regression: infallible residual could not convert error · Issue #86831 · rust-la... - JOYK Joy of Geek, Geek News, Link all geek

Copy link

Member

edited

For the sierra thing, it looks like this is a place where they unwittingly used never type on stable, and that's the core of why it changed. There was actually one place that hit this in the compiler (see the conversation in https://rust-lang.zulipchat.com/#narrow/stream/259160-t-lang.2Fproject-never-type/topic/Coercions.20of.20other.20uninhabited.20types/near/220030226) but that was a place using #![feature(never_type)] so when the previous crater run didn't hit it anywhere else I'd thought it wasn't a thing that would happen in stable :/

The code that broke is

let total_size_usize = group_size_usize
    .checked_mul(info.groups.len())
    .ok_or_else(host_memory_space_overlow)?;

where the function is

pub fn host_memory_space_overlow() -> ! {
    panic!("Memory address space overlow")
}

It actually works fine if it's changed to

.ok_or_else(|| host_memory_space_overlow())?;

I'll send them a PR. (Edit: zakarumych/sierra#3 -- which is now merged.)

Aside: This is one of those spots that I think could be nice with let-else, since it would emphasize that the function called diverges:

let Some(total_size_usize) = group_size_usize
    .checked_mul(info.groups.len())
else { host_memory_space_overlow() };

The dhall one is known, and I sent an (incorrect) PR Nadrieril/dhall-rust#213 for it, and it was fixed back in March with Nadrieril/dhall-rust@846c14f

So twelf should also be fixed as of bnjjj/twelf@3ad9fc5 (thought it's not clear to me what version that would be)

regression: infallible residual could not convert error · Issue #86831 · rust-la...