repotoire 0.7.0

//! Insecure-crypto detector (CWE-327, CWE-328).
//!
//! Detects use of cryptographically broken or weak primitives (MD5, SHA1,
//! DES, RC4, ECB-mode ciphers, ...). Unlike eval/command-injection/pickle,
//! the danger signal here is the **algorithm name** more than the argument
//! shape, which makes the AST migration's value the ability to reliably
//! extract that name from a call expression rather than fishing it out of
//! line-text with regex.
//!
//! # Architecture
//!
//! Two scan paths, picked by file language (mirrors `eval_detector.rs`,
//! `command_injection.rs`, and `pickle_detector.rs`):
//!
//! 1. **AST path** (Python, JS, TS, JSX, TSX): walks the tree-sitter
//!    parse tree looking for **call expressions** whose callee names a
//!    known weak primitive, or whose first string-literal argument names
//!    one. Argument-shape classifiers descend through transparent
//!    wrappers (`await`, ternary, TS `as`/`!`/`<T>x`/`satisfies`,
//!    parentheses) so `crypto.createHash(await getAlgo())` is correctly
//!    classified as `UnknownAlgoFromVariable`, not `Unknown`.
//!
//!      - **Python**:
//!        - `hashlib.md5(...)`, `hashlib.sha1(...)`, `hashlib.md4(...)`
//!        - `hashlib.new('md5'|'sha1'|...)` — string-arg algorithm name
//!        - `Crypto.Cipher.DES.new(...)`, `Crypto.Cipher.AES.new(key, AES.MODE_ECB, ...)`
//!        - `cryptography.hazmat.primitives.hashes.MD5()` etc.
//!        - **Bare-identifier from-import**: `from hashlib import md5;
//!          md5(data)` — resolved via per-file from-import alias map
//!          (this is the first AST detector to do bare-call from-import
//!          resolution; the same pattern can be applied to eval/command
//!          /pickle in follow-up work).
//!
//!      - **JavaScript / TypeScript**:
//!        - `crypto.createHash('md5')`, `crypto.createCipheriv('aes-128-ecb', ...)`,
//!          `crypto.createSign('RSA-MD5')`. Algorithm is the first string-literal
//!          argument.
//!        - `require('crypto').createHash(...)` and
//!          `(await import('crypto')).createHash(...)` — receiver descent
//!          through `require()`/`import()` (B1 lesson from CommandInjection
//!          commit `3c88328e`).
//!
//! 2. **Line path** (Java, C/C++, Ruby, PHP, Go): for languages without a
//!    tree-sitter grammar in our dispatch list, a small line-based regex
//!    scanner matches the canonical forms (`Cipher.getInstance("DES")`,
//!    `MessageDigest.getInstance("MD5")`, `mcrypt_encrypt(MCRYPT_DES,
//!    ...)`, `Digest::MD5`, `des.NewCipher(...)`, ...).
//!
//! The classifier emits one of:
//!
//!   - `BrokenAlgoIdentifier(name)` — `hashlib.md5(...)` (attribute name).
//!   - `BrokenAlgoStringLiteral(name)` — `crypto.createHash('md5')`.
//!   - `WeakModeCombo(algo, mode)` — `crypto.createCipheriv('aes-128-ecb', ...)`.
//!   - `UnknownAlgoFromVariable` — `crypto.createHash(userChoice)`.
//!   - `SafeAlgo` — `hashlib.sha256(...)` etc. → skipped.
//!
//! Severity table:
//!
//!   - Broken algo (md5/sha1/md4/des/rc4/3des/tripledes) → **High**.
//!   - ECB mode (any cipher) → **High** (ECB is broken regardless of
//!     cipher choice — pattern leakage).
//!   - DES (any mode) → **High**.
//!   - Variable-algo (`createHash(userChoice)`) → **Low** (we can't tell
//!     statically; informational).
//!   - Route-handler boost: if the call sits inside a request-handler /
//!     route function (matched the same way as eval/command/pickle),
//!     severity is bumped one tier up.

use crate::detectors::ast_fingerprint::parse_root_ext;
use crate::detectors::ast_walk::AstWalkCtx;
use crate::detectors::base::{Detector, DetectorConfig};
use crate::detectors::fast_search::*;
use crate::detectors::security::ast_helpers::{
    collect_named_args, node_text, receiver_chain_label as receiver_chain_label_shared,
    unwrap_callee,
};
use crate::detectors::security::scan_inputs::{ScanAstInputs, ScanInputs};
use crate::graph::GraphQueryExt;
use crate::models::{Finding, Severity};
use crate::parsers::lightweight::Language;
use anyhow::Result;
use regex::Regex;
use std::collections::HashMap;
use std::path::{Path, PathBuf};
use std::sync::LazyLock;
use tracing::{debug, info};

// ---------------------------------------------------------------------------
// Single source of truth for supported file extensions.
// ---------------------------------------------------------------------------

/// Extensions this detector processes. AST-eligible extensions
/// (Python + JS/TS/JSX/TSX) flow through `scan_file_ast`; the rest fall
/// through to the line scanner.
const SUPPORTED_EXTS: &[&str] = &[
    // AST path
    "py", "js", "ts", "jsx", "tsx", // Line path
    "rb", "java", "go", "php", "c", "cc", "cpp", "h", "hpp", "cs",
];

/// Subset of `SUPPORTED_EXTS` that has tree-sitter grammars in our
/// dispatch list.
const AST_EXTS: &[&str] = &["py", "js", "ts", "jsx", "tsx"];

// ---------------------------------------------------------------------------
// Algorithm classification
// ---------------------------------------------------------------------------

/// Lowercased name → broken / weak / safe.
fn classify_algo_name(name: &str) -> AlgoKind {
    let n = name.to_lowercase();
    let n = n.replace([' ', '_', '-'], "");
    if matches!(
        n.as_str(),
        "md5"
            | "md4"
            | "md2"
            | "sha"
            | "sha1"
            | "sha0"
            | "ripemd"
            | "ripemd128"
            | "ripemd160"
            | "des"
            | "rc4"
            | "rc2"
            | "arc4"
            | "blowfish"
            | "3des"
            | "tripledes"
    ) {
        return AlgoKind::Broken(name.to_string());
    }
    if matches!(
        n.as_str(),
        "sha256"
            | "sha384"
            | "sha512"
            | "sha224"
            | "sha3"
            | "sha3256"
            | "sha3384"
            | "sha3512"
            | "blake2"
            | "blake2b"
            | "blake2s"
            | "blake3"
            | "aes"
            | "aes128"
            | "aes192"
            | "aes256"
            | "chacha20"
            | "chacha20poly1305"
            | "xchacha20poly1305"
    ) {
        return AlgoKind::Safe;
    }
    AlgoKind::Unknown
}

