Fixes a grammar error found by @tschuett.

Addresses review comments.
Updates C++2b status on the index page.

The in C++11 deprecated throw() has been removed in C++20. Remove the test using this feature.

Thanks for the review.

• Rebased.
• Ran clang-format.
• Added more unit tests at @cjdb's suggestion.

• Addresses review comments.
  • Add some fundamental types to the unit test
  • Ran clang-format on the patch
• Rebased to test against a recent main before landing the patch

Thanks for the review. I'll address your comments and run a CI build before landing this revision.

Rebase to trigger CI. The last build failed, but so did main.

Rebase to trigger CI. The last build failed, but so did main.

In D95248#2516233, @tschuett wrote:

Do what ever you want, but a lot sounds a bit cheap. Or give some concrete examples?

In D93912#2515767, @WimLeflere wrote:

The paper is fully implemented.

With the caveat that std::string::contains is not constexpr yet.
std::string_view::contains is constexpr

• Removed unimplemented part of the synopsis
• Ran clang-format on the patch
• Rebased to trigger CI again

@WimLeflere I just noticed that during review we didn't mention the https://libcxx.llvm.org/docs/Cxx2bStatus.html should be updated. Does this commit fully implement the paper?

• Addresses review comments:
  • Removed unimplemented part of the synopsis
  • Removed cpp version check in format.cpp
• Replace c++ version number 2a with 20 where appropriate. This reflects changes done in main by unrelated commits.
• Rebased to trigger CI again.

In D93912#2479938, @WimLeflere wrote:

The macro should be added to libcxx/utils/generate_feature_test_macro_components.py, then run the script to generate the proper macros and tests.

Thanks!
Should I check in the (changed) generated files?

In D93912#2474869, @WimLeflere wrote:

Where are the feature test macros defined?
__cpp_lib_string_contains should be added

Thanks, LGTM.

One minor question, otherwise LGTM.

It seems the CI build fails since it can't process the dependencies properly. I'll rebase and trigger an CI build once the dependencies have landed.

Replace a std:: with _VSTD::

Improve the formatting as requested by @mclow.lists.
Disabled the automatic formatting to make sure the formatting won't be updated
automatically.

Thanks for the review!

Disable unit test when localization is disabled.

LGTM, once it passes CI.

Do we need to discuss the wanted formatting in this patch or on Slack/dev-ml?

Addresses review comments.

Thanks for the review!

Fixes a GCC shadow warning.

Rebased.

Rebased.

Rebased.

@ldionne Can you have a look?

A few minor nits, but otherwise LGTM.

Implement all new header tests as suggested by @curdeius in D92214.

Implement all new header tests as suggested by @curdeius in D92214.

Implement all new header tests as suggested by @curdeius.

Updated the Apple ABI list and synopsis.

Thanks for the review.

Thanks for the patch.
One minor nit, other than that LGTM.

Rebase to trigger CI to generate the Apple ABI list.

Disable the unit tests on clang-8, they fail and we don't support these old compilers for libc++.
Switch to the new status page
Adjusted the unit tests modified in D61170 to work in constexpr evaluation. Since constexpr string isn't supported yet the std::string tests are disabled in constant evaluation context.

Disable the Apple unit tests as suggested by @curdeius.
Formatted the manually written code with the .clang-format settings as proposed in D92229.

Thanks @curdeius for looking into this! The lack of proper clang-format rules for libc++ also bothered me a bit. Ut didn't bother me enough to look into it yet.

In D92214#2420187, @curdeius wrote:

To fix Apple back-deployment build, you'll need to mark your test with sth like:

Properly guard the code in format.cpp to only be used in C++20 mode.
Update the x86_64 ABI list.
Note the Apple ABI list is still out of date.

Thanks for the review!

@ldionne the patch passes the CI build and I addressed all previous review comments. When you have time can you have a look?

Reopened since I reverted to commit. It fails on several build bots using LLVM 8.

Move the enable_borrowed_range to a helper header to avoid including <ranges> in <span> and <string_view> as suggested by @miscco.

Rebase to trigger CI.

Rebase to trigger CI build.

In D91986#2412189, @cjdb wrote:

@CaseyCarter had this to say back in D74291.

There's tons of coverage here for class types and reference types. What about void, function types, pointers, arrays, pointers-to-members? (I don't know about clang/libc++, but the MSVC STL tests for type traits tend to _be_ the tests for compiler intrinsics used to implement those traits, so we consequently cover everything under the sun.)

I'd suggest we apply this here as well.

Rebase to trigger CI.

Rebase to trigger CI.

Thanks for the review.
I've switched to arc, so that should solve the CI issues.

Rebase to trigger the bildkite.

This should complete the implementation of P1645
Additional changes

• Since constexpr vector isn't available some test have a workaround to be tested without this feature. The changes are marked with a fixme to be removed once the support is available.
• Enables additional unit tests in libcxx/test/std/numerics/numeric.ops/inclusive.scan/inclusive_scan_op.pass.cpp. These were disabled when added, but that seems an oversight.
• Changes the changed numeric functions to all use _LIBCPP_INLINE_VISIBILITY and removed the unneeded inline.

In D90569#2385641, @ldionne wrote:

In D90569#2371916, @Mordante wrote:

Is there a reason why their signatures differ or is it an oversight?
Shall I fix them in this patch?

I believe those are oversights, and the blame seems to confirm this. Good catch, and please fix them in this patch. You don't need inline though, just _LIBCPP_INLINE_VISIBILITY. These functions are already inline (i.e. linkonce-ODR) because they are templates.