I'm compiling a nexus one android kernel from source as found on HTCs developer website. I've obtained an ARM tool chain by DLing the android NDK from the android dev site. I am able to run make clean and make defconfig without incident, but when I run make, it only gets so fare before running into compiler errors.
Currently i see the following:
$MY_DIR/nexus_one/arch/arm/include/asm/glue.h:156:29: error: '#' is not followed by a macro parameter
The offending line is:
1 /*
2 * Instruction Fault Status Register. (New register as of ARMv6)
3 * If processor has IFSR then set value, else set translation fault
4 */
5 #if defined(CONFIG_CPU_ABRT_EV7) || defined(CONFIG_CPU_ABRT_EV6)
6 # define CPU_PABORT_IFSR(reg) mrc p15, 0, reg, cr5, cr0, 1 #asm macro;
7 #else
8 # define CPU_PABORT_IFSR(reg) mov reg, #5 #asm macro;
9 #endif
Specifically, line 8 above is what hoses the compiler. Apparently you can't have that second # sign, but i'm not really sure whats going on in this code, and it looks pretty important so i don't want to touch it.
I'm guessing i'm compiling with the wrong tool chain maybe? Or perhaps i have configured things wrong somehow? Does any one have any idea what this is all about?
thanks,
brian
I strongly recommend you to use CodeSourcery toolchain for Linux for compiling Linux kernel.
CodeSourcery GNU Toolchain for ARM Processors: 2008q1-126 version
Turns out it was nothing to do with the specific toolchain. The # sign needed 'escaping' of some sort. The solution was as follows:
/* this is needed to get the mov reg, 5 macro to compile and function correctly */
#define hash_hackery #
#define f(x) x
#if defined(CONFIG_CPU_ABRT_EV7) || defined(CONFIG_CPU_ABRT_EV6)
# define CPU_PABORT_IFSR(reg) mrc p15, 0, reg, cr5, cr0, 1 #asm macro;
#else
# define CPU_PABORT_IFSR(reg) mov reg, f(hash_hackery)5 #asm macro;
#endif
This post was very informative in finding the answer.
Related
I've been trying to get a Go application running on Android 10 and hit a snag. I'm pretty sure I now know why I'm hitting this snag, but not sure how to fix it. First, simple things work with cross compilation. For example,
package main
import "fmt"
func main() {
fmt.Println("Hello, Android!")
}
will run just fine compiled with GOARM=arm64 GOOS=linux go build
But now if I add some networking code:
package main
import (
"fmt"
"net/http"
"io/ioutil"
)
func main() {
fmt.Println("Hello, Android!")
client := &http.Client{}
resp, err := client.Get("https://www.google.com")
if err != nil {
fmt.Println(err)
return
}
bodyBytes, err := ioutil.ReadAll(resp.Body)
if err != nil {
fmt.Println(err)
return
}
fmt.Printf("%s", bodyBytes)
}
And compile the same way it will run, but I'll get an error like:
Get "https://www.google.com": dial tcp: lookup www.google.com on [::1]:53: read udp [::1]:47618->[::1]:53: read: connection refused
Initially I thought this was some networking config on my Android device, but after lots of messing around I'm pretty sure it has to do with the fact that Android networking is "different". Maybe this is a bionic vs libc thing (?) but at least one other person seems to have hit the same snag and solved by building using the toolchain that comes with the NDK.
Does anyone know how to actually get this to work though?
Initially I was hoping it was as easy as GOOS=android GOARCH=arm64 go build but that fails with:
# command-line-arguments
/usr/lib/golang/pkg/tool/linux_amd64/link: running gcc failed: exit status 1
/usr/bin/ld: /tmp/go-link-384615804/go.o: Relocations in generic ELF (EM: 183)
<bunch of these>
/usr/bin/ld: /tmp/go-link-384615804/go.o: error adding symbols: file in wrong format
collect2: error: ld returned 1 exit status
Some Googling turned up https://github.com/golang/go/issues/30216 - which says you need to explicitly enable CGO. But GOARCH=arm64 GOOS=android CGO_ENABLED=1 go build still fails, this time with:
# runtime/cgo
gcc_android.c:6:10: fatal error: android/log.h: No such file or directory
6 | #include <android/log.h>
| ^~~~~~~~~~~~~~~
compilation terminated.
