// RUN: %clang_cc1 -std=c++1z -verify=ref,both %s -fcxx-exceptions -triple=x86_64-linux-gnu // RUN: %clang_cc1 -std=c++1z -verify=expected,both %s -fcxx-exceptions -triple=x86_64-linux-gnu -fexperimental-new-constant-interpreter // both-no-diagnostics namespace EvalOrder { template struct lvalue { T t; constexpr T &get() { return t; } }; struct UserDefined { int n = 0; constexpr UserDefined &operator=(const UserDefined&) { return *this; } constexpr UserDefined &operator+=(const UserDefined&) { return *this; } constexpr void operator<<(const UserDefined&) const {} constexpr void operator>>(const UserDefined&) const {} constexpr void operator+(const UserDefined&) const {} constexpr void operator[](int) const {} }; constexpr UserDefined ud; struct NonMember {}; constexpr void operator+=(NonMember, NonMember) {} constexpr void operator<<(NonMember, NonMember) {} constexpr void operator>>(NonMember, NonMember) {} constexpr void operator+(NonMember, NonMember) {} constexpr NonMember nm; constexpr void f(...) {} // Helper to ensure that 'a' is evaluated before 'b'. struct seq_checker { bool done_a = false; bool done_b = false; template constexpr T &&a(T &&v) { done_a = true; return (T &&)v; } template constexpr T &&b(T &&v) { if (!done_a) throw "wrong"; done_b = true; return (T &&)v; } constexpr bool ok() { return done_a && done_b; } }; // SEQ(expr), where part of the expression is tagged A(...) and part is // tagged B(...), checks that A is evaluated before B. #define A sc.a #define B sc.b #define SEQ(...) static_assert([](seq_checker sc) { void(__VA_ARGS__); return sc.ok(); }({})) // Longstanding sequencing rules. SEQ((A(1), B(2))); SEQ((A(true) ? B(2) : throw "huh?")); SEQ((A(false) ? throw "huh?" : B(2))); SEQ(A(true) && B(true)); SEQ(A(false) || B(true)); // From P0145R3: // Rules 1 and 2 have no effect ('b' is not an expression). // Rule 3: a->*b SEQ(A(ud).*B(&UserDefined::n)); SEQ(A(&ud)->*B(&UserDefined::n)); // Rule 4: a(b1, b2, b3) SEQ(A(f)(B(1), B(2), B(3))); // Rule 5: b = a, b @= a SEQ(B(lvalue().get()) = A(0)); SEQ(B(lvalue().get()) = A(ud)); SEQ(B(lvalue().get()) += A(0)); SEQ(B(lvalue().get()) += A(ud)); SEQ(B(lvalue().get()) += A(nm)); // Rule 6: a[b] constexpr int arr[3] = {}; SEQ(A(arr)[B(0)]); SEQ(A(+arr)[B(0)]); SEQ(A(0)[B(arr)]); SEQ(A(0)[B(+arr)]); SEQ(A(ud)[B(0)]); // Rule 7: a << b SEQ(A(1) << B(2)); SEQ(A(ud) << B(ud)); SEQ(A(nm) << B(nm)); // Rule 8: a >> b SEQ(A(1) >> B(2)); SEQ(A(ud) >> B(ud)); SEQ(A(nm) >> B(nm)); // No particular order of evaluation is specified in other cases, but we in // practice evaluate left-to-right. // FIXME: Technically we're expected to check for undefined behavior due to // unsequenced read and modification and treat it as non-constant due to UB. SEQ(A(1) + B(2)); SEQ(A(ud) + B(ud)); SEQ(A(nm) + B(nm)); SEQ(f(A(1), B(2))); #undef SEQ #undef A #undef B }