备份一下手工书写Android JNI工程时mk的模板，以及uses VFP register arguments, output does not问题

Android JNI开发，手工建立MK模板：

Application.mk

APP_ABI := armeabi-v7aAPP_STL := gnustl_staticAPP_CPPFLAGS := -frtti -fexceptions -std=c++11APP_PLATFORM := android-15

Android.mk

LOCAL_PATH := $(call my-dir)#print message#$(warning "the value of LOCAL_PATH is $(LOCAL_PATH)")include $(CLEAR_VARS)LOCAL_MODULE  := native-libLOCAL_C_INCLUDES = $(LOCAL_PATH)/3rdparty/includeLOCAL_SRC_FILES  := native-lib.cpp #\#3rdparty/src/aaidsys.c \#3rdparty/src/queue.c \#3rdparty/src/video.c \#3rdparty/src/ysqfd.c \#3rdparty/src/yuvproc.c#LOCAL_CFLAGS := -mfloat-abi=hard -D_NDK_MATH_NO_SOFTFP=1LOCAL_CPPFLAGS := -s -mfloat-abi=softfp -mfpu=neon -D_NDK_MATH_NO_SOFTFP=1####include the 3rdparty(method1)#####LOCAL_LDLIBS += -llog -L$(LOCAL_PATH)/3rdparty/lib/ -lysqfd32####include the 3rdparty(method2)#####LOCAL_SHARED_LIBRARIES := ysqfd32include $(BUILD_SHARED_LIBRARY)####the 3rdparty####include $(CLEAR_VARS)#$(warning "the value of LOCAL_PATH is $(LOCAL_PATH)") LOCAL_MODULE := ysqfd32LOCAL_SRC_FILES  := 3rdparty/lib/libysqfd32.soinclude $(PREBUILT_SHARED_LIBRARY)

本来想用一下，于仕琪博士的人脸检测库，无奈怎么也链接不上库~~~~

D:/Android/android-sdk-windows/ndk-bundle/toolchains/arm-linux-androideabi-4.9/prebuilt/windows-x86_64/lib/gcc/arm-linux-androideabi/4.9.x/../../../../arm-linux-androideabi/bin\ld: error: ../../../../src/main/jniLibs/armeabi-v7a/libysqfd32.so uses VFP register arguments, output does not

网搜所有的答案，都不能解决问题，只好求助NDK社区（https://github.com/android-ndk/ndk/），得到答复：

https://github.com/android-ndk/ndk/issues/552#event-1308445210

Like I said in the other bug, you don't. Android does not use the hard float calling convention. You do not need -mfloat-abi=hard to use floating point instructions. It only affects the calling convention.

就是根本不支持啊，开始的时候我也纳闷，怎么就没有一个宏能够改变默认的浮点数使用方式？

虽然得到google人员的肯定回复，但还是有个疑问：NDK编译工具下的hard是干什么用的？

toolchains\arm-linux-androideabi-4.9\prebuilt\windows-x86_64\lib\gcc\arm-linux-androideabi\4.9.x\armv7-a\hard\*

这些库难道不是为了支持hard float？

还是另找出路：arm-linux-gnueabihf 貌似可以解决我的问题，待实践~~~~

找到原因了，Hard VFP是不支持NDK交叉编译工具的，只能使用板载系统工具本地编译，不过这样arm-linux-gnueabihf又有什么用呢？？？

其实是支持交叉编译的，使用 arm-linux-gnueabihf 编译后的程序里面有

Tag_ABI_VFP_args: VFP registers

普通的交叉编译工具是没有这个的！！！

How to Detect if an ARM ELF Binary is Hard-float (ARMHF) or Soft-float (ARMEL)

https://www.cnx-software.com/2013/04/22/how-to-detect-if-an-arm-elf-binary-is-hard-float-armhf-or-soft-float-armel/

以下是一个小程序：

@echo offset PATH=F:\gcc-linaro-arm-linux-gnueabihf-4.8-2013.10_win32\bin;D:\Android\android-sdk-windows\platform-tools;%PATH%echo "mfloat compiling......"arm-linux-gnueabihf-gcc.exe -v -static mfloat.c -o demorem arm-linux-gnueabihf-gcc.exe -v -ffast-math mfloat.c -o demopauserem arm-linux-gnueabihf-readelf.exe -a demopauserem arm-linux-gnueabihf-readelf.exe -a libysqfd32.sopauseadb push demo /data/local/tmp/adb shell chmod 777 /data/local/tmp/demoadb shell ls -l /data/local/tmp/adb shell /data/local/tmp/demopause

mfloat.c

#include <stdio.h>int main(void){double a,b,c;a = 23.543;b = 323.234;c = b/a;printf("the 13/2 = %f\n", c);printf("hello world !\n");return 0;}