static_assert case?

Asked Jul 04 '20 at 20:49

Active Jul 04 '20 at 21:06

Viewed 43 times

I have the following code , MSVC and clang/gcc disagree, MSVC accepts it. I presume MSFT is wrong(godbolt does not have newest compiler, but I do and Intellisense is also complaining while compiler compiles it even with permissive-).

#include <cstdint>
#include <tuple>
using namespace std;
void bla() {
  constexpr int64_t val1 = 1;
  constexpr int64_t val2 = val1 - 1;
  constexpr tuple<int64_t> tup1(val1);
  constexpr tuple<int64_t> tup2(val2);
  const double d1 = std::get<0>(tup1);
  const double d2 = std::get<0>(tup2);
  static_assert(d1 != d2);
}

note: this is mostly theoretical question, I can easily make those variables constexpr.

edited Jul 04 '20 at 21:06

asked Jul 04 '20 at 20:49

NoSenseEtAl

28,205
28
128
277

2

What do you mean by "accepts it"? Does the static assert not fire? – Nicol Bolas Jul 04 '20 at 20:52
2

I think you mean `(d1 != d2)`. Seems like msvc bug, clang/gcc say `d1` and `d2` need to be constexpr, which is true. – cigien Jul 04 '20 at 20:53
1

You should really add the errors from the compilers. – bolov Jul 04 '20 at 20:54
@bolov there is the link in question to godbolt – NoSenseEtAl Jul 04 '20 at 21:04
@NicolBolas I meant compiles it, will update the Q – NoSenseEtAl Jul 04 '20 at 21:04
@cigien yeah, messed up while reducing example – NoSenseEtAl Jul 04 '20 at 21:05
Also, there's a conformance viewer in godbolt that makes it easier to see the comparisons. – cigien Jul 04 '20 at 21:05
@NicolBolas I messed up the example when reducing it, now it should be fine – NoSenseEtAl Jul 04 '20 at 21:08
1

your question should be self sufficient. The godbolt link is nice, but we shouldn't go to another site just to see what the error is. – bolov Jul 04 '20 at 21:19
@cigen I agree it is a duplicate, thank you. It looks MSVC is just buggy(or advanced since I consider this useful). – NoSenseEtAl Jul 04 '20 at 21:29

Is clang/gcc or MSVC correct in this constexpr/static_assert case?

0 Answers0