Question

I’m studying UNIX programming and was experimenting with read/write system calls.
I have a file with a pair of integer:

4 5

and I wrote this code to read the numbers:

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

typedef struct prova {
    int first;
    int second;
} prova_t;

int main(void) {
    int fd;
prova_t origin;
prova_t result;
ssize_t bytes_read;
size_t nbytes;

fd = open("file.bin", O_WRONLY | O_CREAT);
origin.first = 24;
origin.second = 3;
write(fd, &origin, sizeof(prova_t));
close(fd);


fd = open("file.bin", O_RDONLY);
nbytes = sizeof(prova_t);
/* 1.BAD */
bytes_read = read(fd, &result, nbytes);
write(STDOUT_FILENO, &(result.first), sizeof(int));
write(STDOUT_FILENO, &(result.second), sizeof(int));
close(fd);

    /* 2.GOOD */
    nbytes = sizeof(int);
    bytes_read = read(fd, &(result.first), nbytes);
    write(STDOUT_FILENO, &(result.first), bytes_read);
    bytes_read = read(fd, &(result.second), nbytes);
    write(STDOUT_FILENO, &(result.second), bytes_read);

    return 0;
}

In my first attempt I tried to read the whole struct from file and write its members to stdout. In this way, along with the numbers, I get some weird characters

4 5
E�^�

In my second attempt I read the numbers one by one and there were no problems in the output.

Is there any way to read and write the struct using the first method?

Edit: I updated the code to reflect suggestion from other users but still getting strange characters instead of numbers

Answer 1

First, let’s do a hex dump to see what is really stored in the file.

hexdump -C b.txt or od -t x2 -t c b.txt are two examples (od is for octal dump, more common, less pretty output in my opinion)

00000000  34 20 35 0a                                       |4 5.|
00000004

That’s is what the file looks like if it was a created as an ASCII text file (such as using a text editor like vi). You can use man ascii to double check the hexadecimal values.

Now if you had a binary file that only contains two 8-bit bytes, in the system’s native byte ordering (e.g. little-endian for x86, big endian for MIPS, PA-RISC, 680×0) then the hexdump would look like:

00000000  04  05                                            |..|
00000004

Here is the code to both create (open & write) a binary file, and read it back.

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <stdint.h>     /* uint32_t */
#include <sys/stat.h>
#include <fcntl.h>
#include <string.h>
#include <errno.h>

/* User has read & write perms, group and others have read permission */ 
const mode_t mode = S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH;

typedef struct prova {
   uint32_t first;
   uint32_t second;
} prova_t;

#define FILENAME "file.b"

/* 'Safe' write */
int safewrite( int fd, const void *p, size_t want) {
   int ret;

   errno = 0;
   while (want) {
      ret = write(fd, (uint8_t *)p, want);
      if (ret <= 0) {
         if (errno != EINTR && errno != EAGAIN) {
            return -1;
         }
         errno = 0;
         continue;
      }
      want -= ret;
      p = (uint8_t*) p + ret;
   }
   return 0;
}

int saferead(int fd, const void *p, size_t want) {
   int ret;

   errno = 0;
   while (want) {
      ret = read(fd, (uint8_t*)p, want);
      if( ret == 0 )
         return -1;  /* EOF */
      if (ret <= 0) {
         if( errno != EINTR && errno != EAGAIN ) {
            return -1;
         }
         errno = 0;
         continue;
      }
      want -= ret;
      p = (uint8_t*) p + ret;
   }
   return 0;
}


int main(int argc, char **argv) {
   int fd;
   prova_t result;
   size_t nbytes;

   /* Create file */
   fd = creat(FILENAME, mode);
   if (fd < 0) {
      fprintf(stderr, "Unable to open " FILENAME ": %s\n",
            strerror(errno));
      exit(EXIT_FAILURE);
   }
   nbytes = sizeof(prova_t);

   result.first = 4;
   result.second = 5;

   if (0 != safewrite(fd, &result, nbytes)) {
      fprintf(stderr, "Unable to write to " FILENAME ": %s\n",
            strerror(errno));
      exit(EXIT_FAILURE);
   }

   close(fd);
   fd = -1;

   /* Reopen and read from binary file */
   fd = open(FILENAME, O_RDONLY);
   nbytes = sizeof(prova_t);

   if (0 != saferead(fd, &result, nbytes)) {
      fprintf(stderr, "Unable to read file \"" FILENAME "\": %s\n",
            strerror(errno));
      exit(EXIT_FAILURE);
   }
   close(fd);

   printf( "Read: %d %d (%#.02x%.02x)\n",
         result.first, result.second,
         result.first, result.second);

   return EXIT_SUCCESS;
}

Now the data file contents look like:

00000000  04 00 00 00 05 00 00 00                           |........|
00000008

Because the integers were specified as 32-bit integers (32-bits / 8 bits per byte = 4 bytes). I’m using a 64-bit system (little endian, x86), so I wanted to be explicit so the your results should match, assuming little-endian.