After #106856, CARGO_CFG_TARGET_HAS_ATOMIC{,_LOAD_STORE} become empty strings

[taiki-e]

Code

I ran cargo build -vv with this code:

# Cargo.toml
[package]
name = "repro"
version = "0.1.0"
edition = "2021"

// build.rs
fn main() {
    dbg!(std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC").ok());
    dbg!(std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC_LOAD_STORE").ok());
    dbg!(std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC_EQUAL_ALIGNMENT").ok());
}

I expected to see this happen:

[repro 0.1.0] [build.rs:2] std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC").ok() = Some(
[repro 0.1.0]     "128,16,32,64,8,ptr",
[repro 0.1.0] )
[repro 0.1.0] [build.rs:3] std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC_LOAD_STORE").ok() = Some(
[repro 0.1.0]     "128,16,32,64,8,ptr",
[repro 0.1.0] )
[repro 0.1.0] [build.rs:4] std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC_EQUAL_ALIGNMENT").ok() = Some(
[repro 0.1.0]     "128,16,32,64,8,ptr",
[repro 0.1.0] )

Instead, this happened:

[repro 0.1.0] [build.rs:2] std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC").ok() = Some(
[repro 0.1.0]     "",
[repro 0.1.0] )
[repro 0.1.0] [build.rs:3] std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC_LOAD_STORE").ok() = Some(
[repro 0.1.0]     "",
[repro 0.1.0] )
[repro 0.1.0] [build.rs:4] std::env::var("CARGO_CFG_TARGET_HAS_ATOMIC_EQUAL_ALIGNMENT").ok() = Some(
[repro 0.1.0]     "128,16,32,64,8,ptr",
[repro 0.1.0] )

It seems that cfg(target_has_atomic{,_load_store}) without value that is added in #106856 is overriding cfg(target_has_atomic{,_load_store}) with value.

This environment variable still seems to be rarely used in the ecosystem, but profiler_builtins uses it.

rust/library/profiler_builtins/build.rs

Line 67 in 231bcd1

if env::var_os("CARGO_CFG_TARGET_HAS_ATOMIC")

Mentioning @vadorovsky who authored #106856

Version it worked on

It most recently worked on: nightly-2023-01-27

$ rustc +nightly-2023-01-27 --print cfg | grep target_has_atomic
target_has_atomic="128"
target_has_atomic="16"
target_has_atomic="32"
target_has_atomic="64"
target_has_atomic="8"
target_has_atomic="ptr"
target_has_atomic_equal_alignment="128"
target_has_atomic_equal_alignment="16"
target_has_atomic_equal_alignment="32"
target_has_atomic_equal_alignment="64"
target_has_atomic_equal_alignment="8"
target_has_atomic_equal_alignment="ptr"
target_has_atomic_load_store="128"
target_has_atomic_load_store="16"
target_has_atomic_load_store="32"
target_has_atomic_load_store="64"
target_has_atomic_load_store="8"
target_has_atomic_load_store="ptr"

Version with regression

nightly-2023-01-28
rustc --version --verbose:

rustc 1.69.0-nightly (ef982929c 2023-01-27)
binary: rustc
commit-hash: ef982929c0b653436a6ea6892a2a839fba7c8b57
commit-date: 2023-01-27
host: aarch64-apple-darwin
release: 1.69.0-nightly
LLVM version: 15.0.7

$ rustc +nightly-2023-01-28 --print cfg | grep target_has_atomic
target_has_atomic
target_has_atomic="128"
target_has_atomic="16"
target_has_atomic="32"
target_has_atomic="64"
target_has_atomic="8"
target_has_atomic="ptr"
target_has_atomic_equal_alignment="128"
target_has_atomic_equal_alignment="16"
target_has_atomic_equal_alignment="32"
target_has_atomic_equal_alignment="64"
target_has_atomic_equal_alignment="8"
target_has_atomic_equal_alignment="ptr"
target_has_atomic_load_store
target_has_atomic_load_store="128"
target_has_atomic_load_store="16"
target_has_atomic_load_store="32"
target_has_atomic_load_store="64"
target_has_atomic_load_store="8"
target_has_atomic_load_store="ptr"

@rustbot modify labels: +regression-from-stable-to-nightly -regression-untriaged

[vadorovsky]

Thanks for the report! I will try to fix that soon.

[vadorovsky]

@taiki-e Sorry for following up so late. I finally managed to debug it properly and I think it's an issue which should be fixed in Cargo rather than rustc. I created the issue and I will make a PR today:

rust-lang/cargo#11789

Please let me know if you agree with that approach. I think we should support a mix of configs with and without value - that works nicely for differentiating between "check for atomix for type X" and "check for atomics in general" in rustc.

[taiki-e]

Thanks. That approach looks reasonable to me.

[taiki-e]

Fixed in rust-lang/cargo#11790. Thanks @vadorovsky.