Receive multiple messages. This does not provide the socket addresses or the control messages. It does not use any of the input-scattering that recvmmsg offers, meaning that a single datagram is never split across noncontiguous memory. It supplies NULL for the timeout argument. All of the messages must have the same maximum size. All resulting byte arrays have been explicitly pinned. In addition to bytearrays corresponding to each datagram, this also provides the maximum msg_len that recvmmsg wrote back out. This is provided so that users of MSG_TRUNC can detect when bytes were dropped from the end of a message (although it does let the user figure out which message had bytes dropped).

Receive multiple messages. This is similar to uninterruptibleReceiveMultipleMessageA. However, it also provides the sockaddrs of the remote endpoints. These are written in contiguous memory to a bytearray of length max_num_msgs * expected_sockaddr_sz. The sockaddrs must all be expected to be of the same length. This function provides a sockaddr size check that is non-zero when any sockaddr had a length other than the expected length. This can be used to detect if the sockaddr array has one or more corrupt sockaddrs in it. All byte arrays returned by this function are pinned.

The values in the returned tuple are:

• Error-checking number for sockaddr size. Non-zero indicates that at least one sockaddr required a number of bytes other than the expected number.
• Pinned bytearray with all of the sockaddrs in it as a array of structures.
• The size of the largest message received. If MSG_TRUNC is used this lets the caller know if one or more messages were truncated.
• The message data of each message.

The sockaddrs bytearray and the unlifted array of messages are guaranteed to have the same number of elements.

If enabled, the user supplies an IP header in front of the user data. Valid only for SOCK_RAW sockets.

The MSG_DONTWAIT receive flag or send flag.

The MSG_TRUNC receive flag.

The MSG_CTRUNC receive flag.

The SOCK_CLOEXEC receive flag or send flag.

The SOCK_NONBLOCK receive flag or send flag.

Linux extends the type argument of socket to allow setting two socket flags on socket creation: SOCK_CLOEXEC and SOCK_NONBLOCK. It is advisable to set SOCK_CLOEXEC on when opening a socket on linux. For example, we may open a TCP Internet socket with:

uninterruptibleSocket internet (applySocketFlags closeOnExec stream) defaultProtocol

To additionally open the socket in nonblocking mode (e.g. with SOCK_NONBLOCK):

uninterruptibleSocket internet (applySocketFlags (closeOnExec <> nonblocking) stream) defaultProtocol

The size of a udphdr struct.

The size of an iphdr struct.

This poke function requires the user to pack the version and the internet header length (IHL), each 4 bits, into a single 8-bit word. The version should be in the most significant bits. This function will marshal the value appropriately depending on the platform's bit-endianness.