Handle EINTR error in waitpid call for POSIX

[robo-monk]

Using a debugger such as GDB or CodeLLDB on a program that uses nob.h and depends something like this:

if (!nob_cmd_run(&cmd)) {
    fatal("Failed to run command");
  };

causes the debugging process to detach because of:

[ERROR] could not wait on command (pid 16237): Interrupted system call

This happens when a signal interrupts the waitpid() call, setting errno to EINTR.

This fix adds a retry logic for waitpid() to correctly handle EINTR, ensuring the command wait loop resumes normally instead of treating it as a fatal error.

[rexim]

Could you please provide more detailed instructions to reproduce the problem? I want to verify that the fix works on my side. Thank you!

[robo-monk]

Hi Tsoding, happy new year!

1. main.c

#include <signal.h>
#include <unistd.h>

#define NOB_IMPLEMENTATION
#include "nob.h"

  // add a breakpoint here
void breakpoint() {
  nob_log(NOB_INFO, "34 + 35");
} 


int main(void) {
  Nob_Cmd cmd = {0};
  nob_cmd_append(&cmd, "sleep", "2");

  nob_log(NOB_INFO, "Running command...");
  bool ok = nob_cmd_run(&cmd);

  nob_log(NOB_INFO, "Command ran: %s", ok ? "true" : "false");

  nob_cmd_free(cmd);

  breakpoint();

  return ok ? 0 : 1;
}

2. compile && attach debugger

cc -g -o w main.c
lldb ./w

3. add demo breakpoint and start execution

(lldb) b breakpoint
Breakpoint 1: 2 locations.
(lldb) r
Process 93430 launched: '/Users/monk/Documents/random_shit/ccomptime/w' (arm64)
[INFO] Running command...
[INFO] CMD: sleep 2
[ERROR] could not wait on command (pid 93500): Interrupted system call
[INFO] Command ran: false
Process 93430 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
    frame #0: 0x0000000100007040 w`breakpoint at waitpid_eintr.c:9:3
   6
   7      // add a breakpoint here
   8    void breakpoint() {
-> 9      nob_log(NOB_INFO, "34 + 35");
   10   }
   11
   12
Target 0: (w) stopped.

---

The fix hopefully transforms the above into something like this:

(lldb) target create "./w"
Current executable set to '/Users/monk/Documents/random_shit/ccomptime/w' (arm64).
(lldb) b breakpoint
Breakpoint 1: 2 locations.
(lldb) r
Process 92350 launched: '/Users/monk/Documents/random_shit/ccomptime/w' (arm64)
[INFO] Running command...
[INFO] CMD: sleep 2
[INFO] Command ran: true
Process 92350 stopped
* thread #1, queue = 'com.apple.main-thread', stop reason = breakpoint 1.1
    frame #0: 0x0000000100007040 w`breakpoint at waitpid_eintr.c:9:3
   6
   7      // add a breakpoint here
   8    void breakpoint() {
-> 9      nob_log(NOB_INFO, "34 + 35");
   10   }
   11
   12
Target 0: (w) stopped.
(lldb) c
Process 92350 resuming
[INFO] 34 + 35
Process 92350 exited with status = 0 (0x00000000)