From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <SRS0=pMmJ=3S=autistici.org=andrea.monaco@sourceware.org>
Received: from confino.investici.org (confino.investici.org [93.190.126.19])
	by sourceware.org (Postfix) with ESMTPS id 7D4CA3854562
	for <gdb@sourceware.org>; Fri, 18 Nov 2022 19:54:17 +0000 (GMT)
DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 7D4CA3854562
Authentication-Results: sourceware.org; dmarc=pass (p=reject dis=none) header.from=autistici.org
Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=autistici.org
Received: from mx1.investici.org (unknown [127.0.0.1])
	by confino.investici.org (Postfix) with ESMTP id 4NDSCp6jLxz117t
	for <gdb@sourceware.org>; Fri, 18 Nov 2022 19:54:14 +0000 (UTC)
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=autistici.org;
	s=stigmate; t=1668801254;
	bh=w4QDZjgkPAjnlvi+02X0sosRYEq3pSq5kXg1TY0BWK8=;
	h=From:To:Subject:Date:From;
	b=WepGgisLAx+RUN6gNBiFUkAyouznk2ZPwgB8IN5dC1WvO7eHCSveL2IOWfYDel9tT
	 vP4dIhbLbIxngzP9gEk05s5hQuaWBgWRAcHcKBTiU8QK2s1SOXxf94NGyJDkYQ5uRc
	 eglTKju7ZpiCoioVc5Ddth7wXvHY9gflp1woQ1pk=
Received: from [93.190.126.19] (mx1.investici.org [93.190.126.19]) (Authenticated sender: andrea.monaco@autistici.org) by localhost (Postfix) with ESMTPSA id 4NDSCp5pqFz117m
	for <gdb@sourceware.org>; Fri, 18 Nov 2022 19:54:14 +0000 (UTC)
From: Andrea Monaco <andrea.monaco@autistici.org>
To: gdb@sourceware.org
Subject: Program with pseudo-tty behaving differently in single-step
Date: Fri, 18 Nov 2022 20:54:13 +0100
Message-ID: <87a64om42i.fsf@autistici.org>
MIME-Version: 1.0
Content-Type: text/plain
X-Spam-Status: No, score=-3.1 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,SPF_HELO_PASS,SPF_PASS,TXREP autolearn=ham autolearn_force=no version=3.4.6
X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org
List-Id: <gdb.sourceware.org>


I have a program that behaves very differently when single-stepping and
when running freely in gdb.

The first part of the program opens a pseudo-terminal pair in the usual
way.  This pair includes a master (that sort of simulates the user: you
can read from master and write to master like a human would do) and a
slave (that the child process uses as a terminal for stdin and stdout,
receiving what you send to master and sending back to master).

The child process runs bc (a calculator).  The parent sends the line
"1+1\n" and reads back the subsequent line.


#define _GNU_SOURCE


#include <stdlib.h>
#include <stdio.h>
#include <fcntl.h>
#include <sys/types.h>
#include <unistd.h>
#include <string.h>


char buffer [1024];


int master, slave;


ssize_t
readl (int fd, char *buf)
{
  ssize_t i = 0;

  do
    {
      read (fd, buf+i, 1);
    }
  while (buf [i++] != '\n');

  return i;
}


int
main (int argc, char *argv [])
{
  int i;
  char *name;
  pid_t child;

  master = getpt ();

  if (master < 0)
    {
      perror ("getpt() failed");
      return 1;
    }

  if (grantpt (master) < 0 || unlockpt (master) < 0)
    {
      perror ("grantpt() or unlockpt() failed");
      return 1;
    }

  name = ptsname (master);
  
  if (!name)
    return 1;

  slave = open (name, O_RDWR);
  
  if (slave < 0)
    {
      perror ("could not open slave");
      return 1;
    }

  /* until now, we opened a pseudo-tty pair in the standard way */

  child = fork ();

  if (child < 0)
    {
      perror ("fork failed");
      return 1;
    }

  if (!child)
    {
      /* in the child, we close stdin and stdout and substitute them
	 both with the slave pseudo-terminal for input and output */
      
      close (1); 
      close (0);

      if (dup (slave) == -1 || dup (slave) == -1)
	{
	  perror ("dup failed");
	  return 1;
	}
      
      execlp ("bc", "bc", NULL);
      
      perror ("could not exec bc");
    }

  /* in the parent, we skip the startup message of bc */
  for (i = 0; i < 4; i++)
    {
      readl (master, buffer);
    }

  write (master, "1+1\n", strlen ("1+1\n"));
  readl (master, buffer); /* we skip the echo line of 1+1 */
  
  read (master, buffer, 3);  /* now we read the result line */

  buffer [3] = '\0';
  
  printf ("%s", buffer);
  
  return 0;
}


In gdb, when I do a run I get "1+1" as output.  When I single-step, I
get 2 from the last printf instead, which is the desired output.

Can anyone replicate this?  Is there a known explanation?



Andrea Monaco