common/

filter.rs

//! Pattern-based file filtering for include/exclude operations
//!
//! This module provides glob pattern matching for filtering files during copy, link, and remove operations.
//!
//! # Pattern Syntax
//!
//! - `*` matches anything except `/`
//! - `**` matches anything including `/` (crosses directories)
//! - `?` matches a single character (except `/`)
//! - `[...]` character classes
//! - Leading `/` anchors to source root
//! - Trailing `/` matches only directories
//!
//! # Examples
//!
//! ```
//! use common::filter::{FilterSettings, FilterResult};
//! use std::path::Path;
//!
//! let mut settings = FilterSettings::default();
//! settings.add_exclude("*.log").unwrap();
//! settings.add_exclude("target/").unwrap();
//!
//! // .log files are excluded
//! assert!(matches!(
//!     settings.should_include(Path::new("debug.log"), false),
//!     FilterResult::ExcludedByPattern(_)
//! ));
//!
//! // other files are included
//! assert!(matches!(
//!     settings.should_include(Path::new("main.rs"), false),
//!     FilterResult::Included
//! ));
//! ```

use anyhow::{Context, anyhow};
use serde::{Deserialize, Deserializer, Serialize, Serializer};
use std::path::Path;

/// A compiled filter pattern with metadata about its original form
#[derive(Debug, Clone)]
pub struct FilterPattern {
    /// original pattern string for dry-run explain output
    pub original: String,
    /// compiled glob matcher
    matcher: globset::GlobMatcher,
    /// pattern ends with / (matches only directories)
    pub dir_only: bool,
    /// pattern starts with / (anchored to root)
    pub anchored: bool,
}

impl FilterPattern {
    /// Parse a pattern string into a FilterPattern
    pub fn parse(pattern: &str) -> Result<Self, anyhow::Error> {
        if pattern.is_empty() {
            return Err(anyhow!("empty pattern is not allowed"));
        }
        let original = pattern.to_string();
        let dir_only = pattern.ends_with('/');
        let anchored = pattern.starts_with('/');
        // strip leading/trailing markers for glob compilation
        let pattern_str = pattern.trim_start_matches('/').trim_end_matches('/');
        if pattern_str.is_empty() {
            return Err(anyhow!(
                "pattern '{}' results in empty glob after stripping / markers",
                pattern
            ));
        }
        // build glob with appropriate settings
        let glob = globset::GlobBuilder::new(pattern_str)
            .literal_separator(true) // * doesn't match /
            .build()
            .with_context(|| format!("invalid glob pattern: {}", pattern))?;
        let matcher = glob.compile_matcher();
        Ok(Self {
            original,
            matcher,
            dir_only,
            anchored,
        })
    }
    /// Check if this pattern contains path separators (excluding leading/trailing / markers).
    /// Path patterns require full path matching, while simple patterns can match filenames.
    fn is_path_pattern(&self) -> bool {
        // strip leading / (anchored marker) and trailing / (dir-only marker)
        let core = self.original.trim_start_matches('/').trim_end_matches('/');
        core.contains('/')
    }
    /// Check if this pattern matches the given path
    pub fn matches(&self, relative_path: &Path, is_dir: bool) -> bool {
        // directory-only patterns only match directories
        if self.dir_only && !is_dir {
            return false;
        }
        if self.anchored {
            // anchored patterns match from the root only
            self.matcher.is_match(relative_path)
        } else {
            // non-anchored patterns can match any component or the full path
            // first try full path match
            if self.matcher.is_match(relative_path) {
                return true;
            }
            // for non-anchored patterns, also try matching against just the filename
            // unless it's a path pattern (in which case full path match is required)
            if !self.is_path_pattern()
                && let Some(file_name) = relative_path.file_name()
                && self.matcher.is_match(Path::new(file_name))
            {
                return true;
            }
            false
        }
    }
}

/// Result of checking whether a path should be included
#[derive(Debug, Clone)]
pub enum FilterResult {
    /// path should be processed
    Included,
    /// path was excluded because include patterns exist but none matched
    ExcludedByDefault,
    /// path was excluded by a specific pattern
    ExcludedByPattern(String),
}

/// Settings for filtering files based on include/exclude patterns
#[derive(Debug, Clone, Default)]
pub struct FilterSettings {
    /// patterns for files to include (if non-empty, only matching files are included)
    pub includes: Vec<FilterPattern>,
    /// patterns for files to exclude
    pub excludes: Vec<FilterPattern>,
}

