Casey Carter

> Based on observing various MSVC versions, while newer versions of MSVC (since 16.9) do support `[[no_unique_address]]`, it seems to me like it has no effect, while `[[msvc::no_unique_address]]` does have...

Per [[reduce]/5](http://eel.is/c++draft/reduce#5) we are required to diagnose this program as ill-formed now since the requirement is a "Mandates" instead of the previous "Requires". I suspect there may be other similar...

> The convertibility requirements are bogus but most of the callability requirements are real, I think? `binary_op(init, init)` requires `binary_op` to accept two lvalues of type `T`, a pattern which...

> > > The convertibility requirements are bogus but most of the callability requirements are real, I think? > > > > > > `binary_op(init, init)` requires `binary_op` to accept...

> It isn't strictly necessary to convert `binary_op(init, init)` to `T` to evaluate `GENERALIZED_SUM`. However, it grants STL permission to spare intermediate results to `T`s during (parallel) evaluation whenever desirable...

To be even more pedantic, we should always use direct-non-list-initialization.

Is it time for us to have our first `/permissive-`-only library "conformance improvement"? People use `/permissive-` because they want to write standard-conforming code, they are likely willing to add a...

#3924 deprecated the contents of these headers. I'm leaving this issue open as a place to discuss when and how we should remove them.

This was a deliberate change in the IS via https://cplusplus.github.io/LWG/issue3646. We're never going to be able to change the fact that the Standard Library uses unsigned integral types inappropriately, hopefully...

Because potentially-narrowing conversions in C++ should not happen implicitly.