Open BoxyUwU opened 2 weeks ago
This was probably regressed w/ some f16
/f128
stuff cc @tgross35
Hm, when did this last work? I can't get it to compile successfully even on older versions https://rust.godbolt.org/z/ExcMY36bf
edit: slightly more like the published test in https://docs.rs/crate/puffpastry/0.1.0/source/src/activation.rs, same results https://rust.godbolt.org/z/crWhqYq6h
The puffpastry
crate's bound includes T: ValidNumber<T>
, which includes T: ... + From<f64> + ...
. Adding T: Copy + From<f64>
to the example makes it compile on stable https://rust.godbolt.org/z/EbKTjzM9v
Ah, then #123830 also should have removed this line https://github.com/rust-lang/rust/pull/123830/files#diff-154dff23a2c2f4fb97915229fd9c27d73293940250088020bfd2074094186ce4L173, which was added in https://github.com/rust-lang/rust/pull/122470.
Was this crate the only regression of this kind? If that is the case, maybe we would be okay leaving it based on discussion at https://github.com/rust-lang/rust/issues/123831.
@tgross35 should the changes in #123830 / ~#123831~ #124728 be in relnotes
? I'm thinking if there are changes here useful to be announced. Thanks
@apiraino I think the only things maybe relevant for release notes would be conversions that involve a stable type, which could have an effect on inference. These would be:
From<f16> for f64
added in #122470, removed in #12380, added back in #12473 (1.80 current nightly). Already had a crater run, no regressionsFrom<f32> for f128
, added in #122470 (1.79 current beta). No known regressionsFrom<f64> for f128
, added in #122470 (1.79 current beta). RegressionsItem 3 is the subject of this issue, we'll need a decision about whether to accept the regression or remove the impl.
The remaining conversions from #122470 shouldn't affect anything without nightly so I don't think they need release notes.