-2
double fun(int i)
{
 volatile double d[1] = {3.14};
 volatile long int a[2];
 a[i] = 1073741824;
 return d[0];
}

fun(0) ➙ 3.14

fun(1) ➙ 3.14

fun(2) ➙ 3.1399998664856

fun(3) ➙ 2.00000061035156

fun(4) ➙ 3.14, then segmentation fault

can someone explain to me whats going on in this example and why segfault dont show up when calling func(2)? and why return value not always 3.14?

user4464936
  • 153
  • 1
  • 12

1 Answers1

5

You declare an array of 2 entries:

volatile long int a[2];

You then access this array with an index:

a[i] = 1073741824;

Since array-indexing in C starts at 0, any index larger than 1 would yield undefined behavior.

In your program, you are calling the function with indexes i=2, i=3 and i=4.

Each one of these function calls independently yields undefined behavior.

Undefined behavior means that anything can happen.

The fact that you're experiencing a segmentation fault only upon fun(4) is a mere coincidence.

Supplemental:

Disclaimer: the following analysis does not imply that the same behavior is to be expected on every platform, or even on every execution; it is merely in order to explain the "odd" return-values observed.

  • The 4-byte hexadecimal value of 1073741824 is 0x40000000
  • The 8-byte hexadecimal value of 3.14 is 0x40091EB851EB851F (in compliance with IEEE 754)
  • By writing 0x40000000 into a[2], you have overridden the 4 least significant bytes in the 8-byte value of d[0] (without causing a segmentation fault), thus changing it from 0x40091EB851EB851F to 0x40091EB840000000, which stands for the (IEEE 754) floating-point value of 3.1399998664856
  • By writing 0x40000000 into a[3], you have overridden the 4 most significant bytes in the 8-byte value of d[0] (without causing a segmentation fault), thus changing it from 0x40091EB851EB851F to 0x4000000051EB851F, which stands for the (IEEE 754) floating-point value of 2.0000006103516
barak manos
  • 29,648
  • 10
  • 62
  • 114
  • @user4464936: See supplemental. – barak manos Feb 02 '16 at 09:49
  • Why did you not search for dups? This 'array bounds exceeded but no segfault' is posted every other day:( – Martin James Feb 02 '16 at 10:11
  • @MartinJames: Dunno... At first, I just wrote a comment. Then, some other user suggested to make it an answer. Then, I noticed I could explain to OP the "odd" return-values that he or she were observing (with a **big disclaimer** that it's not there in order to explain the general undefined behaviour). I suppose that posting an answer to a duplicate question couldn't do much harm. If it's a complete duplicate, then crown-wisdom will ultimately get it closed. If it's partially duplicate (or not at all), then it will eventually remain as a legitimate post. – barak manos Feb 02 '16 at 10:26