I messed around with a lot of other combinations and the closest I got was pointing Go to compilers in my NDK installation. Basically something like GOARCH=arm64 GOOS=android CC=$NDKROOT/toolchains/arm-linux-androideabi-4.9/prebuilt/linux-x86_64/bin/arm-linux-androideabi-gcc CGO_ENABLED=1 CGO_CFLAGS="--sysroot $NDKROOT/sysroot -I$NDKROOT/sysroot/usr/include/arm-linux-androideabi/" go build. But this failed with:
# runtime/cgo
In file included from _cgo_export.c:4:0:
cgo-gcc-export-header-prolog:25:14: error: size of array '_check_for_64_bit_pointer_matching_GoInt' is negative
There's something broken in the way I'm pointing to the compiler - for example I won't find any header files (stdlib.h) without specifying sysroot, and even then it won't find the architecture specific includes (asm/types.h) without specifying the additional include path. So the error is probably a mismatch between architecture and compiler I'm just not sure where or how to fix it.
If you've read all this, thanks! Some additional info:
go version: 1.14.13 linux/amd64
NDK version: r16b
I've also looked at things like gomobile but that seems more geared towards building Android applications (like apks) although maybe I'm missing something here.
Figured it out! Well, got my specific example working but I'm not really sure why this works. The answer came basically from https://github.com/golang/go/issues/20755 and you have to:
Build a standalone version of the NDK (the comment in the Python script that does this says it packages it in a more convenient way for other tools to use)
Rebuild Go's standard library using the cross compiler in your NDK
Build your application using the NDK cross compiler.
So for my example:
$ $(NDKROOT)/build/tools/make_standalone_toolchain.py --arch arm64 --api 24 --install-dir $(HOME)/ndk_standalone
$ CC=$(HOME)/ndk_standalone/bin/clang CXX=$(HOME)/ndk_standalone/bin/clang GOOS=android GOARCH=arm64 go install std
$ CC=$(HOME)/ndk_standalone/bin/clang CXX=$(HOME)/ndk_standalone/bin/clang GOOS=android GOARCH=arm64 CGO_ENABLED=1 go build
And networking works!
I'm not really sure why (or if) clang makes a difference as opposed to invoking the compiler directly. Maybe I'll play around with this more. Also, I had to still enabled CGO explicitly when compiling my application (but not for the go std library) otherwise linking failed.
I am building ffmpeg and stuck in an unusual spot. Inside libavutil we have float_dsp.h and float_dsp.c files. Inside these file there is a declaration of a methond which is:
void (*butterflies_float)(float *av_restrict v1, float *av_restrict v2, int len);
float (*scalarproduct_float)(const float *v1, const float *v2, int len);
when building and compiling this file i get this error and have no idea what to do. I think problem is somewhere else but again have no idea at all. Error is
jni/ffmpeg/libavcodec/../libavutil/float_dsp.h:150: error: expected ';', ',' or ')' before 'v1'
jni/ffmpeg/libavcodec/../libavutil/float_dsp.h:161: error: expected ';' before 'float'
Anybody who wants to help please take a step forward because i have not really got any support on ffmpeg during couple of weeks now.
Regards
This may be caused by:
not set GCC to C99 mode where restrict keyword is supported
restrict was redefined by some #define
restrict is not supported for particular architecture but this is unlikely
GCC is set to C++ mode where restrict keyword is not supported GCC according http://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html.
There are other supported forms by GCC like __restrict and __restrict__.
ffmpeg configure script sets macro av_restrict to just restrict which can be checked in produced config.h file.
Fix could be to change configure detection code and remove invalid case restrict:
--- ./configure.orig 2014-01-15 18:53:59.000000000 +0100
+++ ./configure 2014-03-13 17:50:45.754442028 +0100
## -3896,7 +3896,7 ##
EOF
_restrict=
-for restrict_keyword in restrict __restrict__ __restrict; do
+for restrict_keyword in __restrict__ __restrict; do
check_cc <<EOF && _restrict=$restrict_keyword && break
void foo(char * $restrict_keyword p);
EOF
I got error when i tried to compile inline assembly with qadd command.