/// Recognise an `algo-mode` style label (`aes-128-ecb`, `des-cbc`, ...)
/// and return `(algo, mode)` if the trailing token is a known mode.
fn split_algo_mode(name: &str) -> Option<(String, String)> {
    let lower = name.to_lowercase();
    const MODES: &[&str] = &["ecb", "cbc", "ctr", "ofb", "cfb", "gcm", "ccm", "xts"];
    for mode in MODES {
        for sep in ['-', '_', '/'] {
            let needle = format!("{sep}{mode}");
            if let Some(idx) = lower.rfind(&needle) {
                let algo = &lower[..idx];
                let algo_canonical: String =
                    algo.chars().take_while(|c| c.is_alphabetic()).collect();
                let algo_canonical = if algo_canonical.is_empty() {
                    algo.to_string()
                } else {
                    algo_canonical
                };
                return Some((algo_canonical, (*mode).to_string()));
            }
        }
    }
    None
}

#[derive(Debug, Clone, PartialEq, Eq)]
enum AlgoKind {
    Broken(String),
    Safe,
    Unknown,
}

/// Classification of a detected weak-crypto call.
#[derive(Debug, Clone, PartialEq, Eq)]
enum WeakCryptoCallKind {
    /// `hashlib.md5(...)` — attribute / identifier names a broken algo.
    BrokenAlgoIdentifier(String),
    /// `crypto.createHash('md5')` — string-literal arg names a broken algo.
    BrokenAlgoStringLiteral(String),
    /// `crypto.createCipheriv('aes-128-ecb', ...)` — algo+mode combo where
    /// the mode is weak (ECB) regardless of the underlying cipher.
    #[allow(dead_code)]
    WeakModeCombo { algo: String, mode: String },
    /// `crypto.createHash(userChoice)` — algo is a non-literal expression.
    UnknownAlgoFromVariable,
}

impl WeakCryptoCallKind {
    fn algo_label(&self) -> String {
        match self {
            WeakCryptoCallKind::BrokenAlgoIdentifier(n)
            | WeakCryptoCallKind::BrokenAlgoStringLiteral(n) => n.clone(),
            WeakCryptoCallKind::WeakModeCombo { algo, mode } => format!("{algo}/{mode}"),
            WeakCryptoCallKind::UnknownAlgoFromVariable => "<dynamic>".to_string(),
        }
    }
}

/// Map a `WeakCryptoCallKind` to a base severity.
fn severity_for(kind: &WeakCryptoCallKind) -> Severity {
    match kind {
        WeakCryptoCallKind::BrokenAlgoIdentifier(_)
        | WeakCryptoCallKind::BrokenAlgoStringLiteral(_) => Severity::High,
        WeakCryptoCallKind::WeakModeCombo { .. } => {
            // DES is broken at any mode; ECB mode is broken at any cipher.
            // Either way: High.
            Severity::High
        }
        WeakCryptoCallKind::UnknownAlgoFromVariable => Severity::Low,
    }
}

// ---------------------------------------------------------------------------
// Detector
// ---------------------------------------------------------------------------

pub struct InsecureCryptoDetector {
    repository_path: PathBuf,
    max_findings: usize,
}

impl InsecureCryptoDetector {
    crate::detectors::detector_new!(50);

    fn relative_path(&self, path: &Path) -> PathBuf {
        crate::detectors::detector_relative_path(&self.repository_path, path)
    }

    /// AST-first scanner. Walks the tree once, emitting findings for
    /// every call expression whose callee or string argument names a
    /// weak primitive.
    fn scan_file_ast(&self, inputs: &ScanAstInputs<'_>) -> Vec<Finding> {
        let path = inputs.path();
        let content = inputs.content();
        let ext = inputs.ext();
        let lang = inputs.lang;
        let cached_tree = inputs.cached_tree;
        let mut findings = vec![];
        if content.contains('\0') || content.len() > 500_000 {
            return findings;
        }

        let owned;
        let root = match cached_tree {
            Some(tree) => tree.root_node(),
            None => match parse_root_ext(content, lang, ext) {
                Some(t) => {
                    owned = t;
                    owned.root_node()
                }
                None => return findings,
            },
        };

        let bytes = content.as_bytes();
        let lines: Vec<&str> = content.lines().collect();

        // Build per-file from-import alias map (Python only). Lets us
        // resolve `from hashlib import md5; md5(data)` to a `hashlib.md5`
        // call without any whole-program analysis.
        let alias_map = if matches!(lang, Language::Python) {
            collect_python_from_imports(root, bytes)
        } else {
            HashMap::new()
        };
        // Per-file Python module-alias map: resolves
        // `import hashlib as hl; hl.md5(data)` by mapping the
        // attribute-receiver text `hl` back to the canonical
        // module name `hashlib` before matching.
        let module_aliases = if matches!(lang, Language::Python) {
            collect_python_module_aliases(root, bytes)
        } else {
            HashMap::new()
        };

        let mut sites: Vec<CryptoSite> = Vec::new();
        let ctx = AstWalkCtx {
            lang,
            source: bytes,
        };
        let aliases = PythonAliases::new(&alias_map, &module_aliases);
        collect_crypto_sites(&ctx, root, &aliases, &mut sites);

        for site in sites {
            if findings.len() >= self.max_findings {
                break;
            }
            let line_idx = site.call_node.start_position().row;
            if let Some(line) = lines.get(line_idx) {
                let prev = if line_idx > 0 {
                    Some(lines[line_idx - 1])
                } else {
                    None
                };
                if crate::detectors::is_line_suppressed(line, prev) {
                    continue;
                }
            }
            let snippet = lines.get(line_idx).map(|s| s.trim()).unwrap_or("");
            let line_num = (line_idx + 1) as u32;

            findings.push(self.build_finding(path, line_num, &site.kind, snippet, ext));
        }

        findings
    }

    /// Legacy line scanner for non-AST languages (Java, C/C++, Go, Ruby,
    /// PHP, C#).
    fn scan_file_line(&self, inputs: &ScanInputs<'_>) -> Vec<Finding> {
        let path = inputs.path;
        let content = inputs.content;
        let ext = inputs.ext;
        let mut findings = vec![];
        if content.len() > 500_000 {
            return findings;
        }
        let lines: Vec<&str> = content.lines().collect();
        for (i, line) in lines.iter().enumerate() {
            if findings.len() >= self.max_findings {
                break;
            }
            let prev = if i > 0 { Some(lines[i - 1]) } else { None };
            if crate::detectors::is_line_suppressed(line, prev) {
                continue;
            }
            let trimmed = line.trim_start();
            if trimmed.starts_with('#') || trimmed.starts_with("//") || trimmed.starts_with('*') {
                continue;
            }
            if let Some(kind) = match_line_crypto(line, ext) {
                let line_num = (i + 1) as u32;
                findings.push(self.build_finding(path, line_num, &kind, line.trim(), ext));
            }
        }
        findings
    }

