Any shim rand_core 4.3 crate around rand_core 0.5?

[burdges]

Is there going to be a shim rand_core 4.3 crate around rand_core 0.5?

It's only the Error type that broke in https://github.com/rust-random/rand/pull/800/files#diff-52e6ceb949d7fa383058092b498596b0R21 so doing a shim is non-trivial, and might break other things, right?

[burdges]

I'd missed that kind and msg are marked pub in the old error, so the only viable shim would be a wrapper type, which itself could appear elsewhere.

pub struct RngCore5As4<R: NewRngCore>(pub  R);

impl <R: NewRngCore> OldRngCore for RngCore5As4<R> {
    fn next_u32(&mut self) -> u32 { self.0.next_u32() }
    fn next_u64(&mut self) -> u64  { self.0.next_u64() }
    fn fill_bytes(&mut self, dest: &mut [u8])  { self.0.fill_bytes(dest) }
    fn try_fill_bytes(&mut self, dest: &mut [u8]) -> Result<(), OldError> {
        self.0.try_fill_bytes(dest),map_err(map_error)
    }
}

fn map_error(err: NewErrorNew) -> OldError {
    let kind = ErrorKind::Unavailable;
    let msg = "Unknown error from another rand_core version";
    #[cfg(not(feature="std"))]
    OldError::new(kind,msg)
    #[cfg(feature="std")]
    OldError::with_casue(kind,msg,err.take_inner())
 }

Anyone running on nightly could do this themselves with rust-lang/cargo#4953 and stable users need two crates for the reexport trick.

[dhardy]

No. See #819.

[burdges]

Right I figured out the Error types break. :) Individual crates could however use RngCore5As4 be rand_code 0.5 themselves while using rand_core 0.4 dependencies. I do not suggest including RngCore5As4 in rand_core of course, more some one off hack for anyone who needs it.