the source that calls ctx.createBuffer.

Yes, there is a null check performed on ctx a half dozen lines before the error is thrown. But, below that, immediately before the error is thrown, there appears to be an opportunity to call self.unload which can set ctx to null.

After placing some log traces in the Howler code I now see the cause of the problem: The call to self.unload (line 292, v2.0.2) sets self.ctx to null. unload then calls setupAudioContext which normally assigns a new audio context to self.ctx, but in rare cases setupAudioContext will throw an error from native code due to a lack of audio resources. That means self.ctx is left set to null when unload returns, which in turn, causes self._scratchBuffer = self.ctx.createBuffer(1, 1, 22050) (line 297, v2.0.2) to throw an unhandled error.

It appears the problem can be fixed by adding a null check after the call to self.unload (line 292, v2.0.2):

if (!self._mobileUnloaded && self.ctx.sampleRate !== 44100) {
  self._mobileUnloaded = true;
  self.unload();

  if (!self.ctx) {
    return;
  }
}

BTW, The bug is still there :(.

Yes, fixed with #1363, waiting for a merge

Hello, we're currently experiencing this error in some very specific situations on iOS Safari. As a temp fix in addition to the MR above, I'm setting self.noAudio=true and returning, because otherwise Howler attempts to call ctx.createGain in Howler.load

I am getting lots of null errors from inside the unlock function, different to the error above.
null is not an object (evaluating 'e.ctx.createBufferSource')

somehow ctx is valid when the touch handlers are added but not later when executed.

Maybe an up here? My pr has been opened for half a year now..

We're also seeing this issue pop up sporadically, would be awesome #1363 could get merged or another similar fix.

@goldfire and/or any other contributors able to weigh in here?

While it might not be the same 'root cause', we are also seeing the following, which feels very similar to this issue:

TypeError null is not an object (evaluating 'e.ctx.suspend') 
    ../node_modules/howler/dist/howler.js:498:17 

I haven't done any deeper tracing to identify the specifics of it at this stage, as it's only affecting a small number of Safari users at this stage.

While it may not be related to this, when googling for the error I stumbled across the following, which sounds similar (though probably not as useful as the explicit tracing done in #654 (comment)):

• TypeError: null is not an object (evaluating 'o.ctx.createGain') #1021

@0xdevalias I also see the suspend issue. Always Safari.