Manish Goregaokar

I think that would be fine! Simplifies our life, and the typedness doesn't get us _much_.

A simple optimization is that we can store empty strings in DecimalSymbolStrs for "∞" and "NaN", which means the only cost is the two index bytes that get added to...

@younies I'm going to work on adding data for these strings, and potentially adding formatter types

Some discussion between @sffc and I: All locales except en-US-posix use ∞. Some locales translate NaN to "not a number" in that language. My argument is that displaying NaN as...

More discussion about the signature of the format function: We shouldn't have a trait for this because `fdf.format(decimal.into())` is common. We can do one of: - `format(WithNan)` and have `.into()`...

@sffc not really in favor of reusing the type as a writeable. @jedel1043 yes, it would

I suspect it's the use of std::thread_local in the recursion counter. I think folks should just `#[cfg_attr(feature = "std", derive(Arbitrary))]` or something (or `#[cfg(feature = "arbitrary)] extern crate std;`). I...

Recursion guard was added in https://github.com/rust-fuzz/arbitrary/pull/109

r? clippy rerolling again, I'm wrapping up vacation and will be on airplanes for a while

@tarcieri the vulns are present in both crates independently: the derive macro doesn't enforce `C, packed`, and the manual impls in the zerovec crate also don't have `C, packed`.