Even though std::float64_t is invalid in C, we can say "the type of 0.0f64" or just such a literal in both C and C++ (when it's supported).

As mentioned in CWG2836, when long double, double, and std::float64_t/_Float64 have the same format, the following example would be valid in C but invalid in C++.

auto x = 0.0f64;
auto y = 0.0l + 0.0f64;
auto p = true ? &x : &y;

Given C++26 has been rebased onto C23, it seems OK to have such an example in Annex C.