#[derive_more::Error] refers to std::

[StackOverflowExcept1on]

I did two simple libraries here: https://github.com/StackOverflowExcept1on/derive-more-issue/

• lib-derive-more
• lib-snafu

Why doesn't the library provide impl core::error::Error for Error?

#![no_std]

#[derive(Debug, Eq, PartialEq, derive_more::Display, derive_more::Error)]
pub enum Error {
    #[display(fmt = "Unable to write result")]
    WriteResult,
}

pub type Result<T, E = Error> = core::result::Result<T, E>;

pub fn foo(a: i32, b: i32) -> Result<i32> {
    a.checked_add(b).ok_or_else(|| Error::WriteResult)
}

error[E0433]: failed to resolve: could not find `std` in the list of imported crates
 --> src/lib.rs:3:54
  |
3 | #[derive(Debug, Eq, PartialEq, derive_more::Display, derive_more::Error)]
  |                                                      ^^^^^^^^^^^^^^^^^^ could not find `std` in the list of imported crates

[JelteF]

The Error trait is not yet in core on stable. Currently it is only in std. So if you are compiling for a no_std environment you currently cannot use Error. It seems there's an unstable feature though to enable Error in core, so maybe we should enable that when the std crate feature is not enabled: rust-lang/rust#103765

[StackOverflowExcept1on]

Could you do an auto-detect using #[cfg]?

[StackOverflowExcept1on]

In the snafu crate, they did a trick with conditional compilation and snafu::Error that turned out to be compatible with std::error::Error
https://github.com/shepmaster/snafu/blob/main/src/lib.rs#L354

[tyranron]

Seems reasonable to have in under feature flag.

[JelteF]

@StackOverflowExcept1on I created a PR that adds support for the Error derive in no_std environments in #268. It also includes some tests, but could you double check that it solves your problem?

[StackOverflowExcept1on]

Looks good, but for example snafu has a polyfill for core::error::Error that can be used at stable rust in no_std.

[tyranron]

@StackOverflowExcept1on is having a polyfill critical for you? We don't want deviate much from what std/core have.

[StackOverflowExcept1on]

@tyranron this is not critical, but it would be useful for stable crates

[JelteF]

I looked into using the same polyfill approach as snafu. But I don't that fits well with the ideas behind this crate, especially since the Error trait is on the path to being stabilized (although that will probably take some more time). Every Error derive crate creating its own Error polyfill is also really bad for the general ecosystem, since none of them will be able to work together. It might be better to actually have add a crate feature which uses the snafu polyfill.

[StackOverflowExcept1on]

btw rust-lang/rust#125951