Question

I have a function that counts how many times a words appears on a file. Now for some reason it gets stack smashing detected error, and i dont see the error.

Here is the code:

int contar_palabra(const char *nombre_file, const char *palabra){

/*variables locales*/
FILE *file; 
char buffer[50]; 
int bytes_leidos, contador = 0, comparar, cerrar, i;    

/*Abrimos el archivo para lectura*/
file = fopen(nombre_file, "r");

/*Verificamos que se haya abierto correctamente*/
if (file == NULL){
  printf("No se pudo abrir el archivo \n");
  perror(nombre_file);
  exit(EXIT_FAILURE);
}

/*Procedemos a contar cuantas veces aparece una palabra*/
while (!feof(file)){

  bytes_leidos = fscanf(file, "%s", buffer);

  if (bytes_leidos > 0){

    /*Hacemos la comparacion*/
    comparar = strcmp(buffer, palabra);
    if (comparar == 0)
      contador++;
  }
  else if(errno == EOF)     
        printf("Error al leer alguna palabra de %s \n", nombre_file);
  else if (bytes_leidos == EOF)
    break;
}

cerrar = fclose(file);

if(cerrar == EOF){
  printf("Error: no se pudo cerra el archivo.");
}

printf("antes de retornar contador \n");
return contador; 

}

I used valgrind trying to identify the error, and the log file gave me this:

  ==2252== Memcheck, a memory error detector
  ==2252== Copyright (C) 2002-2010, and GNU GPL'd, by Julian Seward et al.
  ==2252== Using Valgrind-3.6.1 and LibVEX; rerun with -h for copyright info
  ==2252== Command: ./pargrep viejo.txt
  ==2252== Parent PID: 1756
  ==2252== 
  ==2252== 
  ==2252== HEAP SUMMARY:
  ==2252==     in use at exit: 55 bytes in 1 blocks
  ==2252==   total heap usage: 7 allocs, 6 frees, 1,389 bytes allocated
  ==2252== 
  ==2252== 55 bytes in 1 blocks are still reachable in loss record 1 of 1
  ==2252==    at 0x4026864: malloc (vg_replace_malloc.c:236)
  ==2252==    by 0x40B878B: __libc_message (libc_fatal.c:138)
  ==2252==    by 0x413D09F: __fortify_fail (fortify_fail.c:32)
  ==2252==    by 0x413D049: __stack_chk_fail (stack_chk_fail.c:29)
  ==2252==    by 0x8049142: contar_palabra (in /home/alessandro/OS/Proyecto2/prueba1.0/pargrep)
  ==2252==    by 0x80489D4: main (in /home/alessandro/OS/Proyecto2/prueba1.0/pargrep)
  ==2252== 
  ==2252== LEAK SUMMARY:
  ==2252==    definitely lost: 0 bytes in 0 blocks
  ==2252==    indirectly lost: 0 bytes in 0 blocks
  ==2252==      possibly lost: 0 bytes in 0 blocks
  ==2252==    still reachable: 55 bytes in 1 blocks
  ==2252==         suppressed: 0 bytes in 0 blocks
  ==2252== 
  ==2252== For counts of detected and suppressed errors, rerun with: -v
  ==2252== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 15 from 8)

Whats strange is that it prints a message before returning.
I really dont see the error, appreciate the help.
thanks in advance

Answer 1

Using plain %s as a format specifier to fscanf is very dangerous as you have no protection against writing beyond the end of the buffer which is only 50 bytes in your case. Consider providing a width modifier to the format specifier, although this will be handling an accurate count when the width limit is reached more complex.

You may find that reading character by character (fgetc) or reading fixed buffers (fread) and detecting the word separators by hand yields simpler code.