Error: cannot honor width suffix -- `qadd r7,r7,r1'
I know that qadd is supported in ARMv5TE
These ARM instructions are available in ARMv6 and above, and E
variants of ARMv5T.
C/asm code:
inline int __qadd(int a, int b) {
__asm__ (
"qadd %0, %1, %2" : "=r" (a) : "r" (a), "r" (b));
return a;
}
My cpu features is:
LOGI("__ARM_ARCH__='%d'", __ARM_ARCH__);
LOGI("__ARM_HAVE_5TE='%d'", __ARM_HAVE_5TE);
Output:
__ARM_ARCH__='5'
__ARM_HAVE_5TE='1'
I have next compiler flags:
LOCAL_CFLAGS += -std=c99 -ffast-math -march=armv5te
Besides i have tried replace add instead of qadd - nicely works but with qadd code not compiles.
What i'm doing wrong? Who can provide worked example of qadd command in assembly?
Solution here No qsort_r for Android (or how to disable force Thumb to use CLZ in Android ARM code)
In your Android.mk file, add ".arm" to the filenames and they will get
compiled as ARM mode instead of Thumb mode (e.g. sort.c.arm). I've had
mixed Thumb/ARM code in an Android native library and it worked fine.
Question is closed.
Android ndk told me "undefined reference to pthread_atfork" while I was trying to cross compile libpcsclite for Android
Actually I already knew there is no pthread_atfork function in Bionic library of Android (Android docs), but pthread_atfork must be called in pcsclite library. Is there any solution in this case?
It looks like pthread_atfork() was not added to bionic libc until after android-10/v2.3.6/Gingerbread. If you want to target older versions of Android, you either have to include pthread-atfork.c or the newer pthread_atfork.cpp in your project, or not use pthread_atfork() at all.
-DANDROID_PLATFORM=android-21
this value defaults to the lowest API level supported by the NDK in use
in my Context,use android-9 default.
if you execute shell below on your ${ANDROID-NDK} DIR,
find_path=$1
grep_str=$2
SOS=`find ${find_path} -name *.so`
for item in $SOS
do
echo ======= $item
objdump -Tt $item |grep ${grep_str}
done
sh echo.sh . pthread_atfork
the output:
======= ./platforms/android-9/arch-mips/usr/lib/libc.so
======= ./platforms/android-12/arch-x86/usr/lib/libc.so
00009420 g F .text 0000000a pthread_atfork
00009420 g DF .text 0000000a LIBC pthread_atfork
you will found this message:pthread_atfork is defined on libc.so with android-12 but not android-9.
so when you cross compile with cmake or other tools,you need add -DANDROID_PLATFORM=android-21 [just only gt android-9]
you can try this command with cmake
# on you project_root
mkdir build
cd build
cmake -DCMAKE_TOOLCHAIN_FILE=/cross/github/jni/android-ndk-r14b/build/cmake/android.toolchain.cmake -DANDROID_TOOLCHAIN=clang -DANDROID_ABI=armeabi-v7a -DCMAKE_ANDROID_NDK_TOOLCHAIN_VERSION=clang -DANDROID_PLATFORM=android-21 ..
Look at this: http://comments.gmane.org/gmane.comp.handhelds.android.ndk/15982
I believe function pthread_atfork is implemented in Bionic library (however, it's not documented).
I think all you have to do is to declare it to use it.
Usual caution: If you use undocumented methods, they may not work properly or may be deprecated in the future.
I have a really simple helloworld.cpp program
#include <iostream>
using namespace std;
int main ()
{
cout << "Hello World!";
return 0;
}
And I'm trying to compile it for android x86 with the cross-compiler from the toolchain:
/Users/me/android-ndk-r8/toolchains/x86-4.4.3/prebuilt/darwin-x86/bin/i686-android-linux-g++ helloworld.cpp -L "/Users/me/android-ndk-r8/sources/cxx-stl/stlport/libs/x86/" -lstlport_static
However, I get errors:
helloworld.cpp:2:20: error: iostream: No such file or directory
Any idea why?
