python/cpython · error · ValueError

host/port and sock can not be specified at the same time

Error message

host/port and sock can not be specified at the same time

What it means

Raised by create_connection when both host/port and a pre-connected sock are supplied. asyncio connects either by resolving host+port itself or by adopting the caller's socket — doing both is contradictory (which address would it dial?), so it raises ValueError immediately.

Source

Thrown at Lib/asyncio/base_events.py:1122

        if ssl_handshake_timeout is not None and not ssl:
            raise ValueError(
                'ssl_handshake_timeout is only meaningful with ssl')

        if ssl_shutdown_timeout is not None and not ssl:
            raise ValueError(
                'ssl_shutdown_timeout is only meaningful with ssl')

        if sock is not None:
            _check_ssl_socket(sock)

        if happy_eyeballs_delay is not None and interleave is None:
            # If using happy eyeballs, default to interleave addresses by family
            interleave = 1

        if host is not None or port is not None:
            if sock is not None:
                raise ValueError(
                    'host/port and sock can not be specified at the same time')

            infos = await self._ensure_resolved(
                (host, port), family=family,
                type=socket.SOCK_STREAM, proto=proto, flags=flags, loop=self)
            if not infos:
                raise OSError('getaddrinfo() returned empty list')

            if local_addr is not None:
                laddr_infos = await self._ensure_resolved(
                    local_addr, family=family,
                    type=socket.SOCK_STREAM, proto=proto,
                    flags=flags, loop=self)
                if not laddr_infos:
                    raise OSError('getaddrinfo() returned empty list')
            else:
                laddr_infos = None

View on GitHub (pinned to bc6749cc3b)

Solutions

Pass either host/port OR sock, never both.
If you created the socket only to control the local address, close it and use local_addr= instead.
Audit call sites where host/port have non-None defaults but sock is also threaded through.

Example fix

// before
sock = socket.socket()
await loop.create_connection(proto, 'example.com', 443, sock=sock)

// after
sock.close()
transport, proto = await loop.create_connection(
    proto, 'example.com', 443, local_addr=('192.0.2.5', 0))

Defensive patterns

Strategy: validation

Validate before calling

if (host is not None or port is not None) and sock is not None:
    raise ConfigError('pass host/port or sock, not both')

Prevention

Model connection targets as a union type: either (host, port) or sock — never both in one code path.
Use local_addr= for source-address control instead of pre-binding sockets.
Audit default arguments like host='localhost' that survive into sock-mode calls.

When it happens

Trigger: loop.create_connection(proto, 'example.com', 443, sock=my_socket). Happens when code builds a socket for bind/local-address control and forgets that create_connection accepts local_addr for that, or when a fallback path accidentally leaves host set.

Common situations: Migrating from manual socket.socket + connect code to create_connection and passing both; refactors where host/port defaults ('localhost', 80) are always applied even in sock mode.

Related errors

AI-assisted analysis of python/cpython@bc6749cc3b (2026-08-14). Data as JSON: /api/errors/a1afe85557e1e9b6. Report an issue: GitHub.