impl FilterSettings {
    /// Create new empty filter settings
    pub fn new() -> Self {
        Self::default()
    }
    /// Add an include pattern
    pub fn add_include(&mut self, pattern: &str) -> Result<(), anyhow::Error> {
        self.includes.push(FilterPattern::parse(pattern)?);
        Ok(())
    }
    /// Add an exclude pattern
    pub fn add_exclude(&mut self, pattern: &str) -> Result<(), anyhow::Error> {
        self.excludes.push(FilterPattern::parse(pattern)?);
        Ok(())
    }
    /// Check if this filter has any patterns
    pub fn is_empty(&self) -> bool {
        self.includes.is_empty() && self.excludes.is_empty()
    }
    /// Check if this filter has any include patterns
    pub fn has_includes(&self) -> bool {
        !self.includes.is_empty()
    }
    /// Determine if a root item (the source itself) should be included based on filter patterns.
    ///
    /// This is a specialized version of `should_include` for root items (the source directory
    /// or file being copied). Anchored patterns (starting with `/`) are skipped because they
    /// are meant to match paths INSIDE the source root, not the source root itself.
    ///
    /// For root files, only non-anchored simple patterns (no `/`) can directly match the name.
    /// For root directories, they're always traversed if include patterns exist (to find
    /// matching content inside).
    ///
    /// For example, pattern `/bar` on source `foo/` should match `foo/bar`, not filter out `foo`.
    pub fn should_include_root_item(&self, name: &Path, is_dir: bool) -> FilterResult {
        // check excludes first - only non-anchored, simple patterns apply to root items
        // (patterns with path separators like `bar/baz` match content inside, not the root)
        // note: trailing `/` is a dir-only marker, not a path separator
        for pattern in &self.excludes {
            if !pattern.anchored && !pattern.is_path_pattern() && pattern.matches(name, is_dir) {
                return FilterResult::ExcludedByPattern(pattern.original.clone());
            }
        }
        // if there are include patterns...
        if !self.includes.is_empty() {
            // for root files, check if any non-anchored simple pattern matches
            if !is_dir {
                for pattern in &self.includes {
                    if !pattern.anchored
                        && !pattern.is_path_pattern()
                        && pattern.matches(name, false)
                    {
                        return FilterResult::Included;
                    }
                }
                // no simple pattern matched the root file
                return FilterResult::ExcludedByDefault;
            }
            // for root directories, always traverse to find matching content inside
            // (both anchored patterns like /bar and path patterns like bar/*.txt
            // need us to traverse the directory to find matches)
            return FilterResult::Included;
        }
        // no includes and not excluded = included
        FilterResult::Included
    }
    /// Determine if a path should be included based on filter patterns
    ///
    /// # Precedence
    /// - If only excludes: include everything except matches
    /// - If only includes: include only matches (exclude everything else by default)
    /// - If both: excludes take priority (excludes checked first, then includes)
    ///
    /// # Directory handling
    /// Directories are traversed when include patterns exist if they could potentially contain
    /// matching files. For non-anchored patterns (like `*.txt`), all directories are traversed.
    /// For anchored patterns (like `/bar`), only directories matching the pattern prefix are traversed.
    pub fn should_include(&self, relative_path: &Path, is_dir: bool) -> FilterResult {
        // check excludes first - if matched, path is excluded
        for pattern in &self.excludes {
            if pattern.matches(relative_path, is_dir) {
                return FilterResult::ExcludedByPattern(pattern.original.clone());
            }
        }
        // if there are include patterns, at least one must match
        if !self.includes.is_empty() {
            // first check if this path matches any include pattern
            for pattern in &self.includes {
                if pattern.matches(relative_path, is_dir) {
                    return FilterResult::Included;
                }
            }
            // for directories that don't directly match, check if they could contain matches
            if is_dir {
                for pattern in &self.includes {
                    if self.could_contain_matches(relative_path, pattern) {
                        return FilterResult::Included;
                    }
                }
            }
            return FilterResult::ExcludedByDefault;
        }
        // no includes specified and not excluded = included
        FilterResult::Included
    }
    /// Check if a path directly matches any include pattern (not just could_contain_matches).
    /// Used to determine if an empty directory should be kept.
    ///
    /// Returns true if:
    /// - No include patterns exist (everything is directly included)
    /// - At least one include pattern directly matches the path
    ///
    /// Returns false if:
    /// - Include patterns exist but none match the path (directory was only traversed)
    pub fn directly_matches_include(&self, relative_path: &std::path::Path, is_dir: bool) -> bool {
        if self.includes.is_empty() {
            return true; // no includes = everything directly included
        }
        for pattern in &self.includes {
            if pattern.matches(relative_path, is_dir) {
                return true;
            }
        }
        false
    }
    /// Check if a directory could potentially contain files matching the pattern
    pub fn could_contain_matches(&self, dir_path: &Path, pattern: &FilterPattern) -> bool {
        // non-anchored simple patterns (no path separators) can match anywhere
        if !pattern.anchored && !pattern.is_path_pattern() {
            return true;
        }
        // extract the non-wildcard prefix from the pattern
        // e.g., "/src/**" -> "src", "src/foo/**/*.rs" -> "src/foo", "**/*.rs" -> ""
        let pattern_path = pattern
            .original
            .trim_start_matches('/')
            .trim_end_matches('/');
        let prefix = Self::extract_literal_prefix(pattern_path);
        let dir_str = dir_path.to_string_lossy();
        // if no literal prefix (pattern starts with wildcard like "**/*.rs"),
        // it can match anywhere
        if prefix.is_empty() {
            return true;
        }
        // empty dir_path (root) is always an ancestor of any prefix
        if dir_str.is_empty() {
            return true;
        }
        // check if dir_path could lead to matches:
        // 1. dir_path is an ancestor of prefix (e.g., "src" for prefix "src/foo")
        // 2. dir_path equals the prefix
        // 3. dir_path is a descendant of prefix (e.g., "src/foo/bar" for prefix "src")
        // case 1 & 2: prefix starts with dir_path
        if prefix.starts_with(&*dir_str) {
            let after_dir = &prefix[dir_str.len()..];
            // dir_path is ancestor if followed by '/' or is exact match
            if after_dir.is_empty() || after_dir.starts_with('/') {
                return true;
            }
        }
        // case 3: dir_path is descendant of prefix
        if let Some(after_prefix) = dir_str.strip_prefix(prefix)
            && (after_prefix.is_empty() || after_prefix.starts_with('/'))
        {
            return true;
        }
        false
    }
    /// Extract the literal (non-wildcard) prefix from a pattern.
    /// Returns the portion before any wildcard characters (*, ?, [), trimmed to complete path components.
    /// Examples:
    /// - "src/**" -> "src"
    /// - "src/foo/**/*.rs" -> "src/foo"
    /// - "**/*.rs" -> ""
    /// - "bar" -> "bar" (no wildcards = entire pattern is literal)
    /// - "bar/*.txt" -> "bar"
    /// - "*.txt" -> ""
    fn extract_literal_prefix(pattern: &str) -> &str {
        // find first wildcard character
        let wildcard_pos = pattern.find(['*', '?', '[']).unwrap_or(pattern.len());
        // if no wildcards, entire pattern is literal
        if wildcard_pos == pattern.len() {
            return pattern;
        }
        // if wildcard at start, no prefix
        if wildcard_pos == 0 {
            return "";
        }
        // find the last '/' before the wildcard to get a complete path component
        let prefix = &pattern[..wildcard_pos];
        match prefix.rfind('/') {
            Some(pos) => &pattern[..pos],
            None => {
                // no '/' before wildcard - pattern like "src*.txt" has no usable prefix
                ""
            }
        }
    }
    /// Parse filter settings from a file
    ///
    /// # File Format
    /// ```text
    /// # comments supported
    /// --include *.rs
    /// --include Cargo.toml
    /// --exclude target/
    /// --exclude *.log
    /// ```
    pub fn from_file(path: &Path) -> Result<Self, anyhow::Error> {
        let content = std::fs::read_to_string(path)
            .with_context(|| format!("failed to read filter file: {:?}", path))?;
        Self::parse_content(&content)
    }
    /// Build filter settings from CLI arguments. Either reads patterns from a
    /// filter file, builds them from --include/--exclude lists, or returns
    /// `None` when no filtering was requested.
    ///
    /// The file path and the include/exclude lists are mutually exclusive: the
    /// CLI layer enforces this via clap's `conflicts_with_all`, and this helper
    /// returns an error if a non-clap caller passes both.
    pub fn from_args(
        filter_file: Option<&std::path::Path>,
        include: &[String],
        exclude: &[String],
    ) -> Result<Option<Self>, anyhow::Error> {
        if filter_file.is_some() && (!include.is_empty() || !exclude.is_empty()) {
            return Err(anyhow!(
                "filter_file is mutually exclusive with include/exclude patterns"
            ));
        }
        if let Some(path) = filter_file {
            return Ok(Some(Self::from_file(path)?));
        }
        if include.is_empty() && exclude.is_empty() {
            return Ok(None);
        }
        let mut settings = Self::new();
        for p in include {
            settings.add_include(p)?;
        }
        for p in exclude {
            settings.add_exclude(p)?;
        }
        Ok(Some(settings))
    }
    /// Parse filter settings from a string (filter file format)
    pub fn parse_content(content: &str) -> Result<Self, anyhow::Error> {
        let mut settings = Self::new();
        for (line_num, line) in content.lines().enumerate() {
            let line = line.trim();
            // skip empty lines and comments
            if line.is_empty() || line.starts_with('#') {
                continue;
            }
            let line_num = line_num + 1; // 1-based for error messages
            if let Some(pattern) = line.strip_prefix("--include ") {
                let pattern = pattern.trim();
                settings
                    .add_include(pattern)
                    .with_context(|| format!("line {}: invalid include pattern", line_num))?;
            } else if let Some(pattern) = line.strip_prefix("--exclude ") {
                let pattern = pattern.trim();
                settings
                    .add_exclude(pattern)
                    .with_context(|| format!("line {}: invalid exclude pattern", line_num))?;
            } else {
                return Err(anyhow!(
                    "line {}: invalid syntax '{}', expected '--include PATTERN' or '--exclude PATTERN'",
                    line_num,
                    line
                ));
            }
        }
        Ok(settings)
    }
}

