Android NDK 学习之

分享到: file syntax specification


This document describes the syntax of build file written to describe your C and C++ source files to the Android NDK. To understand what follows, it is assumed that you have read theOVERVIEW file that explains their role and usage.


An file is written to describe your sources to the build system. More specifically:

  • The file is really a tiny GNU Makefile fragment that will be parsed one or more times by the build system. As such, you should try to minimize the variables you declare there and do not assume that anything is not defined during parsing.

  • The file syntax is designed to allow you to group your sources into 'modules'. A module is one of the following:

    • A static library.
    • A shared library.
    • A standalone executable.

Only shared libraries will be installed/copied to your application package. Static libraries can be used to generate shared libraries though.

You can define one or more modules in each file, and you can use the same source file in several modules.

  • The build system handles many details for you. For example, you don't need to list header files or explicit dependencies between generated files in your The NDK build system will compute these automatically for you.

This also means that, when updating to newer releases of the NDK, you should be able to benefit from new toolchain/platform support without having to touch your files.

Note that the syntax is very close to the one used in files distributed with the full open-source Android platform sources. While the build system implementation that uses them is different, this is an intentional design decision made to allow reuse of 'external' libraries' source code easier for application developers.

Simple example:

Before describing the syntax in details, let's consider the simple "hello JNI" example, i.e. the files under:


Here, we can see:

  • The src directory containing the Java sources for the sample Android project.

  • The jni directory containing the native source for the sample, i.e. jni/hello-jni.c

    This source file implements a simple shared library that implements a native method that returns a string to the VM application.

  • The jni/ file that describes the shared library to the NDK build system. Its content is:


    ---------- cut here ------------------
    LOCAL_PATH := $(call my-dir)

    include $(CLEAR_VARS)

    LOCAL_MODULE    := hello-jni
    LOCAL_SRC_FILES := hello-jni.c

    ---------- cut here ------------------

Now, let's explain these lines:

    LOCAL_PATH := $(call my-dir)

An file must begin with the definition of the LOCAL_PATH variable. It is used to locate source files in the development tree. In this example, the macro function 'my-dir', provided by the build system, is used to return the path of the current directory (i.e. the directory containing the file itself).

    include $(CLEAR_VARS)

The CLEAR_VARS variable is provided by the build system and points to a special GNU Makefile that will clear many LOCAL_XXX variables for you (e.g. LOCAL_MODULELOCAL_SRC_FILES,LOCAL_STATIC_LIBRARIES, etc...), with the exception of LOCAL_PATH. This is needed because all build control files are parsed in a single GNU Make execution context where all variables are global.

    LOCAL_MODULE := hello-jni

The LOCAL_MODULE variable must be defined to identify each module you describe in your The name must be unique and not contain any spaces. Note that the build system will automatically add proper prefix and suffix to the corresponding generated file. In other words, a shared library module named 'foo' will generate ''.

IMPORTANT NOTE: If you name your module 'libfoo', the build system will not add another 'lib' prefix and will generate as well. This is to support files that originate from the Android platform sources, would you need to use these.

    LOCAL_SRC_FILES := hello-jni.c

The LOCAL_SRC_FILES variables must contain a list of C and/or C++ source files that will be built and assembled into a module. Note that you should not list header and included files here, because the build system will compute dependencies automatically for you; just list the source files that will be passed directly to a compiler, and you should be good.

Note that the default extension for C++ source files is '.cpp'. It is however possible to specify a different one by defining the variable LOCAL_CPP_EXTENSION. Don't forget the initial dot (i.e. '.cxx' will work, but not 'cxx').


The BUILD_SHARED_LIBRARY is a variable provided by the build system that points to a GNU Makefile script that is in charge of collecting all the information you defined in LOCAL_XXX variables since the latest 'include $(CLEAR_VARS)' and determine what to build, and how to do it exactly. There is also BUILD_STATIC_LIBRARY to generate a static library.

There are more complex examples in the samples directories, with commented files that you can look at.


