Daniel Prilik

chiming in to mention I'd also be very interested in seeing this implemented. until Rust gets support for language-level safe transmute, being able to hook into the various third-party safe...

I agree that taking a direct dependency on either of the crates wouldn't be the best idea - the safe transmute ecosystem continues to be in-flux, so best not to...

> * There might be some trouble with opaque types as they are exposed as byte-arrays by bindgen but could contain non-pod values. For example, if a type with a...

Just ran into this myself. ```rust type ThisIsAVeryLongTypeNameThatWillCauseWrappingToKickIn = usize; #[derive(serde_derive::Serialize)] pub struct Foo( #[serde(rename = "ThisIsAVeryLongTypeNameThatWillCauseWrappingToKickIn")] pub ThisIsAVeryLongTypeNameThatWillCauseWrappingToKickIn, ); ``` Is this something that'll get fixed via https://github.com/rust-lang/rustfmt/pull/5529 or...

...though, after toying around with the feature for a bit, I will mention that it results in quite a lot of re-complation when switching between fuzzzing and fuzzing-repro flows. Maybe...

...yeah, the constant recompilation really isn't great. I would really prefer reconsidering the `--cfg fuzzer_repro` approach, and swapping it out with an env-var instead. I don't think the performance difference...

Yeah, I would've been very surprised if the generated code ended up any worse than that. The lack of any/all panics is especially good to see! > 3\. It cannot...

I agree with @jstarks - when I proposed the bare `#[bitfield]` syntax, I didn't fully reason though the footgun implicit `align(1)` would have. Don't get me wrong - having `align(1)`...

A few thoughts: - I totally agree that you should pick one repr backend and stick with it - having two different codegen backends seems like a bad idea from...

> But yes, enforcing an explicit size (via `bytes`, `ty`, or the other syntax) might be a good idea. Indeed. Allowing bare `#[bitfield]` with align=1, bytes=1 semantics seem like the...