thanks! LGTM. i must have been looking at a reference that's missing that requires...

One thing I have thought about in our march towards finishing the lexically ordered containers is that we should be able to write a test utils header with one or more templated test functions that will work for most or all of these containers. We can possibly even add (in another header?) some helpful types like weakly ordered ints, types that compare with user orderings, etc. so we don't have to keep writing them over and over. This feels fairly sane to me but I don't know what the general vibe is on how DRY is too DRY for our test suite, and maybe it's better to just copy-paste those parts of the code each time.

Sorry! Something has gone horribly wrong with my usage of arcanist and it diffed against a very old version of main. i will try to repair the reviewers list.

• Merge branch 'main' into doc2

• Merge remote-tracking branch 'upstream/main' into doc2

• Revert "use a version of the link that does not currently 404 on wg21.link"

• fix typo in "chrono::time_point"
• fix broken formatting for the "basic_string" hyperlink
• link to differentials of completed project items
• Merge branch 'main' into doc2

• use a version of the link that does not currently 404 on wg21.link

okay, every comment is either resolved or involves work that should go into another differential.

i suspect we can ship this differential as-is without fiddling with test_comparisons.h any further, and any further changes we want to make to that file should be done in another differential.

• re-add possibly less confusing assert to testComparisons that does not reference a parameter it doesn't have
• Merge branch 'main' into variant
• add sfinae tests for variant::operator<=>
• variant::operator<=> does NOT require three_way_comparable for its types in the standard

• simplify TEST_NODISCARD definition per feedback

• assert return value types and simplify order value expressions
• add TEST_NODISCARD macro
• use TEST_NODISCARD
• formatting

I think the random access iterator optimization is probably pretty important here. it may be doable to eventually derive this optimization from very simple iterator types like int*, but there are plenty of examples of very non-trivial iterator types where this may not be the case. iterators over order-statistic trees can satisfy every requirement of random_access_iterator while being somewhat complex.

• correctly wrap testOrder and testComparisons in asserts
• fix nits
• run git clang-format on all changes
• manually clang-format the added function
• manually fix spacing of multiline macros
• add partial_ordering::unordered test cases & utils support
• refactor testComparisons and clean up includes
• add nodiscard vis a vis D131364
• order things sensibly because this is C++

Thanks! looks good to me

• also use _LIBCPP_HIDE_FROM_ABI in monostate

• modernize variant three_way test
• _LIBCPP_INLINE_VISIBILITY --> _LIBCPP_HIDE_FROM_ABI, remove unnecessary inlines

• add comment in test for tested operator
• don't conditionally include <compare>
• use runtime-and-static assert pattern for monostate test
• use test macros for monostate relops

Thanks for looking!

• add missing include for monostate

Updated SpaceshipProjects and added the missing monostate operator as well

• delete dead test function
• use a result type alias
• also implement std::monostate::operator<=>
• update spaceship projects

Feeling like getting something done this weekend

What's the status of this one? Looks stalled, and I understand it's still a blocker for make_{unique,shared}_for_overwrite in D90111

• Revert "add reference to P1614 in the spaceship RST"

• correct section name for move_iterator

• Revert "move [charconv.syn] to Formatting"
• add reference to P1614 in the spaceship RST

• add link to coroutine differential
• move [charconv.syn] to Formatting

This seems like a good ideal to work towards to me, as the organization of these issues is a bit messy.

• add missing comments

• also test size-incompatible < in all versions

• reduce the scope of incompatible-size tuple comparison tests

• rebase

• work around a GCC segfault bug

• move and standardize the test for synth-three-way

• don't cast-to-void at top-level scope
• adl-proofing
• address comments

• address comments