fixed realloc(0) may alloc min unit memory on some allocator

[penneryu]

On some memory allocators, such as hardened_malloc, realloc(0) will allocate a minimum unit of memory (16 bytes)

The hardened_malloc is used on grapheneos, a common operating system on android.

h_malloc.c

static inline void *allocate_small(unsigned arena, size_t requested_size) {
    struct size_info info = get_size_info(requested_size);
    size_t size = likely(info.size) ? info.size : 16;

    struct size_class *c = &ro.size_class_metadata[arena][info.class];
    size_t slots = get_slots(info.class);
    size_t slab_size = get_slab_size(slots, size);
    ...

[saghul]

lre_realloc calls js_realloc_rt, which looks as follows:

void *js_realloc_rt(JSRuntime *rt, void *ptr, size_t size)
{
    size_t old_size;
    JSMallocState *s;

    if (!ptr) {
        if (size == 0)
            return NULL;
        return js_malloc_rt(rt, size);
    }
    if (unlikely(size == 0)) {
        js_free_rt(rt, ptr);
        return NULL;
    }
    old_size = rt->mf.js_malloc_usable_size(ptr);
    s = &rt->malloc_state;
    /* When malloc_limit is 0 (unlimited), malloc_limit - 1 will be SIZE_MAX. */
    if (s->malloc_size + size - old_size > s->malloc_limit - 1)
        return NULL;

    ptr = rt->mf.js_realloc(s->opaque, ptr, size);
    if (!ptr)
        return NULL;

    s->malloc_size += rt->mf.js_malloc_usable_size(ptr) - old_size;
    return ptr;
}

Since we already call js_free_rt if size if 0, I think your case is already handled, isn't it?

[penneryu]

Yes indeed, I missed this, I will close this PR