/// Time-based filter for matching entries by age (mtime/btime).
///
/// Used in addition to glob filters to skip entries that are not yet old enough.
/// Each threshold is interpreted as a minimum age — an entry matches when its
/// timestamp is at least the threshold ago (i.e. the timestamp is "before"
/// `now - threshold`). When both fields are set, both conditions must hold (AND).
///
/// `created_before` uses the file's birth time (`std::fs::Metadata::created()`).
/// Some Linux filesystems do not expose btime; in that case [`Self::matches`]
/// returns an error rather than silently treating the file as a match.
#[derive(Debug, Clone, Default)]
pub struct TimeFilter {
    /// minimum age based on mtime; entry matches when mtime is at least this old
    pub modified_before: Option<std::time::Duration>,
    /// minimum age based on btime; entry matches when btime is at least this old
    pub created_before: Option<std::time::Duration>,
}

/// Outcome of evaluating a [`TimeFilter`] against a metadata entry.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum TimeFilterResult {
    /// entry passes both configured time thresholds (or no thresholds are set)
    Matched,
    /// entry's mtime is too recent to satisfy `modified_before`
    TooNewModified,
    /// entry's btime is too recent to satisfy `created_before`
    TooNewCreated,
    /// entry fails both `modified_before` and `created_before`
    TooNewBoth,
}

/// Why an entry was skipped by a [`TimeFilter`] — the subset of [`TimeFilterResult`]
/// that does not include `Matched`. Constructed via `TimeFilterResult::as_skip_reason`.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum TimeSkipReason {
    /// entry's mtime is too recent
    TooNewModified,
    /// entry's btime is too recent
    TooNewCreated,
    /// both mtime and btime are too recent
    TooNewBoth,
}

impl TimeFilterResult {
    /// Returns the skip reason when this result represents a skip, or `None` when matched.
    pub fn as_skip_reason(self) -> Option<TimeSkipReason> {
        match self {
            TimeFilterResult::Matched => None,
            TimeFilterResult::TooNewModified => Some(TimeSkipReason::TooNewModified),
            TimeFilterResult::TooNewCreated => Some(TimeSkipReason::TooNewCreated),
            TimeFilterResult::TooNewBoth => Some(TimeSkipReason::TooNewBoth),
        }
    }
}

fn parse_duration_arg(flag: &str, value: &str) -> anyhow::Result<std::time::Duration> {
    humantime::parse_duration(value).map_err(|err| {
        anyhow!(
            "{} value {:?} is not a valid duration: {}\n\
             Hint: use suffixes like 1y, 6months, 30d, 12h, 30m, 45s. \
             Note that 'M' means months and 'm' means minutes.",
            flag,
            value,
            err
        )
    })
}

/// Reject `--created-before` on musl builds, where birth time (btime) is unreadable via
/// `std::fs::Metadata::created()` so every entry would fail evaluation and be skipped. No-op on
/// glibc. `tool` names the binary for the error message.
pub fn reject_created_before_on_musl(
    tool: &str,
    created_before: Option<&str>,
) -> anyhow::Result<()> {
    #[cfg(target_env = "musl")]
    if created_before.is_some() {
        return Err(anyhow!(
            "--created-before is not supported on musl builds: birth time (btime) is not \
             readable via std::fs::Metadata::created() under musl, so every entry would \
             fail evaluation and be skipped. Use --modified-before instead, or rebuild \
             {tool} against glibc."
        ));
    }
    #[cfg(not(target_env = "musl"))]
    let _ = (tool, created_before);
    Ok(())
}

impl TimeFilter {
    /// Build a `TimeFilter` from the raw CLI duration strings used by `rrm`/`rchm`
    /// (`--modified-before` / `--created-before`). Returns `Ok(None)` when neither is set.
    /// Durations use humantime syntax; note `M` means months and `m` means minutes.
    pub fn from_cli_args(
        modified_before: Option<&str>,
        created_before: Option<&str>,
    ) -> anyhow::Result<Option<TimeFilter>> {
        let modified_before = modified_before
            .map(|v| parse_duration_arg("--modified-before", v))
            .transpose()?;
        let created_before = created_before
            .map(|v| parse_duration_arg("--created-before", v))
            .transpose()?;
        if modified_before.is_none() && created_before.is_none() {
            return Ok(None);
        }
        Ok(Some(TimeFilter {
            modified_before,
            created_before,
        }))
    }
    /// Returns true when no time thresholds are configured.
    pub fn is_empty(&self) -> bool {
        self.modified_before.is_none() && self.created_before.is_none()
    }
    /// Evaluate this filter against `metadata`.
    ///
    /// Returns:
    /// - `Ok(TimeFilterResult::Matched)` when the entry passes all configured thresholds.
    /// - `Ok(TimeFilterResult::TooNew*)` when one or both thresholds are not yet met.
    /// - `Err(_)` when `created_before` is configured and the underlying `created()`
    ///   call fails (e.g., a filesystem that does not expose birth time).
    pub fn matches(&self, metadata: &std::fs::Metadata) -> anyhow::Result<TimeFilterResult> {
        let mtime = if self.modified_before.is_some() {
            Some(
                metadata
                    .modified()
                    .context("failed to read mtime from metadata")?,
            )
        } else {
            None
        };
        let btime = if self.created_before.is_some() {
            Some(
                metadata
                    .created()
                    .context("failed to read birth time (created) from metadata")?,
            )
        } else {
            None
        };
        Ok(self.evaluate(mtime, btime, std::time::SystemTime::now()))
    }
    /// Pure-logic evaluation of this filter against raw timestamps.
    ///
    /// The timestamps are only inspected when the corresponding threshold is configured.
    /// When a threshold is configured but the timestamp is `None`, the entry is treated as
    /// "too new" for that axis (a `None` timestamp has no age signal, so it cannot satisfy
    /// an age threshold). This helper is for deterministic unit testing of the AND logic;
    /// prefer [`Self::matches`] for real callers.
    fn evaluate(
        &self,
        mtime: Option<std::time::SystemTime>,
        btime: Option<std::time::SystemTime>,
        now: std::time::SystemTime,
    ) -> TimeFilterResult {
        let modified_too_new = self
            .modified_before
            .is_some_and(|threshold| mtime.is_none_or(|t| !is_at_least_age(now, t, threshold)));
        let created_too_new = self
            .created_before
            .is_some_and(|threshold| btime.is_none_or(|t| !is_at_least_age(now, t, threshold)));
        match (modified_too_new, created_too_new) {
            (false, false) => TimeFilterResult::Matched,
            (true, false) => TimeFilterResult::TooNewModified,
            (false, true) => TimeFilterResult::TooNewCreated,
            (true, true) => TimeFilterResult::TooNewBoth,
        }
    }
}

