resolver.py/ThreadResolver: address[3] does not exist

[dilyanpalauzov]

Describe the bug

As outlined at pimutils/vdirsyncer#916, I use python 3.8 with aiohttp 3.8 branch (v3.7.4.post0-209-g22f8be99) . In aiohttp/resolver.py:ThreadedResolver the line

if family == socket.AF_INET6 and address[3]: # type: ignore[misc]

fails, as address has only two elements and address[3] is out of range.

The values of adress are:

AddressFamily.AF_INET ('142.250.185.238', 443)
AddressFamily.AF_INET6 (10, b'\x01\xbb\x00\x00\x00\x00*\x00\x14P@\x01\x08\x13')

The system is not connected to IPv6, no routes, no iPv6 addressese etc, but the kernel understands IPv6. I fixed it locally with:

diff --git a/aiohttp/resolver.py b/aiohttp/resolver.py
index 2161c8fa..4e57b319 100644
--- a/aiohttp/resolver.py
+++ b/aiohttp/resolver.py
@@ -38,16 +38,7 @@ class ThreadedResolver(AbstractResolver):
 
         hosts = []
         for family, _, proto, _, address in infos:
-            if family == socket.AF_INET6 and address[3]:  # type: ignore[misc]
-                # This is essential for link-local IPv6 addresses.
-                # LL IPv6 is a VERY rare case. Strictly speaking, we should use
-                # getnameinfo() unconditionally, but performance makes sense.
-                host, _port = socket.getnameinfo(
-                    address, socket.NI_NUMERICHOST | socket.NI_NUMERICSERV
-                )
-                port = int(_port)
-            else:
-                host, port = address[:2]
+            host, port = address[:2]
             hosts.append(
                 {
                     "hostname": hostname,

To Reproduce

See above.

Expected behavior

Shall not read address[3], when address[3] does not exist.

Logs/tracebacks

Not applicable.

Python Version

$ python --version

Python 3.8.8+

aiohttp Version

$ python -m pip show aiohttp

3.8.0a0

multidict Version

$ python -m pip show multidict
5.1.0

yarl Version

$ python -m pip show yarl

1.6.3

OS

Linux from Scratch, python 3.8.8+

Related component

Client

Additional context

No response

Code of Conduct

• I agree to follow the aio-libs Code of Conduct

[Hanaasagi]

resolve use the getaddrinfo to get the socket infomation.

aiohttp/aiohttp/resolver.py

Lines 27 to 39 in 60f5727

	async def resolve(
	self, hostname: str, port: int = 0, family: int = socket.AF_INET
	) -> List[Dict[str, Any]]:
	infos = await self._loop.getaddrinfo(
	hostname,
	port,
	type=socket.SOCK_STREAM,
	family=family,
	flags=socket.AI_ADDRCONFIG,
	)
	hosts = []
	for family, _, proto, _, address in infos:

From the Python doc, the sockaddr part of getaddrinfo function' return value is a 4-tuple if the family is AF_INET6.

The function returns a list of 5-tuples with the following structure:

(family, type, proto, canonname, sockaddr)

In these tuples, family, type, proto are all integers and are meant to be passed to the socket() function. canonname will be a string representing the canonical name of the host if AI_CANONNAME is part of the flags argument; else canonname will be empty. sockaddr is a tuple describing a socket address, whose format depends on the returned family (a (address, port) 2-tuple for AF_INET, a (address, port, flow info, scope id) 4-tuple for AF_INET6), and is meant to be passed to the socket.connect() method.

The behavior is strange, not as stated in the doc. 😕

Ref: https://docs.python.org/3.8/library/socket.html#socket.getaddrinfo

[webknjaz]

The behavior is strange, not as stated in the doc.

IIRC, it returns the structure from the underlying kernel syscall. FWIW I think this has been fixed by #5176 on master, 3.8 and 3.7.

[webknjaz]

Duplicate of #5156. Please post any other observations or/and a reproducer there.