runtime error SIGFPE . Why?

Question

I am trying to solve a problem that is I have to found a probability of a correct sentence. We assume that only one order of a sentence is correct. lets a sentence " eat you rice ". We can rearrange this sentence by six types. Such as: "eat you rice", "eat rice you", "you eat rice", "you rice eat", "rice you eat", "rice eat you"

So the result is: 1/6 .

We also assume that each sentence consists with some words not exceed 10 words and the number of alphabet of each word is less than 21.

I always get runtime error SIGFPE when i submitted (local programming practice website) it.

My code is given below:

#include<stdio.h>
#include<string.h>
int t[11];
char line[11][21];

int set_line()
{
    int j=1;
    char tmp[21],m;
    scanf("%s",tmp);
    strcpy(line[0],tmp);

    while(getchar() != '\n' )
    {
        scanf("%s",tmp);
        strcpy(line[j],tmp);
        j++;
    }


    return j;
}

void ib()
{
    int i;
    t[0]=1;
    t[1]=1;
    t[2]=2;
    for(i=3;i<11;i++)
    {
        t[i]=t[i-1] * i;
    }

}


int main()
{
    int t_line,word_cnt,n=1,i,j,tmp,p=1;
    scanf("%d",&t_line);
    ib();

    while(t_line--)
    {

        word_cnt=set_line();
        tmp=word_cnt;
        for(i=0;i<word_cnt;i++)
        {
            if((strcmp(line[i],"01"))==0)
                continue;
            for(j=i+1;j<word_cnt;j++)
            {
                if((strcmp(line[i],line[j]))==0)
                {
                    n++;
                    strcpy(line[j],"01");
                }

            }
            if(n>1)
            {
                p=(t[tmp] / (t[tmp-n]*t[n])) * p;
                tmp=tmp-n;
            }
            n=1;
        }
        p=t[tmp]*p;
        printf("1/%d\n",p);
        p=1;
    }
    return 0;
}

Answer 1

My guess about the source of the SIGFPE was wrong, at least on the Mac where I tested. The crash occurred in a call to strcpy() inside set_line() .

Error check your inputs to behave sanely when there is no (valid) input. I got the crash to start with, but resolved it by simply doing the error checking:

int set_line(void)
{
    int j = 1;
    char tmp[21];
    if (scanf("%s", tmp) != 1)
        return 0;
    strcpy(line[0], tmp);

    while (getchar() != '\n')
    {
        if (scanf("%s", tmp) != 1)
            break;
        strcpy(line[j], tmp);
        j++;
    }

    return j;
}

and in main() :

    if (scanf("%d", &t_line) != 1)
    {
        printf("Invalid number in data\n");
        return 1;
    }

The output I got from the data file:

4
monolexis
parsimony rules
rice is nice
cats and dogs too

was:

1/1
1/2
1/6
1/24

This looks plausible as the first four factorials (this is a convoluted way of calculating them).

---

I'm not keen on the input code as amended. The first stage of improvement would lose tmp and the strcpy() operations (and also protect against over-long words — but not against too many words in a line):

int set_line(void)
{
    int j = 0;

    if (scanf("%20s", line[j++]) != 1)
        return 0;

    while (getchar() != '\n')
    {
        if (scanf("%20s", line[j]) != 1)
            break;
        j++;
    }

    return j;
}

But I don't like 'count plus data' programming, especially in a case like this where you don't need to know the count to store all the incoming data. Let computers do the counting — they're good at numbers. All you're doing with the count is reading one line at a time and then splitting it into words until you've reached the counted number of lines. It would be better to structure the code to simply read one line at a time and process that line, up until there isn't another line to read. The set_line() function gets given a string — a line of data — to process, and the main() code simply loops while there's another line to read. Granted, using %n to step through the input data is an advanced use of sscanf() , but it is also an extremely useful one in circumstances like this.

The revised code is protected against both over-long words and against too many words in a line, and looks like:

int set_line(char *data)
{
    int j;
    int offset = 0;
    int eow;    // end of word

    for (j = 0; j < 11 && sscanf(data + offset, "%20s%n", line[j], &eow) == 1; j++)
        offset += eow;

    return j;
}

…


int main(void)
{
    char data[11*21];
    int word_cnt, n = 1, i, j, tmp, p = 1;
    ib();

    while (fgets(data, sizeof(data), stdin) != 0)
    {
        word_cnt = set_line(data);
        tmp = word_cnt;
        …

For the same input file, this now produces:

1/1
1/1
1/2
1/6
1/24

because the count is now counted as a single word line of input.