/// Returns true if `timestamp` is at least `age` old relative to `now`.
/// Treats clock skew (timestamp in the future) as "not old enough".
fn is_at_least_age(
    now: std::time::SystemTime,
    timestamp: std::time::SystemTime,
    age: std::time::Duration,
) -> bool {
    match now.duration_since(timestamp) {
        Ok(elapsed) => elapsed >= age,
        Err(_) => false,
    }
}

/// Data transfer object for FilterSettings serialization.
/// Used for passing filter settings across process boundaries (e.g., to rcpd).
#[derive(Serialize, Deserialize)]
struct FilterSettingsDto {
    includes: Vec<String>,
    excludes: Vec<String>,
}

impl Serialize for FilterSettings {
    fn serialize<S: Serializer>(&self, serializer: S) -> Result<S::Ok, S::Error> {
        let dto = FilterSettingsDto {
            includes: self.includes.iter().map(|p| p.original.clone()).collect(),
            excludes: self.excludes.iter().map(|p| p.original.clone()).collect(),
        };
        dto.serialize(serializer)
    }
}

impl<'de> Deserialize<'de> for FilterSettings {
    fn deserialize<D: Deserializer<'de>>(deserializer: D) -> Result<Self, D::Error> {
        let dto = FilterSettingsDto::deserialize(deserializer)?;
        let mut settings = FilterSettings::new();
        for pattern in dto.includes {
            settings
                .add_include(&pattern)
                .map_err(serde::de::Error::custom)?;
        }
        for pattern in dto.excludes {
            settings
                .add_exclude(&pattern)
                .map_err(serde::de::Error::custom)?;
        }
        Ok(settings)
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    #[test]
    fn test_pattern_basic_glob() {
        let pattern = FilterPattern::parse("*.rs").unwrap();
        assert!(pattern.matches(Path::new("foo.rs"), false));
        assert!(pattern.matches(Path::new("main.rs"), false));
        assert!(!pattern.matches(Path::new("foo.txt"), false));
        // simple patterns match against filename, so src/foo.rs matches via its filename
        assert!(pattern.matches(Path::new("src/foo.rs"), false));
    }
    #[test]
    fn test_pattern_double_star() {
        let pattern = FilterPattern::parse("**/*.rs").unwrap();
        assert!(pattern.matches(Path::new("src/foo.rs"), false));
        assert!(pattern.matches(Path::new("a/b/c/d.rs"), false));
        // ** can match zero segments, so **/*.rs matches foo.rs
        assert!(pattern.matches(Path::new("foo.rs"), false));
    }
    #[test]
    fn test_pattern_question_mark() {
        let pattern = FilterPattern::parse("file?.txt").unwrap();
        assert!(pattern.matches(Path::new("file1.txt"), false));
        assert!(pattern.matches(Path::new("fileA.txt"), false));
        assert!(!pattern.matches(Path::new("file12.txt"), false));
        assert!(!pattern.matches(Path::new("file.txt"), false));
    }
    #[test]
    fn test_pattern_character_class() {
        let pattern = FilterPattern::parse("[abc].txt").unwrap();
        assert!(pattern.matches(Path::new("a.txt"), false));
        assert!(pattern.matches(Path::new("b.txt"), false));
        assert!(pattern.matches(Path::new("c.txt"), false));
        assert!(!pattern.matches(Path::new("d.txt"), false));
    }
    #[test]
    fn test_pattern_anchored() {
        let pattern = FilterPattern::parse("/src").unwrap();
        assert!(pattern.anchored);
        // matches only at root level
        assert!(pattern.matches(Path::new("src"), true));
        assert!(!pattern.matches(Path::new("foo/src"), true));
    }
    #[test]
    fn test_pattern_dir_only() {
        let pattern = FilterPattern::parse("build/").unwrap();
        assert!(pattern.dir_only);
        // only matches directories
        assert!(pattern.matches(Path::new("build"), true));
        assert!(!pattern.matches(Path::new("build"), false)); // file named build
    }
    #[test]
    fn test_include_only_mode() {
        let mut settings = FilterSettings::new();
        settings.add_include("*.rs").unwrap();
        settings.add_include("Cargo.toml").unwrap();
        assert!(matches!(
            settings.should_include(Path::new("main.rs"), false),
            FilterResult::Included
        ));
        assert!(matches!(
            settings.should_include(Path::new("Cargo.toml"), false),
            FilterResult::Included
        ));
        assert!(matches!(
            settings.should_include(Path::new("README.md"), false),
            FilterResult::ExcludedByDefault
        ));
    }
    #[test]
    fn test_exclude_only_mode() {
        let mut settings = FilterSettings::new();
        settings.add_exclude("*.log").unwrap();
        settings.add_exclude("target/").unwrap();
        assert!(matches!(
            settings.should_include(Path::new("main.rs"), false),
            FilterResult::Included
        ));
        match settings.should_include(Path::new("debug.log"), false) {
            FilterResult::ExcludedByPattern(p) => assert_eq!(p, "*.log"),
            other => panic!("expected ExcludedByPattern, got {:?}", other),
        }
        match settings.should_include(Path::new("target"), true) {
            FilterResult::ExcludedByPattern(p) => assert_eq!(p, "target/"),
            other => panic!("expected ExcludedByPattern, got {:?}", other),
        }
    }
    #[test]
    fn test_include_then_exclude() {
        let mut settings = FilterSettings::new();
        settings.add_include("*.rs").unwrap();
        settings.add_exclude("test_*.rs").unwrap();
        // regular .rs files are included
        assert!(matches!(
            settings.should_include(Path::new("main.rs"), false),
            FilterResult::Included
        ));
        // test_*.rs files are excluded even though they match *.rs
        match settings.should_include(Path::new("test_foo.rs"), false) {
            FilterResult::ExcludedByPattern(p) => assert_eq!(p, "test_*.rs"),
            other => panic!("expected ExcludedByPattern, got {:?}", other),
        }
        // non-.rs files are excluded by default
        assert!(matches!(
            settings.should_include(Path::new("README.md"), false),
            FilterResult::ExcludedByDefault
        ));
    }
    #[test]
    fn test_filter_file_basic() {
        let content = r#"
# this is a comment
--include *.rs
--include Cargo.toml

