boreal-cli 1.0.0

#![allow(missing_docs)]
#![allow(unused_results)]
#![allow(unused_crate_dependencies)]
#![allow(clippy::pedantic)]

use std::io::{BufRead, Write};
use std::path::Path;
use std::{fs, io::BufReader};

use assert_cmd::Command;
use predicates::prelude::*;
use tempfile::{NamedTempFile, TempDir};

fn test_file(contents: &[u8]) -> NamedTempFile {
    let mut file = NamedTempFile::new().unwrap();
    file.write_all(contents).unwrap();
    file
}

enum CmdKind {
    #[cfg(feature = "serialize")]
    Save,
    #[cfg(feature = "serialize")]
    Load,
    ListModules,
}

fn boreal_cmd() -> Command {
    Command::cargo_bin("boreal").unwrap()
}

fn test_scan<F>(options: &[&str], rules_files: &[&Path], input: &Path, test: F)
where
    F: Fn(&mut Command),
{
    // First test the scan subcommand
    let mut cmd = boreal_cmd();
    cmd.arg("scan");
    for opt in options {
        cmd.arg(opt);
    }
    for file in rules_files {
        cmd.arg("-f").arg(file);
    }
    cmd.arg(input);
    test(&mut cmd);

    // Then test the yr subcommand
    let mut cmd = boreal_cmd();
    cmd.arg("yr");
    for opt in options {
        cmd.arg(opt);
    }
    for file in rules_files {
        cmd.arg(file);
    }
    cmd.arg(input);
    test(&mut cmd);
}

fn test_cmd<F>(kind: CmdKind, test: F)
where
    F: Fn(&mut Command),
{
    match kind {
        #[cfg(feature = "serialize")]
        CmdKind::Save => {
            test(boreal_cmd().arg("save"));
        }
        #[cfg(feature = "serialize")]
        CmdKind::Load => {
            test(boreal_cmd().arg("yr").arg("-C"));
            test(boreal_cmd().arg("load"));
        }
        CmdKind::ListModules => {
            test(boreal_cmd().arg("yr").arg("-M"));
            test(boreal_cmd().arg("list-modules"));
        }
    }
}

#[test]
fn test_no_arguments() {
    // Some arguments are required to do anything
    boreal_cmd().arg("yr").assert().failure();
    boreal_cmd().arg("scan").assert().failure();
}

#[test]
fn test_invalid_path() {
    // Invalid path to rule
    test_scan(
        &[],
        &[Path::new("do_not_exist")],
        Path::new("input"),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains(
                    "Cannot read rules file do_not_exist: ",
                ))
                .failure();
        },
    );

    // Invalid path to input
    let rule_file = test_file(b"");
    test_scan(&[], &[rule_file.path()], Path::new("bad_input"), |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(predicate::str::contains("Cannot scan bad_input"))
            .failure();
    });
}

#[test]
fn test_scan_file() {
    let rule_file = test_file(
        br#"
rule my_rule {
    strings:
        $a = "abc"
    condition:
        any of them
}"#,
    );

    let input = test_file(b"aaa");
    // Not matching
    test_scan(
        &["--profile", "memory"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert().stdout("").stderr("").success();
        },
    );

    let input = test_file(b"zeabce");
    // Matching
    test_scan(&[], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("my_rule {}\n", input.path().display()))
            .stderr("")
            .success();
    });
}

#[test]
// Need super user to run on linux and macos
#[cfg_attr(unix, ignore)]
fn test_scan_process() {
    let rule_file = test_file(
        br#"
rule process_scan {
    strings:
        $a = "PAYLOAD_ON_STACK"
    condition:
        $a
}"#,
    );

    let proc = BinHelper::run("stack");
    let pid = proc.pid();

    // Not matching
    test_scan(
        &[],
        &[rule_file.path()],
        Path::new(&pid.to_string()),
        |cmd| {
            cmd.assert()
                .stdout(format!("process_scan {}\n", pid))
                .stderr("")
                .success();
        },
    );
}

#[test]
#[cfg(target_os = "linux")]
fn test_scan_process_not_found() {
    let rule_file = test_file(b"rule process_scan { condition: true }");

    // First, find an unused PID. Lets take a very big number, and have
    // some retry code until we find a proper one.
    let mut pid = 999_999_999;
    while Path::new("proc").join(pid.to_string()).exists() {
        pid += 1;
    }

    // Not matching
    test_scan(
        &[],
        &[rule_file.path()],
        Path::new(&pid.to_string()),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(format!("Cannot scan {}: unknown process\n", pid))
                .failure();
        },
    );
}

#[test]
fn test_scan_file_with_process_name() {
    // Test that scanning a file with an integer name works, and does not
    // attempt to scan a process.
    let rule_file = test_file(
        br#"
rule is_file {
    strings:
        $a = "buzo"
    condition:
        any of them
}"#,
    );

    let temp = TempDir::new().unwrap();
    let file = temp.path().join("1");
    fs::write(file, "gabuzomeu").unwrap();
    // Matching
    test_scan(&[], &[rule_file.path()], Path::new("1"), |cmd| {
        cmd.current_dir(temp.path())
            .assert()
            .stdout("is_file 1\n")
            .stderr("")
            .success();
    });
}

#[test]
fn test_rule_error() {
    let rule_file = test_file(
        br#"rule a {
        strings: $a = /[z-a]/
        condition: $a
    }"#,
    );

    let input = test_file(b"");
    test_scan(&[], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(
                predicate::str::contains("error")
                    .and(predicate::str::contains("invalid regex class range")),
            )
            .failure();
    });
}

#[test]
fn test_rule_warning() {
    let rule_file = test_file(
        br#"rule rule_with_warning {
            condition:
                "a"
        }
        "#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    // Warning is OK and rule is eval'ed
    test_scan(&[], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("rule_with_warning {path}\n"))
            .stderr(
                predicate::str::contains("warning").and(predicate::str::contains(
                    "implicit cast from a bytes value to a boolean",
                )),
            )
            .success();
    });

