Why doesn't this program crash on a Mac ?

It should, as I'm using setrlimit() to limit memory usage...
On a GNU/Linux system it does crash.



#include <sys/types.h>
#include <sys/time.h>
#include <sys/resource.h>
#include <stdio.h>
#include <stdlib.h>

int main(void)
{
struct rlimit lim;
int i;

if(getrlimit(RLIMIT_AS, &lim) == -1) return 1;
printf("%d\n", lim.rlim_cur);

lim.rlim_cur = lim.rlim_max = 100 * 1024;
if(setrlimit(RLIMIT_AS, &lim) == -1) return 2;

if(getrlimit(RLIMIT_AS, &lim) == -1) return 1;
printf("%d\n", lim.rlim_cur);

char *v = malloc(100000000);
for (i=1;i<1000000;i++)
v[i]=(v[i-1]+3)%200; // pour pas que le malloc se fasse optimiser,
on sait jamais
printf("%d\n",42);

}


Thanks !

--
Thomas Deniau

In article ,
thomas.RemoveThis@deniau.org (Thomas Deniau) wrote:

> It should, as I'm using setrlimit() to limit memory usage...
> On a GNU/Linux system it does crash.

I think any time you're making the statement that a piece of code
"should" crash you're into kind of weird space. I'd argue that
abstractly any crash is to be avoided if possible and point out that
absolutely nothing I can find in the documentation for setrlimit
requires a process exceeding its limits to crash (or respond in any
specific way).

The short answer is that it didn't crash because malloc (part of the C
standard library, not system call as setrlimit is) was - at the time it
was invoked - able to satisfy the request. Why it was able to do that is
an implementation detail; presumably it already had a pool reserved.
It's not guaranteed to succeed or to not succeed, which is why it's
important to check that malloc returned something other than NULL.