netsky 0.1.4

//! `netsky doctor [--brief|--quiet]` — one-command health smoke test.
//!
//! Runs a batch of non-destructive read-only checks grouped by section.
//! Exits 0 on green + yellow, 1 on red. `--brief` prints only the summary
//! line; `--quiet` is exit-code only.
//!
//! PUSH vs PULL: the escalation marker is the push signal ("system is
//! unwell, come look now"); doctor is the pull signal ("what is the
//! current state").

use std::fs;
use std::os::unix::fs::PermissionsExt;
use std::path::Path;
use std::process::Command;
use std::time::{SystemTime, UNIX_EPOCH};

use netsky_core::config::Config;
use netsky_core::consts::{
    AGENT0_NAME, AGENTINFINITY_NAME, CLAUDE, DISK_MIN_MB_DEFAULT, ENV_DISK_MIN_MB,
    ENV_TICKER_INTERVAL, LAUNCHD_LABEL, NETSKY_BIN, TICKER_INTERVAL_DEFAULT_S, TICKER_LOG_PATH,
    TICKER_SESSION, TMUX_BIN,
};
use netsky_core::paths::{
    agent0_hang_marker, agent0_inbox_dir, agentinfinity_ready_marker, agentinit_escalation_marker,
    agentinit_failures_file, handoff_archive_dir, is_netsky_source_tree, resolve_netsky_dir,
    ticker_missing_count_file,
};
use netsky_sh::{tmux, which};

#[derive(Clone, Copy, PartialEq, Eq)]
enum Verdict {
    Green,
    Yellow,
    Red,
}

struct Doc {
    brief: bool,
    quiet: bool,
    red: u32,
    yellow: u32,
}

pub fn run(brief: bool, quiet: bool) -> netsky_core::Result<()> {
    let mut d = Doc {
        brief,
        quiet,
        red: 0,
        yellow: 0,
    };

    d.section("tmux");
    for sess in [AGENT0_NAME, AGENTINFINITY_NAME] {
        if tmux::session_is_alive(sess) {
            d.check(sess, Verdict::Green, "session alive");
        } else if tmux::has_session(sess) {
            d.check(sess, Verdict::Red, "session pane exited");
        } else {
            d.check(sess, Verdict::Red, "session MISSING");
        }
    }
    check_ticker(&mut d);

    d.section("channels");
    check_inbox(&mut d);

    d.section("markers");
    check_markers(&mut d);

    d.section("watchdog");
    check_watchdog(&mut d);

    d.section("disk");
    check_disk(&mut d);

    d.section("archives");
    check_archives(&mut d);

    d.section("binaries");
    for bin in [CLAUDE, NETSKY_BIN, TMUX_BIN] {
        match which(bin) {
            Some(p) => d.check(bin, Verdict::Green, &p.display().to_string()),
            None => d.check(
                bin,
                Verdict::Red,
                "not on PATH — see ONBOARDING.md or run /onboard",
            ),
        }
    }

    d.section("config");
    check_config(&mut d);

    d.section("hooks");
    check_hooks(&mut d);

    d.section("launchctl");
    check_launchctl(&mut d);

    let red = d.red;
    let yellow = d.yellow;
    d.summary(red, yellow);

    if red > 0 {
        netsky_core::bail!("doctor: red")
    } else {
        Ok(())
    }
}

// ---- checks ---------------------------------------------------------------

fn check_inbox(d: &mut Doc) {
    let inbox = agent0_inbox_dir();
    match fs::read_dir(&inbox) {
        Ok(entries) => {
            let n = entries
                .filter_map(|e| e.ok())
                .filter(|e| e.path().extension().and_then(|s| s.to_str()) == Some("json"))
                .count();
            let (v, det) = match n {
                0 => (Verdict::Green, "empty (drained)".to_string()),
                1..=4 => (Verdict::Yellow, format!("{n} envelope(s) pending")),
                _ => (
                    Verdict::Red,
                    format!("{n} envelopes backing up — MCP poll stalled?"),
                ),
            };
            d.check("agent0 inbox", v, &det);
        }
        Err(_) => d.check(
            "agent0 inbox",
            Verdict::Red,
            &format!("directory missing: {}", inbox.display()),
        ),
    }
}

fn check_markers(d: &mut Doc) {
    let ready = agentinfinity_ready_marker();
    if ready.exists() {
        d.check("agentinfinity-ready", Verdict::Green, "present");
    } else {
        d.check(
            "agentinfinity-ready",
            Verdict::Red,
            "marker missing; agentinit phase-2 never completed",
        );
    }