    // Warning is considered an error
    test_scan(
        &["--fail-on-warnings"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(
                    predicate::str::contains("warning").and(predicate::str::contains(
                        "implicit cast from a bytes value to a boolean",
                    )),
                )
                .failure();
        },
    );

    // Ignore warnings
    test_scan(&["-w"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("rule_with_warning {path}\n"))
            .stderr("")
            .success();
    });
}

#[test]
fn test_rule_include() {
    let temp = TempDir::new().unwrap();

    let rule_b = temp.path().join("b.yar");
    fs::write(
        rule_b,
        r#"
rule included {
    strings:
        $ = "abc"
    condition:
        any of them
}
"#,
    )
    .unwrap();

    let rule_a = temp.path().join("a.yar");
    fs::write(
        &rule_a,
        r#"
include "b.yar"

rule includer {
    strings:
        $ = "xyz"
    condition:
        included and any of them
}
"#,
    )
    .unwrap();

    // Match on the included
    let input = test_file(b"abc");
    test_scan(&[], &[&rule_a], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("included {}\n", input.path().display()))
            .success();
    });

    // Match on both
    let input = test_file(b"xyz abc");
    test_scan(&[], &[&rule_a], input.path(), |cmd| {
        cmd.assert()
            .stdout(
                predicate::str::contains(format!("included {}", input.path().display())).and(
                    predicate::str::contains(format!("includer {}", input.path().display())),
                ),
            )
            .success();
    });
}

#[test]
fn test_rule_include_error() {
    let temp = TempDir::new().unwrap();

    let rule_b = temp.path().join("b.yar");
    fs::write(
        &rule_b,
        r#"
rule included {
    strings:
        $ = /[z-a]/
    condition:
        any of them
}
"#,
    )
    .unwrap();

    let rule_a = temp.path().join("a.yar");
    fs::write(&rule_a, r#"include "b.yar""#).unwrap();

    // Match on the included
    let input = test_file(b"");
    test_scan(&[], &[&rule_a], input.path(), |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(
                // Contains the path to the faulty file
                // Need canonicalize as the path is canonicalized in boreal, and this causes
                // differences on Windows.
                predicate::str::contains(rule_b.canonicalize().unwrap().display().to_string())
                    .and(
                        // And the proper string that caused the error
                        predicate::str::contains("z-a"),
                    )
                    .and(
                        // And the error message
                        predicate::str::contains("invalid regex class range"),
                    ),
            )
            .failure();
    });
}

#[test]
#[cfg(unix)]
fn test_rule_dir() {
    use std::os::unix::fs::symlink;

    let rule_file = test_file(b"rule a { condition: true }");

    let temp = TempDir::new().unwrap();
    let temp2 = TempDir::new().unwrap();

    // a and b in temp
    let file_a = temp.path().join("a");
    fs::write(&file_a, "").unwrap();
    let file_b = temp.path().join("b");
    fs::write(&file_b, "").unwrap();

    // c in temp/subdir
    let subdir = temp.path().join("subdir");
    fs::create_dir(&subdir).unwrap();
    let file_c = subdir.join("c");
    fs::write(&file_c, "").unwrap();

    // d.yar and e.yar in temp2
    let file2_d = temp2.path().join("d");
    fs::write(&file2_d, "").unwrap();
    let file2_e = temp2.path().join("e");
    fs::write(&file2_e, "").unwrap();

    // symlink temp/d.yar to temp2/d.yar
    let file_d = temp.path().join("d");
    symlink(&file2_d, &file_d).unwrap();
    // symlink temp/subdir/e.yar to temp2/e.yar
    let file_e = subdir.join("e");
    symlink(&file2_e, &file_e).unwrap();

    let match_str = |input: &Path| predicate::str::contains(format!("a {}", input.display()));

    // Non recursive
    // Add some threads to instrument the code
    test_scan(
        &["--threads", "20"],
        &[rule_file.path()],
        temp.path(),
        |cmd| {
            cmd.assert()
                .stdout(
                    // match on "a", "b" and "d"
                    match_str(&file_a)
                        .and(match_str(&file_b))
                        .and(match_str(&file_c).not())
                        .and(match_str(&file_d))
                        .and(match_str(&file_e).not()),
                )
                .stderr("")
                .success();
        },
    );

    // Non recursive and non follow symlinks
    test_scan(&["-N"], &[rule_file.path()], temp.path(), |cmd| {
        cmd.assert()
            .stdout(
                // match on "a" and "b"
                match_str(&file_a)
                    .and(match_str(&file_b))
                    .and(match_str(&file_c).not())
                    .and(match_str(&file_d).not())
                    .and(match_str(&file_e).not()),
            )
            .stderr("")
            .success();
    });

    // Recursive
    test_scan(&["-r"], &[rule_file.path()], temp.path(), |cmd| {
        cmd.assert()
            .stdout(
                // match on all
                match_str(&file_a)
                    .and(match_str(&file_b))
                    .and(match_str(&file_c))
                    .and(match_str(&file_d))
                    .and(match_str(&file_e)),
            )
            .stderr("")
            .success();
    });

    // recursive and non follow symlinks
    // Add some threads to instrument the code
    test_scan(
        &["--recursive", "--no-follow-symlinks", "-p", "2"],
        &[rule_file.path()],
        temp.path(),
        |cmd| {
            cmd.assert()
                .stdout(
                    // match on "a", "b" and "c"
                    match_str(&file_a)
                        .and(match_str(&file_b))
                        .and(match_str(&file_c))
                        .and(match_str(&file_d).not())
                        .and(match_str(&file_e).not()),
                )
                .stderr("")
                .success();
        },
    );
}

#[test]
fn test_skip_larger() {
    let rule_file = test_file(b"rule a { condition: true }");

    let temp = TempDir::new().unwrap();

    // a is size 0
    let file_a = temp.path().join("a");
    fs::write(&file_a, "").unwrap();

