mirror of
https://github.com/ElementsProject/lightning.git
synced 2024-11-20 10:39:49 +01:00
5142dc81f6
We weren't writing out the length of a nested subtype's dynamicarraylenght, now we do. The trick is to iterate through the fields on a subtype (since the length field is added separately) and to also iterate down through the otherfield values as we 'descend'
232 lines
11 KiB
Python
232 lines
11 KiB
Python
#! /usr/bin/python3
|
|
from pyln.proto.message import MessageNamespace, Message
|
|
import pytest
|
|
import io
|
|
|
|
|
|
def test_fundamental():
|
|
ns = MessageNamespace()
|
|
ns.load_csv(['msgtype,test,1',
|
|
'msgdata,test,test_byte,byte,',
|
|
'msgdata,test,test_u16,u16,',
|
|
'msgdata,test,test_u32,u32,',
|
|
'msgdata,test,test_u64,u64,',
|
|
'msgdata,test,test_chain_hash,chain_hash,',
|
|
'msgdata,test,test_channel_id,channel_id,',
|
|
'msgdata,test,test_sha256,sha256,',
|
|
'msgdata,test,test_signature,signature,',
|
|
'msgdata,test,test_point,point,',
|
|
'msgdata,test,test_short_channel_id,short_channel_id,',
|
|
])
|
|
|
|
mstr = """test
|
|
test_byte=255
|
|
test_u16=65535
|
|
test_u32=4294967295
|
|
test_u64=18446744073709551615
|
|
test_chain_hash=0102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f20
|
|
test_channel_id=0102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f20
|
|
test_sha256=0102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f20
|
|
test_signature=0102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f202122232425262728292a2b2c2d2e2f303132333435363738393a3b3c3d3e3f40
|
|
test_point=0201030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f2021
|
|
test_short_channel_id=1x2x3"""
|
|
m = Message.from_str(ns, mstr)
|
|
|
|
# Same (ignoring whitespace differences)
|
|
assert m.to_str().split() == mstr.split()
|
|
|
|
|
|
def test_static_array():
|
|
ns = MessageNamespace()
|
|
ns.load_csv(['msgtype,test1,1',
|
|
'msgdata,test1,test_arr,byte,4'])
|
|
ns.load_csv(['msgtype,test2,2',
|
|
'msgdata,test2,test_arr,short_channel_id,4'])
|
|
|
|
for test in [["test1 test_arr=00010203", bytes([0, 1] + [0, 1, 2, 3])],
|
|
["test2 test_arr=[0x1x2,4x5x6,7x8x9,10x11x12]",
|
|
bytes([0, 2]
|
|
+ [0, 0, 0, 0, 0, 1, 0, 2]
|
|
+ [0, 0, 4, 0, 0, 5, 0, 6]
|
|
+ [0, 0, 7, 0, 0, 8, 0, 9]
|
|
+ [0, 0, 10, 0, 0, 11, 0, 12])]]:
|
|
m = Message.from_str(ns, test[0])
|
|
assert m.to_str() == test[0]
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue() == test[1]
|
|
assert Message.read(ns, io.BytesIO(test[1])).to_str() == test[0]
|
|
|
|
|
|
def test_subtype():
|
|
ns = MessageNamespace()
|
|
ns.load_csv(['msgtype,test1,1',
|
|
'msgdata,test1,test_sub,channel_update_timestamps,4',
|
|
'subtype,channel_update_timestamps',
|
|
'subtypedata,'
|
|
+ 'channel_update_timestamps,timestamp_node_id_1,u32,',
|
|
'subtypedata,'
|
|
+ 'channel_update_timestamps,timestamp_node_id_2,u32,'])
|
|
|
|
for test in [["test1 test_sub=["
|
|
"{timestamp_node_id_1=1,timestamp_node_id_2=2}"
|
|
",{timestamp_node_id_1=3,timestamp_node_id_2=4}"
|
|
",{timestamp_node_id_1=5,timestamp_node_id_2=6}"
|
|
",{timestamp_node_id_1=7,timestamp_node_id_2=8}]",
|
|
bytes([0, 1]
|
|
+ [0, 0, 0, 1, 0, 0, 0, 2]
|
|
+ [0, 0, 0, 3, 0, 0, 0, 4]
|
|
+ [0, 0, 0, 5, 0, 0, 0, 6]
|
|
+ [0, 0, 0, 7, 0, 0, 0, 8])]]:
|
|
m = Message.from_str(ns, test[0])
|
|
assert m.to_str() == test[0]
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue() == test[1]
|
|
assert Message.read(ns, io.BytesIO(test[1])).to_str() == test[0]