This is the list of variables you should either rely on or define in an You can define other variables for your own usage, but the NDK build system reserves the following variable names:

  • Names that begin with LOCAL_ (e.g. LOCAL_MODULE)
  • Names that begin with PRIVATE_, NDK_ or APP_ (used internally)
  • Lower-case names (used internally, e.g. my-dir)

If you need to define your own convenience variables in an file, we recommend using the MY_ prefix, for a trivial example:

    ---------- cut here ------------------
    MY_SOURCES := foo.c
    ifneq ($(MY_CONFIG_BAR),)
      MY_SOURCES += bar.c

    ---------- cut here ------------------

So, here we go:

NDK-provided variables:

These GNU Make variables are defined by the build system before your file is parsed. Note that under certain circumstances the NDK might parse your several times, each with different definition for some of these variables.


Points to a build script that undefines nearly all LOCAL_XXX variables listed in the "Module-description" section below. You must include the script before starting a new module, e.g.:

        include $(CLEAR_VARS)


Points to a build script that collects all the information about the module you provided in LOCAL_XXX variables and determines how to build a target shared library from the sources you listed. Note that you must have LOCAL_MODULE and LOCAL_SRC_FILES defined, at a minimum before including this file. Example usage:

        include $(BUILD_SHARED_LIBRARY)

note that this will generate a file named lib$(LOCAL_MODULE).so.


A variant of BUILD_SHARED_LIBRARY that is used to build a target static library instead. Static libraries are not copied into your project/packages but can be used to build shared libraries (seeLOCAL_STATIC_LIBRARIES and LOCAL_WHOLE_STATIC_LIBRARIES described below). Example usage:

        include $(BUILD_STATIC_LIBRARY)

Note that this will generate a file named lib$(LOCAL_MODULE).a.


Points to a build script used to specify a prebuilt shared library. Unlike BUILD_SHARED_LIBRARY and BUILD_STATIC_LIBRARY, the value of LOCAL_SRC_FILES must be a single path to a prebuilt shared library (e.g. foo/, instead of a source file.

You can reference the prebuilt library in another module using the LOCAL_PREBUILTS variable (see PREBUILTS for more information).


This is the same as PREBUILT_SHARED_LIBRARY, but for a static library file instead. See PREBUILTS for more.


Name of the target CPU architecture as it is specified by the full Android open-source build. This is 'arm' for any ARM-compatible build, independent of the CPU architecture revision.


Name of the target Android platform when this is parsed. For example, 'android-3' correspond to Android 1.5 system images. For a complete list of platform names and corresponding Android system images, read STABLE APIS.


Name of the target CPU+ABI when this is parsed. Four values are supported at the moment:

        For ARMv5TE

        For ARMv7

        For i686

        For mips32 (r1)

NOTE: Up to Android NDK 1.6_r1, this variable was simply defined as 'arm'. However, the value has been redefined to better match what is used internally by the Android platform.

For more details about architecture ABIs and corresponding compatibility issues, please read CPU ARCH ABIS

Other target ABIs will be introduced in future releases of the NDK and will have a different name. Note that all ARM-based ABIs will have 'TARGET_ARCH' defined to 'arm', but may have different 'TARGET_ARCH_ABI'


The concatenation of target platform and ABI, it really is defined as $(TARGET_PLATFORM)-$(TARGET_ARCH_ABI) and is useful when you want to test against a specific target system image for a real device.

By default, this will be 'android-3-armeabi'

(Up to Android NDK 1.6_r1, this used to be 'android-3-arm' by default)

NDK-provided function macros:

The following are GNU Make 'function' macros, and must be evaluated by using '$(call )'. They return textual information.


Returns the path of the last included Makefile, which typically is the current's directory. This is useful to define LOCAL_PATH at the start of your as with:

          LOCAL_PATH := $(call my-dir)

IMPORTANT NOTE: Due to the way GNU Make works, this really returns the path of the last included Makefile during the parsing of build scripts. Do not call my-dir after including another file.