    // b is size 100
    let file_b = temp.path().join("b");
    fs::write(&file_b, [0; 100]).unwrap();

    // c is size 1024
    let file_c = temp.path().join("c");
    fs::write(&file_c, [0; 1024]).unwrap();

    let match_file = |input: &Path| predicate::str::contains(format!("a {}", input.display()));
    let skip_file =
        |input: &Path| predicate::str::contains(format!("skipping {}", input.display()));

    // default will match the 3 files
    test_scan(&[], &[rule_file.path()], temp.path(), |cmd| {
        cmd.assert()
            .stdout(
                // match on "a", "b" and "d"
                match_file(&file_a)
                    .and(match_file(&file_b))
                    .and(match_file(&file_c)),
            )
            .stderr("")
            .success();
    });

    // Limit to 1024, will skip c
    test_scan(
        &["--skip-larger", "512"],
        &[rule_file.path()],
        temp.path(),
        |cmd| {
            cmd.assert()
                .stdout(
                    // match on "a" and "b"
                    match_file(&file_a)
                        .and(match_file(&file_b))
                        .and(match_file(&file_c).not()),
                )
                .stderr(
                    skip_file(&file_a)
                        .not()
                        .and(skip_file(&file_b).not())
                        .and(skip_file(&file_c)),
                )
                .success();
        },
    );

    // Limit to 10, will skip all but a
    test_scan(&["-z", "10"], &[rule_file.path()], temp.path(), |cmd| {
        cmd.assert()
            .stdout(
                // match on all
                match_file(&file_a)
                    .and(match_file(&file_b).not())
                    .and(match_file(&file_c).not()),
            )
            .stderr(
                skip_file(&file_a)
                    .not()
                    .and(skip_file(&file_b))
                    .and(skip_file(&file_c)),
            )
            .success();
    });
}

#[test]
fn test_print_module_data() {
    let rule_file = test_file(
        br#"
import "pe"
rule a {
    condition: false
}
"#,
    );

    let input = Path::new(env!("CARGO_MANIFEST_DIR"))
        .join("..")
        .join("boreal")
        .join("tests")
        .join("assets")
        .join("libyara")
        .join("data")
        .join("mtxex.dll");
    test_scan(&["-D"], &[rule_file.path()], &input, |cmd| {
        cmd.assert()
            .stdout(
                predicate::str::starts_with("pe\n")
                    // Static values
                    // Integer
                    .and(predicate::str::contains("DLL = 8192 (0x2000)"))
                    // Function
                    .and(predicate::str::contains("section_index[function]"))
                    // Dynamic values
                    // Integer
                    .and(predicate::str::contains("base_of_code = 4096 (0x1000)"))
                    // Undef
                    .and(predicate::str::contains("base_of_data[undef]"))
                    // Array
                    .and(predicate::str::contains(
                        r#"
        data_directories
            [0]
                size = 220 (0xdc)
"#,
                    ))
                    // Empty array
                    .and(predicate::str::contains("delayed_import_details = []"))
                    // Bytes printable
                    .and(predicate::str::contains(r#"dll_name = "mtxex.dll""#))
                    // Bytes non-printable
                    .and(predicate::str::contains(
                        "[\"ProductName\"] = { 4d6963726f736f6674ae2057696e646f7773ae204f7065\
                                           726174696e672053797374656d }",
                    ))
                    // Struct
                    .and(predicate::str::contains(
                        r#"
        image_version
            major = 10 (0xa)
            minor = 0 (0x0)
"#,
                    ))
                    // Dictionary
                    .and(predicate::str::contains(
                        r#"
        version_info
            ["CompanyName"] = "Microsoft Corporation"
"#,
                    )),
            )
            .stderr("")
            .success();
    });
}

#[test]
fn test_print_string_stats() {
    let rule_file = test_file(
        br#"
rule a {
    strings:
        $a = "ab<de>g"
        $b = { 01 ( FE | EF ) }
        $c = /foo\d??barbaz/ fullword
        $d = /.{10}/ fullword
    condition:
        any of them
}
"#,
    );

    let stats = r#"
  $a = "ab<de>g"
    literals: ["ab<de>g"]
    atoms: ["<de>"]
    atoms quality: 84
    algo: Literals
  $b = { 01 ( FE | EF ) }
    literals: [{ 01fe }, { 01ef }]
    atoms: [{ 01fe }, { 01ef }]
    atoms quality: 44
    algo: Literals
  $c = /foo\d??barbaz/ fullword
    literals: ["barbaz"]
    atoms: ["rbaz"]
    atoms quality: 80
    algo: Atomized { NonGreedy { reverse: Dfa, forward: none } }
  $d = /.{10}/ fullword
    literals: []
    atoms: []
    atoms quality: 0
    algo: Raw
"#;

    let input = test_file(b"");
    test_scan(
        &["--string-stats"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!(
                    "default:a (from {}){}",
                    rule_file.path().display(),
                    stats
                ))
                .stderr("")
                .success();
        },
    );
}

#[test]
fn test_print_scan_stats() {
    let rule_file = test_file(
        br#"
rule a {
    strings:
        $a = "abc"
    condition:
        any of them
}
"#,
    );

    let input = test_file(b"abc");
    test_scan(
        &["--scan-stats"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(
                    predicate::str::is_match(
                        r"Evaluation \{
    no_scan_eval_duration: .*,
    ac_duration: .*,
    fetch_memory_duration: .*,
    ac_confirm_duration: .*,
    nb_ac_matches: .*,
    rules_eval_duration: .*,
    raw_regexes_eval_duration: .*,
    memory_scanned_size: .*,
    nb_memory_chunks: .*,
\}
",
                    )
                    .unwrap(),
                )
                .stderr("")
                .success();
        },
    );
}

