How can a Unix program display output on screen even when stdout and stderr are redirected?

Question

I was running a program (valgrind, actually) on my Ubuntu machine, and had redirected both stdout and stderr to different files. I was surprised to see a short message appea

Answer 1

The message is most probably from GCC's stack protector feature or from glib itself. If it's from GCC, it is output using the fail() function, which directly opens /dev/tty:

fd = open (_PATH_TTY, O_WRONLY);

_PATH_TTY is not really standard, but SingleUnix actually demands that /dev/tty exists.

Answer 2

Here is some sample code that does exactly what was asked (thanks to earlier answers pointing me in the right direction). Both are compiled with g++, and will print a message to the screen even when stdout and stderr are redirected.

For Linux (Ubuntu 14):

#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
#include <string.h>

int main( int, char *[]) {

    printf("This goes to stdout\n");

    fprintf(stderr, "This goes to stderr\n");

    int ttyfd = open("/dev/tty", O_RDWR);
    const char *msg = "This goes to screen\n";
    write(ttyfd, msg, strlen(msg));
}

For Windows 7, using MinGW:

#include <stdio.h>
#include <fcntl.h>
#include <string.h>
#include <conio.h>

void writeConsole( const char *s) {
    while( *s) {
        putch(*(s++));
    }
}

int main( int, char *[]) {  
    printf("This goes to stdout\n");

    fprintf(stderr, "This goes to stderr\n");

    writeConsole( "This goes to screen\n");
}

Answer 3

It is not written by valgrind but rather glibc and your ./myprogram is using glibc:

#define _PATH_TTY   "/dev/tty"

/* Open a descriptor for /dev/tty unless the user explicitly
   requests errors on standard error.  */
const char *on_2 = __libc_secure_getenv ("LIBC_FATAL_STDERR_");
if (on_2 == NULL || *on_2 == '\0')
  fd = open_not_cancel_2 (_PATH_TTY, O_RDWR | O_NOCTTY | O_NDELAY);

if (fd == -1)
  fd = STDERR_FILENO;

...
written = WRITEV_FOR_FATAL (fd, iov, nlist, total);

Below are some relevant parts of glibc:

void
__attribute__ ((noreturn))
__stack_chk_fail (void)
{
  __fortify_fail ("stack smashing detected");
}

void
__attribute__ ((noreturn))
__fortify_fail (msg)
     const char *msg;
{
  /* The loop is added only to keep gcc happy.  */
  while (1)
    __libc_message (2, "*** %s ***: %s terminated\n",
            msg, __libc_argv[0] ?: "<unknown>");
}


/* Abort with an error message.  */
void
__libc_message (int do_abort, const char *fmt, ...)
{
  va_list ap;
  int fd = -1;

  va_start (ap, fmt);

#ifdef FATAL_PREPARE
  FATAL_PREPARE;
#endif

  /* Open a descriptor for /dev/tty unless the user explicitly
     requests errors on standard error.  */
  const char *on_2 = __libc_secure_getenv ("LIBC_FATAL_STDERR_");
  if (on_2 == NULL || *on_2 == '\0')
    fd = open_not_cancel_2 (_PATH_TTY, O_RDWR | O_NOCTTY | O_NDELAY);

  if (fd == -1)
    fd = STDERR_FILENO;

  ...
  written = WRITEV_FOR_FATAL (fd, iov, nlist, total);