The DBC loader reads in and remembers the endian flag of signals, calling it "byte_order". It then completely ignores it and makes no use of it at all.
This means it will not cope with big-endian messages - which is most of them where I work!

If the scaling factor of a signal uses a negative gain or, more commonly, offset, the load_dbc method's regex (like I said elsewhere, BAD IDEA) fails.

1. Index shift takes into account only message id and not the dlc.
2. Payload byte should be referenced like this: msg[index: index + 2] and not msg[index: index + 1].

The init method of USBtinException uses the "super" operator with no arguments, which throws an exception on at least some versions of Python. It should instead say:
def __init__(self, *args, **kwargs): super(USBtinException, self).__init__(*args, **kwargs)

Running PyUSBTin on Linux but with a DBC file written by a Windows program (Kvaser DB Editor) results in only the first frame (message) being loaded but all the defined signals being allocated to it.
I'd provide a patch but the "parser" uses regex's - which is a really bad idea nearly all the time - so it's rather hard to figure out how to fix it.

Thanks for writing this project. But when i using this API, i found out there something wrong with close_can_channel function, it trapped into a deadloop and never get channel closed.

The author of the filtermask module appears to have been writing in a language other than Python.
The class defines two __init__ methods in an attempt to create overloaded constructors like in C++ or Java. This does not work in Python - the second definition merely overwrites the first and the first is thrown away.
Once in the __init__ method, the first thing that happens is the 'registers' member is set to be an empty list of zero length. Then it tries to set the first four elements of this zero-length list, which don't exist because it's zero length!

In the getattribute method of CANMessage, around line 270 of canmessage.py:
if (factor, offset) == (1, 0): return value else: return (value - offset) * factor
This is wrong. The offset in a DBC file is defined in engineering units and should be added after the multiplication
return (value * factor ) + offset
The equivalent error also exists in the setattr method

See the Kvaser help page