// Test when some inputs in a dir cannot be read
#[test]
#[cfg(unix)]
fn test_input_cannot_read() {
    use std::os::unix::fs::{OpenOptionsExt, PermissionsExt};

    let rule_file = test_file(b"rule bee { condition: true }");

    let temp = TempDir::new().unwrap();
    let child = temp.path().join("child");
    let _file = fs::OpenOptions::new()
        .create(true)
        .truncate(true)
        .write(true)
        .mode(0o000)
        .open(&child)
        .unwrap();

    let subdir = temp.path().join("subdir");
    fs::create_dir(&subdir).unwrap();
    fs::set_permissions(&subdir, fs::Permissions::from_mode(0o000)).unwrap();

    test_scan(&["-r"], &[rule_file.path()], temp.path(), |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(
                predicate::str::contains(format!("Cannot scan file {}", child.display())).and(
                    predicate::str::contains(format!(
                        "IO error for operation on {}",
                        subdir.display()
                    )),
                ),
            )
            // Still successful, since some other files in the directory may have been scanned
            .success();
    });
}

#[test]
fn test_module_names() {
    test_cmd(CmdKind::ListModules, |cmd| {
        cmd.assert()
            .stdout(
                predicate::str::contains("math\n")
                    .and(predicate::str::contains("string\n"))
                    .and(predicate::str::contains("time\n")),
            )
            .stderr("")
            // Still successful, since some other files in the directory may have been scanned
            .success();
    });
}

#[test]
#[cfg(feature = "memmap")]
fn test_no_mmap() {
    let rule_file = test_file(
        br#"
rule first {
    strings:
        $a = "abc"
    condition:
        any of them
}
rule second {
    strings:
        $a = "xyz"
    condition:
        any of them
}"#,
    );

    let input = test_file(b"xyabcz");
    // Not matching
    test_scan(&["--no-mmap"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("first {}\n", input.path().display()))
            .stderr("")
            .success();
    });
}

#[test]
fn test_console_log() {
    let rule_file = test_file(
        br#"
import "console"

rule logger {
    condition:
        console.log("this is ", "a log")
}"#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    test_scan(&[], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("this is a log\nlogger {path}\n"))
            .stderr("")
            .success();
    });

    // Logs can be disabled with the -q flag
    test_scan(&["-q"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("logger {path}\n"))
            .stderr("")
            .success();
    });
}

#[test]
fn test_invalid_fragmented_scan_mode() {
    test_scan(
        &["--fragmented-scan-mode", "bad_value"],
        &[Path::new("rules.yar")],
        Path::new("input"),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains(
                    "invalid value 'bad_value' for \
            '--fragmented-scan-mode <legacy|fast|singlepass>\': invalid value",
                ))
                .failure();
        },
    );
}

#[test]
fn test_invalid_compiler_profile() {
    test_scan(
        &["--profile", "bad_value"],
        &[Path::new("rules.yar")],
        Path::new("input"),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains(
                    "invalid value 'bad_value' for \
            '--profile <speed|memory>\': invalid value",
                ))
                .failure();
        },
    );
}

#[test]
fn test_tags() {
    let rule_file = test_file(
        br#"
rule notag {
    condition:
        true
}
rule tag1: first {
    condition:
        true
}
rule tag3: first second third {
    condition:
        true
}
"#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    // Test print tags
    test_scan(&["-g"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                "notag [] {path}\n\
             tag1 [first] {path}\n\
             tag3 [first,second,third] {path}\n"
            ))
            .stderr("")
            .success();
    });

    // Test filter by tag
    test_scan(&["-t", "first"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("tag1 {path}\ntag3 {path}\n"))
            .stderr("")
            .success();
    });
    test_scan(&["--tag=third"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("tag3 {path}\n"))
            .stderr("")
            .success();
    });
    test_scan(&["-t", ""], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert().stdout("").stderr("").success();
    });
}

#[test]
fn test_identifier() {
    let rule_file = test_file(
        br#"
rule first { condition: true }
rule second { condition: true }
"#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    // Test filter by identifier
    test_scan(&["-i", "first"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("first {path}\n"))
            .stderr("")
            .success();
    });
    test_scan(
        &["--identifier=second"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!("second {path}\n"))
                .stderr("")
                .success();
        },
    );
    test_scan(
        &["--identifier=third"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert().stdout("").stderr("").success();
        },
    );
}

#[test]
fn test_print_meta() {
    let rule_file = test_file(
        br#"
rule first: tag {
    meta:
        integer = -15
        string = "d mol"
        test = true
    condition:
        true
}
rule second: tag {
    condition:
        true
}
rule third: tag {
    meta:
        value = "ok"
    condition:
        true
}
rule fourth { condition: true }
"#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    // Test print meta
    test_scan(&["-m"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                "first [integer=-15,string=\"d mol\",test=true] {path}\n\
             second [] {path}\n\
             third [value=\"ok\"] {path}\n\
             fourth [] {path}\n"
            ))
            .stderr("")
            .success();
    });

    // Test print meta + tag
    test_scan(
        &["-g", "--print-meta"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!(
                    "first [tag] [integer=-15,string=\"d mol\",test=true] {path}\n\
             second [tag] [] {path}\n\
             third [tag] [value=\"ok\"] {path}\n\
             fourth [] [] {path}\n"
                ))
                .stderr("")
                .success();
        },
    );
}

#[test]
fn test_print_string_matches() {
    let rule_file = test_file(
        br#"
rule my_rule {
    strings:
        $a = /<.{1,5}?>/
        $b = "abc"
    condition:
        // The rule can be evaluated without scanning the strings.
        // This ensures that printing strings forces the computation of the
        // string matches.
        true or any of them
}"#,
    );

    let input = test_file(
        b"<a>\n
<<abc>d>\n
<\x01\x02>\n
<a\tv>\n
<a\xFF>>\n
",
    );
    let path = input.path().display();

    // Test match data only
    test_scan(&["-s"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                r#"my_rule {path}
0x0:$a: <a>
0x5:$a: <<abc>
0x6:$a: <abc>
0xf:$a: <\x01\x02>
0x15:$a: <a\tv>
0x1c:$a: <a\xff>
0x7:$b: abc
"#
            ))
            .stderr("")
            .success();
    });