For example, consider the following example:

          LOCAL_PATH := $(call my-dir)

          ... declare one module

          include $(LOCAL_PATH)/foo/``

          LOCAL_PATH := $(call my-dir)

          ... declare another module

The problem here is that the second call to my-dir will define LOCAL_PATH to $PATH/foo instead of $PATH, due to the include that was performed before that.

For this reason, it's better to put additional includes after everything else in an, as in:

          LOCAL_PATH := $(call my-dir)

          ... declare one module

          LOCAL_PATH := $(call my-dir)

          ... declare another module

          # extra includes at the end of the ``
          include $(LOCAL_PATH)/foo/``

If this is not convenient, save the value of the first my-dir call into another variable, for example:

          MY_LOCAL_PATH := $(call my-dir)

          LOCAL_PATH := $(MY_LOCAL_PATH)

          ... declare one module

          include $(LOCAL_PATH)/foo/``

          LOCAL_PATH := $(MY_LOCAL_PATH)

          ... declare another module


Returns a list of located in all sub-directories of the current 'my-dir' path. For example, consider the following hierarchy:


If sources/foo/ contains the single line:

          include $(call all-subdir-makefiles)

Then it will include automatically sources/foo/lib1/ and sources/foo/lib2/

This function can be used to provide deep-nested source directory hierarchies to the build system. Note that by default, the NDK will only look for files in sources/*/


Returns the path of the current Makefile (i.e. where the function is called).


Returns the path of the parent Makefile in the inclusion tree, i.e. the path of the Makefile that included the current one.


Guess what...


A function that allows you to find and include the of another module by name. A typical example is:

        $(call import-module,<name>)

And this will look for the module tagged in the list of directories referenced by your NDK_MODULE_PATH environment variable, and include its automatically for you.

Read IMPORT-MODULE for more details.

Module-description variables:

The following variables are used to describe your module to the build system. You should define some of them between an 'include $(CLEAR_VARS)' and an 'include $(BUILD_XXXXX)'. As written previously, $(CLEAR_VARS) is a script that will undefine/clear all of these variables, unless explicitly noted in their description.


This variable is used to give the path of the current file. You MUST define it at the start of your, which can be done with:

        LOCAL_PATH := $(call my-dir)

This variable is not cleared by $(CLEAR_VARS) so only one definition per is needed (in case you define several modules in a single file).


This is the name of your module. It must be unique among all module names, and shall not contain any space. You MUST define it before including any $(BUILD_XXXX) script.

By default, the module name determines the name of generated files, e.g. for a shared library module named . However you should only refer to other modules with their 'normal' name (e.g.) in your NDK build files (either or

You can override this default with LOCAL_MODULE_FILENAME (see below)


This variable is optional, and allows you to redefine the name of generated files. By default, module will always generate a static library named lib.a or a shared library named, which are standard Unix conventions.

You can override this by defining LOCAL_MODULE_FILENAME, For example:

          LOCAL_MODULE := foo-version-1
          LOCAL_MODULE_FILENAME := libfoo

*NOTE(: You should not put a path or file extension in your LOCAL_MODULE_FILENAME, these will be handled automatically by the build system.


This is a list of source files that will be built for your module. Only list the files that will be passed to a compiler, since the build system automatically computes dependencies for you.

Note that source files names are relative to LOCAL_PATH and you can use path components, e.g.:

        LOCAL_SRC_FILES := foo.c \

Absolute file paths are also supported:

        LOCAL_SRC_FILES := /home/user/mysources/foo.c

or on Windows:

        LOCAL_SRC_FILES := c:/Users/user/sources/foo.c

Avoiding absolute file paths is recommended, this makes your easy to reuse on a different machine / system.

NOTE: Always use Unix-style forward slashes (/) in build files. Windows-style back-slashes will not be handled properly.