--exclude target/
--exclude *.log
"#;
        let settings = FilterSettings::parse_content(content).unwrap();
        assert_eq!(settings.includes.len(), 2);
        assert_eq!(settings.excludes.len(), 2);
    }
    #[test]
    fn test_filter_file_comments() {
        let content = "# only comments\n# and empty lines\n\n";
        let settings = FilterSettings::parse_content(content).unwrap();
        assert!(settings.is_empty());
    }
    #[test]
    fn test_filter_file_syntax_error() {
        let content = "invalid line without prefix";
        let result = FilterSettings::parse_content(content);
        assert!(result.is_err());
        let err = result.unwrap_err().to_string();
        assert!(err.contains("line 1"));
        assert!(err.contains("invalid syntax"));
    }
    #[test]
    fn test_empty_pattern_error() {
        let result = FilterPattern::parse("");
        assert!(result.is_err());
    }
    #[test]
    fn test_is_empty() {
        let empty = FilterSettings::new();
        assert!(empty.is_empty());
        let mut with_include = FilterSettings::new();
        with_include.add_include("*.rs").unwrap();
        assert!(!with_include.is_empty());
        let mut with_exclude = FilterSettings::new();
        with_exclude.add_exclude("*.log").unwrap();
        assert!(!with_exclude.is_empty());
    }
    #[test]
    fn test_has_includes() {
        let empty = FilterSettings::new();
        assert!(!empty.has_includes());
        let mut with_include = FilterSettings::new();
        with_include.add_include("*.rs").unwrap();
        assert!(with_include.has_includes());
        let mut with_exclude = FilterSettings::new();
        with_exclude.add_exclude("*.log").unwrap();
        assert!(!with_exclude.has_includes());
        let mut with_both = FilterSettings::new();
        with_both.add_include("*.rs").unwrap();
        with_both.add_exclude("*.log").unwrap();
        assert!(with_both.has_includes());
    }
    #[test]
    fn test_filename_match_for_simple_patterns() {
        // simple patterns (no /) should match the filename anywhere in the path
        let pattern = FilterPattern::parse("*.rs").unwrap();
        assert!(pattern.matches(Path::new("foo.rs"), false));
        assert!(pattern.matches(Path::new("src/foo.rs"), false)); // matches filename
        // nested paths also match via filename
        assert!(pattern.matches(Path::new("a/b/c/foo.rs"), false));
    }
    #[test]
    fn test_path_pattern_requires_full_match() {
        // patterns with / require the full path to match
        let pattern = FilterPattern::parse("src/*.rs").unwrap();
        assert!(pattern.matches(Path::new("src/foo.rs"), false));
        assert!(!pattern.matches(Path::new("foo.rs"), false));
        assert!(!pattern.matches(Path::new("other/src/foo.rs"), false));
    }
    #[test]
    fn test_double_star_matches_nested_paths() {
        // **/*.rs should match files at any depth
        let pattern = FilterPattern::parse("**/*.rs").unwrap();
        assert!(pattern.matches(Path::new("foo.rs"), false));
        assert!(pattern.matches(Path::new("src/foo.rs"), false));
        assert!(pattern.matches(Path::new("src/lib/foo.rs"), false));
        assert!(pattern.matches(Path::new("a/b/c/d/e.rs"), false));
    }
    #[test]
    fn test_anchored_pattern_matches_only_at_root() {
        // /src should match only at root, not nested
        let pattern = FilterPattern::parse("/src").unwrap();
        assert!(pattern.matches(Path::new("src"), true));
        assert!(!pattern.matches(Path::new("foo/src"), true));
        assert!(!pattern.matches(Path::new("a/b/src"), true));
    }
    #[test]
    fn test_nested_directory_pattern() {
        // src/lib/ should match only that specific nested path
        let pattern = FilterPattern::parse("src/lib/").unwrap();
        assert!(pattern.matches(Path::new("src/lib"), true));
        assert!(!pattern.matches(Path::new("lib"), true));
        assert!(!pattern.matches(Path::new("other/src/lib"), true));
    }
    #[test]
    fn test_dir_only_simple_pattern_matches_at_any_level() {
        // target/ (dir-only) should match "target" at any level, like simple patterns
        // the trailing / is just a dir-only marker, not a path separator
        let pattern = FilterPattern::parse("target/").unwrap();
        assert!(pattern.dir_only);
        assert!(!pattern.anchored);
        // should match at root
        assert!(pattern.matches(Path::new("target"), true));
        // should match nested (filename matching)
        assert!(pattern.matches(Path::new("foo/target"), true));
        assert!(pattern.matches(Path::new("a/b/target"), true));
        // should NOT match files (dir-only)
        assert!(!pattern.matches(Path::new("target"), false));
        assert!(!pattern.matches(Path::new("foo/target"), false));
    }
    #[test]
    fn test_dir_only_pattern_could_contain_matches() {
        // target/ should allow traversal into any directory since it can match anywhere
        let mut settings = FilterSettings::new();
        settings.add_include("target/").unwrap();
        let pattern = &settings.includes[0];
        // should return true for any directory since target/ can match at any level
        assert!(settings.could_contain_matches(Path::new("foo"), pattern));
        assert!(settings.could_contain_matches(Path::new("a/b"), pattern));
        assert!(settings.could_contain_matches(Path::new("src"), pattern));
    }
    #[test]
    fn test_precedence_exclude_overrides_include() {
        // when both include and exclude match, exclude wins (excludes checked first)
        let mut settings = FilterSettings::new();
        settings.add_include("*.rs").unwrap();
        settings.add_exclude("test_*.rs").unwrap();
        // file matching both patterns should be excluded
        match settings.should_include(Path::new("test_main.rs"), false) {
            FilterResult::ExcludedByPattern(p) => assert_eq!(p, "test_*.rs"),
            other => panic!("expected ExcludedByPattern, got {:?}", other),
        }
        // file matching only include should be included
        assert!(matches!(
            settings.should_include(Path::new("main.rs"), false),
            FilterResult::Included
        ));
    }
    #[test]
    fn test_should_include_root_item_non_anchored_exclude() {
        // non-anchored exclude patterns should apply to root items
        let mut settings = FilterSettings::new();
        settings.add_exclude("*.log").unwrap();
        // root file matching exclude pattern is excluded
        match settings.should_include_root_item(Path::new("debug.log"), false) {
            FilterResult::ExcludedByPattern(p) => assert_eq!(p, "*.log"),
            other => panic!("expected ExcludedByPattern, got {:?}", other),
        }
        // root file not matching exclude pattern is included
        assert!(matches!(
            settings.should_include_root_item(Path::new("main.rs"), false),
            FilterResult::Included
        ));
    }
    #[test]
    fn test_should_include_root_item_anchored_exclude_skipped() {
        // anchored exclude patterns should NOT apply to root items
        let mut settings = FilterSettings::new();
        settings.add_exclude("/target/").unwrap();
        // root directory "target" should NOT be excluded (pattern is anchored)
        assert!(matches!(
            settings.should_include_root_item(Path::new("target"), true),
            FilterResult::Included
        ));
    }
    #[test]
    fn test_should_include_root_item_non_anchored_include() {
        // non-anchored include patterns should apply to root items
        let mut settings = FilterSettings::new();
        settings.add_include("*.rs").unwrap();
        // root file matching include pattern is included
        assert!(matches!(
            settings.should_include_root_item(Path::new("main.rs"), false),
            FilterResult::Included
        ));
        // root file not matching include pattern is excluded by default
        assert!(matches!(