    // Test match length only
    test_scan(&["-L"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                r#"my_rule {path}
0x0:3:$a
0x5:6:$a
0x6:5:$a
0xf:4:$a
0x15:5:$a
0x1c:4:$a
0x7:3:$b
"#
            ))
            .stderr("")
            .success();
    });

    // Test both
    test_scan(
        &["--print-strings", "--print-string-length"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!(
                    r#"my_rule {path}
0x0:3:$a: <a>
0x5:6:$a: <<abc>
0x6:5:$a: <abc>
0xf:4:$a: <\x01\x02>
0x15:5:$a: <a\tv>
0x1c:4:$a: <a\xff>
0x7:3:$b: abc
"#
                ))
                .stderr("")
                .success();
        },
    );
}

#[test]
fn test_print_string_xor_key() {
    let rule_file = test_file(
        br#"
rule my_rule {
    strings:
        $a = "aaa" ascii wide xor
    condition:
        // The rule can be evaluated without scanning the strings.
        // This ensures that printing xor keys forces the computation of the
        // string matches.
        true or any of them
}"#,
    );

    let input = test_file(b"aaa\nbbb\nabcccba\nB#B#B#\n");
    let path = input.path().display();

    // Test xor only
    test_scan(&["-X"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                r#"my_rule {path}
0x0:$a:xor(0x00,aaa)
0x4:$a:xor(0x03,aaa)
0xa:$a:xor(0x02,aaa)
0x10:$a:xor(0x23,a\x00a\x00a\x00)
"#
            ))
            .stderr("")
            .success();
    });

    // With match data
    test_scan(&["-Xs"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                r#"my_rule {path}
0x0:$a:xor(0x00,aaa): aaa
0x4:$a:xor(0x03,aaa): bbb
0xa:$a:xor(0x02,aaa): ccc
0x10:$a:xor(0x23,a\x00a\x00a\x00): B#B#B#
"#
            ))
            .stderr("")
            .success();
    });
}

#[test]
fn test_timeout() {
    let rule_file = test_file(
        br#"
rule too_long {
    condition:
        for all i in (0..9223372036854775807) : (
            for all j in (0..9223372036854775807) : (
                for all k in (0..9223372036854775807) : (
                    for all l in (0..9223372036854775807) : (
                        i + j + k + l >= 0
                    )
                )
            )
        )
}"#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    test_scan(&["-a", "1"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(format!("Cannot scan {path}: timeout\n"))
            .failure();
    });
}

#[test]
fn test_print_namespace() {
    let rule_file = test_file(
        br#"
rule first { condition: true }
"#,
    );

    let input = test_file(b"");
    let path = input.path().display();

    test_scan(&["-e"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("default:first {path}\n"))
            .stderr("")
            .success();
    });
}

#[test]
fn test_define_symbol() {
    let input = test_file(b"");
    let path = input.path().display();

    let rule_file = test_file(
        br#"
rule symbols {
    condition:
        symbol_float + 4.3 == 2.5 and
        symbol_int + 5 == 2 and
        symbol_true == true and
        symbol_false == false and
        symbol_str == "a_string" and
        symbol_str2 == "2.5a" and
        symbol_empty == ""
}
"#,
    );

    test_scan(
        &[
            "-d",
            "symbol_float=-1.8",
            "--define=symbol_int=-3",
            "--define",
            "symbol_true=true",
            "--define",
            "symbol_false=false",
            "--define",
            "symbol_str=a_string",
            "--define",
            "symbol_str2=2.5a",
            "--define",
            "symbol_empty=",
        ],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!("symbols {path}\n"))
                .stderr("")
                .success();
        },
    );

    // Test a bad define
    test_scan(&["-d", "name"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(predicate::str::contains(
                "invalid value 'name' for '--define <VAR=VALUE>': \
            missing '=' delimiter",
            ))
            .failure();
    });

    // Test a mismatched type
    let bad_rule = test_file(
        br#"
rule bad_symbols {
    condition:
        symbol_float == "-1.8"
}
"#,
    );
    test_scan(
        &["-d", "symbol_float=-1.8"],
        &[bad_rule.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains("expressions have invalid types"))
                .failure();
        },
    );
}

#[test]
fn test_scan_list() {
    let rule_file = test_file(b"rule a { condition: true }");

    // dir1
    //   a
    //   b
    // dir2
    //   subdir
    //     c
    let dir1 = TempDir::new().unwrap();
    let file_a = dir1.path().join("a");
    fs::write(&file_a, "").unwrap();
    let file_b = dir1.path().join("b");
    fs::write(&file_b, "").unwrap();

    let dir2 = TempDir::new().unwrap();
    let subdir = dir2.path().join("subdir");
    fs::create_dir(&subdir).unwrap();
    let file_c = subdir.join("c");
    fs::write(&file_c, "").unwrap();

    let match_str = |input: &Path| predicate::str::contains(format!("a {}", input.display()));

    // dir1 + c, will match the 3 filse
    let list = test_file(format!("{}\n{}\n", dir1.path().display(), file_c.display()).as_bytes());
    test_scan(&["--scan-list"], &[rule_file.path()], list.path(), |cmd| {
        cmd.assert()
            .stdout(
                match_str(&file_a)
                    .and(match_str(&file_b))
                    .and(match_str(&file_c)),
            )
            .stderr("")
            .success();
    });

    // a + dir2, but not recursive, will match only a
    let list = test_file(format!("{}\n{}\n", file_a.display(), dir2.path().display()).as_bytes());
    test_scan(&["--scan-list"], &[rule_file.path()], list.path(), |cmd| {
        cmd.assert()
            .stdout(
                match_str(&file_a)
                    .and(match_str(&file_b).not())
                    .and(match_str(&file_c).not()),
            )
            .stderr("")
            .success();
    });

