Profiler error at fscanf caused by fscanf

Question

I am writing a function in C, that is supposed to input information about students and their grades from a file. However, upon running the programme at times the tests show profiler error at fscanf by fscanf, even though the output is as expected.

/*structure for grade*/
struct Grade {
  char subject[51];
  int grade;
};
/*structure for student*/
struct Student {
  char name[21], surname[21];
  int no_grades;
  struct Grade grades[100];
};

int input_students(struct Student *students, int n) {
  int i = 0, no, j, first;
  char name[21], surname[21], subject[51];
  FILE *database = fopen("input.txt", "r");
  if (database == NULL) {
    printf("Error while opening input.txt");
    return 0;
  }
  while (/*i < n && */fscanf(database, "%20s %20s %50s %d\n", name, surname, subject,
                         &no) == 4) {
    first = 1;
    for (j = 0; j < i; j++) {
      if (strcmp(students[j].name, name) == 0 &&
          strcmp(students[j].surname, surname) == 0) {
        first = 0;
        if (students[j].no_grades >= 100) { /*limit for number of grades per student is 100*/
          break;
        }
        strcpy(students[j].grades[students[j].no_grades].subject,
               predmet);
        students[j].grades[students[j].no_grades].grade = no;
        students[j].no_grades++;
        break;
      }
    }
    if (first && i < n) {
      strcpy(students[i].name, name);
      strcpy(students[i].surname, surname);
      students[i].no_grades = 0;
      strcpy(students[i].grades[students[i].no_grades].subject, subject);
      students[i].grades[students[i].no_grades].grade = no;
      students[i].no_grades++;
      i++;
    }
  }
  /*while (fscanf(database, "%20s %20s %50s %d\n", name, surname, subject,
                         &no) == 4) {
    for (j = 0; j < i; j++) {
      if (strcmp(students[j].name, name) == 0 &&
          strcmp(students[j].surname, surname) == 0) {
        if (students[j].no_grades >= 100) {
          break;
        }
        strcpy(students[j].grades[students[j].no_grades].subject,
               subject);
        students[j].grades[students[j].no_grades].grade = no;
        students[j].no_grades++;
        break;
      }
    }
  }*/

  fclose(database);
  return i;
}

/update, comments changing the code to only one loop, same thing happens/ Return value is the number of students whose information was read from the file, while i < n.

The profiler message indicates a mistake in the line with fscanf.

==3928== Invalid write of size 1
==3928== at 0x37154582C7: _IO_vfscanf (in /lib64/libc-2.12.so)
==3928== by 0x371546465A: __isoc99_fscanf (in /lib64/libc-2.12.so)
==3928== by 0x400AD5: input_students (main.c:22)
==3928== by 0x401394: main (main.c:174)
==3928== Address 0x7feff2e00 expected vs actual:
==3928== Expected: stack array "name" of size 21 in frame 2 back from here
==3928== Actual: stack array "surname" of size 21 in frame 2 back from here
==3928== Actual: is 32 before Expected

I can't seem to be able to find what is the cause of this. I have double-checked that each string ends with '\0', I have run the programme through debugger and it stops where it should, I have not seen it going over and working with uninitalised values, and the values it takes are as expected.

I have translated the name of the function and variables as it wasn't originally in English, therefore the problem can not be in the name of the function overlapping with the name of a library function but I will change it here.

Test code. It's an autotest.

int i, j, no_students;
struct Student students[10];
FILE* database = fopen("input.txt", "w");
fputs("Pero Peric Osnove_racunarstva 8", database);
fputc(10, database);
fputs("Suljo Suljic Osnove_racunarstva 9", database);
fputc(10, database);
fputs("Pero Peric Inzenjerska_matematika_1 6", database);
fclose(database);

no_students=input_students(students, 10);
printf("%d\n", no_students);
for (i=0; i<no_students; i++) {
    printf("%s %s ", students[i].name, studenti[i].surname);
    for (j=0; j<students[i].no_grades; j++)
        printf("%s %d ", students[i].grades[j].subject, students[i].grades[j].grade);
    printf("\n");
}

Output for this test.

2
Pero Peric Osnove_racunarstva 8 Inzenjerska_matematika_1 6 
Suljo Suljic Osnove_racunarstva 9

When the test code is run within the main.c, it does not show any compiler errors, nor does debugger show segmentation error.

Example for input.txt, created by this autotest.

Pero Peric Osnove_racunarstva 8
Suljo Suljic Osnove_racunarstva 9
Pero Peric Inzenjerska_matematika_1 6

This would be the information about the profiler, bare in mind I am still only a student

==17000== exp-sgcheck, a stack and global array overrun detector
==17000== NOTE: This is an Experimental-Class Valgrind Tool
==17000== Copyright (C) 2003-2012, and GNU GPL'd, by OpenWorks Ltd et al.
==17000== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
==17000== Command: outputP9YqwH

Answer 1

Ok, here is what I think is happening:

no_students=input_students(students, 10);

so you are expecting 10 students, yet you have 4. So simplicity reason we will have file pointer database pint to the row 0 (instead of address of 0). So you have this:

row  name   surname subject                  grade_cnt
0    Pero   Peric   Osnove_racunarstva       8\n
1    Suljo  Suljic  Osnove_racunarstva       9\n
3    Pero   Peric   Inzenjerska_matematika_1 6<END OF FILE>

NOTE: that on the end of row 3 might be end of file.

so you first loop goes and reads row 0, 1 ... 3 and tries to read row 4. There is no row 4, so it fails. But what might be happening is that database is pointing to row 4 when first loop is done skipping EOF and reading beyond end of file buffer.

There are two ways of testing for this:

1. input empty line after 4th line
2. close and reopen file between while() loops.

The other reason that you might be skipping beyond file pointer is this:

fscanf(database, "%20s %20s %50s %d\n", name, surname, subject, &no) == 4)

notice \n after %d what happens if that \n is gone? fscanf() is looking for '\n'.

Try adding empty line to the end of input file.

Or removing '\n'.

fscanf(database, "%20s %20s %50s %d", name, surname, subject, &no) == 4)

Another reason behind this is numbers, %20s:

fscanf(database, "%s %s %s %d", name, surname, subject, &no) == 4)

Please anyone let me know in comments if I am mistaking.

due to the fact that fscanf reads 20 characters for name.

EDIT: Now I am sure that I am right: Read this. scanf and fscanf should be used with caution, in best case should be avoided.

Answer 2

Solution was rather simple, instead of using four variables, name, surname, subject and no. It is way simpler to declare a new variable

struct Student s;

and use it for input.

Not only does it solve the problem with fscanf(), but the rest of the code is simpler and easier to read.

As for the reason why it works, that would need further research.

Here is the revisited code

int input_students(struct Student *students, int n) {
  int i = 0, no, j, first;
  char name[21], surname[21], subject[51];
  FILE *database = fopen("input.txt", "r");
  if (database == NULL) {
    printf("Error while opening input.txt");
    return 0;
  }
  while (fscanf(database, "%20s %20s %50s %d\n", s.name, s.surname, s.grades[0].subject,
                         &s.grades[0].grade) == 4) {
    first = 1;
    for (j = 0; j < i; j++) {
      if (strcmp(students[j].name, name) == 0 &&
          strcmp(students[j].surname, surname) == 0) {
        first = 0;
        if (students[j].no_grades >= 100) { /*limit for number of grades per student is 100*/
          break;
        }
        students[j].grades[students[j].no_grades] = s.grades[0];
        students[j].no_grades++;
        break;
      }
    }
    if (first && i < n) {
      students[i] = s;
      i++;
    }

  fclose(database);
  return i;
}