1 =========================================
2 Libc++ 14.0.0 (In-Progress) Release Notes
3 =========================================
9 Written by the `Libc++ Team <https://libcxx.llvm.org>`_
13 These are in-progress notes for the upcoming libc++ 14 release.
14 Release notes for previous releases can be found on
15 `the Download Page <https://releases.llvm.org/download.html>`_.
20 This document contains the release notes for the libc++ C++ Standard Library,
21 part of the LLVM Compiler Infrastructure, release 14.0.0. Here we describe the
22 status of libc++ in some detail, including major improvements from the previous
23 release and new feature work. For the general LLVM release notes, see `the LLVM
24 documentation <https://llvm.org/docs/ReleaseNotes.html>`_. All LLVM releases may
25 be downloaded from the `LLVM releases web site <https://llvm.org/releases/>`_.
27 For more information about libc++, please see the `Libc++ Web Site
28 <https://libcxx.llvm.org>`_ or the `LLVM Web Site <https://llvm.org>`_.
30 Note that if you are reading this file from a Git checkout or the
31 main Libc++ web page, this document applies to the *next* release, not
32 the current one. To see the release notes for a specific release, please
33 see the `releases page <https://llvm.org/releases/>`_.
35 What's New in Libc++ 14.0.0?
36 ============================
41 - There's initial support for the C++20 header ``<format>``. The implementation
42 is incomplete. Some functions are known to be inefficient; both in memory
43 usage and performance. The implementation is considered experimental and isn't
44 considered ABI stable.
46 - There's a new CMake option ``LIBCXX_ENABLE_UNICODE`` to disable Unicode
47 support in the ``<format>`` header. This only affects the estimation of the
48 output width of the format functions.
50 - Support for building libc++ on top of a C Standard Library that does not support ``wchar_t`` was
51 added. This is useful for building libc++ in an embedded setting, and it adds itself to the various
52 freestanding-friendly options provided by libc++.
54 - ``_LIBCPP_DEBUG`` equals to ``1`` enables the randomization of unspecified
55 behavior of standard algorithms (e.g. equal elements in ``std::sort`` or
56 randomization of both sides of partition for ``std::nth_element``)
58 - Floating-point support for ``std::to_chars`` support has been added.
59 Thanks to Stephan T. Lavavej and Microsoft for providing their implementation
65 - The functions ``std::atomic<T*>::fetch_(add|sub)`` and
66 ``std::atomic_fetch_(add|sub)`` no longer accept a function pointer. While
67 this is technically an API break, the invalid syntax isn't supported by
68 libstdc++ and MSVC STL. See https://godbolt.org/z/49fvzz98d.
70 - The call of the functions ``std::atomic_(add|sub)(std::atomic<T*>*, ...)``
71 with the explicit template argument ``T`` are now ill-formed. While this is
72 technically an API break, the invalid syntax isn't supported by libstdc++ and
73 MSVC STL. See https://godbolt.org/z/v9959re3v.
75 Due to this change it's now possible to call these functions with the
76 explicit template argument ``T*``. This allows using the same syntax on the
77 major Standard library implementations.
78 See https://godbolt.org/z/oEfzPhTTb.
80 Calls to these functions where the template argument was deduced by the
81 compiler are unaffected by this change.
83 - The functions ``std::allocator<T>::allocate`` and
84 ``std::experimental::pmr::polymorphic_allocator<T>::allocate`` now throw
85 an exception of type ``std::bad_array_new_length`` when the requested size
86 exceeds the maximum supported size, as required by the C++ standard.
87 Previously the type ``std::length_error`` was used.
89 - Removed the nonstandard methods ``std::chrono::file_clock::to_time_t`` and
90 ``std::chrono::file_clock::from_time_t``; neither libstdc++ nor MSVC STL
91 had such methods. Instead, in C++20, you can use ``std::chrono::file_clock::from_sys``
92 and ``std::chrono::file_clock::to_sys``, which are specified in the Standard.