    // When recursive, will match c
    test_scan(
        &["--scan-list", "-r"],
        &[rule_file.path()],
        list.path(),
        |cmd| {
            cmd.assert()
                .stdout(
                    match_str(&file_a)
                        .and(match_str(&file_b).not())
                        .and(match_str(&file_c)),
                )
                .stderr("")
                .success();
        },
    );

    // Empty
    let list = test_file(b"");
    test_scan(&["--scan-list"], &[rule_file.path()], list.path(), |cmd| {
        cmd.assert().stdout("").stderr("").success();
    });

    // Do these tests only on linux, as the error messages can depend on the OS.
    if cfg!(target_os = "linux") {
        // path is a directory
        // On linux, the open on a dir works but the read fails, making
        // this a great test to test the read failure case.
        test_scan(&["--scan-list"], &[rule_file.path()], dir1.path(), |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains(format!(
                    "cannot read from scan list {}: Is a directory",
                    dir1.path().display()
                )))
                .failure();
        });

        // path does not exist
        test_scan(
            &["--scan-list"],
            &[rule_file.path()],
            Path::new("invalid_path"),
            |cmd| {
                cmd.assert()
                    .stdout("")
                    .stderr(predicate::str::contains(
                        "cannot open scan list invalid_path: No such file or directory",
                    ))
                    .failure();
            },
        );
    }
}

#[test]
fn test_negate() {
    let rule_file = test_file(
        br#"
rule a { condition: true }
rule b { condition: false }
rule c { condition: true }
rule d { condition: false }
"#,
    );

    let input = test_file(b"");
    // Not matching
    test_scan(&["-n"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                "b {}\nd {}\n",
                input.path().display(),
                input.path().display()
            ))
            .stderr("")
            .success();
    });
}

#[test]
fn test_count() {
    let rule_file = test_file(
        br#"
rule a {
    strings:
        $ = "abc"
    condition:
        any of them
}
rule b {
    strings:
        $ = "def"
    condition:
        any of them
}
"#,
    );

    let temp = TempDir::new().unwrap();
    let file_a = temp.path().join("a");
    fs::write(&file_a, "abc").unwrap();
    let file_b = temp.path().join("b");
    fs::write(&file_b, "abcdef").unwrap();
    let file_c = temp.path().join("c");
    fs::write(&file_c, "").unwrap();

    // Test against a single file
    test_scan(&["-c"], &[rule_file.path()], &file_a, |cmd| {
        cmd.assert()
            .stdout(format!("{}: 1\n", file_a.display()))
            .stderr("")
            .success();
    });

    // Test in combination with the negate flag
    test_scan(&["-cn"], &[rule_file.path()], &file_c, |cmd| {
        cmd.assert()
            .stdout(format!("{}: 2\n", file_c.display()))
            .stderr("")
            .success();
    });

    // Test against a directory
    test_scan(&["-c"], &[rule_file.path()], temp.path(), |cmd| {
        cmd.assert()
            .stdout(
                predicate::str::contains(format!("{}: 1", file_a.display()))
                    .and(predicate::str::contains(format!("{}: 2", file_b.display())))
                    .and(predicate::str::contains(format!("{}: 0", file_c.display()))),
            )
            .stderr("")
            .success();
    });
}

#[test]
#[cfg_attr(unix, ignore)]
fn test_count_process() {
    let rule_file = test_file(
        br#"
rule process_scan {
    strings:
        $a = "PAYLOAD_ON_STACK"
    condition:
        $a
}"#,
    );

    let proc = BinHelper::run("stack");
    let pid = proc.pid();

    // Not matching
    test_scan(
        &["--count"],
        &[rule_file.path()],
        Path::new(&pid.to_string()),
        |cmd| {
            cmd.assert()
                .stdout(format!("{pid}: 1\n"))
                .stderr("")
                .success();
        },
    );
}

#[test]
#[cfg_attr(unix, ignore)]
fn test_count_limit() {
    let rule_file = test_file(
        br#"
rule a { condition: true }
rule b { condition: true }
rule c { condition: true }
"#,
    );

    let input = test_file(b"");
    test_scan(&["-l", "2"], &[rule_file.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!(
                "a {}\nb {}\n",
                input.path().display(),
                input.path().display(),
            ))
            .stderr("")
            .success();
    });

    // Also works with a process
    let proc = BinHelper::run("stack");
    let pid = proc.pid();

    // Not matching
    test_scan(
        &["--max-rules=2"],
        &[rule_file.path()],
        Path::new(&pid.to_string()),
        |cmd| {
            cmd.assert()
                .stdout(format!("a {pid}\nb {pid}\n"))
                .stderr("")
                .success();
        },
    );
}

#[test]
fn test_max_strings_per_rule() {
    let rule_file = test_file(
        br#"
rule a {
    strings:
        $a = "aaa"
        $b = "bbb"
        $c = "ccc"
    condition:
        any of them
}
"#,
    );

    let input = test_file(b"");
    test_scan(
        &["--max-strings-per-rule=2"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(
                    predicate::str::contains("error").and(predicate::str::contains(
                        "the rule contains more than 2 strings",
                    )),
                )
                .failure();
        },
    );
}

#[test]
fn test_string_max_nb_matches() {
    let rule_file = test_file(
        br#"
rule a {
    strings:
        $a = "a"
    condition:
        any of them
}
"#,
    );

    let warning = "warning: string $a in rule default:a reached the maximum number of matches";

    // Default behavior is to print the warning and continue
    let input = test_file(b"aaaa");
    test_scan(
        &["--string-max-nb-matches=2"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!("a {}\n", input.path().display()))
                .stderr(predicate::str::contains(warning))
                .success();
        },
    );

    // We can ignore warnings with a flag
    test_scan(
        &["--string-max-nb-matches=2", "--no-warnings"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!("a {}\n", input.path().display()))
                .stderr("")
                .success();
        },
    );