    let esc = agentinit_escalation_marker();
    if esc.exists() {
        d.check(
            "agentinit-escalation",
            Verdict::Red,
            &format!("ESCALATION FIRED — see {}", esc.display()),
        );
    } else {
        d.check(
            "agentinit-escalation",
            Verdict::Green,
            "absent (no escalation)",
        );
    }

    let hang = agent0_hang_marker();
    if hang.exists() {
        d.check(
            "agent0-hang-suspected",
            Verdict::Red,
            &format!("HANG SUSPECTED — see {}", hang.display()),
        );
    } else {
        d.check(
            "agent0-hang-suspected",
            Verdict::Green,
            "absent (pane is progressing)",
        );
    }

    let fails = agentinit_failures_file();
    match fs::read_to_string(&fails) {
        Ok(s) => {
            let n = s.lines().filter(|l| !l.trim().is_empty()).count();
            let (v, det) = match n {
                0 => (Verdict::Green, "empty".to_string()),
                1..=2 => (Verdict::Yellow, format!("{n} recent failure(s) in window")),
                _ => (
                    Verdict::Red,
                    format!("{n} failures in window (at/over threshold)"),
                ),
            };
            d.check("agentinit-failures", v, &det);
        }
        Err(_) => d.check("agentinit-failures", Verdict::Green, "no failures recorded"),
    }
}

fn check_watchdog(d: &mut Doc) {
    let log = Path::new(TICKER_LOG_PATH);
    if !log.exists() {
        d.check(
            "watchdog.out.log",
            Verdict::Red,
            &format!("log not present at {}", log.display()),
        );
        return;
    }
    let age = age_seconds(log).unwrap_or(u64::MAX);
    let (v, det) = match age {
        a if a <= 300 => (Verdict::Green, format!("{a}s ago")),
        a if a <= 600 => (Verdict::Yellow, format!("{a}s ago (>5min)")),
        a => (
            Verdict::Red,
            format!("{a}s ago (>10min — watchdog stalled?)"),
        ),
    };
    d.check("watchdog.out.log mtime", v, &det);

    if let Some(last) = last_line(log) {
        let truncated: String = last.chars().take(70).collect();
        d.check("watchdog last entry", Verdict::Green, &truncated);
    }

    // Tick-rate over last 10 minutes.
    let interval = std::env::var(ENV_TICKER_INTERVAL)
        .ok()
        .and_then(|v| v.parse::<u64>().ok())
        .unwrap_or(TICKER_INTERVAL_DEFAULT_S);
    let expected = 600u64.saturating_div(interval).max(1);
    let half = expected / 2;
    if let Ok(content) = fs::read_to_string(log) {
        let cutoff_s = unix_now().saturating_sub(600);
        let count = content
            .lines()
            .filter(|l| tick_line_after(l, cutoff_s))
            .count() as u64;
        let detail = format!("{count} tick(s) in last 10min (expected ~{expected})");
        let v = if count >= half {
            Verdict::Green
        } else if count > 0 {
            Verdict::Yellow
        } else {
            Verdict::Red
        };
        d.check("watchdog tick-rate", v, &detail);
    }
}

fn check_ticker(d: &mut Doc) {
    if tmux::has_session(TICKER_SESSION) {
        d.check(TICKER_SESSION, Verdict::Green, "session alive");
        return;
    }

    let missing = fs::read_to_string(ticker_missing_count_file())
        .ok()
        .and_then(|s| s.trim().parse::<u32>().ok())
        .unwrap_or(0);
    let detail = match missing {
        0 => "session missing; watchdog will self-heal on the next miss".to_string(),
        1 => "missing for 1 tick; watchdog will respawn on the second miss".to_string(),
        n => format!("missing for {n} ticks; watchdog self-heal pending"),
    };
    d.check(TICKER_SESSION, Verdict::Yellow, &detail);
}

fn check_disk(d: &mut Doc) {
    let min_mb = std::env::var(ENV_DISK_MIN_MB)
        .ok()
        .and_then(|v| v.parse::<u64>().ok())
        .unwrap_or(DISK_MIN_MB_DEFAULT);

