lib: cgen: make _bf_round_next_power_of_2 correct on 64-bit systems

[pzmarzly]

size_t can be 32-bit or 64-bit integer. We can use a conditional macro to support both.

This was raised by Claude PR reviewer in unrelated PR.

This function is currently only used to calculate _BF_LOG_MAP_SIZE (which is small static number), and img_size (which is being removed by #487), so there isn't any danger at the moment. But let's avoid a future surprise.

---

Also, on both 32-bit and 64-bit platforms, it gives incorrect result for 2**0. Should be 1, is 0. Fixing.

[github-actions]

Claude review of PR #494 (e205435)

Clean, minimal fix. The #if SIZE_MAX > 0xFFFFFFFFU guard is the correct portable way to conditionally include the 64-bit shift — SIZE_MAX is available via the existing <stdint.h> include, the comparison correctly distinguishes 32-bit vs 64-bit size_t, and column-0 preprocessor directive placement matches the project convention (e.g. src/libbpfilter/bpf.c:21). Commit message follows the component: subcomponent: description format.

Suggestions

• Guard against value == 0 — src/libbpfilter/cgen/program.c:68 — Pre-existing edge case where _bf_round_next_power_of_2(0) returns 0 via unsigned wrap-around. Current callers are safe, but a guard would make the helper defensive for future use.

Workflow run

[github-actions]

Claude: suggestion: Pre-existing: _bf_round_next_power_of_2(0) returns 0 due to unsigned wrap-around (0 - 1 wraps to SIZE_MAX, OR-shifts fill all bits, then ++value wraps back to 0). Since you're already touching this function, consider adding a guard:

static inline size_t _bf_round_next_power_of_2(size_t value)
{
    if (value == 0)
        return 1;

    value--;
    ...

Neither current call site can pass 0 (sizeof(struct bf_log) * _BF_LOG_MAP_N_ENTRIES is always positive), so this is not blocking — just a defensive improvement for future callers.

[pzmarzly]

That's true, and it is something I stumbled upon already when using this function for bf_hashset (where initial capacity was 0).

2**0 == 1, but here _bf_round_next_power_of_2(0) == 0.

I'll address this here in this PR as well, but I'll let reviewer decide if this should be merged together.

[qdeslandes]

LGTM!