    // Or we can abort on warnings
    test_scan(
        &["--string-max-nb-matches=2", "--fail-on-warnings"],
        &[rule_file.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains(warning))
                .success();
        },
    );
}

#[test]
fn test_yr_invalid_number_of_arguments() {
    // Not enough arguments
    boreal_cmd()
        .arg("yr")
        .arg("rules_file")
        .assert()
        .stderr(predicate::str::contains("Invalid number of arguments"))
        .failure();

    #[cfg(feature = "serialize")]
    {
        // -C means a single rules_file
        boreal_cmd()
            .arg("yr")
            .arg("-C")
            .arg("rules_file")
            .arg("rules_file2")
            .arg("input")
            .assert()
            .stderr(predicate::str::contains(
                "Only a single rules path must be passed",
            ))
            .failure();
    }
}

#[test]
fn test_multiple_rules_files() {
    let rule_a = test_file(b"rule a { condition: true }");
    let rule_b = test_file(b"rule b { condition: true }");
    let rule_c = test_file(b"rule c { condition: true }");

    let input = test_file(b"");
    let input_path = input.path().display().to_string();

    test_scan(&[], &[rule_a.path(), rule_b.path()], input.path(), |cmd| {
        cmd.assert()
            .stdout(format!("a {input_path}\nb {input_path}\n"))
            .success();
    });
    test_scan(
        &[],
        &[rule_a.path(), rule_b.path(), rule_c.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout(format!("a {input_path}\nb {input_path}\nc {input_path}\n"))
                .success();
        },
    );

    // Name collision will make compilation fail
    test_scan(
        &[],
        &[rule_a.path(), rule_b.path(), rule_a.path()],
        input.path(),
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr(predicate::str::contains(
                    "rule `a` is already declared in this namespace",
                ))
                .failure();
        },
    );
}

#[test]
fn test_rules_files_namespace() {
    let temp = TempDir::new().unwrap();
    let rule_a = temp.path().join("rule_a");
    fs::write(&rule_a, b"rule a { condition: true }").unwrap();
    let rule_b = temp.path().join("rule:b");
    fs::write(&rule_b, b"rule b { condition: true }").unwrap();

    let input = temp.path().join("input");
    fs::write(&input, b"").unwrap();
    let input_path = input.display().to_string();

    // Specify namespaces using the ':' spearator
    test_scan(
        &["-e"],
        &[Path::new("ns1:rule_a"), Path::new("ns2:rule_a")],
        &input,
        |cmd| {
            cmd.current_dir(temp.path())
                .assert()
                .stdout(format!("ns1:a {input_path}\nns2:a {input_path}\n"))
                .success();
        },
    );

    // A filepath has a ':' in it: it should not be taken as a namespace if possible
    test_scan(
        &["-e"],
        &[
            Path::new("ns1:rule_a"),
            Path::new("rule:b"),
            Path::new("ns2:rule:b"),
        ],
        &input,
        |cmd| {
            cmd.current_dir(temp.path())
                .assert()
                .stdout(format!(
                    "ns1:a {input_path}\ndefault:b {input_path}\nns2:b {input_path}\n"
                ))
                .success();
        },
    );

    // When a filepath that has a ':' does not exist, it should fail properly
    test_scan(&["-e"], &[Path::new("rule:c")], &input, |cmd| {
        cmd.current_dir(temp.path())
            .assert()
            .stdout("")
            .stderr(predicate::str::contains("Cannot read rules file c:"))
            .failure();
    });
    test_scan(&["-e"], &[Path::new("ns1:rule:c")], &input, |cmd| {
        cmd.current_dir(temp.path())
            .assert()
            .stdout("")
            .stderr(predicate::str::contains("Cannot read rules file rule:c:"))
            .failure();
    });
}

#[test]
fn test_ignored_arguments() {
    let rule_file = test_file(b"rule a { strings: $ = /abc/ condition: any of them }");
    let input = test_file(b"abc");

    boreal_cmd()
        .arg("yr")
        .arg("-E")
        .arg("--fast-scan")
        .arg("--stack-size=300")
        .arg(rule_file.path())
        .arg(input.path())
        .assert()
        .stdout(format!("a {}\n", input.path().display()))
        .stderr("")
        .success();
}

#[test]
#[cfg(feature = "serialize")]
fn test_save_load() {
    let rule_file = test_file(
        br#"
import "console"

rule a {
    condition:
        console.log("foo")
}
"#,
    );

    let temp = TempDir::new().unwrap();
    let save_path = temp.path().join("serialized_rules");
    let input = temp.path().join("input");
    fs::write(&input, b"a").unwrap();

    // Save into the file
    test_cmd(CmdKind::Save, |cmd| {
        cmd.arg("-f")
            .arg(rule_file.path())
            .arg(&save_path)
            .assert()
            .success();
    });

    // Now reload
    test_cmd(CmdKind::Load, |cmd| {
        cmd.arg(&save_path)
            .arg(&input)
            .assert()
            .stdout(format!("foo\na {}\n", input.display()))
            .stderr("")
            .success();
    });
}

#[test]
#[cfg(feature = "serialize")]
fn test_invalid_load() {
    let temp = TempDir::new().unwrap();
    let input = temp.path().join("input");
    fs::write(&input, b"a").unwrap();

    // Non existing path
    let fake_path = temp.path().join("non_existing");
    test_cmd(CmdKind::Load, |cmd| {
        cmd.arg(&fake_path)
            .arg(&input)
            .assert()
            .stdout("")
            .stderr(predicate::str::contains(format!(
                "Unable to read from {}",
                fake_path.display()
            )))
            .failure();
    });

    // Failure during deserialization
    test_cmd(CmdKind::Load, |cmd| {
        cmd.arg(&input)
            .arg(&input)
            .assert()
            .stdout("")
            .stderr(predicate::str::contains("Unable to deserialize rules"))
            .failure();
    });
}

