multi: avoid a theoretical 32-bit wrap

[bagder]

If Curl_uint32_tbl_count() at some future point actually can return UINT_MAX, this fixes the math to not wrap.

[Copilot]

Pull request overview

Adjusts curl_multi_get_handles() handle-count arithmetic to avoid 32-bit wrap when sizing the returned handle array.

Changes:

• Change local count type from unsigned int to size_t in curl_multi_get_handles().
• Use count + 1 with a wider type for the allocation sizing expression.

Comments suppressed due to low confidence (1)

lib/multi.c:3732

• i remains unsigned int while count is now size_t. To keep types consistent with the allocated array size and avoid any potential truncation/overflow if count can exceed UINT_MAX, consider making i a size_t as well (and adjust related asserts accordingly).

    unsigned int i = 0;
    uint32_t mid;

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

count is now size_t, but the allocation size expression sizeof(struct Curl_easy *) * (count + 1) can still overflow/wrap on 32-bit when count is large (e.g. UINT32_MAX from Curl_uint32_tbl_count() equals SIZE_MAX). Consider adding an explicit overflow check (for both (count + 1) and the multiplication) and returning NULL on overflow to avoid undersized allocations and potential out-of-bounds writes.