Thread: Problem using sscanf fgets and overflow checking

It seems the matter is a bit more complex. Even i was surprised as it had never occurred to me a need to check for arithmetic overflows. I found the following very informative, although a bit advanced page.
http://www.fefe.de/intof.html
I think you need the simple solutions at the start...

1. if input line is longer than the buffer provided - fgets will not store \n character - so just check its present in the resulting buffer
2. to check that n1 and n2 are out of range - scan them with sscanf, check the return value to be sure that both values are parsed, check the position with %n to see that the parsing stopped at \n character, then sprintf them back to the temp buffer and compare the 2 buffers to see that numbers are parsed ok
3. do not forget - you need 2*n2

To xuftugulus:
True this advanced but very useful, thank you. I even bookmarked the page
But unfortunately it doesn't help my question.

To vart
Thank you very much for your reply.

1. if input line is longer than the buffer provided - fgets will not store \n character - so just check its present in the resulting buffer

I figured that out just before I read your comment and was very glad you confirmed it and I added an extra step to check that there is nothing other than two integers. I posted the code I wrote so far below (commented) so whoever have a similar question can benefit from it and I will be glad if you can skim through it.

3. do not forget - you need 2*n2

Thank you

2. to check that n1 and n2 are out of range - scan them with sscanf, check the return value to be sure that both values are parsed, check the position with %n to see that the parsing stopped at \n character, then sprintf them back to the temp buffer and compare the 2 buffers to see that numbers are parsed ok

Unfortunately, I still can't check n1 and n2 for over/underflow. I understood the idea that I have to reconstruct the input into a string using the variables "n1" and "n2" right after executing sscanf and then use strcmp on the two strings.
So it would be something like this
sprintf(tempstr, "%ld %ld", n1, n2);
Now the problems I faced with this method is how to reconstruct the string to be exactly the same? For instance how to I deal with white space? I might be able to trim the two strings from all white spaces before comparison but even then how do I deal with inputs like?
> 00000001111 00003
or this
> +123 --3
n1 and n2 wouldn't obviously have all the zeros neither the positive and double negation signs stored in.

While reading the man page of sprintf I saw this modifier you talked about %n and though it would be useful. I still don't know how to use it though (even after an extensive google search)?
I get a Bus Error with
sprintf(tempstr, "%ld %ld %n", n1, n2, %n3); /*n3 is an int*/


As promised here is my code below:

Code:

#include <stdio.h>
#include <limits.h>
#include <inttypes.h>
#include <string.h>


#define MAX 50 /* MAX is 50 since I can assume that no legal input is more than 50 chars  excluding the \n */

long n1, n2;
int c;
char str[MAX+2];
/*  This is the string where I will store the input. 
I added 2 to MAX to leave room for the new line character as well as for the (null) character
which is added automatically after fgets execution. */

int main(void)
{	
	if (fgets(str, MAX + 2, stdin) == str) /* To check if fgets ran..successfully..*/
	{
		if (str[strlen(str)-1] == '\n')
/* I subtracted 1 from strlen to get to the character before the null which should be \n */
		{
			if (sscanf(str, "%ld %ld %1s", &n1, &n2, %c) == 2)
			{
				if(c == '\0')
				{
					if n1 and n2 is not over/underflow /* I am stuck with this part */
					{
						if (n1 > LONG_MAX - (2*n2))
							printf("overflow\n");
						if (n1 < LONG_MIN - (2*n2))
							printf("underflow\n");
						else
							printf("%d\n", n1 + (2*n2));
					}
				}
				else
				{
					printf("illegal input\n");
				}
			}
			else
			{
				printf("illegal input\n");
			}
		}
		else
		{
			printf("illegal input\n");
		}
	}
	if (fgets(str, MAX + 2, stdin) == '\0') 
/* This means that we have reached EOF so we need to check for illegal input of more than 50 char differently
 becuase there might not be a \n */
 	{
 		if(strlen(str) <= MAX)
 		{
 			/* I will call a function with the above code which starts with if n1 and n2 ... */
 		}
		else
		{
			printf("illegal input\n");
		}
 	}
 	else
 	{
		printf("illegal input\n");
 	}
 	return 0;
}

One option would be to NOT use sscanf(), but rather parse the string yourself, and use strtol() - as this allows for overflow checking (it clamps the number to the range MIN_LONG..MAX_LONG, and sets errno to "ERANGE" [I think]) - it also gives you a pointer back to the character where it stopped parsing, which you can check if it's at the end of the string or not [e.g. on a space, '\n' or '\0' character, depending on the circumstance) - you could also use the "parsed to" pointer to find the next string to parse, as strtol() will allow whitespace at the beginning of the string.

--
Mats

%n is for scanf not sprintf

Code:

{
	long n1,n2;
	int position;
	char buffer[] = "1234567890123 5678\n";
	int ret = sscanf(buffer, "%ld %ld%n", &n1,&n2,&position);

	size_t bufLen = strlen(buffer);

	if(ret != 2 || buffer[position] != '\n' || bufLen != (size_t)position+1)
	{
		/*incorrect input */
		return -1;
	}
	else
	{
		char temp[52];
		sprintf(temp, "%ld %ld\n", n1,n2);

		if(strcmp(buffer,temp))
		{
			/* wrong parse */
			return -2;
		}

	}
	return 0;
}