This is an optional variable that can be defined to indicate the file extension(s) of C++ source files. They must begin with a dot. The default is '.cpp' but you can change it. For example:

          LOCAL_CPP_EXTENSION := .cxx

Since NDK r7, you can list several extensions in this variable, as in:

          LOCAL_CPP_EXTENSION := .cxx .cpp .cc


This is an optional variable that can be defined to indicate that your code relies on specific C++ features. To indicate that your code uses RTTI (RunTime Type Information), use the following:

          LOCAL_CPP_FEATURES := rtti

To indicate that your code uses C++ exceptions, use:

          LOCAL_CPP_FEATURES := exceptions

You can also use both of them with (order is not important):

          LOCAL_CPP_FEATURES := rtti features

The effect of this variable is to enable the right compiler/linker flags when building your modules from sources. For prebuilt binaries, this also helps declare which features the binary relies on to ensure the final link works correctly.

It is recommended to use this variable instead of enabling -frtti and -fexceptions directly in your LOCAL_CPPFLAGS definition.


An optional list of paths, relative to the NDK root directory, which will be appended to the include search path when compiling all sources (C, C++ and Assembly). For example:

          LOCAL_C_INCLUDES := sources/foo

Or even:

          LOCAL_C_INCLUDES := $(LOCAL_PATH)/../foo

These are placed before any corresponding inclusion flag in LOCAL_CFLAGS / LOCAL_CPPFLAGS

The LOCAL_C_INCLUDES path are also used automatically when launching native debugging with ndk-gdb.


An optional set of compiler flags that will be passed when building C and C++ source files.

This can be useful to specify additional macro definitions or compile options.

IMPORTANT: Try not to change the optimization/debugging level in your, this can be handled automatically for you by specifying the appropriate information in your, and will let the NDK generate useful data files used during debugging.

NOTE: In android-ndk-1.5_r1, the corresponding flags only applied to C source files, not C++ ones. This has been corrected to match the full Android build system behaviour. (You can useLOCAL_CPPFLAGS to specify flags for C++ sources only now).

It is possible to specify additional include paths with LOCAL_CFLAGS += -I<path>, however, it is better to use LOCAL_C_INCLUDES for this, since the paths will then also be used during native debugging with ndk-gdb.


An alias for LOCAL_CPPFLAGS. Note that use of this flag is obsolete as it may disappear in future releases of the NDK.


An optional set of compiler flags that will be passed when building C++ source files only. They will appear after the LOCAL_CFLAGS on the compiler's command-line.

NOTE: In android-ndk-1.5_r1, the corresponding flags applied to both C and C++ sources. This has been corrected to match the full Android build system. (You can use LOCAL_CFLAGS to specify flags for both C and C++ sources now).


The list of static libraries modules that the current module depends on.

If the current module is a shared library or an executable, this will force these libraries to be linked into the resulting binary.

If the current module is a static library, this simply tells that another other module that depends on the current one will also depend on the listed libraries.


The list of shared libraries modules this module depends on at runtime. This is necessary at link time and to embed the corresponding information in the generated file.


A variant of LOCAL_STATIC_LIBRARIES used to express that the corresponding library module should be used as "whole archives" to the linker. See the GNU linker's documentation for the --whole-archive flag.

This is generally useful when there are circular dependencies between several static libraries. Note that when used to build a shared library, this will force all object files from your whole static libraries to be added to the final binary. This is not true when generating executables though.


The list of additional linker flags to be used when building your shared library or executable. This is useful to pass the name of specific system libraries with the '-l' prefix. For example, the following will tell the linker to generate a module that links to /system/lib/ at load time:

        LOCAL_LDLIBS := -lz

See STABLE-APIS for the list of exposed system libraries you can linked against with this NDK release.

NOTE: This is ignored for static libraries, and ndk-build will print a warning if you define it in such a module.


The list of other linker flags to be used when building your shared library or executable. For example, the following will use the ld.bfd linker on ARM/X86 GCC 4.6+ where is the default

        LOCAL_LDFLAGS += -fuse-ld=bfd

NOTE: This is ignored for static libraries, and ndk-build will print a warning if you define it in such a module.


