Reading Time: < 1 minute

Points To Catch

  • As for the control flow: setjmp returns twice, and longjmp never returns.
  • When you call setjmp for the first time, to store the environment, it returns zero,
  • And then when you call longjmp, the control flow passes to return from setjmp with the value provided in the argument.
  • Use cases are generally cited as “error handling”, and “don’t use these functions”.

Note: setjmp needn’t actually be functions; it may well be a macro. longjmp is a function, though.

Here’s a little control flow example:

Example

#include <stdio.h>
#include <setjmp.h>

jmp_buf env;

void foo()
{
    longjmp(&env, 10);                      +---->----+
}                                           |         |
                                            |         |
int main()              (entry)---+         ^         V
{                                 |         |         |
    if(setjmp(&env) == 0)         | (= 0)   |         | (= 10)
    {                             |         ^         |
        foo();                    +---->----+         |
    }                                                 +---->----+
    else                                                        |
    {                                                           |
        return 0;                                               +--- (end)
    }
}

Important Notes:

  • You cannot pass 0 to longjmp. If you do, 1 is returned by setjmp.
  • You must not return from the function that called setjmp before the corresponding longjmp. In other words, longjmp must only be called above setjmp in the call stack.
  • You cannot actually store the result of setjmp. If you want to return in several different ways, you can use a switch, though:
    switch (setjmp(&env))
    {
    case 0:   // first call
    case 2:   // returned from longjmp(&env, 2)
    case 5:   // returned from longjmp(&env, 5)
    // etc.
    }
Awesome
Awesome Interesting Useful Cool Boring Sucks