    let avail_mb = match Command::new("df").args(["-Pk", "/tmp"]).output() {
        Ok(o) if o.status.success() => String::from_utf8_lossy(&o.stdout)
            .lines()
            .nth(1)
            .and_then(|line| line.split_whitespace().nth(3))
            .and_then(|s| s.parse::<u64>().ok())
            .map(|kb| kb / 1024),
        _ => None,
    };
    match avail_mb {
        Some(mb) if mb >= min_mb * 2 => d.check(
            "/tmp free",
            Verdict::Green,
            &format!("{mb}MB (threshold {min_mb}MB)"),
        ),
        Some(mb) if mb >= min_mb => d.check(
            "/tmp free",
            Verdict::Yellow,
            &format!("{mb}MB (close to {min_mb}MB threshold)"),
        ),
        Some(mb) => d.check(
            "/tmp free",
            Verdict::Red,
            &format!("{mb}MB < {min_mb}MB threshold"),
        ),
        None => d.check("/tmp free", Verdict::Yellow, "df parse failed"),
    }
}

fn check_archives(d: &mut Doc) {
    let archive = handoff_archive_dir();
    if archive.exists() {
        if fs::metadata(&archive).is_ok_and(|m| !m.permissions().readonly()) {
            let n = fs::read_dir(&archive)
                .map(|es| {
                    es.filter_map(|e| e.ok())
                        .filter(|e| e.path().extension().and_then(|s| s.to_str()) == Some("json"))
                        .count()
                })
                .unwrap_or(0);
            d.check(
                "handoff archive",
                Verdict::Green,
                &format!("{n} envelope(s); writable"),
            );
        } else {
            d.check("handoff archive", Verdict::Red, "exists but not writable");
        }
    } else if let Some(parent) = archive.parent()
        && fs::metadata(parent).is_ok_and(|m| !m.permissions().readonly())
    {
        d.check(
            "handoff archive",
            Verdict::Green,
            "not yet created (parent writable; created on next restart)",
        );
    } else {
        d.check("handoff archive", Verdict::Red, "parent dir not writable");
    }
}

fn check_config(d: &mut Doc) {
    // NETSKY_DIR resolution. Binary-only installs are green and get a
    // note so the operator knows the source tree is optional.
    let dir = resolve_netsky_dir();
    if is_netsky_source_tree(&dir) {
        d.check("netsky dir", Verdict::Green, &dir.display().to_string());
    } else {
        d.check(
            "netsky dir",
            Verdict::Green,
            &format!(
                "binary mode at {} (prompts, addenda, and notes live under ~/.netsky)",
                dir.display()
            ),
        );
    }

    // netsky.toml presence + parse health. Missing is GREEN (steady
    // state today; loader treats it as "use env + defaults"). Present-
    // but-malformed is RED — a parse error means env precedence still
    // works but anyone editing the file is going to be confused.
    let toml_path = netsky_core::config::netsky_toml_path();
    match Config::load_from(&toml_path) {
        Ok(None) => d.check(
            "netsky.toml",
            Verdict::Green,
            &format!("absent at {} (env + defaults active)", toml_path.display()),
        ),
        Ok(Some(cfg)) => {
            let owner = cfg
                .owner
                .as_ref()
                .and_then(|o| o.name.as_deref())
                .unwrap_or("(no [owner].name)");
            let machine = cfg
                .netsky
                .as_ref()
                .and_then(|n| n.machine_id.as_deref())
                .unwrap_or("(no [netsky].machine_id)");
            let addendum = cfg
                .addendum
                .as_ref()
                .and_then(|a| a.agent0.as_deref())
                .unwrap_or("(no [addendum].agent0; defaults to 0.md)");
            d.check(
                "netsky.toml",
                Verdict::Green,
                &format!(
                    "loaded from {} (owner={owner}, machine={machine}, addendum.agent0={addendum})",
                    toml_path.display()
                ),
            );
        }
        Err(e) => d.check(
            "netsky.toml",
            Verdict::Red,
            &format!("parse failed at {}: {e}", toml_path.display()),
        ),
    }
}