    fn build_finding(
        &self,
        path: &Path,
        line_num: u32,
        kind: &WeakCryptoCallKind,
        snippet: &str,
        _ext: &str,
    ) -> Finding {
        let (title, suggested_fix, cwe) = match kind {
            WeakCryptoCallKind::BrokenAlgoIdentifier(name)
            | WeakCryptoCallKind::BrokenAlgoStringLiteral(name) => {
                let upper = name.to_uppercase();
                if matches!(
                    upper.as_str(),
                    "MD5" | "SHA1" | "SHA-1" | "MD4" | "MD2" | "SHA"
                ) {
                    (
                        format!("Weak hash algorithm ({upper})"),
                        "Use SHA-256 / SHA-3 / BLAKE3 instead. For password hashing use Argon2 or scrypt.".to_string(),
                        "CWE-328",
                    )
                } else {
                    (
                        format!("Weak cipher algorithm ({upper})"),
                        "Use AES-GCM or ChaCha20-Poly1305 instead.".to_string(),
                        "CWE-327",
                    )
                }
            }
            WeakCryptoCallKind::WeakModeCombo { algo, mode } => (
                format!(
                    "Weak cipher mode ({}/{})",
                    algo.to_uppercase(),
                    mode.to_uppercase()
                ),
                "ECB mode leaks plaintext patterns. Use AES-GCM or ChaCha20-Poly1305.".to_string(),
                "CWE-327",
            ),
            WeakCryptoCallKind::UnknownAlgoFromVariable => (
                "Cryptographic algorithm chosen at runtime".to_string(),
                "Validate the algorithm against an allowlist (SHA-256+, AES-GCM, ChaCha20-Poly1305).".to_string(),
                "CWE-327",
            ),
        };

        let description = format!(
            "**Insecure cryptographic primitive**\n\n\
             **Algorithm**: `{}`\n\n\
             **Location**: {}:{}\n\n\
             **Code**:\n```\n{}\n```\n\n\
             Cryptographically broken primitives (MD5, SHA-1, DES, RC4, ECB mode) can be \
             exploited via collision attacks, key recovery, or plaintext-pattern leakage. \
             Replace them with modern primitives.",
            kind.algo_label(),
            path.display(),
            line_num,
            snippet,
        );

        Finding {
            id: String::new(),
            detector: "InsecureCryptoDetector".to_string(),
            severity: severity_for(kind),
            title,
            description,
            affected_files: vec![self.relative_path(path)],
            line_start: Some(line_num),
            line_end: Some(line_num),
            suggested_fix: Some(suggested_fix),
            estimated_effort: Some("30 minutes".to_string()),
            category: Some("security".to_string()),
            cwe_id: Some(cwe.to_string()),
            why_it_matters: Some(
                "Broken primitives let attackers forge signatures, recover keys, or leak \
                 plaintext patterns."
                    .to_string(),
            ),
            ..Default::default()
        }
    }
}

impl Detector for InsecureCryptoDetector {
    fn name(&self) -> &'static str {
        "insecure-crypto"
    }
    fn description(&self) -> &'static str {
        "Detects weak cryptographic algorithms"
    }

    fn requires_graph(&self) -> bool {
        false
    }

    fn file_extensions(&self) -> &'static [&'static str] {
        SUPPORTED_EXTS
    }

    fn content_requirements(&self) -> crate::detectors::detector_context::ContentFlags {
        crate::detectors::detector_context::ContentFlags::HAS_CRYPTO
    }

    fn detect(
        &self,
        ctx: &crate::detectors::analysis_context::AnalysisContext,
    ) -> Result<Vec<Finding>> {
        let graph = ctx.graph;
        let files = &ctx.as_file_provider();
        debug!("Starting insecure-crypto detection (AST-first)");

        let mut findings: Vec<Finding> = Vec::new();

        for path in files.files_with_extensions(SUPPORTED_EXTS) {
            if findings.len() >= self.max_findings {
                break;
            }

            let path_str = path.to_string_lossy().to_lowercase();
            if crate::detectors::base::is_test_path(&path_str) {
                continue;
            }
            // Skip translation/localization files (French "des" = "of the").
            if path_str.contains("/lang/")
                || path_str.contains("/locale")
                || path_str.contains("/i18n/")
                || path_str.contains("/translations/")
                || path_str.contains("_lang")
                || path_str.contains(".lang.")
            {
                continue;
            }
            if path_str.contains("scripts/") || path_str.contains("/script/") {
                continue;
            }

            let content = match files.content(path) {
                Some(c) => c,
                None => continue,
            };
            // Cheap pre-filter: skip files with no crypto-related keyword.
            if !contains_any(CRYPTO_KEYWORD_FINDERS, &content) {
                continue;
            }

            let ext = path.extension().and_then(|e| e.to_str()).unwrap_or("");
            let scan = ScanInputs::new(path, &content, ext);
            let new_findings = if AST_EXTS.contains(&ext) {
                let cached = files.tree(path);
                let lang = Language::from_path(path);
                let ast_inputs = ScanAstInputs::new(scan, lang, cached.as_deref());
                self.scan_file_ast(&ast_inputs)
            } else {
                self.scan_file_line(&scan)
            };
            findings.extend(new_findings);
        }

        // Severity boost when call appears in a request-handler / route
        // function. Mirrors `eval_detector` / `command_injection` /
        // `pickle_detector`.
        static HANDLER_VERB_RE: LazyLock<Regex> = LazyLock::new(|| {
            Regex::new(r"^(get|post|put|delete|patch|head|options)[A-Z]").expect("valid regex")
        });
        for finding in &mut findings {
            if !matches!(finding.severity, Severity::High | Severity::Medium) {
                continue;
            }
            if let (Some(file_path), Some(line)) =
                (finding.affected_files.first(), finding.line_start)
            {
                let path_str = file_path.to_string_lossy().to_string();
                let i = graph.interner();
                if let Some(func) = graph.find_function_at(&path_str, line) {
                    let raw_name = func.node_name(i);
                    let name_lower = raw_name.to_lowercase();
                    let is_route = name_lower.contains("handler")
                        || name_lower.contains("route")
                        || name_lower.contains("endpoint")
                        || name_lower.contains("view")
                        || name_lower.contains("controller")
                        || name_lower.contains("middleware")
                        || name_lower.contains("request")
                        || name_lower.contains("response")
                        || HANDLER_VERB_RE.is_match(raw_name);
                    if is_route && finding.severity == Severity::High {
                        finding.severity = Severity::Critical;
                    }
                }
            }
        }

        info!(
            "InsecureCryptoDetector found {} potential vulnerabilities",
            findings.len()
        );
        Ok(findings)
    }

    fn bypass_postprocessor(&self) -> bool {
        true
    }
}

impl crate::detectors::RegisteredDetector for InsecureCryptoDetector {
    fn create(init: &crate::detectors::DetectorInit) -> std::sync::Arc<dyn Detector> {
        std::sync::Arc::new(Self::new(init.repo_path))
    }
}