By default, any undefined reference encountered when trying to build a shared library will result in an "undefined symbol" error. This is a great help to catch bugs in your source code.

However, if for some reason you need to disable this check, set this variable to 'true'. Note that the corresponding shared library may fail to load at runtime.

NOTE: This is ignored for static libraries, and ndk-build will print a warning if you define it in such a module.


By default, ARM target binaries will be generated in 'thumb' mode, where each instruction are 16-bit wide. You can define this variable to 'arm' if you want to force the generation of the module's object files in 'arm' (32-bit instructions) mode. E.g.:

        LOCAL_ARM_MODE := arm

Note that you can also instruct the build system to only build specific sources in ARM mode by appending an '.arm' suffix to its source file name. For example, with:

        LOCAL_SRC_FILES := foo.c bar.c.arm

Tells the build system to always compile 'bar.c' in ARM mode, and to build foo.c according to the value of LOCAL_ARM_MODE.

NOTE: Setting APP_OPTIM to 'debug' in your will also force the generation of ARM binaries as well. This is due to bugs in the toolchain debugger that don't deal too well with thumb code.


Defining this variable to 'true' allows the use of ARM Advanced SIMD (a.k.a. NEON) GCC intrinsics in your C and C++ sources, as well as NEON instructions in Assembly files.

You should only define it when targeting the 'armeabi-v7a' ABI that corresponds to the ARMv7 instruction set. Note that not all ARMv7 based CPUs support the NEON instruction set extensions and that you should perform runtime detection to be able to use this code at runtime safely. To learn more about this, please read the documentation at CPU-ARM-NEON and CPU-FEATURES.

Alternatively, you can also specify that only specific source files may be compiled with NEON support by using the '.neon' suffix, as in:

        LOCAL_SRC_FILES = foo.c.neon bar.c zoo.c.arm.neon

In this example, 'foo.c' will be compiled in thumb+neon mode, 'bar.c' will be compiled in 'thumb' mode, and 'zoo.c' will be compiled in 'arm+neon' mode.

Note that the '.neon' suffix must appear after the '.arm' suffix if you use both (i.e. foo.c.arm.neon works, but not foo.c.neon.arm !)


Android NDK r4 added support for the "NX bit" security feature. It is enabled by default, but you can disable it if you really need to by setting this variable to 'true'.

NOTE: This feature does not modify the ABI and is only enabled on kernels targeting ARMv6+ CPU devices. Machine code generated with this feature enabled will run unmodified on devices running earlier CPU architectures.

For more information, see:


By default, NDK compiled code is built with read-only relocations and GOT protection. This instructs the runtime linker to mark certain regions of memory as being read-only after relocation, making certain security exploits (such as GOT overwrites) harder to perform.

It is enabled by default, but you can disable it if you really need to by setting this variable to 'true'.

NOTE: These protections are only effective on newer Android devices ("Jelly Bean" and beyond). The code will still run on older versions (albeit without memory protections).

For more information, see:


By default, NDK compiled code is compiled with format string protection. This forces a compiler error if a non-constant format string is used in a printf style function.

It is enabled by default, but you can disable it if you really need to by setting this variable to 'true'.


Define this variable to record a set of C/C++ compiler flags that will be added to the LOCAL_CFLAGS definition of any other module that uses this one with LOCAL_STATIC_LIBRARIES orLOCAL_SHARED_LIBRARIES.

For example, consider the module 'foo' with the following definition:

          include $(CLEAR_VARS)
          LOCAL_MODULE := foo
          LOCAL_SRC_FILES := foo/foo.c
          include $(BUILD_STATIC_LIBRARY)

And another module, named 'bar' that depends on it as:

          include $(CLEAR_VARS)
          LOCAL_MODULE := bar
          LOCAL_SRC_FILES := bar.c
          LOCAL_CFLAGS := -DBAR=2
          include $(BUILD_SHARED_LIBRARY)

Then, the flags '-DFOO=1 -DBAR=2' will be passed to the compiler when building bar.c.