            settings.should_include_root_item(Path::new("readme.md"), false),
            FilterResult::ExcludedByDefault
        ));
    }
    #[test]
    fn test_should_include_root_item_anchored_include_skipped() {
        // anchored include patterns should NOT apply to root items directly
        // but root directories should still be traversed if anchored patterns exist
        let mut settings = FilterSettings::new();
        settings.add_include("/bar").unwrap();
        // root directory "foo" should be included (so we can traverse to find /bar inside)
        assert!(matches!(
            settings.should_include_root_item(Path::new("foo"), true),
            FilterResult::Included
        ));
        // root file "baz" should be excluded by default (only anchored includes, none match)
        assert!(matches!(
            settings.should_include_root_item(Path::new("baz"), false),
            FilterResult::ExcludedByDefault
        ));
    }
    #[test]
    fn test_should_include_root_item_mixed_patterns() {
        // mix of anchored and non-anchored patterns
        let mut settings = FilterSettings::new();
        settings.add_include("*.rs").unwrap();
        settings.add_include("/bar").unwrap();
        settings.add_exclude("test_*.rs").unwrap();
        // root .rs file is included (non-anchored include)
        assert!(matches!(
            settings.should_include_root_item(Path::new("main.rs"), false),
            FilterResult::Included
        ));
        // root test_*.rs file is excluded (non-anchored exclude)
        match settings.should_include_root_item(Path::new("test_foo.rs"), false) {
            FilterResult::ExcludedByPattern(p) => assert_eq!(p, "test_*.rs"),
            other => panic!("expected ExcludedByPattern, got {:?}", other),
        }
        // root directory "foo" is included (to traverse for /bar)
        assert!(matches!(
            settings.should_include_root_item(Path::new("foo"), true),
            FilterResult::Included
        ));
    }
    #[test]
    fn test_could_contain_matches_anchored_double_star() {
        // /src/** should only match directories under src, not unrelated directories
        let mut settings = FilterSettings::new();
        settings.add_include("/src/**").unwrap();
        let pattern = &settings.includes[0];
        // should return true for ancestors of "src" (we need to traverse to get there)
        assert!(settings.could_contain_matches(Path::new(""), pattern));
        // should return true for "src" itself (it's the prefix)
        assert!(settings.could_contain_matches(Path::new("src"), pattern));
        // should return true for descendants of "src" (matches can be inside)
        assert!(settings.could_contain_matches(Path::new("src/foo"), pattern));
        assert!(settings.could_contain_matches(Path::new("src/foo/bar"), pattern));
        // should return FALSE for unrelated directories
        assert!(!settings.could_contain_matches(Path::new("build"), pattern));
        assert!(!settings.could_contain_matches(Path::new("target"), pattern));
        assert!(!settings.could_contain_matches(Path::new("build/src"), pattern));
    }
    #[test]
    fn test_could_contain_matches_non_anchored_double_star() {
        // **/*.rs should match anywhere (no prefix)
        let mut settings = FilterSettings::new();
        settings.add_include("**/*.rs").unwrap();
        let pattern = &settings.includes[0];
        // should return true for any directory since ** can match anywhere
        assert!(settings.could_contain_matches(Path::new("src"), pattern));
        assert!(settings.could_contain_matches(Path::new("build"), pattern));
        assert!(settings.could_contain_matches(Path::new("any/path"), pattern));
    }
    #[test]
    fn test_could_contain_matches_nested_prefix() {
        // /src/foo/** should have prefix "src/foo"
        let mut settings = FilterSettings::new();
        settings.add_include("/src/foo/**").unwrap();
        let pattern = &settings.includes[0];
        // ancestors of prefix
        assert!(settings.could_contain_matches(Path::new(""), pattern));
        assert!(settings.could_contain_matches(Path::new("src"), pattern));
        // the prefix itself
        assert!(settings.could_contain_matches(Path::new("src/foo"), pattern));
        // descendants of prefix
        assert!(settings.could_contain_matches(Path::new("src/foo/bar"), pattern));
        // unrelated directories
        assert!(!settings.could_contain_matches(Path::new("build"), pattern));
        assert!(!settings.could_contain_matches(Path::new("src/bar"), pattern));
    }
    #[test]
    fn test_extract_literal_prefix() {
        // test the helper function
        assert_eq!(FilterSettings::extract_literal_prefix("src/**"), "src");
        assert_eq!(
            FilterSettings::extract_literal_prefix("src/foo/**"),
            "src/foo"
        );
        assert_eq!(FilterSettings::extract_literal_prefix("**/*.rs"), "");
        assert_eq!(FilterSettings::extract_literal_prefix("*.rs"), "");
        assert_eq!(FilterSettings::extract_literal_prefix("src/*.rs"), "src");
        // no wildcards = entire pattern is literal
        assert_eq!(
            FilterSettings::extract_literal_prefix("src/foo/bar"),
            "src/foo/bar"
        );
        assert_eq!(FilterSettings::extract_literal_prefix("bar"), "bar");
        assert_eq!(FilterSettings::extract_literal_prefix("src[0-9]/*.rs"), "");
    }
    #[test]
    fn test_directly_matches_include_simple_pattern() {
        // simple pattern like *.txt matches file
        let mut settings = FilterSettings::new();
        settings.add_include("*.txt").unwrap();
        // should match files with .txt extension
        assert!(settings.directly_matches_include(Path::new("foo.txt"), false));
        assert!(settings.directly_matches_include(Path::new("bar/foo.txt"), false));
        // should not match other files
        assert!(!settings.directly_matches_include(Path::new("foo.rs"), false));
        // directories don't match file patterns
        assert!(!settings.directly_matches_include(Path::new("txt"), true));
    }
    #[test]
    fn test_directly_matches_include_anchored_pattern() {
        // anchored /foo matches at root only
        let mut settings = FilterSettings::new();
        settings.add_include("/foo").unwrap();
        // should match at root
        assert!(settings.directly_matches_include(Path::new("foo"), true));
        assert!(settings.directly_matches_include(Path::new("foo"), false));
        // should not match nested
        assert!(!settings.directly_matches_include(Path::new("bar/foo"), true));
    }
    #[test]
    fn test_directly_matches_include_empty_includes() {
        // returns true when no includes
        let settings = FilterSettings::new();
        assert!(settings.directly_matches_include(Path::new("anything"), true));
        assert!(settings.directly_matches_include(Path::new("foo/bar"), false));
    }
    #[test]
    fn test_directly_matches_include_path_pattern() {
        // path pattern like src/*.rs
        let mut settings = FilterSettings::new();
        settings.add_include("src/*.rs").unwrap();
        // should match paths in src/
        assert!(settings.directly_matches_include(Path::new("src/foo.rs"), false));
        // should not match at root or other paths
        assert!(!settings.directly_matches_include(Path::new("foo.rs"), false));
        assert!(!settings.directly_matches_include(Path::new("other/foo.rs"), false));
    }
    #[test]
    fn test_directly_matches_include_dir_only_pattern() {
        // directory-only pattern target/
        let mut settings = FilterSettings::new();
        settings.add_include("target/").unwrap();
        // should match directories named target
        assert!(settings.directly_matches_include(Path::new("target"), true));
        assert!(settings.directly_matches_include(Path::new("foo/target"), true));
        // should not match files
        assert!(!settings.directly_matches_include(Path::new("target"), false));
    }

