[Bug binutils/31972] New: strip cause SIGKILL in macOS

[lapla.cogito at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=31972

            Bug ID: 31972
           Summary: strip cause SIGKILL in macOS
           Product: binutils
           Version: 2.42
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: binutils
          Assignee: unassigned at sourceware dot org
          Reporter: lapla.cogito at gmail dot com
  Target Milestone: ---

I would like to report a problem with using strip on macOS (M1 MAX, but I
believe this problem also occurs on x86_64).

I ran into this problem while writing code in Rust. Therefore, I will show you
a procedure that allows you to reproduce a similar event:

````
$ cargo new foo && cd foo
$ cargo run --release
```

This produces the following output:

```
   Compiling foo v0.1.0 (/path_to_binary)
    Finished `release` profile [optimized] target(s) in 0.72s
     Running `target/release/foo`.
zsh: killed cargo run --release
```

Noting that simply `cargo build` runs fine, I thought strip might be the cause.
Then I realized that the output changes when I use build scripts in Rust:

```
$ cat << EOF > build.rs                                                         
fn main() {
    println!("cargo::rerun-if-changed=src/main.rs");
}
EOF
```

After creating build.rs, `cargo run --release` would result in the following
output:

```
error: failed to run custom build command for `foo v0.1.0 (/path_to_binary)`

Caused by:
  process didn't exit successfully: `/path_to_build_script/build-script-build`
(signal: 9, SIGKILL: kill)
```

Both cases (with and without build.rs) don't occur if the standard macOS binary
(/usr/bin/strip) is used as strip instead of binutils from Homebrew. For this
reason, I suspect that this is a bug in the binutils strip.

Moreover, from looking at the source code, it appears that the binutils strip
is not Mach-O Version 2 compliant, and I believe the two problems mentioned
above are most likely caused by this.

I see two main ways to deal with this problem:

- Reject Mach-O Version 2 when it is given
- Modify the strip to work correctly even if Mach-O Version 2 is given

-- 
You are receiving this mail because:
You are on the CC list for the bug.