Exported flags are prepended to your module's LOCAL_CFLAGS so you can easily override them. They are also transitive: if 'zoo' depends on 'bar' which depends on 'foo', then 'zoo' will also inherit all flags exported by 'foo'.

Finally, exported flags are not used when building the module that exports them. In the above example, -DFOO=1 would not be passed to the compiler when building foo/foo.c.


Same as LOCAL_EXPORT_CFLAGS, but for C++ flags only.


Same as LOCAL_EXPORT_CFLAGS, but for C include paths. This can be useful if 'bar.c' wants to include headers that are provided by module 'foo'.


Same as LOCAL_EXPORT_CFLAGS, but for linker flags. Note that the imported linker flags will be appended to your module's LOCAL_LDLIBS though, due to the way Unix linkers work.

This is typically useful when module 'foo' is a static library and has code that depends on a system library. LOCAL_EXPORT_LDLIBS can then be used to export the dependency. For example:

          include $(CLEAR_VARS)
          LOCAL_MODULE := foo
          LOCAL_SRC_FILES := foo/foo.c
          LOCAL_EXPORT_LDLIBS := -llog
          include $(BUILD_STATIC_LIBRARY)

          include $(CLEAR_VARS)
          LOCAL_MODULE := bar
          LOCAL_SRC_FILES := bar.c
          include $(BUILD_SHARED_LIBRARY)

There, will be built with a -llog at the end of the linker command to indicate that it depends on the system logging library, because it depends on 'foo'.


Set this variable to 'true' when your module has a very high number of sources and/or dependent static or shared libraries. This forces the build system to use an intermediate list file, and use it with the library archiver or static linker with the @$(listfile) syntax.

This can be useful on Windows, where the command-line only accepts a maximum of 8191 characters, which can be too small for complex projects.

This also impacts the compilation of individual source files, placing nearly all compiler flags inside list files too.

Note that any other value than 'true' will revert to the default behaviour. You can also define APP_SHORT_COMMANDS in your to force this behaviour for all modules in your project.

NOTE: We do not recommend enabling this feature by default, since it makes the build slower.


Set this variable to 'true' when building static libraries. This will generate a 'thin archive', i.e. a library file (e.g. libfoo.a) which doesn't contain object files, but simply file paths to the actual objects that it should normally contain.

This is useful to reduce the size of your build output. The drawback is that such libraries cannot be moved to a different location (all paths inside them are relative).

Valid values are 'true', 'false' or empty. A default value can be set in your through APP_THIN_ARCHIVE.

NOTE: This is ignored for non-static library modules, or prebuilt static library ones.


Define this variable to a shell command that will be used to filter the assembly files from, or generated from, your LOCAL_SRC_FILES.

When it is defined, the following happens:

 - Any C or C++ source file is generated into a temporary assembly
   file (instead of being compiled into an object file).

 - Any temporary assembly file, and any assembly file listed in
   `LOCAL_SRC_FILES` is sent through the `LOCAL_FILTER_ASM` command
   to generate _another_ temporary assembly file.

 - These filtered assembly files are compiled into object file.

In other words, If you have:

          LOCAL_SRC_FILES  := foo.c bar.S
          LOCAL_FILTER_ASM := myasmfilter

        foo.c --1--> $OBJS_DIR/foo.S.original --2--> $OBJS_DIR/foo.S --3--> $OBJS_DIR/foo.o
        bar.S                                 --2--> $OBJS_DIR/bar.S --3--> $OBJS_DIR/bar.o

Were "1" corresponds to the compiler, "2" to the filter, and "3" to the assembler. The filter must be a standalone shell command that takes the name of the input file as its first argument, and the name of the output file as the second one, as in:

          myasmfilter $OBJS_DIR/foo.S.original $OBJS_DIR/foo.S
          myasmfilter bar.S $OBJS_DIR/bar.S
昵    称: