SIGKILL 9 on aarch64-darwin after 1.83 upgrade

[simonzkl]

Something between 9051466..3497106 is causing builds onaarch64-darwin getting SIGKILL-ed. Anything that requires a build script fails with:

       > error: failed to run custom build command for `proc-macro2 v1.0.87`
       >
       > Caused by:
       >   process didn't exit successfully: `/private/tmp/nix-build-cargo-llvm-cov-0.6.14.drv-0/source/target/release/build/proc-macro2-cc0987ae2e60fb30/build-script-build` (signal: 9, SIGKILL: kill)
       > warning: build failed, waiting for other jobs to finish...
       > error: failed to run custom build command for `serde v1.0.210`
       >
       > Caused by:
       >   process didn't exit successfully: `/private/tmp/nix-build-cargo-llvm-cov-0.6.14.drv-0/source/target/release/build/serde-a79c191d45e158fe/build-script-build` (signal: 9, SIGKILL: kill)
       > error: failed to run custom build command for `rustix v0.38.37`
       >
       > Caused by:
       >   process didn't exit successfully: `/private/tmp/nix-build-cargo-llvm-cov-0.6.14.drv-0/source/target/release/build/rustix-0d75d4ace3e7e5a5/build-script-build` (signal: 9, SIGKILL: kill)
       > error: failed to run custom build command for `anyhow v1.0.89`
       >
       > Caused by:
       >   process didn't exit successfully: `/private/tmp/nix-build-cargo-llvm-cov-0.6.14.drv-0/source/target/release/build/anyhow-a8439441284bc7bc/build-script-build` (signal: 9, SIGKILL: kill)
       > error: failed to run custom build command for `libc v0.2.159`
       >
       > Caused by:
       >   process didn't exit successfully: `/private/tmp/nix-build-cargo-llvm-cov-0.6.14.drv-0/source/target/release/build/libc-df0d4e5857f68e29/build-script-build` (signal: 9, SIGKILL: kill)
       For full logs, run 'nix log /nix/store/vvwpm8hsrgd7ww0pw6lcf62g7ap0fxls-cargo-llvm-cov-0.6.14.drv'.

Downgrading rust-overlay (keeping the same nixpkgs revision) solves the issue.

Possible candidates #196, #192, #190?

[simonzkl]

Narrowed it down to 3497106. The manifest update to Rust 1.83.

[simonzkl]

Possibly related to NixOS/nixpkgs#299606

[simonzkl]

I also tested a cross build with --sandbox and I noticed that 1.83 is now trying to use /usr/bin/strip which is unavailable of course. Reverting to 1.82 again fixes the issue.

Fun rabbithole:

• Fix up setting strip = true in Cargo.toml makes build scripts fail in… rust-lang/rust#130781
• Hard-coded strip path breaks Linux -> Darwin builds rust-lang/rust#131206

[simonzkl]

Horrible binary patch to force it to use strip from PATH:

toolchain.overrideAttrs (prev: {
  buildCommand = ''
    ${prev.buildCommand}

    source ${pkgs.darwin.signingUtils}
    driver=$(echo $out/lib/librustc_driver-*.dylib)
    ${pkgs.bbe}/bin/bbe -e 's#/usr/bin/strip#strip-xxxxxxxx#' $driver > $driver.new
    mv $driver.new $driver
    signIfRequired $driver

    echo "#!${pkgs.bash}/bin/bash" >> $out/bin/strip-xxxxxxxx
    echo 'exec strip "$@"' >> $out/bin/strip-xxxxxxxx
    chmod +x $out/bin/strip-xxxxxxxx
  '';
});

This fixes the issue, though don't actually use this.

[saiintbrisson]

Also related: NixOS/nixpkgs#362388.

[simonzkl]

Not sure what we can really do about this. Some options that come to mind:

• Wait for 1.84 (upstream switched to llvm-strip)
• Use libredirect to redirect /usr/bin/strip
• Binary patch it (a bit difficult to do because of Rust's non-null terminated strings)
• Disable stripping (CARGO_PROFILE_RELEASE_STRIP=false). The stdenv strips the final binary anyway

The least hacky approach might be to suggest users to disable stripping and wait for 1.84.

[saiintbrisson]

The least hacky approach might be to suggest users to disable stripping and wait for 1.84.

This is what I've done. Pinned Rust to 1.82 for now. Not a big deal.

[simonzkl]

No longer an issue in 1.84.