sockaddr_storage_to_addr is documented as being more lenient with the size than it is in practice

[khuey]

The documentation says

The len argument should be the number of bytes in the sockaddr_storage that are actually allocated and valid. It must be at least as large as all the useful parts of the structure.

This suggests that passing in a size larger than what is actually used (say, size_of::<sockaddr_storage>()) should be fine. But in the implementation, both AF_INET and AF_INET6 assert that the size passed is exactly size_of::<sockaddr_in>()/size_of::<sockaddr_in6>(). Ideally IMO the implementation would match the documented behavior.

[asomers]

Yes, I think you're probably right. Doing it as documented shouldn't lead to a memory leak. But do you have any examples of something that creates a sockaddr_in or sockaddr_in6 as a sockaddr_storage with too much space?

[khuey]

When doing strace-like things it's convenient to grab size_of::<sockaddr_storage>() bytes of data and then have the function just work.

[asomers]

That makes sense. Care to submit a patch?

[khuey]

When I have a chance, sure.