93 If you are not using C++20, you should move to it.
95 - The declarations of functions ``declare_reachable``, ``undeclare_reachable``, ``declare_no_pointers``,
96 ``undeclare_no_pointers``, and ``get_pointer_safety`` have been removed not only from C++2b but
97 from all modes. Their symbols are still provided by the dynamic library for the benefit of
98 existing compiled code. All of these functions have always behaved as no-ops.
103 - The C++17 variable templates ``is_error_code_enum_v`` and
104 ``is_error_condition_enum_v`` are now of type ``bool`` instead of ``size_t``.
106 - The C++03 emulation type for ``std::nullptr_t`` has been removed in favor of
107 using ``decltype(nullptr)`` in all standard modes. This is an ABI break for
108 anyone compiling in C++03 mode and who has ``std::nullptr_t`` as part of their
109 ABI. However, previously, these users' ABI would be incompatible with any other
110 binary or static archive compiled with C++11 or later. If you start seeing linker
111 errors involving ``std::nullptr_t`` against previously compiled binaries, this may
112 be the cause. You can define the ``_LIBCPP_ABI_USE_CXX03_NULLPTR_EMULATION`` macro
113 to return to the previous behavior. That macro will be removed in LLVM 15. Please
114 comment `here <https://reviews.llvm.org/D109459>`_ if you are broken by this change
115 and need to define the macro.
117 - On Apple platforms, ``std::random_device`` is now implemented on top of ``arc4random()``
118 instead of reading from ``/dev/urandom``. Any implementation-defined token used when
119 constructing a ``std::random_device`` will now be ignored instead of interpreted as a
120 file to read entropy from.
125 - Building the libc++ shared or static library requires a C++ 20 capable compiler.
126 Consider using a Bootstrapping build to build libc++ with a fresh Clang if you
127 can't use the system compiler to build libc++ anymore.
129 - Historically, there has been numerous ways of building libc++ and libc++abi. This has
130 culminated in over 5 different ways to build the runtimes, which made it impossible to
131 maintain with a good level of support. Starting with this release, the runtimes support
132 exactly two ways of being built, which should cater to all use-cases. Furthermore,
133 these builds are as lightweight as possible and will work consistently even when targeting
134 embedded platforms, which used not to be the case. Please see the documentation on building
135 libc++ to see those two ways of building and migrate over to the appropriate build instructions
138 All other ways to build are deprecated and will not be supported in the next release.
139 We understand that making these changes can be daunting. For that reason, here's a
140 summary of how to migrate from the two most common ways to build:
142 - If you were rooting your CMake invocation at ``<monorepo>/llvm`` and passing ``-DLLVM_ENABLE_PROJECTS=<...>``
143 (which was the previously advertised way to build the runtimes), please simply root your CMake invocation at
144 ``<monorepo>/runtimes`` and pass ``-DLLVM_ENABLE_RUNTIMES=<...>``.
146 - If you were doing two CMake invocations, one rooted at ``<monorepo>/libcxx`` and one rooted at
147 ``<monorepo>/libcxxabi`` (this used to be called a "Standalone build"), please move them to a
148 single invocation like so:
152 $ cmake -S <monorepo>/libcxx -B libcxx-build <LIBCXX-OPTIONS>
153 $ cmake -S <monorepo>/libcxxabi -B libcxxabi-build <LIBCXXABI-OPTIONS>
159 $ cmake -S <monorepo>/runtimes -B build -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi" <LIBCXX-OPTIONS> <LIBCXXABI-OPTIONS>
161 - Support for building the runtimes using the GCC 32 bit multilib flag (``-m32``) has been removed. Support
162 for this had been flaky for a while, and we didn't know of anyone depending on this. Instead, please perform
163 a normal cross-compilation of the runtimes using the appropriate target, such as passing the following to
164 your bootstrapping build:
168 -DLLVM_RUNTIME_TARGETS=i386-unknown-linux
170 - Libc++, libc++abi and libunwind will not be built with ``-fPIC`` by default anymore.
171 If you want to build those runtimes with position independent code, please specify
172 ``-DCMAKE_POSITION_INDEPENDENT_CODE=ON`` explicitly when configuring the build, or
173 ``-DRUNTIMES_<target-name>_CMAKE_POSITION_INDEPENDENT_CODE=ON`` if using the