Note: If there‘s anything you want to see done in the NDK, file a bug! Nothing here is set in stone, and if there’s something that we haven‘t thought of that would be of more use, we’d be happy to adjust our plans for that.
Disclaimer: Everything here is subject to change. The further the plans are in the future, the less stable they will be. Things in the upcoming release are fairly certain, and the second release is quite likely. Beyond that, anything written here is what we would like to accomplish in that release assuming things have gone according to plan until then.
Note: For release timing, see our release schedule on our wiki.
Every NDK release aims to include a new toolchain, new headers, and a new version of libc++.
We also maintain GitHub Projects to track the bugs we intend to fix in any given NDK release.
The NDK and the Android OS use the same toolchain. Android's toolchain team is constantly working on updating to the latest upstream LLVM for the OS. It can take a long time to investigate issues when compiling -- or issues that the newer compiler finds in -- OS code or OEM code, for all 4 supported architectures, so these updates usually take a few months.
Even then, a new OS toolchain may not be good enough for the NDK. In the OS, we can work around compiler bugs by changing our code, but for the NDK we want to make compiler updates cause as little disruption as possible. We also don't want to perform a full compiler update late in the NDK release cycle for the sake of stability.
The aim is that each NDK will have a new toolchain that‘s as up to date as feasible without sacrificing stability, but we err on the side of stability when we have to make a choice. If an NDK release doesn’t include a new compiler, or that compiler isn‘t as new as you’d hoped, trust us --- you wouldn't want anything newer that we have just yet!
Most of the team's work is currently focused outside the NDK proper, so while the NDK release notes may seem a bit sparse, there are still plenty of improvements coming for NDK users:
std::filesystem is now ported and included in the NDK.
NDK r18 made LLD available, r20 made it more usable, and r22 has made it the default.
r22 also migrates from GNU
llvm-strip. No GNU binutils tools are used by the NDK with the default configuration. GNU binutils will be removed in a future release of the NDK.
LLDB is now included alongside the toolchain in the NDK, and
ndk-gdb supports the
--lldb option. GDB will be removed in a future release of the NDK.
The following projects are listed in order of their current priority.
Note that some of these projects do not actually affect the contents of the NDK package. The samples, documentation, etc are all NDK work but are separate from the NDK package. As such they will not appear in any specific release, but are noted here to show where the team's time is being spent.
Right now only 32-bit Arm uses the LLVM unwinder for exception handling, and the remaining architectures rely on libgcc. Using the same unwinder across all architectures will give more predictable behavior, and consolidating on a single unwinder reduces duplicated upkeep costs so we can work on other improvements.
These two libraries provide the runtime support the compiler relies on. While similar, LLVM‘s
libclang_rt.builtins includes many things that libgcc does not. We’ve been working around inconsistencies for a while by using both, but it would be best to just finish the migration.
One blocker here is that most architectures currently rely on libgcc for exception handling support. Once we‘ve migrated to LLVM’s unwinder for all architectures this will be easier.
We've switched to LLD and the rest of the LLVM tools by default as of r22. We should consolidate on the supported tools after they have been given some soak time to discover and fix any remaining issues.
CMake added their own NDK support about the same time we added our toolchain file. The two often conflict with each other, and a toolchain file is a messy way to implement this support. However, fully switching to the integrated support puts NDK policy decisions (default options, NDK layout, etc) fully into the hands of CMake, which makes them impossible to update without the user also updating their CMake version.
We will reorganize our toolchain file to match the typical implementation of a CMake platform integration (like
$CMAKE/Modules/Platform/Android-*.cmake) and CMake will be modified to load the implementation from the NDK rather than its own.
Preferably, most of this work will actually involve improving Clang‘s defaults so that the toolchain file doesn’t need to do anything aside from setting up toolchain paths.
See Issue 463 for discussion.
The following projects are things we intend to do, but have not yet been scheduled into the sections above.
We should probably add basic doc comments to the bionic headers:
char*point to? Who owns it? Are errors -1 (as for most functions) or
Should these be in the NDK API reference too? If so, how will we keep them from swamping the “real” NDK API?
vim is ready, Android Studio now supports doxygen comments (but seems to have gained a new man page viewer that takes precedence), and Visual Studio Code has nothing but feature requests.
Beyond writing the documentation, we also should invest some time in improving the presentation of the NDK API reference on developer.android.com.
The samples are low-quality and don't necessarily cover interesting/difficult topics.
The NDK has long included
gtest and clang supports various sanitiziers, but there are things we can do to improve the state of testing/code quality:
NDK APIs are C-only for ABI stability reasons.
Complaints about basic JNI handling are common. We should make libnativehelper available as an AAR.
Before we can take on maintenance for additional packages we need to improve the tooling for ndkports. Automation for package updates, testing, and the release process would make it possible to expand.
For serious i18n,
icu4c is too big too bundle, and non-trivial to use the platform. We have a C API wrapper prototype, but we need to make it easily available for NDK users.
We still need to update libc++ twice: once for the platform, and once for the NDK. We also still have two separate test runners.
iOS developers are used to using weak symbols to refer to function that may be present in their equivalent of
targetSdkVersion but not in their
minSdkVersion. We could potentially do something similar. See issue 1003.
By Q2 2019 Clang may have a complete enough implementation of the modules TS and Android may have a Clang with those changes available.
At least for the current spec (which is in the process of merging with the Clang implementation, so could change), the NDK will need to:
Full history is available, but this section summarizes major changes in recent releases.
We shipped Prefab and the accompanying support for the Android Gradle Plugin to support native dependencies. AGP 4.0 includes the support for importing these packages, and 4.1 includes the support for creating AARs that support them.
We also maintain a few packages as part of ndkports. Currently just curl, OpenSSL, and JsonCpp.
Updated Clang, LLD, libc++, make, and GDB. Much better LLD behavior on Windows. 32-bit Windows support removed. Neon by default for all API levels. OpenMP now available as both a static and shared library.
Updated Clang and libc++, added Q APIs. Improved out-of-the-box Clang behavior.
Reorganized the toolchain packaging and modified Clang so that standalone toolchains are now unnecessary. Clang can now be invoked directly from its installed location in the NDK.
C++ compilation defaults to C++14.
Removed GCC and gnustl/stlport. Added lld.
compile_commands.json for better tooling support.
Defaulted to libc++.
Removed ARMv5 (armeabi), MIPS, and MIPS64.
Fixed libandroid_support, libc++ now the recommended STL (but still not the default).
Removed non-unified headers.
Defaulted to unified headers (opt-out).
Removed support for API levels lower than 14 (Android 4.0).
Added unified headers (opt-in).
Removed support for API levels lower than 9 (Android 2.3).