// ---------------------------------------------------------------------------
// Pre-filter
// ---------------------------------------------------------------------------

/// Cheap pre-filter: does this file contain any crypto keyword?
///
/// Each callee name we match downstream MUST be covered by one of the
/// substrings here, otherwise the AST scan never runs on a file
/// containing only that callee. Mirrors the `pickle_detector` audit.
///
/// Cross-checked against every matcher below:
///   - `hashlib.md5` / `hashlib.new` → FIND_HASHLIB or FIND_MD5_LOWER.
///   - `from hashlib import md5; md5(data)` → FIND_HASHLIB.
///   - `Crypto.Cipher.DES.new` → FIND_CRYPTO_CIPHER or FIND_DES_UPPER.
///   - `cryptography.hazmat.primitives.hashes.MD5()` → FIND_HAZMAT or
///     FIND_MD5_UPPER.
///   - `crypto.createHash('md5')` → FIND_CREATE_HASH.
///   - `crypto.createCipheriv('aes-128-ecb', ...)` → FIND_CREATE_CIPHER
///     or FIND_ECB.
///   - `crypto.createSign('RSA-MD5')` → FIND_CREATE_SIGN.
///   - `MessageDigest.getInstance("MD5")` → FIND_MESSAGE_DIGEST or
///     FIND_MD5_UPPER.
///   - `Cipher.getInstance("DES/ECB/PKCS5Padding")` →
///     FIND_CIPHER_GETINSTANCE or FIND_DES_UPPER or FIND_ECB.
static CRYPTO_KEYWORD_FINDERS: &[&LazyLock<memchr::memmem::Finder<'static>>] = &[
    &FIND_HASHLIB,
    &FIND_CRYPTO_CIPHER,
    &FIND_HAZMAT,
    &FIND_CREATE_HASH,
    &FIND_CREATE_CIPHER,
    &FIND_CREATE_DECIPHER,
    &FIND_CREATE_SIGN,
    &FIND_REQUIRE_CRYPTO,
    &FIND_REQUIRE_CRYPTO_DQ,
    &FIND_MD5_UPPER,
    &FIND_SHA1_UPPER,
    &FIND_MD4_UPPER,
    &FIND_DES_UPPER,
    &FIND_RC4_UPPER,
    &FIND_RC2_UPPER,
    &FIND_BLOWFISH,
    &FIND_ECB,
    &FIND_MD5_LOWER,
    &FIND_SHA1_LOWER,
    &FIND_MESSAGE_DIGEST,
    &FIND_CIPHER_GETINSTANCE,
];

// ---------------------------------------------------------------------------
// AST walking
// ---------------------------------------------------------------------------

struct CryptoSite<'a> {
    call_node: tree_sitter::Node<'a>,
    kind: WeakCryptoCallKind,
}

fn collect_crypto_sites<'a>(
    ctx: &AstWalkCtx<'a>,
    node: tree_sitter::Node<'a>,
    aliases: &PythonAliases<'_>,
    out: &mut Vec<CryptoSite<'a>>,
) {
    if let Some(site) = match_crypto_site(node, ctx.source, ctx.lang, aliases) {
        out.push(site);
    }
    let mut cursor = node.walk();
    for child in node.children(&mut cursor) {
        collect_crypto_sites(ctx, child, aliases, out);
    }
}

fn match_crypto_site<'a>(
    node: tree_sitter::Node<'a>,
    source: &'a [u8],
    lang: Language,
    aliases: &PythonAliases<'_>,
) -> Option<CryptoSite<'a>> {
    match (node.kind(), lang) {
        ("call", Language::Python) => match_python_crypto_call(node, source, aliases),
        ("call_expression", Language::JavaScript | Language::TypeScript) => {
            match_js_crypto_call(node, source)
        }
        _ => None,
    }
}

// ---------------------------------------------------------------------------
// Python from-import alias resolution
// ---------------------------------------------------------------------------

use super::python_imports::{
    collect_python_from_imports, collect_python_module_aliases, PythonAliases,
};

// ---------------------------------------------------------------------------
// Python
// ---------------------------------------------------------------------------

/// Names that are broken at the *attribute* level (`hashlib.md5`,
/// `hashlib.sha1`, ...). Lowercase.
const PY_BROKEN_HASH_ATTRS: &[&str] = &["md5", "sha1", "md4", "md2", "sha"];

/// Match a Python `call` node against the weak-crypto API list.
fn match_python_crypto_call<'a>(
    node: tree_sitter::Node<'a>,
    source: &'a [u8],
    aliases: &PythonAliases<'_>,
) -> Option<CryptoSite<'a>> {
    let func = node.child_by_field_name("function")?;
    let func = unwrap_callee(func);
    let args = node.child_by_field_name("arguments")?;
    let arg_nodes = collect_named_args(args);

    let kind = match func.kind() {
        "attribute" => {
            let obj = func.child_by_field_name("object")?;
            let attr = func.child_by_field_name("attribute")?;
            let attr_text = node_text(attr, source)?.to_string();
            let raw_label = receiver_chain_label(obj, source);
            // Resolve `import hashlib as hl; hl.md5(...)` — if the
            // receiver is an aliased module, swap the label for the
            // canonical module name before matching. The map's identity
            // entries (`{"hashlib": "hashlib"}`) make this lookup safe
            // even for unaliased imports; an unknown receiver falls
            // through to the literal label.
            //
            // Look up against the case-sensitive identifier text first
            // (the import map preserves source case), then fall back to
            // the lowercased label `receiver_chain_label` produces.
            let obj_text = node_text(obj, source).unwrap_or("");
            let obj_label = aliases
                .modules
                .get(obj_text)
                .or_else(|| aliases.modules.get(raw_label.as_str()))
                .cloned()
                .unwrap_or(raw_label);

