#!/bin/bash

# 核心配置
CRASH_LOG_DIR="./crash_logs"
RUN_DURATION=2
GDB_TIMEOUT=$((RUN_DURATION + 2))  # 延长超时，确保GDB完全初始化
MAX_LOOP=0
# 强制捕获所有致命信号
CRASH_SIGNALS=("SIGSEGV" "SIGABRT" "SIGILL" "SIGFPE" "SIGBUS")

# 创建日志目录
mkdir -p ${CRASH_LOG_DIR}
rm -f ${CRASH_LOG_DIR}/*  # 清空旧日志（可选）

# 清理进程函数
cleanup() {
    pkill -9 gdb >/dev/null 2>&1
    pkill -9 pPowerManger >/dev/null 2>&1
    sleep 0.2
}

# 检查是否是真崩溃（从日志+系统日志双重验证）
is_real_crash() {
    local log_file=$1
    local loop=$2
    # 1. 检查GDB日志中的崩溃关键字
    if grep -qiE "segmentation fault|sigsegv|crash detected|backtrace" "${log_file}"; then
        return 0
    fi
    # 2. 检查系统dmesg中的段错误记录（终极验证）
    if dmesg | grep -i "pPowerManger.*segfault" | grep -i "loop_${loop}"; then
        return 0
    fi
    return 1
}

# 打印系统级崩溃日志（dmesg）
print_system_crash_log() {
    local loop=$1
    echo -e "\033[33m===== 系统级崩溃日志（dmesg）=====\033[0m"
    dmesg | grep -A 5 -B 2 "pPowerManger.*segfault" | tail -10
    echo -e "\033[33m===================================\033[0m"
}

echo -e "🔍 启动pPowerManger稳定性测试（强制捕获段错误）\n"
echo "📂 崩溃日志目录: ${CRASH_LOG_DIR}"
echo "⏱ 程序运行时长: ${RUN_DURATION}秒"
echo "📌 捕获信号: ${CRASH_SIGNALS[*]}"
echo -e "========================================\n"

cleanup  # 初始清理
LOOP_COUNT=0

while true; do
    LOOP_COUNT=$((LOOP_COUNT + 1))
    if [ ${MAX_LOOP} -ne 0 ] && [ ${LOOP_COUNT} -gt ${MAX_LOOP} ]; then
        echo -e "\n✅ 测试完成（${MAX_LOOP}次循环无崩溃）"
        cleanup
        exit 0
    fi

    echo -e "===== 第 ${LOOP_COUNT} 次循环 ====="
    CRASH_LOG="${CRASH_LOG_DIR}/crash_loop_${LOOP_COUNT}.log"
    GDB_CMD="${CRASH_LOG_DIR}/gdb_cmds_${LOOP_COUNT}.gdb"

    # 生成强化版GDB脚本：强制捕获所有信号，优先初始化信号处理
    cat > ${GDB_CMD} << EOF
# 核心：GDB启动后先设置信号捕获，再启动程序
set args pPowerManger.moos
set timeout ${GDB_TIMEOUT}
set print pretty on
set print frame-arguments all
set logging file ${CRASH_LOG}.gdb_raw
set logging on  # 开启GDB原始日志（双重保障）

# 强制捕获所有致命信号
EOF

# 为每个崩溃信号添加捕获规则（追加到GDB脚本）
for sig in "${CRASH_SIGNALS[@]}"; do
    cat >> ${GDB_CMD} << EOF
handle ${sig} stop print pass nostop  # 即使子线程崩溃也捕获
EOF
done

# 追加程序启动和崩溃处理逻辑
cat >> ${GDB_CMD} << EOF
# 延迟10ms启动程序，确保GDB完全初始化
sleep 0.01
run

# 崩溃后强制打印所有信息（无论是否主线程）
if \$exitcode != 0 || \$signal_number != 0
    echo "\n====================================="
    echo "            CRASH DETECTED"
    echo "====================================="
    echo "Loop: ${LOOP_COUNT}"
    echo "Exit code: \$exitcode"
    echo "Signal: \$signal_name (\$signal_number)"
    echo "====================================="
    thread apply all bt full  # 打印所有线程的调用栈（关键！）
    info threads
    info proc mappings  # 打印内存映射（排查内存越界）
    info locals
    info registers
endif

set logging off
quit
EOF

    # 核心：用nohup+GDB启动，避免输出丢失；同时记录进程ID
    echo "启动程序（GDB强化模式）..."
    nohup timeout ${GDB_TIMEOUT} gdb -batch -x ${GDB_CMD} ./pPowerManger > ${CRASH_LOG} 2>&1 &
    GDB_PID=$!
    wait ${GDB_PID}  # 等待GDB执行完成

    # 双重验证：检查是否真崩溃
    if is_real_crash ${CRASH_LOG} ${LOOP_COUNT}; then
        # 捕获到真崩溃
        echo -e "\033[31m❌ 第 ${LOOP_COUNT} 次循环捕获到段错误！\033[0m"
        echo -e "\033[31m📄 崩溃日志: ${CRASH_LOG}\033[0m"
        print_system_crash_log ${LOOP_COUNT}
        
        # 打印最关键的调用栈（所有线程）
        echo -e "\033[31m===== 所有线程崩溃调用栈 =====\033[0m"
        grep -A 50 "thread apply all bt full" ${CRASH_LOG} | grep -v "No stack frame"
        echo -e "\033[31m==============================\033[0m"
        
        cleanup
        echo -e "\n🛑 测试终止，完整日志在 ${CRASH_LOG}"
        exit 1
    else
        # 无崩溃（正常退出/启动失败）
        if grep -qi "Program exited normally" ${CRASH_LOG}; then
            echo "✅ 程序正常运行${RUN_DURATION}秒后退出"
        else
            echo "⚠️  程序启动失败（非段错误），继续循环"
            # 保留启动失败日志（方便排查）
            mv ${CRASH_LOG} ${CRASH_LOG}.start_fail
        fi
        rm -f ${GDB_CMD} ${CRASH_LOG}.gdb_raw
    fi

    cleanup
    echo -e "第 ${LOOP_COUNT} 次循环完成\n"
    sleep 0.5
done