Hadrien G.

I suspect that ~/.rustup might be a symlink, which would explain why caching it doesn't work while caching its content does. But indeed, I ended up using a similar strategy...

I believe that [as currently written](https://github.com/withoutboats/failure/blob/master/failure-1.X/src/backtrace/internal.rs), your code can already run is_backtrace_enabled more than once. This is unrelated to acquire/release ordering and remains true even under sequentially consistent execution: -...

@alexcrichton As an aside, I would respectfully disagree with your decision of using SeqCst as a default. Personally, when I meet SeqCst ordering in someone's code, what I get out...

Besides the crates.io website, it would also be nice if the `cargo` CLI could (perhaps optionally) integrate some security warning features, such as what `cargo-audit` provides today. As a crate...

This will bump rustc requirements quite significantly, but hopefully for the last time in a while. Will let the current rustc release age a bit before merging this.

@dependabot rebase

Blocked on EDM rework and Stepper rework.

Just in case you haven't seen this blog post pass by, here's [a nice overview/comparison](https://blog.yoshuawuyts.com/error-handling-survey/) of the various available Rust error handling crates.

> Most of the discussion so far is about the general concept and implementability of an "atomic per byte memcpy", but I'd also like some feedback on the API design...

It's probably best to experiment at the library level first, and that in turn requires const generics if you're not in the mood for fighting typenum.