[libcxx-commits] [libcxx] [libc++] Fix search_n with an invalid count (PR #223747)

[A. Jiang via libcxx-commits]

================
@@ -79,14 +80,19 @@ __find_longest_suffix(_RAIter __first, _RAIter __last, const _ValueT& __value, _
 template <class _AlgPolicy, class _Pred, class _Iter, class _SizeT, class _Type, class _Proj, class _DiffT>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 std::pair<_Iter, _Iter> __search_n_random_access_impl(
     _Iter __first, _SizeT __count_in, const _Type& __value, _Pred& __pred, _Proj& __proj, _DiffT __size) {
-  auto __last  = __first + __size;
-  auto __count = static_cast<_DiffT>(__count_in);
-
-  if (__count == 0)
+  // Trivially match if count is non-positive. This must be checked before we perform arithmetic on the count.
+  if (__count_in <= 0)
     return std::make_pair(__first, __first);
-  if (__size < __count)
+
+  auto __last = __first + __size;
+
+  // If count is larger than the sequence size, no match is possible. Compare as unsigned to prevent
+  // truncation between _DiffT and _SizeT.
+  if (std::__to_unsigned_like(__size) < std::__to_unsigned_like(__count_in))
----------------
frederick-vs-ja wrote:

IIUC we've obtained an integer per the `__convert_to_integral` call in `search_n`, so it should be fine to use `__to_unsigned_like` here.

Also, `_DiffT` is the difference type of the iterator, so `__to_unsigned_like` is also fine for it.

https://github.com/llvm/llvm-project/pull/223747