            // hashlib.md5 / hashlib.sha1 / hashlib.md4 ...
            if obj_label == "hashlib"
                && PY_BROKEN_HASH_ATTRS.contains(&attr_text.to_lowercase().as_str())
            {
                WeakCryptoCallKind::BrokenAlgoIdentifier(attr_text)
            }
            // hashlib.new('md5')
            else if obj_label == "hashlib" && attr_text == "new" {
                let first_arg = arg_nodes.first().copied();
                match first_arg.and_then(|n| python_string_literal_value(n, source)) {
                    Some(s) => match classify_algo_name(&s) {
                        AlgoKind::Broken(n) => WeakCryptoCallKind::BrokenAlgoStringLiteral(n),
                        AlgoKind::Safe => return None,
                        AlgoKind::Unknown => return None,
                    },
                    None => match first_arg {
                        Some(n) if !is_static_string(n) => {
                            WeakCryptoCallKind::UnknownAlgoFromVariable
                        }
                        _ => return None,
                    },
                }
            }
            // Crypto.Cipher.DES.new(...) / Crypto.Cipher.AES.new(key, AES.MODE_ECB, ...)
            // Also PyCrypto-imported `DES.new(...)` after `from Crypto.Cipher import DES`.
            else if attr_text == "new" {
                let recv_full = node_text(obj, source).unwrap_or("");
                let last = recv_full.rsplit('.').next().unwrap_or(recv_full);
                let last_lower = last.to_lowercase();
                match classify_algo_name(last) {
                    AlgoKind::Broken(_) => {
                        WeakCryptoCallKind::BrokenAlgoIdentifier(last.to_string())
                    }
                    _ if last_lower == "aes" => {
                        // AES is fine — but check for ECB mode in the
                        // second positional argument.
                        if let Some(mode_arg) = positional_arg(&arg_nodes, 1) {
                            let mode_text = node_text(mode_arg, source).unwrap_or("");
                            if mode_text.to_uppercase().ends_with("MODE_ECB") {
                                WeakCryptoCallKind::WeakModeCombo {
                                    algo: "aes".to_string(),
                                    mode: "ecb".to_string(),
                                }
                            } else {
                                return None;
                            }
                        } else {
                            return None;
                        }
                    }
                    _ => return None,
                }
            }
            // cryptography.hazmat.primitives.hashes.MD5()
            else if matches!(attr_text.to_uppercase().as_str(), "MD5" | "SHA1" | "MD4")
                && obj_label == "hashes"
            {
                WeakCryptoCallKind::BrokenAlgoIdentifier(attr_text)
            } else {
                return None;
            }
        }
        // Bare-identifier call: only interesting if it's a from-import alias.
        "identifier" => {
            let name = node_text(func, source)?.to_string();
            if let Some(module) = aliases.imports.get(&name) {
                if module == "hashlib"
                    && PY_BROKEN_HASH_ATTRS.contains(&name.to_lowercase().as_str())
                {
                    WeakCryptoCallKind::BrokenAlgoIdentifier(name)
                } else {
                    return None;
                }
            } else {
                return None;
            }
        }
        _ => return None,
    };

    Some(CryptoSite {
        call_node: node,
        kind,
    })
}

fn positional_arg<'a>(args: &[tree_sitter::Node<'a>], idx: usize) -> Option<tree_sitter::Node<'a>> {
    let mut count = 0;
    for a in args {
        if a.kind() == "keyword_argument" {
            continue;
        }
        if count == idx {
            return Some(*a);
        }
        count += 1;
    }
    None
}

fn is_static_string(node: tree_sitter::Node<'_>) -> bool {
    matches!(node.kind(), "string" | "concatenated_string")
}

fn python_string_literal_value(node: tree_sitter::Node<'_>, source: &[u8]) -> Option<String> {
    if node.kind() != "string" {
        return None;
    }
    let raw = node_text(node, source)?;
    let bytes = raw.as_bytes();
    let mut i = 0;
    while i < bytes.len() && bytes[i].is_ascii_alphabetic() {
        i += 1;
    }
    let after_prefix = &raw[i..];
    let inner = after_prefix
        .strip_prefix("\"\"\"")
        .and_then(|s| s.strip_suffix("\"\"\""))
        .or_else(|| {
            after_prefix
                .strip_prefix("'''")
                .and_then(|s| s.strip_suffix("'''"))
        })
        .or_else(|| {
            after_prefix
                .strip_prefix('"')
                .and_then(|s| s.strip_suffix('"'))
        })
        .or_else(|| {
            after_prefix
                .strip_prefix('\'')
                .and_then(|s| s.strip_suffix('\''))
        })?;
    Some(inner.to_string())
}

// ---------------------------------------------------------------------------
// JavaScript / TypeScript
// ---------------------------------------------------------------------------

fn match_js_crypto_call<'a>(
    node: tree_sitter::Node<'a>,
    source: &'a [u8],
) -> Option<CryptoSite<'a>> {
    let func = node.child_by_field_name("function")?;
    let args = node.child_by_field_name("arguments")?;
    let arg_nodes = collect_named_args(args);
    let func = unwrap_callee(func);

    let kind = match func.kind() {
        "member_expression" => {
            let obj = func.child_by_field_name("object")?;
            let prop = func.child_by_field_name("property")?;
            let prop_text = node_text(prop, source)?.to_string();
            let recv = receiver_chain_label(obj, source);

            // Only inspect calls whose receiver looks like the `crypto`
            // module (bare alias `crypto`, `Crypto`, or resolved through
            // `require('crypto')` / `import('crypto')`).
            let is_crypto_recv =
                recv == "crypto" || recv == "Crypto".to_lowercase() || recv == "node:crypto";

            if !is_crypto_recv {
                return None;
            }

            match prop_text.as_str() {
                "createHash" | "createHmac" => {
                    classify_string_algo_arg(&arg_nodes, source, /*allow_mode=*/ false)?
                }
                "createCipheriv" | "createDecipheriv" | "createCipher" | "createDecipher" => {
                    classify_string_algo_arg(&arg_nodes, source, /*allow_mode=*/ true)?
                }
                "createSign" | "createVerify" => {
                    classify_string_algo_arg(&arg_nodes, source, /*allow_mode=*/ false)?
                }
                _ => return None,
            }
        }
        _ => return None,
    };

