Support [[msvc::no_unique_address]] with cl.

[lukester1975]

Hi

Very nice library!

Was wondering if you would consider supporting [[msvc::no_unique_address]] with cl? [[no_unique_address]] is silently ignored.

Also, clang-cl noisily complains about both: https://reviews.llvm.org/D110485

Something like this works for me - though maybe making it configurable is preferable, to avoid ABI breakage?

Regards

Luke.

[codeinred]

Thank you so much for your PR! On line 11 you don't provide a definition for TUPLET_NO_UNIQUE_ADDRESS - is this the issue you were referring to with clang complaining about it on windows?

[lukester1975]

Yeah sadly there are two cases:

• cl silently ignores [[no_unique_address]] but supports their own [[msvc::no_unique_address]]. It seems like the next breaking ABI version will enable real [[no_unique_address]] support, from what I've read.
• clang-cl nosily ignores both (with "unknown attribute" warnings). It appears they will be adding support for the msvc specific attribute "Now that Microsoft has decided to define [[msvc::no_unique_address]] with a particular ABI, it would be reasonable for us to support that spelling" (from the link above), but it's not there yet!

I've reworked the PR to use __has_cpp_attribute so as and when support is added it should just be correct. Note that __has_cpp_attribute(msvc::no_unique_address) itself is currently wrong hence the additional old-style check for (cl && !clang)...

[lukester1975]

So that commit was just to quieten the warnings - I didn't want to use -Wno-missing-attributes (too coarse) and clang-cl doesn't support the pragma to disable only the [[no_unique_address]] warning.

Anyway, I though I'd check what effect [[msvc::no_unique_address] has on tuplet's layout - and the answer was very surprising (to me, anyway!). I've added another commit on this branch with some tests if you're interested, but in summary it doesn't allow an element's alignment padding to be used for the next element, irrespective of whether it's an aggregate or not.

So there's no reduction in tuple size but at least the warnings are gone...

[codeinred]

Hi! I'm really sorry it took me so long to get back to you! School was pretty busy this week. I've merged your changes! I appreciate your help with the project!