[mbgrydeland]

XNU has honoured MSG_NOSIGNAL on send since it adopted the POSIX.1-2008 flag, and libc defines it for Apple targets.

What does this PR do

Adds MSG_NOSIGNAL as a valid bit flag on struct MsgFlags on apple_targets.

Checklist:

• I have read CONTRIBUTING.md
• I have written necessary tests and rustdoc comments
• A change log has been added if this PR modifies nix's API

[2570165831]

I tested this on real hardware, in case it helps review.

macOS 27.2, arm64. A small C program closes the peer, then calls send() in a child with SIGPIPE at its default disposition:

socket	flags	result
AF_UNIX stream pair	none	child killed by SIGPIPE
AF_UNIX stream pair	MSG_NOSIGNAL	send returns EPIPE, no signal
TCP over loopback	none	child killed by SIGPIPE
TCP over loopback	MSG_NOSIGNAL	send returns EPIPE, no signal

So the flag is honoured there, and the value matches libc (0x80000).

On "since when": in the published XNU sources the flag first appears in xnu-7195 (macOS 11 / iOS 14), both in bsd/sys/socket.h and in the SIGPIPE check in bsd/kern/uipc_syscalls.c. It is absent from xnu-6153 (macOS 10.15). I only read the sources for that; I have not run anything on a system older than 27. It may be worth a line in the doc comment, since x86_64-apple-darwin binaries can still target older systems, where an unknown flag bit would presumably be ignored and the signal still raised.

Happy to share the test program if useful.

[mbgrydeland]

Force-pushed rebase on master

[mbgrydeland]

Force pushed rebase on master

[mbgrydeland]

Had to trigger a rerun to get past an apparent flaky test, that as far as I can see is not related to the changes from this PR. CI is all green now.