    Some(CryptoSite {
        call_node: node,
        kind,
    })
}

/// Inspect the first argument of a JS `crypto.createX(...)` call and
/// return the appropriate `WeakCryptoCallKind`.
fn classify_string_algo_arg(
    args: &[tree_sitter::Node<'_>],
    source: &[u8],
    allow_mode: bool,
) -> Option<WeakCryptoCallKind> {
    let first = args.first().copied()?;
    let first = unwrap_arg(first);
    if let Some(s) = js_string_literal_value(first, source) {
        // Case 1: explicit algo+mode label like "aes-128-ecb".
        if allow_mode {
            if let Some((algo, mode)) = split_algo_mode(&s) {
                let algo_class = classify_algo_name(&algo);
                if mode.eq_ignore_ascii_case("ecb") || matches!(algo_class, AlgoKind::Broken(_)) {
                    return Some(WeakCryptoCallKind::WeakModeCombo { algo, mode });
                }
                if matches!(algo_class, AlgoKind::Safe) {
                    return None;
                }
            }
        }
        // Case 2: simple algo name like "md5" or composite like "RSA-MD5".
        let lower = s.to_lowercase();
        for token in lower.split(['-', '_', '/']) {
            match classify_algo_name(token) {
                AlgoKind::Broken(_) => {
                    return Some(WeakCryptoCallKind::BrokenAlgoStringLiteral(s));
                }
                AlgoKind::Safe => {
                    return None;
                }
                AlgoKind::Unknown => continue,
            }
        }
        match classify_algo_name(&s) {
            AlgoKind::Broken(_) => Some(WeakCryptoCallKind::BrokenAlgoStringLiteral(s)),
            _ => None,
        }
    } else {
        // Non-literal — algorithm chosen at runtime.
        Some(WeakCryptoCallKind::UnknownAlgoFromVariable)
    }
}

/// Descend through transparent wrappers (`await`, parens, TS type ops).
fn unwrap_arg(mut node: tree_sitter::Node<'_>) -> tree_sitter::Node<'_> {
    loop {
        match node.kind() {
            "parenthesized_expression"
            | "await_expression"
            | "as_expression"
            | "type_assertion_expression"
            | "non_null_expression"
            | "satisfies_expression" => {
                let next = node.named_child(0);
                match next {
                    Some(n) => node = n,
                    None => return node,
                }
            }
            _ => return node,
        }
    }
}

// ---------------------------------------------------------------------------
// Helpers shared with eval / command_injection / pickle.
// `unwrap_callee`, `collect_named_args`, `node_text` live in `ast_helpers`
// (imported above). Only `receiver_chain_label` (which uses this
// detector's `call_expression_module_label`) and `js_string_literal_value`
// are kept local.
// ---------------------------------------------------------------------------

/// Lowercased "receiver label" for a JS/TS or Python member-call
/// receiver. Mirrors `pickle_detector::receiver_chain_label`.
///
/// Resolves `require('crypto')` / `(await import('crypto'))` to the
/// canonical lowercased module name, so:
///
///   - `crypto.createHash(...)` → label `"crypto"`
///   - `require('crypto').createHash(...)` → label `"crypto"`
///   - `(await import('crypto')).createHash(...)` → label `"crypto"`
///   - `self.hashlib.md5(...)` → label `"hashlib"`
///
/// Implementation: delegate to the shared
/// [`receiver_chain_label`](crate::detectors::security::ast_helpers::receiver_chain_label)
/// passing this detector's [`call_expression_module_label`] as the
/// resolver — that's the only piece that varies between detectors (it
/// names the dangerous crypto module(s) for *this* detector, e.g.
/// `crypto`).
fn receiver_chain_label(node: tree_sitter::Node<'_>, source: &[u8]) -> String {
    receiver_chain_label_shared(node, source, Some(&call_expression_module_label))
}

/// If `node` is `require('MODULE')` or `import('MODULE')` for a known
/// crypto module, return the canonical lowercased module name.
fn call_expression_module_label(
    node: tree_sitter::Node<'_>,
    source: &[u8],
) -> Option<&'static str> {
    debug_assert_eq!(node.kind(), "call_expression");
    let func = node.child_by_field_name("function")?;
    let func_text = node_text(func, source)?;
    let is_require_or_import =
        matches!(func.kind(), "identifier" | "import") && matches!(func_text, "require" | "import");
    if !is_require_or_import {
        return None;
    }
    let args = node.child_by_field_name("arguments")?;
    let arg_nodes = collect_named_args(args);
    let first = arg_nodes.first()?;
    let module = js_string_literal_value(*first, source)?;
    match module.as_str() {
        "crypto" | "node:crypto" => Some("crypto"),
        _ => None,
    }
}

fn js_string_literal_value(node: tree_sitter::Node<'_>, source: &[u8]) -> Option<String> {
    if node.kind() != "string" {
        return None;
    }
    let mut cursor = node.walk();
    let mut buf = String::new();
    let mut saw_fragment = false;
    for child in node.children(&mut cursor) {
        if child.kind() == "string_fragment" {
            if let Some(t) = node_text(child, source) {
                buf.push_str(t);
                saw_fragment = true;
            }
        }
    }
    if saw_fragment {
        return Some(buf);
    }
    let raw = node_text(node, source)?;
    let inner = raw
        .strip_prefix('"')
        .and_then(|s| s.strip_suffix('"'))
        .or_else(|| raw.strip_prefix('\'').and_then(|s| s.strip_suffix('\'')))?;
    Some(inner.to_string())
}

// `node_text` lives in `ast_helpers`; imported above.

// ---------------------------------------------------------------------------
// Line scanner (Java, C/C++, Go, Ruby, PHP, C#)
// ---------------------------------------------------------------------------

/// Recognise weak crypto on a line for languages without AST coverage.
fn match_line_crypto(line: &str, _ext: &str) -> Option<WeakCryptoCallKind> {
    static MESSAGE_DIGEST_RE: LazyLock<Regex> = LazyLock::new(|| {
        Regex::new(r#"MessageDigest\.getInstance\s*\(\s*["']([^"']+)["']"#).expect("valid regex")
    });
    static CIPHER_GETINSTANCE_RE: LazyLock<Regex> = LazyLock::new(|| {
        Regex::new(r#"Cipher\.getInstance\s*\(\s*["']([^"']+)["']"#).expect("valid regex")
    });
    static MCRYPT_RE: LazyLock<Regex> = LazyLock::new(|| {
        Regex::new(r"mcrypt_(?:encrypt|decrypt)\s*\(\s*MCRYPT_(\w+)").expect("valid regex")
    });
    static OPENSSL_ENCRYPT_RE: LazyLock<Regex> = LazyLock::new(|| {
        Regex::new(r#"openssl_(?:encrypt|decrypt)\s*\([^,]*,\s*["']([^"']+)["']"#)
            .expect("valid regex")
    });
    static GO_DES_RE: LazyLock<Regex> = LazyLock::new(|| {
        Regex::new(r"\bdes\.NewCipher\s*\(|\bdes\.NewTripleDESCipher\s*\(").expect("valid regex")
    });
    static GO_RC4_RE: LazyLock<Regex> =
        LazyLock::new(|| Regex::new(r"\brc4\.NewCipher\s*\(").expect("valid regex"));
    static GO_WEAK_HASH_RE: LazyLock<Regex> =
        LazyLock::new(|| Regex::new(r"\b(md5|sha1)\s*\.\s*(?:New|Sum)\s*\(").expect("valid regex"));
    static RUBY_DIGEST_RE: LazyLock<Regex> =
        LazyLock::new(|| Regex::new(r"\bDigest::(MD5|SHA1|MD4|MD2)\b").expect("valid regex"));
    static C_DIRECT_HASH_RE: LazyLock<Regex> = LazyLock::new(|| {
        Regex::new(r"\b(MD5|SHA1|MD4|MD2|DES_set_key|DES_ecb_encrypt|RC4)\s*\(")
            .expect("valid regex")
    });

