Alan Somers
Alan Somers
This is a step backwards for ergonomics, by exposing the raw int values. If there's a specific protocol value that's not available now, then you should add it like how...
Yeah, actually, I do like that version better. What I don't want to do is to abandon type safety in the name of flexibility. However, instead of `impl IntoSockProtocol for...
What are you trying to do anyway? Change an interface's mac address? There is a good reason why the inner `sockaddr_ll` cannot be exposed; changing its `sll_family` field could lead...
I don't like the proposed constructors because they're potentially too limited. On Linux and OSX, the libc structure has 7 fields. On the BSDs, it has between 8 and 10....
> Is there anything preventing you from providing a safe constructor that just takes in the `sockaddr_ll` by move or a similar `From` implementation? Yes. We can't provide a safe...
This PR will break backwards compatibility for callers who pass a `libc::off_t` to Nix, right? What about making these functions generic, like this, and always calling the 64-bit syscall? ```rust...
> > This PR will break backwards compatibility for callers who pass a `libc::off_t` to Nix, right? What about making these functions generic, like this, and always calling the 64-bit...
Yes, I wasn't thinking about signedness. `i64` is definitely better than `u64`. The important questions are: * Are there any platforms where `off_t` is 32-bits and *don't* also have 64-bit...
That's sensible. Patches welcome.
Care to submit a PR?