#[test]
#[cfg(feature = "serialize")]
fn test_invalid_save() {
    let temp = TempDir::new().unwrap();
    let rule = temp.path().join("rule");
    fs::write(&rule, b"rule a { condition: true }").unwrap();

    // Non existing path
    test_cmd(CmdKind::Save, |cmd| {
        cmd.arg("-f")
            .arg(&rule)
            .arg(&rule)
            .assert()
            .stdout("")
            .stderr(predicate::str::contains(format!(
                "File {} already exists",
                rule.display()
            )))
            .failure();
    });

    let bad_rule = temp.path().join("bad_rule");
    fs::write(&bad_rule, b"rule a { ").unwrap();
    // Compilation error
    test_cmd(CmdKind::Save, |cmd| {
        cmd.arg("-f")
            .arg(&bad_rule)
            .arg(&bad_rule)
            .assert()
            .stdout("")
            .stderr(predicate::str::contains("syntax error"))
            .failure();
    });

    #[cfg(unix)]
    {
        use std::os::unix::fs::PermissionsExt;

        let subdir = temp.path().join("subdir");
        fs::create_dir(&subdir).unwrap();
        let out = subdir.join("out");

        // cannot inspect file
        let mut perms = fs::metadata(&subdir).unwrap().permissions();
        perms.set_mode(0o000);
        fs::set_permissions(&subdir, perms).unwrap();
        test_cmd(CmdKind::Save, |cmd| {
            cmd.arg("-f")
                .arg(&rule)
                .arg(&out)
                .assert()
                .stdout("")
                .stderr(predicate::str::contains(format!(
                    "Unable to inspect file {}",
                    out.display()
                )))
                .failure();
        });

        // cannot create file
        let mut perms = fs::metadata(&subdir).unwrap().permissions();
        perms.set_mode(0o555);
        fs::set_permissions(&subdir, perms).unwrap();
        let out_path = subdir.join("out");
        test_cmd(CmdKind::Save, |cmd| {
            cmd.arg("-f")
                .arg(&rule)
                .arg(&out_path)
                .assert()
                .stdout("")
                .stderr(predicate::str::contains(format!(
                    "Unable to create file {}",
                    out.display()
                )))
                .failure();
        });
    }
}

#[test]
#[cfg(feature = "cuckoo")]
fn test_module_data() {
    let temp = TempDir::new().unwrap();

    let rule = temp.path().join("rule");
    fs::write(
        &rule,
        r#"
import "cuckoo"

rule a {
    condition:
        cuckoo.network.host(/bcd/) == 1
}"#,
    )
    .unwrap();

    let data = temp.path().join("data");
    fs::write(&data, r#"{ "network": { "hosts": ["abcde"] } }"#).unwrap();

    let input = temp.path().join("input");
    fs::write(&input, "").unwrap();

    test_scan(
        &[&format!("--module-data=cuckoo={}", data.display())],
        &[&rule],
        &input,
        |cmd| {
            cmd.assert()
                .stdout(format!("a {}\n", input.display()))
                .stderr("")
                .success();
        },
    );
}

#[test]
#[cfg(feature = "cuckoo")]
fn test_invalid_module_data() {
    let temp = TempDir::new().unwrap();

    let rule = temp.path().join("rule");
    fs::write(&rule, "rule a { condition: true }").unwrap();

    let data = temp.path().join("data");
    fs::write(&data, r"{").unwrap();

    let input = temp.path().join("input");
    fs::write(&input, "").unwrap();

    // Invalid module data
    test_scan(&["-x", "name"], &[&rule], &input, |cmd| {
        cmd.assert()
            .stdout("")
            .stderr(predicate::str::contains(
                "invalid value 'name' for '--module-data <MODULE=FILE>': \
            missing '=' delimiter",
            ))
            .failure();
    });

    // Unknown module
    test_scan(
        &[&format!("--module-data=piou={}", data.display())],
        &[&rule],
        &input,
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr("Cannot set data for unsupported module piou\n")
                .failure();
        },
    );

    // Non existing file
    test_scan(&["-x", "cuckoo=non_existing"], &[&rule], &input, |cmd| {
        cmd.assert()
            .stderr(predicate::str::contains(
                "Unable to read cuckoo data from file non_existing",
            ))
            .failure();
    });

    // Cannot parse data
    test_scan(
        &[&format!("--module-data=cuckoo={}", data.display())],
        &[&rule],
        &input,
        |cmd| {
            cmd.assert()
                .stdout("")
                .stderr("The data for the cuckoo module is invalid\n")
                .failure();
        },
    );
}

// Copied in `boreal/tests/it/utils.rs`. Not trivial to share, and won't be
// modified too frequently.
struct BinHelper {
    proc: std::process::Child,
}

impl BinHelper {
    fn run(arg: &str) -> Self {
        // Path to current exe
        let path = std::env::current_exe().unwrap();
        // Path to "deps" dir
        let path = path.parent().unwrap();
        // Path to parent of deps dir, ie destination of build artifacts
        let path = path.parent().unwrap();
        // Now select the bin helper
        let path = path.join(if cfg!(windows) {
            "boreal-test-helpers.exe"
        } else {
            "boreal-test-helpers"
        });
        if !path.exists() {
            panic!(
                "File {} not found. \
                You need to compile the `boreal-test-helpers` crate to run this test",
                path.display()
            );
        }
        let mut child = std::process::Command::new(path)
            .arg(arg)
            .stdout(std::process::Stdio::piped())
            .spawn()
            .unwrap();

        // Accumulate read inputs until the "ready" line is found
        let mut stdout = BufReader::new(child.stdout.take().unwrap());
        let mut buffer = String::new();
        loop {
            buffer.clear();
            stdout.read_line(&mut buffer).unwrap();
            if buffer.trim() == "ready" {
                break;
            }
        }
        Self { proc: child }
    }

    fn pid(&self) -> u32 {
        self.proc.id()
    }
}

impl Drop for BinHelper {
    fn drop(&mut self) {
        drop(self.proc.kill());
        drop(self.proc.wait());
    }
}