    if let Some(cap) = MESSAGE_DIGEST_RE.captures(line) {
        let algo = &cap[1];
        if let AlgoKind::Broken(name) = classify_algo_name(algo) {
            return Some(WeakCryptoCallKind::BrokenAlgoStringLiteral(name));
        }
    }
    if let Some(cap) = CIPHER_GETINSTANCE_RE.captures(line) {
        let spec = &cap[1];
        // Java cipher specs are `algo/mode/padding` (e.g. `DES/ECB/PKCS5Padding`).
        let parts: Vec<&str> = spec.split('/').collect();
        let algo = parts.first().copied().unwrap_or("");
        let mode = parts.get(1).copied().unwrap_or("");
        match classify_algo_name(algo) {
            AlgoKind::Broken(n) => {
                return Some(WeakCryptoCallKind::BrokenAlgoStringLiteral(n));
            }
            _ => {
                if mode.eq_ignore_ascii_case("ecb") {
                    return Some(WeakCryptoCallKind::WeakModeCombo {
                        algo: algo.to_string(),
                        mode: mode.to_string(),
                    });
                }
            }
        }
    }
    if let Some(cap) = MCRYPT_RE.captures(line) {
        let algo = &cap[1];
        if let AlgoKind::Broken(name) = classify_algo_name(algo) {
            return Some(WeakCryptoCallKind::BrokenAlgoStringLiteral(name));
        }
    }
    if let Some(cap) = OPENSSL_ENCRYPT_RE.captures(line) {
        let spec = &cap[1];
        let parts: Vec<&str> = spec.split('-').collect();
        let algo = parts.first().copied().unwrap_or("");
        let mode = parts.last().copied().unwrap_or("");
        if let AlgoKind::Broken(name) = classify_algo_name(algo) {
            return Some(WeakCryptoCallKind::BrokenAlgoStringLiteral(name));
        }
        if mode.eq_ignore_ascii_case("ecb") {
            return Some(WeakCryptoCallKind::WeakModeCombo {
                algo: algo.to_string(),
                mode: mode.to_string(),
            });
        }
    }
    if GO_DES_RE.is_match(line) {
        return Some(WeakCryptoCallKind::BrokenAlgoIdentifier("DES".to_string()));
    }
    if GO_RC4_RE.is_match(line) {
        return Some(WeakCryptoCallKind::BrokenAlgoIdentifier("RC4".to_string()));
    }
    if let Some(cap) = GO_WEAK_HASH_RE.captures(line) {
        return Some(WeakCryptoCallKind::BrokenAlgoIdentifier(
            cap[1].to_uppercase(),
        ));
    }
    if let Some(cap) = RUBY_DIGEST_RE.captures(line) {
        return Some(WeakCryptoCallKind::BrokenAlgoIdentifier(cap[1].to_string()));
    }
    if let Some(cap) = C_DIRECT_HASH_RE.captures(line) {
        let name = cap[1].to_string();
        let canonical = if name.starts_with("DES_") {
            "DES".to_string()
        } else if name.starts_with("RC4") {
            "RC4".to_string()
        } else {
            name
        };
        return Some(WeakCryptoCallKind::BrokenAlgoIdentifier(canonical));
    }
    None
}

// ---------------------------------------------------------------------------
// Tests
// ---------------------------------------------------------------------------

#[cfg(test)]
mod tests {
    use super::*;
    use crate::graph::builder::GraphBuilder;

    fn run(file: &str, content: &str) -> Vec<Finding> {
        let store = GraphBuilder::new().freeze();
        let detector = InsecureCryptoDetector::new("/mock/repo");
        let ctx = crate::detectors::analysis_context::AnalysisContext::test_with_mock_files(
            &store,
            vec![(file, content)],
        );
        detector.detect(&ctx).expect("detection should succeed")
    }

    // -----------------------------------------------------------------
    // Pre-existing tests (preserved through migration).
    // -----------------------------------------------------------------

    #[test]
    fn test_detects_md5_usage() {
        let findings = run(
            "crypto_util.py",
            "import hashlib\n\ndef compute_hash(data):\n    return hashlib.md5(data).hexdigest()\n",
        );
        assert!(!findings.is_empty(), "Should detect hashlib.md5 usage");
    }

    #[test]
    fn test_no_finding_for_sha256() {
        let findings = run(
            "crypto_util.py",
            "import hashlib\n\ndef compute_hash(data):\n    return hashlib.sha256(data).hexdigest()\n",
        );
        assert!(
            findings.is_empty(),
            "sha256 must not fire. Got: {:?}",
            findings.iter().map(|f| &f.title).collect::<Vec<_>>()
        );
    }

    #[test]
    fn test_no_finding_for_class_definition() {
        let findings = run(
            "text.py",
            "from django.db.models import Transform\n\nclass MD5(Transform):\n    function = 'MD5'\n    lookup_name = 'md5'\n",
        );
        assert!(
            findings.is_empty(),
            "class MD5(...) is a class def, not a call. Got: {:?}",
            findings.iter().map(|f| &f.title).collect::<Vec<_>>()
        );
    }

    #[test]
    fn test_detects_java_des_cipher() {
        let findings = run(
            "CryptoUtil.java",
            "import javax.crypto.*;\n\npublic class CryptoUtil {\n    public byte[] encrypt(byte[] data) throws Exception {\n        Cipher cipher = Cipher.getInstance(\"DES\");\n        return cipher.doFinal(data);\n    }\n}\n",
        );
        assert!(
            !findings.is_empty(),
            "Cipher.getInstance(\"DES\") must fire"
        );
    }

    #[test]
    fn test_detects_go_md5_new_and_sum() {
        let findings = run(
            "hash.go",
            r#"
package main
import "crypto/md5"
func hash(payload []byte) {
    h := md5.New()
    _ = h.Sum(payload)
    _ = md5.Sum(payload)
}
"#,
        );

        assert!(findings.iter().any(|f| f.line_start == Some(5)));
        assert!(findings.iter().any(|f| f.line_start == Some(7)));
    }