    mod time_filter_tests {
        use super::*;
        use crate::testutils;

        fn write_with_mtime(path: &std::path::Path, age: std::time::Duration) {
            std::fs::write(path, "x").unwrap();
            let past = filetime::FileTime::from_system_time(std::time::SystemTime::now() - age);
            filetime::set_file_mtime(path, past).unwrap();
        }

        #[test]
        fn is_empty_when_no_thresholds_set() {
            assert!(TimeFilter::default().is_empty());
            let only_mtime = TimeFilter {
                modified_before: Some(std::time::Duration::from_secs(1)),
                created_before: None,
            };
            assert!(!only_mtime.is_empty());
            let only_btime = TimeFilter {
                modified_before: None,
                created_before: Some(std::time::Duration::from_secs(1)),
            };
            assert!(!only_btime.is_empty());
        }

        #[tokio::test]
        async fn matches_returns_matched_when_no_thresholds() {
            let tmp = testutils::create_temp_dir().await.unwrap();
            let path = tmp.join("file");
            write_with_mtime(&path, std::time::Duration::from_secs(0));
            let metadata = std::fs::metadata(&path).unwrap();
            assert_eq!(
                TimeFilter::default().matches(&metadata).unwrap(),
                TimeFilterResult::Matched
            );
        }

        #[tokio::test]
        async fn matches_when_mtime_older_than_threshold() {
            let tmp = testutils::create_temp_dir().await.unwrap();
            let path = tmp.join("file");
            write_with_mtime(&path, std::time::Duration::from_secs(7200));
            let metadata = std::fs::metadata(&path).unwrap();
            let filter = TimeFilter {
                modified_before: Some(std::time::Duration::from_secs(3600)),
                created_before: None,
            };
            assert_eq!(
                filter.matches(&metadata).unwrap(),
                TimeFilterResult::Matched
            );
        }

        #[tokio::test]
        async fn reports_too_new_modified_when_mtime_recent() {
            let tmp = testutils::create_temp_dir().await.unwrap();
            let path = tmp.join("file");
            write_with_mtime(&path, std::time::Duration::from_secs(0));
            let metadata = std::fs::metadata(&path).unwrap();
            let filter = TimeFilter {
                modified_before: Some(std::time::Duration::from_secs(3600)),
                created_before: None,
            };
            assert_eq!(
                filter.matches(&metadata).unwrap(),
                TimeFilterResult::TooNewModified
            );
        }

        /// Exercises `matches()` on the `created_before` axis with real metadata.
        /// The expected outcome depends on whether the filesystem exposes birth time:
        /// - Supported: a freshly-written file has a recent btime → `TooNewCreated`.
        /// - Unsupported: `matches()` must return `Err` (the contract is to surface
        ///   the failure rather than silently treat the file as a match).
        /// Deterministic on either platform — drift in either branch will fail loudly.
        #[tokio::test]
        async fn matches_exercises_btime_deterministically() {
            let tmp = testutils::create_temp_dir().await.unwrap();
            let path = tmp.join("file");
            std::fs::write(&path, "x").unwrap();
            let metadata = std::fs::metadata(&path).unwrap();
            let filter = TimeFilter {
                modified_before: None,
                created_before: Some(std::time::Duration::from_secs(3600)),
            };
            let result = filter.matches(&metadata);
            match metadata.created() {
                Ok(_) => assert_eq!(result.unwrap(), TimeFilterResult::TooNewCreated),
                Err(_) => assert!(
                    result.is_err(),
                    "matches() must return Err when created_before is set but btime is unavailable"
                ),
            }
        }

        #[tokio::test]
        async fn matches_with_zero_threshold_is_always_satisfied() {
            // any file is at least zero seconds old
            let tmp = testutils::create_temp_dir().await.unwrap();
            let path = tmp.join("file");
            write_with_mtime(&path, std::time::Duration::from_secs(0));
            let metadata = std::fs::metadata(&path).unwrap();
            let filter = TimeFilter {
                modified_before: Some(std::time::Duration::from_secs(0)),
                created_before: None,
            };
            assert_eq!(
                filter.matches(&metadata).unwrap(),
                TimeFilterResult::Matched
            );
        }

        /// Focused unit tests for the pure-logic [`TimeFilter::evaluate`] helper.
        /// Using raw timestamps makes AND/OR boundaries deterministic without depending
        /// on platform btime support or filesystem timestamp resolution.
        mod evaluate_and_or_logic {
            use super::*;

            fn now() -> std::time::SystemTime {
                // pick a fixed anchor far in the past so threshold subtraction cannot underflow
                std::time::UNIX_EPOCH + std::time::Duration::from_secs(10_000_000)
            }

            fn age_before(
                t: std::time::SystemTime,
                age: std::time::Duration,
            ) -> std::time::SystemTime {
                t - age
            }

            #[test]
            fn no_thresholds_always_matches_regardless_of_timestamps() {
                let filter = TimeFilter::default();
                assert_eq!(
                    filter.evaluate(None, None, now()),
                    TimeFilterResult::Matched
                );
                assert_eq!(
                    filter.evaluate(
                        Some(age_before(now(), std::time::Duration::from_secs(0))),
                        None,
                        now()
                    ),
                    TimeFilterResult::Matched
                );
            }

            #[test]
            fn and_logic_both_pass_is_matched() {
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(3600)),
                    created_before: Some(std::time::Duration::from_secs(3600)),
                };
                let old = age_before(now(), std::time::Duration::from_secs(7200));
                assert_eq!(
                    filter.evaluate(Some(old), Some(old), now()),
                    TimeFilterResult::Matched
                );
            }

            #[test]
            fn and_logic_only_mtime_passes_reports_created_too_new() {
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(3600)),
                    created_before: Some(std::time::Duration::from_secs(3600)),
                };
                let old = age_before(now(), std::time::Duration::from_secs(7200));
                let recent = age_before(now(), std::time::Duration::from_secs(60));
                assert_eq!(
                    filter.evaluate(Some(old), Some(recent), now()),
                    TimeFilterResult::TooNewCreated
                );
            }

