Compile errors on apple clang about cppassist HOT 6 CLOSED

I just had a look at both code fragments.
For both errors I doesn't see any flaw in the source code; especially I don't understand the error message of the Allocator::value_type and their context.
The first error may be a misinterpretation of AppleClang regarding overloaded member functions for both const and non-const objects or a more restrictive pointer arithmetic.

I'll test on a macOS.

from cppassist.

Regarding the second error:
using value_type = value<T, SIMD_SIZE>;
using allocator = cppassist::aligned_allocator<T, simd_size * sizeof(T)>;
using iterator = typename std::vector<value_type, allocator>::iterator;

The value type for the vector is value_type (which is value<T, SIMD_SIZE>).
The value type for the allocator is T.

T != value<T, SIMD_SIZE), hence the error message. It can be solved syntactically by
using allocator = cppassist::aligned_allocator<value_type, simd_size * sizeof(T)>;

however I'm not sure if this is correct semantically. Should at least read sizeof(value_type), I guess?

from cppassist.

I don't understand the first error at all. The function that is called in non-const, so the used version of data() should also be the non-const version. But it seems like the compiler chooses the const version of data() here. I don't know what to do here except circumventing this possible compiler bug by using a const_cast.

from cppassist.

I just want to add that clang 3.8 (the official clang, not apple clang) has no problem with the constness.

I will provide a const_cast solution.

from cppassist.

Somehow, the first error vanished while fixing the second one on apple clang 8.0.

from cppassist.

true, it's gone. So with luck this was just a very strange follow-up bug.

from cppassist.