ObsidianDragon/src/daemon/daemon_busy_detect.h

#pragma once

// Is the node busy with something that stopping it would throw away? Tracks dragonxd's own log markers,
// in order, for two independent states:
//   - rebuilding: a witness-cache rebuild (BuildWitnessCache). It holds cs_main for up to an hour on a
//     big wallet, its second phase logs nothing and ignores shutdown, and a node killed mid-rebuild redoes
//     it (after replaying everything since its last flush) on the next start;
//   - replaying: the startup replay ("Activating best chain..."). Stopping there is safe — the daemon
//     checks for shutdown between steps and flushes — it just takes a moment.
// A rebuild can start in the middle of a replay, so its end must not clear `replaying`, and the state is
// order-based: it never compares timestamps (POSIX nodes log to stdout without them). Pure and
// header-only, so the test target can feed it lines.

#include <cstdint>
#include <cstdio>
#include <string>

namespace dragonx {
namespace daemon {

/**
 * "YYYY-MM-DD HH:MM:SS ..." (dragonxd's debug.log prefix, which is UTC) -> seconds since the epoch, or 0
 * when the line has no such prefix. Days-from-civil arithmetic, not mktime(): mktime reads local time.
 */
inline std::int64_t parseDaemonLogUtc(const std::string& line)
{
    int y = 0, mo = 0, d = 0, h = 0, mi = 0, s = 0;
    if (line.size() < 19 || std::sscanf(line.c_str(), "%4d-%2d-%2d %2d:%2d:%2d", &y, &mo, &d, &h, &mi, &s) != 6)
        return 0;
    if (mo < 1 || mo > 12 || d < 1 || d > 31 || h > 23 || mi > 59 || s > 60) return 0;
    // Howard Hinnant's days_from_civil.
    const int yy = y - (mo <= 2 ? 1 : 0);
    const int era = (yy >= 0 ? yy : yy - 399) / 400;
    const unsigned yoe = static_cast<unsigned>(yy - era * 400);
    const unsigned doy = (153u * static_cast<unsigned>(mo + (mo > 2 ? -3 : 9)) + 2u) / 5u + static_cast<unsigned>(d) - 1u;
    const unsigned doe = yoe * 365u + yoe / 4u - yoe / 100u + doy;
    const std::int64_t days = static_cast<std::int64_t>(era) * 146097 + static_cast<std::int64_t>(doe) - 719468;
    return days * 86400 + h * 3600 + mi * 60 + s;
}

class DaemonBusyTracker {
public:
    // Feed raw output: any chunking, a line split across two chunks is reassembled.
    void feed(const std::string& chunk)
    {
        for (char c : chunk) {
            if (c == '\n') {
                feedLine(carry_);
                carry_.clear();
            } else if (c != '\r') {
                carry_.push_back(c);
                if (carry_.size() > 64 * 1024) carry_.clear();   // not a log line; don't grow without bound
            }
        }
    }

    void feedLine(const std::string& line)
    {
        // A new daemon process, or the end of the last one: forget everything.
        if (contains(line, "DragonX version ") || contains(line, "Hush version ") || contains(line, "Shutdown: done")) {
            reset();
            return;
        }
        if (contains(line, "init message: Activating best chain")) {
            replaying_ = true;
            replayStartUtc_ = parseDaemonLogUtc(line);
            lastMarker_ = line;
            return;
        }
        if (contains(line, "init message: Full Node Done Loading") || contains(line, "init message: Done loading")) {
            replaying_ = false;
            lastMarker_ = line;
            return;
        }
        const size_t read = line.find("Reading blocks for witness rebuild: ");
        if (read != std::string::npos || contains(line, "Building Witnesses for block ")) {
            if (!rebuilding_) rebuildStartUtc_ = parseDaemonLogUtc(line);
            rebuilding_ = true;
            if (read != std::string::npos) {
                long done = 0, total = 0;
                if (std::sscanf(line.c_str() + read, "Reading blocks for witness rebuild: %ld / %ld", &done, &total) == 2) {
                    readDone_ = done;
                    readTotal_ = total;
                }
            }
            lastMarker_ = line;
            return;
        }
        // The rebuild's end. UpdateTip is logged under cs_main, so one after a rebuild marker means the
        // rebuild has released it (exceptions and non-bulk paths print no end line).
        if (rebuilding_ && (contains(line, "note witness cache(s) to height") ||
                            contains(line, "aborting witness rebuild") ||
                            contains(line, "aborted during witness rebuild") ||
                            contains(line, "aborting building witnesses") ||
                            contains(line, "UpdateTip: "))) {
            rebuilding_ = false;
            readDone_ = readTotal_ = 0;
            rebuildStartUtc_ = 0;
            lastMarker_ = line;
        }
    }

    void reset()
    {
        rebuilding_ = replaying_ = false;
        readDone_ = readTotal_ = 0;
        rebuildStartUtc_ = replayStartUtc_ = 0;
        lastMarker_.clear();
    }