            #[test]
            fn and_logic_only_btime_passes_reports_modified_too_new() {
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(3600)),
                    created_before: Some(std::time::Duration::from_secs(3600)),
                };
                let old = age_before(now(), std::time::Duration::from_secs(7200));
                let recent = age_before(now(), std::time::Duration::from_secs(60));
                assert_eq!(
                    filter.evaluate(Some(recent), Some(old), now()),
                    TimeFilterResult::TooNewModified
                );
            }

            #[test]
            fn and_logic_neither_passes_reports_too_new_both() {
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(3600)),
                    created_before: Some(std::time::Duration::from_secs(3600)),
                };
                let recent = age_before(now(), std::time::Duration::from_secs(60));
                assert_eq!(
                    filter.evaluate(Some(recent), Some(recent), now()),
                    TimeFilterResult::TooNewBoth
                );
            }

            #[test]
            fn threshold_boundary_exactly_at_age_matches() {
                // a timestamp exactly `threshold` ago is considered "at least threshold old"
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(3600)),
                    created_before: None,
                };
                let exact = age_before(now(), std::time::Duration::from_secs(3600));
                assert_eq!(
                    filter.evaluate(Some(exact), None, now()),
                    TimeFilterResult::Matched
                );
            }

            #[test]
            fn future_timestamp_treated_as_too_new() {
                // clock skew: if mtime is AFTER now, treat as "not old enough"
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(1)),
                    created_before: None,
                };
                let future = now() + std::time::Duration::from_secs(3600);
                assert_eq!(
                    filter.evaluate(Some(future), None, now()),
                    TimeFilterResult::TooNewModified
                );
            }

            #[test]
            fn missing_timestamp_when_threshold_configured_is_too_new() {
                // None with a configured threshold has no age signal; treat as too new
                // (used when the OS can't produce the timestamp — see evaluate doc comment)
                let filter = TimeFilter {
                    modified_before: Some(std::time::Duration::from_secs(3600)),
                    created_before: Some(std::time::Duration::from_secs(3600)),
                };
                let old = age_before(now(), std::time::Duration::from_secs(7200));
                assert_eq!(
                    filter.evaluate(Some(old), None, now()),
                    TimeFilterResult::TooNewCreated
                );
                assert_eq!(
                    filter.evaluate(None, Some(old), now()),
                    TimeFilterResult::TooNewModified
                );
                assert_eq!(
                    filter.evaluate(None, None, now()),
                    TimeFilterResult::TooNewBoth
                );
            }
        }

        /// Tests for the skip-reason projection.
        mod skip_reason {
            use super::*;

            #[test]
            fn matched_yields_none() {
                assert_eq!(TimeFilterResult::Matched.as_skip_reason(), None);
            }

            #[test]
            fn too_new_variants_yield_matching_reasons() {
                assert_eq!(
                    TimeFilterResult::TooNewModified.as_skip_reason(),
                    Some(TimeSkipReason::TooNewModified)
                );
                assert_eq!(
                    TimeFilterResult::TooNewCreated.as_skip_reason(),
                    Some(TimeSkipReason::TooNewCreated)
                );
                assert_eq!(
                    TimeFilterResult::TooNewBoth.as_skip_reason(),
                    Some(TimeSkipReason::TooNewBoth)
                );
            }
        }
    }
    mod from_args_tests {
        use super::*;
        use std::sync::atomic::{AtomicU64, Ordering};
        static SEQ: AtomicU64 = AtomicU64::new(0);
        /// RAII guard that writes a uniquely-named filter file under the
        /// system temp dir and removes it when dropped. /tmp policies vary
        /// (systemd-tmpfiles often runs weekly, never on some hosts), so
        /// explicit cleanup keeps tests from leaving junk behind.
        struct TempFilterFile {
            path: std::path::PathBuf,
        }
        impl TempFilterFile {
            fn new(content: &str) -> Self {
                let n = SEQ.fetch_add(1, Ordering::Relaxed);
                let path = std::env::temp_dir()
                    .join(format!("rcp-from-args-test-{}-{n}.txt", std::process::id()));
                std::fs::write(&path, content).unwrap();
                Self { path }
            }
            fn path(&self) -> &std::path::Path {
                &self.path
            }
        }
        impl Drop for TempFilterFile {
            fn drop(&mut self) {
                let _ = std::fs::remove_file(&self.path);
            }
        }
        #[test]
        fn returns_none_when_nothing_specified() {
            let result = FilterSettings::from_args(None, &[], &[]).unwrap();
            assert!(result.is_none());
        }
        #[test]
        fn builds_from_include_only() {
            let include = vec!["*.rs".to_string(), "Cargo.toml".to_string()];
            let settings = FilterSettings::from_args(None, &include, &[])
                .unwrap()
                .expect("should return Some when include is non-empty");
            assert_eq!(settings.includes.len(), 2);
            assert!(settings.excludes.is_empty());
        }
        #[test]
        fn builds_from_exclude_only() {
            let exclude = vec!["*.log".to_string(), "target/".to_string()];
            let settings = FilterSettings::from_args(None, &[], &exclude)
                .unwrap()
                .expect("should return Some when exclude is non-empty");
            assert!(settings.includes.is_empty());
            assert_eq!(settings.excludes.len(), 2);
        }
        #[test]
        fn builds_from_include_and_exclude() {
            let include = vec!["*.rs".to_string()];
            let exclude = vec!["target/".to_string()];
            let settings = FilterSettings::from_args(None, &include, &exclude)
                .unwrap()
                .expect("should return Some");
            assert_eq!(settings.includes.len(), 1);
            assert_eq!(settings.excludes.len(), 1);
        }
        #[test]
        fn loads_from_filter_file() {
            let file = TempFilterFile::new("--include *.rs\n--exclude target/\n");
            let settings = FilterSettings::from_args(Some(file.path()), &[], &[])
                .unwrap()
                .expect("should return Some when filter file is read");
            assert_eq!(settings.includes.len(), 1);
            assert_eq!(settings.excludes.len(), 1);
        }
        #[test]
        fn errors_when_filter_file_combined_with_include() {
            let file = TempFilterFile::new("--include *.rs\n");
            let include = vec!["*.txt".to_string()];
            let err = FilterSettings::from_args(Some(file.path()), &include, &[]).unwrap_err();
            assert!(err.to_string().contains("mutually exclusive"));
        }
        #[test]
        fn errors_when_filter_file_combined_with_exclude() {
            let file = TempFilterFile::new("--include *.rs\n");
            let exclude = vec!["*.log".to_string()];
            let err = FilterSettings::from_args(Some(file.path()), &[], &exclude).unwrap_err();
            assert!(err.to_string().contains("mutually exclusive"));
        }
        #[test]
        fn propagates_invalid_include_pattern() {
            let include = vec!["".to_string()];
            assert!(FilterSettings::from_args(None, &include, &[]).is_err());
        }
        #[test]
        fn propagates_missing_filter_file() {
            let path = std::path::PathBuf::from("/nonexistent/path/filters.txt");
            assert!(FilterSettings::from_args(Some(&path), &[], &[]).is_err());
        }
    }
}