Pushed changes to address @strega-nil-ms's comments, plus minor things I noticed along the way:

• Rename to _Subscript_closure.
• Rename to _Get_iterator_inner.
  • Also drop a couple of unnecessary explicit template arguments.
• Shorten _Iter_sent_difference_type (to avoid wrapping weirdly).
• Remove clang-format suppressions.
• Use C-style comments for /* strengthened */ to avoid wrapping braces on the next line.
• Change _Is_end_noexcept to be a _CONSTEVAL function.

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

I had to push an additional commit to work around an internal compiler assertion, VSO-1732554 "x86chk assertion failure in immediately invoked lambda computing iterator_category: RPF_ISSET(TypeSeenThisDepth) || Parser::Utility::CurrentlyInRdParser(), p1inl.c 5344". The workaround is to use a chain of conditional_ts.

I had to push another commit to work around #1030. The internal build uses native compilers for x86 (eww 🙀 🤮) and /analyze's higher memory consumption exceeds 4 GB for at least the random-access configuration (and is dangerously high for the others).

The really expensive thing is the matrix of 8 permutations. Instead of skipping everything for /analyze, I chose to skip 7 out of 8, to leave at least some coverage. (I'm not too worried about /analyze-specific warnings or problems, given this residual coverage.)

/azp run

Thanks for implementing another major view in this paper! 🤐 🔀 😻