#define _GNU_SOURCE #include #include #include struct foo { uint64_t a; uint64_t b; }; struct bounded_flex_t { uint64_t count; // NOTE: try to move this member somewhere else and you will get: // counted_by.c:13:20: error: flexible array member not at end of struct struct foo flex_array[] __attribute__((counted_by(count))); }; void free_flex(struct bounded_flex_t **flex) { struct bounded_flex_t *ptr = *flex; printf("Freeing a flexible array of %ld members.\n", ptr->count); free(*flex); } __attribute__((noreturn)) void die(const char *const str) { fprintf(stderr, str); exit(EXIT_FAILURE); } int main(int argc, char *argv[]) { uint64_t len = 1; if (argc == 2) { len = (uint64_t) strtol(argv[1], NULL, 10); if (!len) { die("error: bad integer value!\n"); } } else if (argc > 2) { die("error: only pass one argument maximum!\n"); } // That's a mouthful :P struct bounded_flex_t *flex __attribute__((cleanup(free_flex))) = malloc(sizeof(struct bounded_flex_t) + (len * sizeof(struct foo))); if (!flex) { die("error: could not allocate memory for it"); } // NOTE: if this is not the first thing that we do, and we use 'len' for the // for loop below, then we will get a runtime error from the array bound // checker. Hence, pretty much like in Rust :) flex->count = len; uint64_t accumulator = 0; for (uint64_t i = 0; i < flex->count; i++) { flex->flex_array[i] = (struct foo){ .a = accumulator, .b = accumulator + 1}; accumulator++; } for (uint64_t i = 0; i < flex->count; i++) { printf("%ld: .a = %ld, .b = %ld\n", i, flex->flex_array[i].a, flex->flex_array[i].b); } }