[libc++] Disable incomplete features for the LLVM 14 release

[ldionne]

Summary

For the LLVM 14 release, don't ship incomplete features (currently
ranges and format) by default. If some vendors want to ship those
features, they can override the CMake option manually. This is
consistent with what we did for LLVM 13.

Note that this patch is only aimed towards the release/14.x branch.

[Mordante]

libcxx/CMakeLists.txt

136

No objection to this change. However I think the release notes need to be updated. The first bullet regarding <format> contains the wording.

Vendors can still disable this header by turning the CMake option
LIBCXX_ENABLE_INCOMPLETE_FEATURES off.

The current wording implies ON is the default.

[ldionne]

Address review comments, rebase and fix CI by excluding the format symbols from the ABI list by default
(since by default, we're effectively not including those symbols in the dylib).

[Mordante]

libcxx/docs/ReleaseNotes.rst

47 ↗

(On Diff #405808)

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.no_new_in_libcxx.abilist

1044 ↗

(On Diff #405808)

I expect this to break the ABI check target in CMake when LIBCXX_ENABLE_INCOMPLETE_FEATURES is turned on, am I ?
Should we use different ABI lists depending on the state of the LIBCXX_ENABLE_INCOMPLETE_FEATURES flag?

[ldionne]

Address review comments and try to fix GCC build

[ldionne]

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.no_new_in_libcxx.abilist

1044 ↗

(On Diff #405808)

Yes, that will break the ABI check target when LIBCXX_ENABLE_INCOMPLETE_FEATURES=ON, however on the release/14.x branch, that is never the case, so it's no big deal. This is arguably a bit hacky.

In the future, what I want instead is to ship those symbols inside c++experimental.a, and vendors can actually ship that static archive. We'd also add a Clang flag (something like -fexperimental) that controls whether those features are enabled so that users can play around with them. Since the symbols would be in c++experimental.a, we wouldn't put them in the ABI lists until they are moved into the main dylib anyway, and that would be a non issue.

[Mordante]

LGTM after the CI passes.

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.no_new_in_libcxx.abilist

1044 ↗

(On Diff #405808)

That sounds like a good future direction. That also makes it easier for users to decide whether or not to depend on these not complete features. Instead of vendors needing to make the choices.

[• Quuxplusone]

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.no_new_in_libcxx.abilist

1044 ↗

(On Diff #405808)

That seems like a fine future direction, but notice that it doesn't help at all for Ranges (or the templated parts of Format); I think we're converging on the idea that the "ABI-dangerous" part of Ranges is stuff like transform_view, which are header-only.
Vice versa, I don't actually see what would be the problem with unconditionally shipping __ZNSt3__112format_errorD0Ev in our dylib. That's not the ABI-dangerous part of Format. Nothing about the ABI of format_error seems unstable to me.

[ldionne]

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.no_new_in_libcxx.abilist

1044 ↗

(On Diff #405808)

You're right, it's just that we might as well wait to ship all of format before shipping any of it.

I do think that -fexperimental will help for Ranges, because -fexperimental will be documented as being unstable and not to be used for production code. That way, we can ship half-baked stuff and keep changing them until we're confident that they'll be stable.

[ldionne]

Try to fix the GCC build again, less invasively.

[Mordante]

libcxx/lib/abi/arm64-apple-darwin.libcxxabi.v1.stable.exceptions.no_new_in_libcxx.abilist

1044 ↗

(On Diff #405808)

@Quuxplusone having experimental library would make it possible to move more parts of std::format to the dylib. Due to ABI concerns I've avoided doing that. This is something at my todo list and has also been requested by @ldionne.

(Your observation regarding format_error is correct, that's ABI stable.)

[ldionne]

Try to fix the CI -- I think this should be good.

This is kind of nasty, I'll have to cherry-pick some of these changes back onto main.
There's a couple of problems:

1. The benchmarks didn't work when incomplete features were disabled -- that's trivial to fix
2. The modular build is not compatible with various knobs we have in the __config_site. For example, you can't use the modular build with LIBCXX_ENABLE_LOCALIZATION=OFF. A lot of stuff is going to break. In this patch, I am fixing just enough for the modular build to work when incomplete features are disabled, but we should really investigate the situation so that the modular build works with all configuration knobs.

[ldionne]

CI is passing. I'm going to push this because I really want to land this in the 14 release. I'll then cherry-pick bits and pieces back onto main and we can solve some of the underlying issues separately.

[ldionne]

Landed in 7ff79971b98a5143cba81bcaebb2d5a70b3728c0

libcxx/CMakeLists.txt

option(LIBCXX_ENABLE_VENDOR_AVAILABILITY_ANNOTATIONS
    libc++ for any specific vendor, and we disable those annotations. Vendors wishing
    to provide compile-time errors when using features unavailable on some version of
    the shared library they shipped should turn this on and see `include/__availability`
    for more details." OFF)
 option(LIBCXX_ENABLE_INCOMPLETE_FEATURES
     "Whether to enable support for incomplete library features. Incomplete features
     are new library features under development. These features don't guarantee
     ABI stability nor the quality of completed library features. Vendors
-    shipping the library may want to disable this option." ON)
+    shipping the library may want to disable this option." OFF)
 set(LIBCXX_TEST_CONFIG "legacy.cfg.in" CACHE STRING
     "The path to the Lit testing configuration to use when running the tests.
      If a relative path is provided, it is assumed to be relative to '<monorepo>/libcxx/test/configs'.")
 if (NOT IS_ABSOLUTE "${LIBCXX_TEST_CONFIG}")
   set(LIBCXX_TEST_CONFIG "${CMAKE_CURRENT_SOURCE_DIR}/test/configs/${LIBCXX_TEST_CONFIG}")
 endif()
 set(LIBCXX_TEST_PARAMS "" CACHE STRING
     "A list of parameters to run the Lit test suite with.")