    bool rebuilding() const { return rebuilding_; }
    bool replaying() const { return replaying_; }
    bool busy() const { return rebuilding_ || replaying_; }
    long readDone() const { return readDone_; }     // the last Phase-1 "reading blocks" progress, 0 if none
    long readTotal() const { return readTotal_; }
    std::int64_t rebuildStartUtc() const { return rebuildStartUtc_; }   // 0: unknown (no timestamps)
    std::int64_t replayStartUtc() const { return replayStartUtc_; }
    const std::string& lastMarker() const { return lastMarker_; }

private:
    static bool contains(const std::string& s, const char* needle) { return s.find(needle) != std::string::npos; }

    std::string carry_;
    bool rebuilding_ = false;
    bool replaying_ = false;
    long readDone_ = 0;
    long readTotal_ = 0;
    std::int64_t rebuildStartUtc_ = 0;
    std::int64_t replayStartUtc_ = 0;
    std::string lastMarker_;
};

/**
 * The "keep the node running?" guard on quit. Prompt only when this quit would stop the node and it is
 * rebuilding witnesses, or has been alive but too busy to answer RPC for a while (a rebuild that logged
 * no marker). A replay alone never prompts: stopping it is safe. No daemon version suppresses the prompt
 * (v1.3.0 does not checkpoint a witness rebuild; nothing survives a kill).
 */
inline bool shouldConfirmStop(bool stopsDaemon, const DaemonBusyTracker& tracker, bool rpcBusyWhileAlive)
{
    return stopsDaemon && (tracker.rebuilding() || rpcBusyWhileAlive);
}

// Retry schedule after failing to find a pid's banner in debug.log: re-scan fast during the fresh-node grace
// window, then back off 4 s doubling to a 60 s cap. `backoff` is the caller's running state (0 until grace ends).
inline double nextExternalLogRetryDelay(double ageSinceMiss, double& backoff,
                                        double graceWindow = 15.0, double fast = 2.0,
                                        double base = 4.0, double cap = 60.0)
{
    if (ageSinceMiss < graceWindow) return fast;
    backoff = (backoff < base) ? base : (backoff * 2.0 < cap ? backoff * 2.0 : cap);
    return backoff;
}

/**
 * Where to start feeding a DaemonBusyTracker from the last `window` of the debug.log of a node someone else
 * started at `processStartUtc` (UTC seconds). Only lines that process wrote count: a node started with
 * -printtoconsole leaves an old debug.log behind, whose last run may have died mid-rebuild. Returns the
 * offset of that process's startup banner in the window; 0 when its banner is older than the window
 * (`windowStartsMidFile`) and the window's first stamped line postdates the process start; npos when the
 * log isn't that process's.
 */
inline size_t externalLogFeedOffset(const std::string& window, bool windowStartsMidFile, std::int64_t processStartUtc)
{
    size_t lastBanner = std::string::npos;
    std::int64_t lastBannerAt = 0, firstStamp = 0;
    size_t pos = 0;
    while (pos < window.size()) {
        size_t end = window.find('\n', pos);
        if (end == std::string::npos) end = window.size();
        const std::string line = window.substr(pos, end - pos);
        const std::int64_t at = parseDaemonLogUtc(line);
        if (at > 0 && firstStamp == 0) firstStamp = at;
        if (line.find("DragonX version ") != std::string::npos || line.find("Hush version ") != std::string::npos) {
            lastBanner = pos;
            lastBannerAt = at;
        }
        pos = end + 1;
    }
    if (processStartUtc <= 0) return std::string::npos;
    const std::int64_t since = processStartUtc - 5;   // the log's whole seconds vs the process's start time
    if (lastBanner != std::string::npos) return (lastBannerAt > 0 && lastBannerAt >= since) ? lastBanner : std::string::npos;
    if (windowStartsMidFile && firstStamp > 0 && firstStamp >= since) return 0;
    return std::string::npos;
}

// What the "node is starting / not answering" overlay should say, most specific first.
enum class DaemonStatusKind {
    Rebuilding,  // its own log says it's rebuilding note witnesses
    Replaying,   // its own log says it's catching up on blocks it saved before (startup replay)
    InitPhase,   // a recognisable init phase line (loading the block index, verifying, ...)
    Busy,        // it answers the connection but not the request (reply timeout / queue full)
    Launching,   // nothing bound to the RPC port yet
};

inline DaemonStatusKind chooseDaemonStatus(bool hasInitPhaseLine, bool reachableButBusy,
                                           const DaemonBusyTracker& tracker)
{
    if (tracker.rebuilding()) return DaemonStatusKind::Rebuilding;
    if (tracker.replaying()) return DaemonStatusKind::Replaying;
    if (hasInitPhaseLine) return DaemonStatusKind::InitPhase;
    if (reachableButBusy) return DaemonStatusKind::Busy;
    return DaemonStatusKind::Launching;
}

}  // namespace daemon
}  // namespace dragonx