fn check_hooks(d: &mut Doc) {
    let root = match Command::new("git")
        .args(["rev-parse", "--show-toplevel"])
        .output()
    {
        Ok(o) if o.status.success() => {
            let s = String::from_utf8_lossy(&o.stdout).trim().to_string();
            if s.is_empty() {
                return;
            }
            Path::new(&s).to_path_buf()
        }
        _ => return,
    };
    let canonical = root.join(".githooks/pre-push");
    if !canonical.exists() {
        // Not a netsky-style repo; silently skip.
        return;
    }
    let hook = root.join(".git/hooks/pre-push");
    match crate::cmd::hooks::classify(&hook) {
        crate::cmd::hooks::HookState::CanonicalLink => {
            // classify() is pure topology — the symlink points where we
            // want. Health requires the canonical also exist and be
            // executable. A missing/non-exec canonical is a silent
            // "hook present but will crash on invocation" hole.
            if !canonical.exists() {
                d.check(
                    "pre-push hook",
                    Verdict::Yellow,
                    "symlink ok but canonical missing",
                );
            } else if !fs::metadata(&canonical)
                .map(|m| m.permissions().mode() & 0o111 != 0)
                .unwrap_or(false)
            {
                d.check(
                    "pre-push hook",
                    Verdict::Yellow,
                    "symlink ok but canonical not executable",
                );
            } else {
                d.check("pre-push hook", Verdict::Green, "canonical symlink");
            }
        }
        crate::cmd::hooks::HookState::Absent => {
            d.check(
                "pre-push hook",
                Verdict::Red,
                "absent — run `netsky hooks install`",
            );
        }
        crate::cmd::hooks::HookState::Drift => {
            d.check(
                "pre-push hook",
                Verdict::Red,
                "drifted — run `netsky hooks install --force`",
            );
        }
    }
    if std::env::var("SKIP_PREPUSH_CHECK").is_ok_and(|v| v == "1") {
        d.check(
            "SKIP_PREPUSH_CHECK",
            Verdict::Yellow,
            "set in env; hook will be bypassed",
        );
    }
}

fn check_launchctl(d: &mut Doc) {
    if which("launchctl").is_none() {
        return;
    }
    let target = format!("gui/{}/{LAUNCHD_LABEL}", unsafe { getuid() });
    let loaded = Command::new("launchctl")
        .args(["print", &target])
        .stdout(std::process::Stdio::null())
        .stderr(std::process::Stdio::null())
        .status()
        .map(|s| s.success())
        .unwrap_or(false);
    if loaded {
        d.check(LAUNCHD_LABEL, Verdict::Green, "plist loaded");
    } else {
        d.check(
            LAUNCHD_LABEL,
            Verdict::Yellow,
            "plist not loaded (tmux-ticker is primary; this is fallback)",
        );
    }
}

// ---- printing -------------------------------------------------------------

impl Doc {
    fn section(&self, label: &str) {
        if !self.brief && !self.quiet {
            println!();
            println!("{label}");
        }
    }
    fn check(&mut self, name: &str, v: Verdict, detail: &str) {
        match v {
            Verdict::Green => {}
            Verdict::Yellow => self.yellow += 1,
            Verdict::Red => self.red += 1,
        }
        if self.brief || self.quiet {
            return;
        }
        let badge = match v {
            Verdict::Green => "ok",
            Verdict::Yellow => "warn",
            Verdict::Red => "FAIL",
        };
        println!("  [{badge:>4}] {name:<34} {detail}");
    }
    fn summary(&self, red: u32, yellow: u32) {
        if self.quiet {
            return;
        }
        if !self.brief {
            println!();
            print!("summary: ");
        }
        if red > 0 {
            println!("RED — {red} fail, {yellow} warn");
        } else if yellow > 0 {
            println!("YELLOW — {yellow} warn");
        } else {
            println!("GREEN — all checks passed");
        }
    }
}

// ---- util ---------------------------------------------------------------

fn age_seconds(path: &Path) -> Option<u64> {
    let mtime = fs::metadata(path).ok()?.modified().ok()?;
    SystemTime::now()
        .duration_since(mtime)
        .ok()
        .map(|d| d.as_secs())
}

fn last_line(path: &Path) -> Option<String> {
    fs::read_to_string(path)
        .ok()
        .and_then(|s| s.lines().last().map(|s| s.to_string()))
}

fn unix_now() -> u64 {
    SystemTime::now()
        .duration_since(UNIX_EPOCH)
        .map(|d| d.as_secs())
        .unwrap_or(0)
}

/// True when a log line of the form `[watchdog-tick <ISO-TS>] ...` has a
/// timestamp at or after `cutoff_s` (unix seconds). ISO-8601 strings in
/// the log sort lexicographically in chronological order, so we compare
/// against the ISO form of the cutoff without parsing each tick line.
fn tick_line_after(line: &str, cutoff_s: u64) -> bool {
    let Some(rest) = line.strip_prefix("[watchdog-tick ") else {
        return false;
    };
    let Some(end) = rest.find(']') else {
        return false;
    };
    let ts = &rest[..end];
    let cutoff_iso = chrono::DateTime::<chrono::Utc>::from_timestamp(cutoff_s as i64, 0)
        .map(|dt| dt.format("%Y-%m-%dT%H:%M:%SZ").to_string());
    match cutoff_iso {
        Some(c) => ts >= c.as_str(),
        None => false,
    }
}

#[allow(non_snake_case)]
unsafe fn getuid() -> u32 {
    unsafe extern "C" {
        fn getuid() -> u32;
    }
    unsafe { getuid() }
}