Ubuntu编译的二进制在Alpine中调用duplocale()触发段错误的解决办法

问题重现 #

在Ubuntu 22.04上编译以下C代码：

#include <locale.h>
#include <stdio.h>

int main() {
        printf("1\n");
        locale_t oldlocale = uselocale(NULL);
        printf("2\n");
        locale_t duploc = duplocale(oldlocale);
        printf("3\n");
        return 0;
}

Ubuntu上运行正常：

$ gcc -g c.c && ./a.out 
1
2
3

但将二进制复制到Alpine容器中运行时，调用duplocale()触发段错误：

# ./a.out 
1
2
Segmentation fault (core dumped)

环境说明 #

已在Alpine容器中安装libc6-compat等兼容组件，且该代码在Alpine本地编译可正常运行，但实际场景无法采用本地编译方式。问题源于依赖的libjson-c调用duplocale()导致段错误，即使动态链接Alpine版本的json-c也无法解决。

调试信息 #

Alpine中的调试输出 #

process 340 is executing new program: /lib/ld-musl-x86_64.so.1
1
2

Program received signal SIGSEGV, Segmentation fault.
0x00007ffff7f46aa7 in ?? ()
(gdb) bt
#0  0x00007ffff7f46aa7 in ?? ()
#1  0x0000000000000001 in ?? ()
#2  0x00007ffff7f521cd in ?? ()
#3  0xffffffffffffffff in ?? ()
#4  0x00007ffff7fc1600 in __libc_start_init () at ldso/dynlink.c:1621
#5  0x00007ffff7f786d1 in libc_start_main_stage2 (main=0x7ffff7f52189, argc=-1, argv=0x7fffffffebb0)
    at src/env/__libc_start_main.c:95
#6  0x00007ffff7f520c5 in ?? ()
#7  0x00007fffffffebf8 in ?? ()
#8  0x00007ffff7f520a0 in ?? ()
#9  0x0000000000000001 in ?? ()
#10 0x00007fffffffee50 in ?? ()
#11 0x0000000000000000 in ?? ()

带回Ubuntu调试core文件的输出 #

Core was generated by `ld-linux-x86-64.so.2 --argv0 ./a.out --preload /lib/libgcompat.so.0  -- /dm/res'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x00007f1c76999aa7 in ?? ()
(gdb) bt
#0  0x00007f1c76999aa7 in ?? ()
#1  0x0000000000000001 in ?? ()
#2  0x00007f1c769a51cd in ?? ()
#3  0xffffffffffffffff in ?? ()
#4  0x00007f1c76a0e600 in ?? ()
#5  0x00007ffc8621ad98 in ?? ()
#6  0x00007f1c769c56d1 in ?? ()
#7  0x0000000000000000 in ?? ()

注：二进制保留调试信息，未被剥离。

关键线索 #

json-c的CMakeLists.txt中提到：

# uClibc *intentionally* crashes in duplocale(), at least as of:
# https://github.com/ffainelli/uClibc/blob/266bdc1/libc/misc/locale/locale.c#L1322
# So, if it looks like we're compiling for a system like that just disable
# locale handling entirely.

可行解决办法 #

• 禁用json-c的locale支持：若能重新编译依赖的json-c，添加编译选项-DENABLE_LOCALES=OFF，让json-c跳过locale相关逻辑，避免调用duplocale()。
• Hook替换duplocale函数：通过LD_PRELOAD加载自定义共享库，替换duplocale的实现。示例代码：

  #define _GNU_SOURCE
  #include <locale.h>
  #include <dlfcn.h>
  
  locale_t duplocale(locale_t loc) {
      // 直接返回传入的locale，不执行复制操作（需注意线程安全）
      return loc;
  }
  

  编译为共享库：gcc -shared -fPIC hook_duplocale.c -o libhook_duplocale.so，运行时加载：LD_PRELOAD=./libhook_duplocale.so ./your_binary
• 静态链接glibc：在Ubuntu编译时添加-static选项静态链接glibc，让二进制不依赖Alpine的musl libc，但会增大体积，且可能带来其他兼容性问题。
• 更换运行容器：若业务允许，直接使用Ubuntu容器运行，规避libc兼容问题。

内容的提问来源于stack exchange，提问作者hudac