|
|
|
|
# Test missing field logic.
|
|
m = Message.from_str(ns, "test1", incomplete_ok=True)
|
|
assert m.missing_fields()
|
|
|
|
|
|
def test_subtype_array():
|
|
ns = MessageNamespace()
|
|
ns.load_csv(['msgtype,tx_signatures,1',
|
|
'msgdata,tx_signatures,num_witnesses,u16,',
|
|
'msgdata,tx_signatures,witness_stack,witness_stack,num_witnesses',
|
|
'subtype,witness_stack',
|
|
'subtypedata,witness_stack,num_input_witness,u16,',
|
|
'subtypedata,witness_stack,witness_element,witness_element,num_input_witness',
|
|
'subtype,witness_element',
|
|
'subtypedata,witness_element,len,u16,',
|
|
'subtypedata,witness_element,witness,byte,len'])
|
|
|
|
for test in [["tx_signatures witness_stack="
|
|
"[{witness_element=[{witness=3045022100ac0fdee3e157f50be3214288cb7f11b03ce33e13b39dadccfcdb1a174fd3729a02200b69b286ac9f0fc5c51f9f04ae5a9827ac11d384cc203a0eaddff37e8d15c1ac01},{witness=02d6a3c2d0cf7904ab6af54d7c959435a452b24a63194e1c4e7c337d3ebbb3017b}]}]",
|
|
bytes.fromhex('00010001000200483045022100ac0fdee3e157f50be3214288cb7f11b03ce33e13b39dadccfcdb1a174fd3729a02200b69b286ac9f0fc5c51f9f04ae5a9827ac11d384cc203a0eaddff37e8d15c1ac01002102d6a3c2d0cf7904ab6af54d7c959435a452b24a63194e1c4e7c337d3ebbb3017b')]]:
|
|
m = Message.from_str(ns, test[0])
|
|
assert m.to_str() == test[0]
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue().hex() == test[1].hex()
|
|
assert Message.read(ns, io.BytesIO(test[1])).to_str() == test[0]
|
|
|
|
|
|
def test_tlv():
|
|
ns = MessageNamespace()
|
|
ns.load_csv(['msgtype,test1,1',
|
|
'msgdata,test1,tlvs,test_tlvstream,',
|
|
'tlvtype,test_tlvstream,tlv1,1',
|
|
'tlvdata,test_tlvstream,tlv1,field1,byte,4',
|
|
'tlvdata,test_tlvstream,tlv1,field2,u32,',
|
|
'tlvtype,test_tlvstream,tlv2,255',
|
|
'tlvdata,test_tlvstream,tlv2,field3,byte,...'])
|
|
|
|
for test in [["test1 tlvs={tlv1={field1=01020304,field2=5}}",
|
|
bytes([0, 1]
|
|
+ [1, 8, 1, 2, 3, 4, 0, 0, 0, 5])],
|
|
["test1 tlvs={tlv1={field1=01020304,field2=5},tlv2={field3=01020304}}",
|
|
bytes([0, 1]
|
|
+ [1, 8, 1, 2, 3, 4, 0, 0, 0, 5]
|
|
+ [253, 0, 255, 4, 1, 2, 3, 4])],
|
|
["test1 tlvs={tlv1={field1=01020304,field2=5},4=010203,tlv2={field3=01020304}}",
|
|
bytes([0, 1]
|
|
+ [1, 8, 1, 2, 3, 4, 0, 0, 0, 5]
|
|
+ [4, 3, 1, 2, 3]
|
|
+ [253, 0, 255, 4, 1, 2, 3, 4])]]:
|
|
m = Message.from_str(ns, test[0])
|
|
assert m.to_str() == test[0]
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue() == test[1]
|
|
assert Message.read(ns, io.BytesIO(test[1])).to_str() == test[0]
|
|
|
|
# Ordering test (turns into canonical ordering)
|
|
m = Message.from_str(ns, 'test1 tlvs={tlv1={field1=01020304,field2=5},tlv2={field3=01020304},4=010203}')
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue() == bytes([0, 1]
|
|
+ [1, 8, 1, 2, 3, 4, 0, 0, 0, 5]
|
|
+ [4, 3, 1, 2, 3]
|
|
+ [253, 0, 255, 4, 1, 2, 3, 4])
|
|
|
|
|
|
def test_tlv_complex():
