Allow NaN and Inf to pass through Stack floating-point push/get/isInstance

[Copilot]

is_floating_point_representable_by() rejected NaN and Inf with FloatingPointDoesntFitIntoLuaNumber — NaN due to IEEE 754 comparisons always returning false, Inf due to exceeding numeric_limits::max(). Since Lua natively supports both, these values should pass through rather than error.

Changes

• Source/LuaBridge/detail/LuaHelpers.h / Distribution/LuaBridge/LuaBridge.h: Added early-return true for std::isnan/std::isinf in is_floating_point_representable_by(). Dropped constexpr (not valid for these calls in C++17) and added #include <cmath>.

• Tests/Source/StackTests.cpp: Added FloatNaNAndInfPush, DoubleNaNAndInfPush, LongDoubleNaNAndInfPush covering push, isInstance, and get round-trip for quiet NaN, +Inf, and −Inf.

// Previously returned FloatingPointDoesntFitIntoLuaNumber error
luabridge::push(L, std::numeric_limits<float>::quiet_NaN()); // now ok
luabridge::push(L, std::numeric_limits<float>::infinity());  // now ok

// get() and isInstance() also work correctly after the fix
auto result = luabridge::get<float>(L, -1);
assert(result && std::isnan(*result));

The fix is applied uniformly via is_floating_point_representable_by(), so push, get, and isInstance all benefit for float, double, and long double.

---

⌨️ Start Copilot coding agent tasks without leaving your editor — available in VS Code, Visual Studio, JetBrains IDEs and Eclipse.

[kunitoki]

Great work !