NLRI attribute: IP prefix parse error

[millionvoid]

In file: https://github.com/smartbgp/yabgp/blob/master/yabgp/message/attribute/nlri/ipv6_unicast.py
line 45: prefix_addr = str(netaddr.IPAddress(int(binascii.b2a_hex(nlri_data[1:offset]), 16))) \ + '/%s' % prefix_bit_len
It returns an address with extra leading zeros when prefix_bit_len is non-zero.
For example, an original IP prefix a::/96 will turn out to be 0:0: a::/96, because the parser netaddr.IPAddress does not know there are some zeros masked, it will put the missing zeros at the start of the IP address, causing this issue.
One possible correction is prefix_addr = str(netaddr.IPAddress(int(binascii.b2a_hex(nlri_data[1:offset]), 16)<<(full_length-prefix_bit_len))) \ + '/%s' % prefix_bit_len where full_length is 32 for ipv4 and 128 for ipv6.