Published in 0.1.4 & yanked old versions.

FYI: I believe the new logic is still not entirely correct, at least not for arbitrary "header" and "value" types.

In particular, it fails if size_of::<Header>() is not a power of two: https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=b2ba0406e00379e7986511ab3815a133

This probably isn't a significant problem right now because Header == [usize; 2], but IMO this property should either be enforced by a test, or the logic fixed.

@TimNN Thanks for the heads up!

We don't want to define a general-purpose offset_of (we should use memoffset crate for that use case) and it would be sufficient to adopt a way that works correctly for the specific headers we use.

That said, we would at least need to add a comment about the fact that the current implementation also relies on the actual properties of the header.

Ideally I would like to use the same way that std::sync::Arc uses (e.g., padding_needed_for), but that is unstable...
I will look into whether it is possible to emulate it in stable as part of #140.

EDIT: #141 changed it to use the same way as std::sync::Arc.