g++ 3.2.2 realloc bug?

[Jonathan.Bailleul]

Am I ignorant of bug releases and fixes, a bad man-reader, or is it a
bug???

In advance, thanks for your help.


#include <iostream>
#include <string>
#include <math.h>


using namespace std;


int
main(void)
{

int i, size = 5;
int* tab;
assert(tab = (int*) calloc(size, sizeof(int)));
for (i = 0; i < size; i++) tab = i + 1;

for (i = 0, cout << endl; i < size; i++) cout << tab << " ";

size = 10;
assert(tab = (int*) realloc(tab, size));
for (i = 0, cout << endl; i < size; i++) cout << tab << " ";

return EXIT_SUCCESS;
}

//1 2 3 4 5
//1 2 3 4 16 0 139048 0 0 0


// /users/these/bailleul/These/OpenSource/Tests>g++ --version
//g++ (GCC) 3.2.2

 

[Jonathan.Bailleul]

Am I ignorant of bug releases and fixes, a bad man-reader, or is it a
bug???

Nope your program has an error.

assert(tab = (int*) realloc(tab, size));
for (i = 0, cout << endl; i < size; i++) cout << tab << " ";

realloc takes a size in bytes. You forgot to multiply size by sizeof (int).

Ok, I interpreted man at the letter without thinking. That was that...

It's also extremely bad form to put code in an assert that does something.

I don't get it. Is that for the same reason of your following remark?

Your program will fail to invoke the realloc when asserts are disabled.

Can we? My vision of the assert is to immediately interrupt the program
if given condition is not fulfilled.

 

[Ron Natalie]

Ok, I interpreted man at the letter without thinking. That was that...
I don't get it. Is that for the same reason of your following remark?

assert.h has the equivelent of

#ifdef NDEBUG
#define assert(ignore) ((void)0)
#else
#define assert(cond) ...something implementation specific...
#endif

Typically NDEBUG is defined for final (release mode) builds. This
means that all the code that you put in the parameter to assert is
not evaluated in that case.

 

[John Harrison]

Code within an assert will be removed if NDEBUG is not defined.

Sorry, if NDEBUG is defined.

john