c++ - 是否可以在异常(段错误)错误后自动重启 GDB 中的程序？

Question

我如何使用 GDB 在 C++ 中运行程序，以便在出现错误异常(段错误)时重新启动程序(我希望 GDB 自动使用“运行”命令)，同时将任何错误记录到文件中(命令“哪里”)。

有可能吗？

Answer 1

让我向您展示一个示例，在程序崩溃时重新启动程序 3 次。我使用 python 脚本来处理 SIGSEGV ( https://sourceware.org/gdb/onlinedocs/gdb/Events-In-Python.html )。

首先，这是一个 GDB session 的例子:

>gdb -q -x restart.py ./a.out
Reading symbols from /home/a.out...done.
process id: 1700

Program received signal SIGSEGV, Segmentation fault.
0x000000000040060e in c () at main2.cpp:9
9         *ptr = *ptr +1;
#0  0x000000000040060e in c () at main2.cpp:9
#1  0x000000000040062a in main () at main2.cpp:15
process id: 1704

Program received signal SIGSEGV, Segmentation fault.
0x000000000040060e in c () at main2.cpp:9
9         *ptr = *ptr +1;
#0  0x000000000040060e in c () at main2.cpp:9
#1  0x000000000040062a in main () at main2.cpp:15
process id: 1705

Program received signal SIGSEGV, Segmentation fault.
0x000000000040060e in c () at main2.cpp:9
9         *ptr = *ptr +1;
#0  0x000000000040060e in c () at main2.cpp:9
#1  0x000000000040062a in main () at main2.cpp:15
process id: 1706

Program received signal SIGSEGV, Segmentation fault.
0x000000000040060e in c () at main2.cpp:9
9         *ptr = *ptr +1;
#0  0x000000000040060e in c () at main2.cpp:9
#1  0x000000000040062a in main () at main2.cpp:15
(gdb)

每次崩溃都会创建一个名为 crash.file.PID 的文件。这是一个文件示例:

>more crash.file.1860
#0  0x000000000040060e in c () at main2.cpp:9
#1  0x000000000040062a in main () at main2.cpp:15

这是一个 C++ 程序:

#include <stdio.h>
#include <sys/types.h>
#include <unistd.h>

int c()
{
  printf("process id: %d\n", getpid());
  int *ptr =0;
  *ptr = *ptr +1;
  return *ptr;
}

int main()
{
  c();
  return 0;
}

这是一个 python 脚本:

>cat restart.py
#!gdb
import sys
import gdb
import os

number_restarts = 3

def on_stop(sig):
  global number_restarts
  if isinstance(sig, gdb.SignalEvent) and sig.stop_signal == "SIGSEGV":
    crash_file = "crash.file." + str( gdb.selected_inferior().pid)
    gdb.execute("set logging file " + crash_file)
    gdb.execute("set logging on")
    gdb.execute("where")
    gdb.execute("set logging off")
    if (number_restarts > 0):
      number_restarts -= 1
      gdb.execute("set confirm off")
      gdb.execute("kill")
      gdb.execute("run")


gdb.events.stop.connect (on_stop)
gdb.execute("set pagination off")
gdb.execute("run")