|
|
# A real example from the spec.
|
|
ns = MessageNamespace(["msgtype,reply_channel_range,264,gossip_queries",
|
|
"msgdata,reply_channel_range,chain_hash,chain_hash,",
|
|
"msgdata,reply_channel_range,first_blocknum,u32,",
|
|
"msgdata,reply_channel_range,number_of_blocks,u32,",
|
|
"msgdata,reply_channel_range,full_information,byte,",
|
|
"msgdata,reply_channel_range,len,u16,",
|
|
"msgdata,reply_channel_range,encoded_short_ids,byte,len",
|
|
"msgdata,reply_channel_range,tlvs,reply_channel_range_tlvs,",
|
|
"tlvtype,reply_channel_range_tlvs,timestamps_tlv,1",
|
|
"tlvdata,reply_channel_range_tlvs,timestamps_tlv,encoding_type,byte,",
|
|
"tlvdata,reply_channel_range_tlvs,timestamps_tlv,encoded_timestamps,byte,...",
|
|
"tlvtype,reply_channel_range_tlvs,checksums_tlv,3",
|
|
"tlvdata,reply_channel_range_tlvs,checksums_tlv,checksums,channel_update_checksums,...",
|
|
"subtype,channel_update_timestamps",
|
|
"subtypedata,channel_update_timestamps,timestamp_node_id_1,u32,",
|
|
"subtypedata,channel_update_timestamps,timestamp_node_id_2,u32,",
|
|
"subtype,channel_update_checksums",
|
|
"subtypedata,channel_update_checksums,checksum_node_id_1,u32,",
|
|
"subtypedata,channel_update_checksums,checksum_node_id_2,u32,"])
|
|
|
|
binmsg = bytes.fromhex('010806226e46111a0b59caaf126043eb5bbf28c34f3a5e332a1fc7b2b73cf188910f000000670000000701001100000067000001000000006d000001000003101112fa300000000022d7a4a79bece840')
|
|
msg = Message.read(ns, io.BytesIO(binmsg))
|
|
buf = io.BytesIO()
|
|
msg.write(buf)
|
|
assert buf.getvalue() == binmsg
|
|
|
|
|
|
def test_message_constructor():
|
|
ns = MessageNamespace(['msgtype,test1,1',
|
|
'msgdata,test1,tlvs,test_tlvstream,',
|
|
'tlvtype,test_tlvstream,tlv1,1',
|
|
'tlvdata,test_tlvstream,tlv1,field1,byte,4',
|
|
'tlvdata,test_tlvstream,tlv1,field2,u32,',
|
|
'tlvtype,test_tlvstream,tlv2,255',
|
|
'tlvdata,test_tlvstream,tlv2,field3,byte,...'])
|
|
|
|
m = Message(ns.get_msgtype('test1'),
|
|
tlvs='{tlv1={field1=01020304,field2=5}'
|
|
',tlv2={field3=01020304},4=010203}')
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue() == bytes([0, 1]
|
|
+ [1, 8, 1, 2, 3, 4, 0, 0, 0, 5]
|
|
+ [4, 3, 1, 2, 3]
|
|
+ [253, 0, 255, 4, 1, 2, 3, 4])
|
|
|
|
|
|
def test_dynamic_array():
|
|
"""Test that dynamic array types enforce matching lengths"""
|
|
ns = MessageNamespace(['msgtype,test1,1',
|
|
'msgdata,test1,count,u16,',
|
|
'msgdata,test1,arr1,byte,count',
|
|
'msgdata,test1,arr2,u32,count'])
|
|
|
|
# This one is fine.
|
|
m = Message(ns.get_msgtype('test1'),
|
|
arr1='01020304', arr2='[1,2,3,4]')
|
|
buf = io.BytesIO()
|
|
m.write(buf)
|
|
assert buf.getvalue() == bytes([0, 1]
|
|
+ [0, 4]
|
|
+ [1, 2, 3, 4]
|
|
+ [0, 0, 0, 1,
|
|
0, 0, 0, 2,
|
|
0, 0, 0, 3,
|
|
0, 0, 0, 4])
|
|
|
|
# These ones are not
|
|
with pytest.raises(ValueError, match='Inconsistent length.*count'):
|
|
m = Message(ns.get_msgtype('test1'),
|
|
arr1='01020304', arr2='[1,2,3]')
|
|
|
|
with pytest.raises(ValueError, match='Inconsistent length.*count'):
|
|
m = Message(ns.get_msgtype('test1'),
|
|
arr1='01020304', arr2='[1,2,3,4,5]')
|