Testing for when malloc returns NULL

Question

I'm trying to replicate the strdup function in C. It's part of an exercise for school. I'd like to unit test it, including the case where malloc returns NULL and sets errno to ENOMEM.

I'm on OSX 10.8.

I've tried limiting the stacksize, then datasize, realizing malloc allocates on the heap:

limit stacksize 0
limit datasize 0

Confirmation that limiting should have worked:

my-host% limit
cputime         unlimited
filesize        unlimited
datasize        0kB
stacksize       0kB
coredumpsize    0kB
addressspace    unlimited
memorylocked    unlimited
maxproc         709
descriptors     256

However, even with a limit of 0kB for the stack size, I am able to run programs just fine. So I'm thinking there might be a restriction on my host on the minimum stack size.

How else could I test the case where malloc returns NULL?

The issue isn't with stack size, it's the heap. Maybe addressspace sets a heap limit, not sure. — Charlie Burns, Nov 19 '13 at 17:24
Allocate big chunks until malloc fails, then call your strdup. — Charlie Burns, Nov 19 '13 at 17:27
The stack size does not affect how much memory you allocate on the heap. — Jonathan Leffler, Nov 19 '13 at 17:28
To clear up the ***[stack/heap](http://stackoverflow.com/questions/79923/what-and-where-are-the-stack-and-heap)*** confusion (for those of ***us*** needing it) — ryyker, Nov 19 '13 at 17:32
@JonathanLeffler: I realized that, which is why I had tried limiting the datasize. — conradkleinespel, Nov 19 '13 at 17:33
Not quite a duplicate, but the best approach IMO is given here: http://stackoverflow.com/questions/14168388/how-to-replace-default-malloc-by-code — Rob Napier, Nov 19 '13 at 17:41
@KerrekSB: I've tried that, using #define and #undef. I've also thought about redefining malloc, but that would make every function that needs regular malloc bug, no? — conradkleinespel, Nov 19 '13 at 17:42
@conradk: Well, the point of a **unit** test is that you don't have any other functions... — Kerrek SB, Nov 19 '13 at 17:48

score 4 · Accepted Answer · answered Nov 19 '13 at 17:32

I suggest you fake the memory allocation failure. For example, I needed to test memory failure during allocation of a 3D array (lots of memory to clean up), so I used a framework where the function xmalloc() was called to get memory, and the test program controlled when xmalloc() would fail.

static int fail_after = 0;
static int num_allocs = 0;

static void *xmalloc(size_t size)
{
    if (fail_after > 0 && num_allocs++ >= fail_after)
    {
        fputs("Out of memory\n", stdout);
        return 0;
    }
    return malloc(size);
}

...code using xmalloc...

int main(int argc, char **argv)
{
    int no1 = 5;
    int fail_limit = 33;

    if (argc == 2)
        fail_limit = atoi(argv[1]);

    for (fail_after = 0; fail_after < fail_limit; fail_after++)
    {
        printf("Fail after: %d\n", fail_after);
        num_allocs = 0;
        test_allocation(no1);
    }

    return 0;
}

Write your strdup() surrogate to use xmalloc() instead of malloc(). Test it more or less as shown. You can get fancier with the 'out of memory' requirements; this is rudimentary but effective.

score 0 · Answer 2 · answered Dec 20 '13 at 13:34

0

You can test this with the mock object support of the cmocka unit testing framework.

answered Dec 20 '13 at 13:34

asn

798
9
17

Testing for when malloc returns NULL

2 Answers2