Bytes/repeated value not parsed, RangeError

[aman-tiwari]

Ran into this whilst trying to parse the TensorProtos produced by Tensorflow.
For an example, Tensorflow returns the following protobuf (tensor_shape.dim.size is an int64) for a 1x1 tensor of type float that has value 3.0:
dtype: DT_FLOAT tensor_shape { dim { size: 1 } dim { size: 1 } } float_val: 3.0
with binary encoding as:
[8, 1, 18, 8, 18, 2, 8, 1, 18, 2, 8, 1, 42, 4, 0, 0, 64, 64], in hex:
'0x8 0x1 0x12 0x8 0x12 0x2 0x8 0x1 0x12 0x2 0x8 0x1 0x2a 0x4 0x0 0x0 0x40 0x40
When parsing this with protobufjs 6.0.0, I get RangeError: index out of range: 18 + 1 > 18

However, when parsing with protobufjs 5, the message is parsed correctly.

If I modify TensorProto to be

message TensorProto2 {
  DataType dtype = 1;
  TensorShapeProto tensor_shape = 2;
  int32 version_number = 3;

  bytes tensor_content = 4;

  //repeated float float_val = 5 [packed = true];
}

The message is parsed correctly (but missing the floatVal, as it's been commented out)

[dcodeIO]

Seems that there is an issue with packed repeated fields with float / double types.

Can you try to inject the following right before loading your proto?

protobuf.types.packed.double = 1;
protobuf.types.packed.float = 5;

Please let me know if this fixes your issue.

[aman-tiwari]

That seems to fix it!