PSeitz

Wouldn't randomizing stack offset relatively be easy, e.g. by using https://docs.rs/stackalloc/latest/stackalloc/ or https://crates.io/crates/alloca before a test? I regularly run into caching issues, which make my measurements hard to use.

It was requested here: https://www.reddit.com/r/rust/comments/kaj50o/lz4_flex_041_matches_now_the_cpp_lz4_reference/gfbjac7/?context=3 Number of up-votes suggests there is a clear opinion about this. In my use case this is the default, because I usually benchmark against different...

+1 Shouldn't be a problem to throw an error if the cost parameter is used with an unsupported algorithm?

Same here for version 31.0.1650.63

Can I check somehow why it doesn't?

> Can you provide a reproduction script? No, it happens only on my machine :)

The issue only occurs with the nightly compiler, stable works

> Does your nightly toolchain have rustfmt installed? Yes, e.g. this command completes `cargo +nightly fmt --all -- --check`

I have no overrides ``` rustup override list no overrides ``` Did you try to reproduce the issue? formatting does not work (files changed) ``` git clone https://github.com/quickwit-inc/quickwit cd quickwit;cd...

Would the `--async-stack-traces` option in nodejs12 solve this issue? http://thecodebarbarian.com/async-stack-traces-in-node-js-12.html