tests: unix: allow passfd's recvfd_payload() to fail

[kevans]

Summary

We're going to add some tests that exercise failure at one or both of
these points, so add an argument to allow it and return the len -- for
the control message failure, the caller will be verifying that we
received data but no files.

No functional change.

[markj]

I know I'm being annoying, but I really don't like naked bool function parameters. They're expedient but they make function calls harder to read and add complexity to the function itself. It would be cleaner to refactor recvfd_payload(): add a helper function which sets up the msghdr and calls recvmsg(), letting the caller provide a cmsg buffer and letting it decide what to do with the return value. Use that helper directly in your new test case.

Anyway, I don't object to this, I just try to avoid this pattern myself.

[kevans]

In D57425#1318871, @markj wrote:

I know I'm being annoying, but I really don't like naked bool function parameters.

Nah, that's completely fair criticism. Doing so also allows me to better confirrm that the cmsg was empty.

[kevans]

Superceded by D57544