    #[test]
    fn test_detects_go_sha1_sum() {
        let findings = run(
            "hash.go",
            r#"
package main
import "crypto/sha1"
func hash(payload []byte) {
    _ = sha1.Sum(payload)
}
"#,
        );

        assert!(findings.iter().any(|f| f.line_start == Some(5)));
    }

    // -----------------------------------------------------------------
    // Audit / regression tests for the AST-first migration.
    //
    // Cohort 1 (currently-passing-shape, 5)
    // Cohort 2 (audit-shape, 4)
    // Cohort 3 (audit-pending, 3)
    // -----------------------------------------------------------------

    // ----- Cohort 1: currently-passing-shape -----

    #[test]
    fn test_detects_hashlib_md5_python() {
        let findings = run(
            "h.py",
            "import hashlib\ndef hash_pwd(p):\n    return hashlib.md5(p).hexdigest()\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "hashlib.md5 should fire High. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    #[test]
    fn test_detects_hashlib_sha1_python() {
        let findings = run(
            "h.py",
            "import hashlib\ndef hash_pwd(p):\n    return hashlib.sha1(p).hexdigest()\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "hashlib.sha1 should fire High. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    #[test]
    fn test_detects_crypto_createhash_md5_js() {
        let findings = run(
            "h.js",
            "const crypto = require('crypto');\nfunction f(d) { return crypto.createHash('md5').update(d).digest(); }\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "crypto.createHash('md5') should fire High. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    #[test]
    fn test_skips_hashlib_sha256_python() {
        let findings = run(
            "h.py",
            "import hashlib\ndef ok(p):\n    return hashlib.sha256(p).hexdigest()\n",
        );
        assert!(
            findings.is_empty(),
            "hashlib.sha256 must not fire. Got: {:?}",
            findings.iter().map(|f| &f.title).collect::<Vec<_>>()
        );
    }

    #[test]
    fn test_skips_md5_in_comment() {
        let findings = run(
            "h.py",
            "import hashlib\ndef ok(p):\n    # Use hashlib.md5 if needed\n    return hashlib.sha256(p).hexdigest()\n",
        );
        assert!(
            findings.is_empty(),
            "md5 in comment must not fire. Got: {:?}",
            findings.iter().map(|f| &f.title).collect::<Vec<_>>()
        );
    }

    // ----- Cohort 2: audit-shape -----

    /// B1-class: `require('crypto').createHash('md5')` — receiver-chain
    /// label resolves the require() callee to the canonical module name.
    #[test]
    fn test_b1_require_crypto_createhash_md5_js() {
        let findings = run(
            "go.js",
            "function go(data) { return require('crypto').createHash('md5').update(data).digest(); }\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "B1: require('crypto').createHash('md5') must fire. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    /// `class C: def md5(self): pass` — `md5` is a method definition. The
    /// AST sees a function definition, not a `hashlib.md5(...)` call, so
    /// nothing fires.
    #[test]
    fn test_skips_md5_as_method_name_python() {
        let findings = run(
            "c.py",
            "class C:\n    def md5(self, data):\n        return data\n\nc = C()\nresult = c.md5(b'x')\n",
        );
        // `c.md5(...)` must NOT fire — receiver is `c`, not `hashlib`.
        assert!(
            findings.is_empty(),
            "Method name md5 must not be confused with hashlib.md5. Got: {:?}",
            findings.iter().map(|f| &f.title).collect::<Vec<_>>()
        );
    }

    /// `crypto.createCipheriv('aes-128-ecb', ...)` — algo+mode label.
    #[test]
    fn test_aes_ecb_mode_combo_detected_js() {
        let findings = run(
            "c.js",
            "const crypto = require('crypto');\nfunction f(k, iv) { return crypto.createCipheriv('aes-128-ecb', k, iv); }\n",
        );
        assert!(
            !findings.is_empty(),
            "AES/ECB combo must fire. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    /// `let msg = "compute md5 hash"` — string-literal text, no call.
    #[test]
    fn test_skips_md5_in_string_literal() {
        let findings = run(
            "s.js",
            "function doc() { let msg = \"compute md5 hash\"; return msg; }\n",
        );
        assert!(
            findings.is_empty(),
            "md5 inside a string literal must not fire. Got: {:?}",
            findings.iter().map(|f| &f.title).collect::<Vec<_>>()
        );
    }

    // ----- Cohort 3: audit-pending (un-ignored as the migration lands) -----

    /// `from hashlib import md5; md5(data)` — bare-call after from-import.
    /// First detector in the codebase to do this; alias map is built per
    /// file in `collect_python_from_imports`.
    #[test]
    fn test_python_bare_md5_after_from_import() {
        let findings = run(
            "h.py",
            "from hashlib import md5\ndef hash_pwd(p):\n    return md5(p).hexdigest()\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "Bare md5(...) after from hashlib import md5 must fire. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    /// `from Crypto.Cipher import DES; DES.new(key, DES.MODE_ECB)` — DES
    /// is broken (any mode); also AES.new(..., AES.MODE_ECB, ...).
    #[test]
    fn test_pycrypto_des_new_detected() {
        let findings = run(
            "p.py",
            "from Crypto.Cipher import DES\ndef encrypt(data, key):\n    cipher = DES.new(key, DES.MODE_ECB)\n    return cipher.encrypt(data)\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "PyCrypto DES.new must fire High. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    /// `crypto.createHash(userChoice)` — algo is a non-literal expression;
    /// classifier emits `UnknownAlgoFromVariable` → Low.
    #[test]
    fn test_unknown_algo_from_variable_low_severity() {
        let findings = run(
            "v.js",
            "const crypto = require('crypto');\nfunction f(userChoice, d) { return crypto.createHash(userChoice).update(d).digest(); }\n",
        );
        assert!(
            findings.iter().any(|f| matches!(f.severity, Severity::Low)),
            "Variable-algo createHash must produce a Low finding. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }

    /// Audit shape: `import hashlib as hl; hl.md5(data)`.
    ///
    /// `hl.md5(...)` parses as an attribute call whose object text is
    /// `"hl"`, not `"hashlib"`. Without the module-alias resolver the
    /// matcher compares `obj_label == "hashlib"` and misses. Mirrors
    /// `test_python_bare_md5_after_from_import`, but for the
    /// `import M as N` shape.
    #[test]
    fn test_python_aliased_module_hashlib_md5_detected() {
        let findings = run(
            "h.py",
            "import hashlib as hl\ndef hash_pwd(p):\n    return hl.md5(p).hexdigest()\n",
        );
        assert!(
            findings
                .iter()
                .any(|f| matches!(f.severity, Severity::High | Severity::Critical)),
            "Aliased `hl.md5(...)` after `import hashlib as hl` must fire. Got: {:?}",
            findings
                .iter()
                .map(|f| (&f.title, f.severity))
                .collect::<Vec<_>>()
        );
    }
}