Brian Smith

I removed this from the 0.17.15 release milestone since it isn't a regression from 0.17.14.

> I removed this from the 0.17.15 release milestone since it isn't a regression from 0.17.14. If I read the log correctly, the first release with the following commit was...

Also, to clarify for others, until very recently, aarch64-pc-windows was BUILT in CI (on an x86-64 Windows runner) but the tests were not run. Only recently did we start running...

> A workaround, avoiding running the tests on aarch64-pc-windows-msvc 1.66 release, was put into CI in PR [#2538](https://github.com/briansmith/ring/pull/2538). PR #2550 undoes this workaround, replacing it with a minimally-scoped one that...

See also discussion in https://github.com/typst/comemo/issues/3. Note that I'm suggesting adding support for the weaker SipHash 1-3 with 128-bit output, not the default stronger SipHash 2-4 with the default 64-bit output.

> Does rust currently uses 64-bit TypeIds ? * Rust originally used the original SipHash 2-4 with 64-bit output for `TypeId`. I believe somebody demonstrated a collision so it switched...

> Only transmutation by choice? But no bad side consequences that could creep up on people? The `Any` and `TypeId` APIs are easily reachable from anything that "downcasts", e.g. https://doc.rust-lang.org/stable/std/io/struct.Error.html#method.downcast....

See also https://github.com/rust-lang/rust/issues/41215 which seems to imply that it is used in the incremental compilation cache. See also the use in the mold linker: https://github.com/rui314/mold/commit/fa8e95a289f911c0c47409d5848c993cb50c8862. My understanding is that when...

How hard would it be to create a program that takes a source code representation of a `$ty` and produces the bytes that are fed into SipHasher? I am not...

Another motivating example: https://github.com/rust-lang/cargo/issues/6529#issuecomment-2181473376