[PATCH] D26903: [libcxx] Add <variant> tests (but not implementation)

[Casey Carter via cfe-commits]

CaseyCarter added inline comments.


================
Comment at: test/std/utilities/variant/variant.hash/hash.pass.cpp:34
+
+void test_hash_variant()
+{
----------------
This is non-portable. The fact that both your implementation and mine pass this test suggests they both have poor QoI: There's strong opinion in the community that a variant's index should contribute to its hash value, so that e.g. `variant<int, int>{in_place_index<0>, 42}` and `variant<int, int>{in_place_index<1>, 42}` have differing hash values.


================
Comment at: test/std/utilities/variant/variant.helpers/variant_size.pass.cpp:41
+{
+    test<std::variant<>, 0>();
+    test<std::variant<void*>, 1>();
----------------
STL_MSFT wrote:
> Hmm, is this cromulent now that empty variants are forbidden?
It's not forbidden to *name* `variant<>`; it is forbidden to *instantiate* it. The pertinent specialization of `variant_size`:
```
template <class... Types>
struct variant_size<variant<Types...>>
  : integral_constant<size_t, sizeof...(Types)> { };
```
is implicitly specified to not instantiate `variant<Types...>`.


https://reviews.llvm.org/D26903