In the documentation for Android NDK, the following statement is present:
The Android.mk file resides in a subdirectory of your project\'s jni/ directory
/
|
+-- jni/
| |
| +-- Android.mk
| +-- [Application.mk] (optionally)
| +-- main.c
|
+-- AndroidManifest.xml
According to a short guide from here: https://developer.android.com/ndk/guides/concepts
- Create an Android.mk file in the jni/ directory of your project
- ... compile your native code using the ndk-build
$ cd
/ / $ /ndk-build
However, there is a way to define custom NDK project structures using ndk-build
with arguments.
/
|
+-- Android.mk
+-- [Application.mk]
+-- main.c
$ cd
$ ndk-build NDK_PROJECT_PATH=. APP_BUILD_SCRIPT=Android.mk
why there is a need for multiple makefiles when Android.mk will contain the definitions for all modules anyway-that could just as well be placed in Application.mk.
Android.mk
is mandatory and Application.mk
is not.Android.mk
contains module definitions and Application.mk
describes architecture, compiler options, and other "global" options.Android.mk
is orthogonal to Application.mk
by design. If Android.mk
contains 3 modules, and Application.mk
2 ABIs (e.g. arm and x86), then NDK will build (3 * 2) artifacts.Application.mk
appears in Android.mk
, so you can use, for instance, APP_ABI
in Android.mk
to link architecture dependent libraries.Application.mk
is a clean way.Application.mk
is just a design decision, it could've been done differently.