Null java peer callback

[tobrun]

This PR adds null checking to the java peer before calling into a NativeMapView callback.
With the Mapbox#snapshot api I was able to reproduce:

JNI DETECTED ERROR IN APPLICATION: can't call void com.mapbox.mapboxsdk.maps.NativeMapView.onSnapshotReady(android.graphics.Bitmap) on null object

When requesting a snapshot as part of Activity#onPause:

  @Override
  protected void onPause() {
    super.onPause();
    mapboxMap.snapshot(new MapboxMap.SnapshotReadyCallback() {
      @Override
      public void onSnapshotReady(Bitmap snapshot) {
        Timber.e("snapshot ready");
      }
    });
    mapView.onPause();
  }

This could potentially explain crashes as #11268.

[ivovandongen]

Could you make this just if (javaPeer) or if (javaPeer != nulllptr) if you want to make it more explicit

[ivovandongen]

Ditto

[ivovandongen]

@tobrun I'm wondering why the callback is actually fired when the javaPeer is gone. If the java peer is gone, that would be because onDestroy is called on the NativeMapView right? Which would mean that the renderer is also already reset and no callbacks should follow.

Do you have a way to reproduce this with any reliability?

[tobrun]

@ivovandongen

Do you have a way to reproduce this with any reliability?

Yes, with the code shown in OP I was able to reproduce this on demand.
I will add a test to catch any regressions around this.

[tobrun]

When writing a test for this issue, I'm occasionally running into the exception mentioned by the user:

Abort message: 'java_vm_ext.cc:534] JNI DETECTED ERROR IN APPLICATION: use of deleted global reference 0x2'

This is a more logical exception as the one I tried addressing in OP. Been reading up on this issue though don't have an immediate fix for it.

[michaelfomin]

Michael from Wahoo Fitness here, this issue is causing a lot of crashes for us. Just wondering if this fix was scheduled soon?

[tobrun]

@ivovandongen I revisit this PR with a different approach, it introduces a paused state to the map renderer and it resolves the crashes (including #11358 (comment)).

@michaelfomin getting this fixed is a priority, the fix will be backported to a 5.5.x release.

[ivovandongen]

Since onPause and onResume are called on the UI thread and render on the GL thread, I think we should use std::atomic<bool> here for the paused flag.

[tobrun]

@ivovandongen thank you for flagging, fixed up the code to std::atomic<bool>