Check the documentation.html file included with the NDK, under "Standalone Toolchain". It says that if you invoke the compiler in this way you won't be able to "use any C++ STL". However it is possible, as the documentation explains, if you first create a "customized" toolchain installation, using something like the following command:
$NDK/build/tools/make-standalone-toolchain.sh --platform=android-8 --install-dir=/tmp/my-android-toolchain --arch=x86
where $NDK is the path to your NDK directory. Note the --arch=x86 which means that the toolchain is prepared specifically for the x86 Android. This prepares what you need in one directory, including the STL headers and folders. You should then be able to use -lstdc++ to link against the STL (static version), i.e. something like:
/tmp/my-android-toolchain/bin/i686-android-linux-g++ helloworld.cpp -lstdc++
For a more complete explanation, please see the NDK documentation.
The NDK documentation is not entirely accurate, at least not currently. In fact, it states when using the prebuilt toolchain "You won't be able to use any C++ STL (either STLport or the GNU libstdc++) with it.", but this is out of date. I created a small hello world program using the include with the same error. It can be solved without creating your own toolchain though, which is nice if you don't want to have to add one more step to your configuration process and allows you to always use the latest SDK platform without creating a new toolchain every time.
The NDK ships with the source code for several versions of standard C++ libraries: GAbi++, STLport, and GNU STL. Each flavor comes with prebuilt shared and static libs as well. My example below will use stlport.
To use the stand-alone toolchain at its installed location, you can do something like this:
export CXX='$NDK_ROOT/toolchains/arm-linux-androideabi-4.8/prebuilt/darwin-x86_64/bin/arm-linux-androideabi-g++ --sysroot="$NDK_ROOT/platforms/android-19/arch-arm"'
This, for example, would set your CXX compiler to compile ARM on the OS X system using SDK platform level 19. This much you probably already knew. Also, you would want to export your CC, CPP, LD, AR, and RANLIB if you use it. I also personally create an envar for READELF.
To add support for C++ libs, you could do something like follows:
$CXX helloworld.cpp -I$NDK_ROOT/sources/cxx-stl/stlport/stlport -L$NDK_ROOT/sources/cxx-stl/stlport/libs/armeabi -lstlport_shared
Note that this will link the libstlport_shared.so which will now be required at runtime, so you may need to add a relative path to the command above to support that, depending on your APK structure. If you want to just test this simple case, you can just run this on the emulator as follows:
adb push a.out /data
adb push $NDK_ROOT/sources/cxx-stl/stlport/libs/armeabi/libstlport_shared.so /data
adb shell
# su
# cd /data
# chmod 777 a.out
# ./a.out
To get rid of the headache of dealing with shared library paths, you can also statically link the C++ library in by changing "-lstlport_shared" to "-lstlport_static". There are some consequences of doing this, as explained in the NDK Dev Guide. The biggest issue is due to having the static library linked in multiple places, causing:
- memory allocated in one library, and freed in the other would leak or even corrupt the heap.
- exceptions raised in libfoo.so cannot be caught in libbar.so (and may simply crash the program).
- the buffering of std::cout not working properly
A useful tool is also included to see what dependencies your program has, the readelf tool.
To see what other shared libraries your program requires, you can run the following:
$NDK_ROOT/toolchains/arm-linux-androideabi-4.8/prebuilt/darwin-x86_64/bin/arm-linux-androideabi-readelf -a a.out | grep NEEDED
Other cool standard tools:
addr2line - convert a stack trace address to a line of code
nm - Displays the symbol table
objdump - displays info on an object
i call one of the functions from gnustl after it runs fine from prebuilt aosp arm-linux-androideabi-gcc --std=c++11 and after crashing error i cant get a backtrace from logs or reason, my hope is turning to crossbuilt qemu-linux-user to run host compiled i686 binary on the arm, difficulty is interacting with crosshost libs aapt from adt always crashes on any other than platform specific libs, unless kernel module packaged update is possible...