Jason Rhinelander

> Isn't Rust (:rocket:) secure? Yes, but many of the CVEs are in crates that use **unsafe** :fearful: :broken_heart: :no_bicycles: blocks. Realistically, use of an unsafe block should be a...

The main issue here in how return-value-policies propagate (or rather, *don't* propagate) through `std::function` bindings. You can see this in `functional.h`: the `value = [func](...)` lambda simply calls the `py::function`...

This seems a surprisingly hard problem. The main issue is that the C++ `Clone` and the Python `Clone` do fundamentally different things: the former returns a new C++ instance, the...

Agreed that it would be nice. (Retitling).

I don't think there's any particularly good reason why it's not there--other than that no one has proposed/implemented it yet. (As a side note, for tuples in particular, you're better...

Yup. Meanwhile there are other places that already *do* accept arbitrary arguments--the arguments to `make_tuple`, for example. So you're right that's it's not a stretch (and indeed expected) to get...

The documentation in this PR is quite misleading: > :param seed: [:class:`bytes`] Random 32-byte value for generating Ed25519 private key > :param secret_key: [:class:`bytes`] A previously generated 64-byte Ed25519 private...

> Neither libsodium nor pynacl mention in the docs that this is possible. You have to read the source code to know that. I didn't know that when I started...

The referenced other project of the same name appears to use a random compiled binary embedded in the source code without any source code provided to set the dmg icon,...

I should also point out that some versions of `send()` *do* have a no-flags default, so it seems weird that the common version taking a `message_t` does not: ```C++ detail::send_result_t...