building the android beagle toolchain -- getting around the issues
Robert P. J. Day
a number of observations regarding building the android toolchain if
anyone wanted to try that. at the moment, it doesn't build according
to what's in the README, and here's why, and here's how to fix it.
(this will, not surprisingly, go up on my wiki in short order.)
if you want the official embinux toolchain for android for the
beagle, you can go here:
http://www.embinux.com/download_toolchain.php
where what you get is a tarball from which you can allegedly *build*
the toolchain. sadly, there are issues, and here's how to deal with
them.
the README instructs one to use the configuration option
"--target=arm-eabi". however, the configure file itself contains this
snippet regarding gdb:
=====
# We only allow the following 3 targets:
# 1. arm-android-eabi (primary target)
# 2. arm-eabi (to be phased out.)
# 3. arm-newlib-eabi (for gcc testing)
# 4.i[3456]86-*-linux-gnu, x86-*-linux-gnu (for x86 targets)
GDB_TARGET=$target
case $target in
arm-android-eabi | arm-unknown-eabi)
BUILD_ANDROID_GCC=yes
GDB_TARGET=arm-elf-linux
;;
arm-newlib-elf)
BUILD_ANDROID_GCC=no
;;
i[3456]86-*-linux-gnu | x86_64-*-linux-gnu)
BUILD_ANDROID_GCC=yes
;;
*)
{ { echo "$as_me:$LINENO: error: Unsupported target $target." >&5
echo "$as_me: error: Unsupported target $target." >&2;}
{ (exit 1); exit 1; }; };
esac
=====
which suggests that the primary target is now "arm-android-eabi",
but also that the comments don't match the case statement --
"arm-newlib-eabi" versus "arm-newlib-elf". i'm just going to go with
"arm-android-eabi", and somebody should tweak that configure file just
a bit so that the comments match the code. :-) onward.
if you go with the default binutils of 2.17 and gdb of 6.6, the
build will fail with this type of error for both of them:
http://www.mail-archive.com/bug-bi...@gnu.org/msg04261.html
because the older scripts think that a makeinfo version of, say, 4.13,
is *older* than a version of 4.4. newer versions of the software fix
this so the fix is to upgrade to newer binutils and gdb source. the
README explains how to configure a newer software component:
=====
For any --with-XXX-version=YYY, a sub-directory called XXX-YYY must be
present in the source level directory. For example,
--with-gcc-version=4.3.1
requires that gcc-4.3.1-android to be present and contain the source
code of a buildable gcc version.
=====
the README is wrong. let's RTFS (the "configure" file):
=====
macro to check that we have a package. A package x exists if any one
# of the following exists:
# 1. x/
# 2. tarballs/x.tar.gz
# 3. tarballs/x.tgz
# 4. tarballs/x.tar.bz2
=====
so you need to have the new source package in the source directory
either as a straight directory (binutils-2.19.1) or as a tar ball
underneath a "tarballs" directory. take your pick. i went with a
symlink to an extensive source tarball directory of mine.
finally, the build of gdb-6.8 will fail thusly:
=====
cc1: warnings being treated as errors
/home/rpjday/beagle/android/toolchain_build/temp-src/gdb-6.8/gdb/eval.c:
In function ‘evaluate_subexp_standard’:
/home/rpjday/beagle/android/toolchain_build/temp-src/gdb-6.8/gdb/eval.c:1705:
error: ‘subscript_array’ may be used uninitialized in this function
/home/rpjday/beagle/android/toolchain_build/temp-src/gdb-6.8/gdb/eval.c:1644:
note: ‘subscript_array’ was declared here
make[3]: *** [eval.o] Error 1
make[3]: Leaving directory
`/home/rpjday/beagle/android/toolchain_build/gdb-6.8/gdb'
make[2]: *** [all-gdb] Error 2
make[2]: Leaving directory
`/home/rpjday/beagle/android/toolchain_build/gdb-6.8'
make[1]: *** [all] Error 2
make[1]: Leaving directory
`/home/rpjday/beagle/android/toolchain_build/gdb-6.8'
make: *** [stmp-build-target-gdb] Error 2
[rpjday@localhost toolchain_build]$ makeinfo --version
makeinfo (GNU texinfo) 4.13
=====
which requires that you initially configure with "--disable-werror" so
that simple warnings don't blow you out of your build. and that lets
the build complete. in short:
$ .../configure \
--target=arm-android-eabi \
--with-binutils-version=2.19.1 \
--with-gdb-version=6.8 \
--disable-werror
builds successfully. haven't tested it yet, but at least it builds.
rday
--
========================================================================
Robert P. J. Day Waterloo, Ontario, CANADA
Linux Consulting, Training and Annoying Kernel Pedantry.
Web page: http://crashcourse.ca
Linked In: http://www.linkedin.com/in/rpjday
Twitter: http://twitter.com/rpjday
========================================================================
Rupesh Gujare
The toolchain you are mentioning is not from embinux. looking at source tar ball name (android-toolchain-20081019.tar.bz2 ), it seems that you have downloaded it from following link.
http://android.git.kernel.org/pub/android-toolchain-20081019.tar.bz2
I will like to clarify that embinux had never released a toolchain for android. And the one released at following link is not meant to compile android build.
http://www.embinux.com/download_toolchain.php
Regards,
Rupesh Gujare
Robert P. J. Day
> Robert,
>
> The toolchain you are mentioning is not from embinux. looking
> at source tar ball name (android-toolchain-20081019.tar.bz2 ), it
> seems that you have downloaded it from following link.
>
> http://android.git.kernel.org/pub/android-toolchain-20081019.tar.bz2
>
> I will like to clarify that embinux had never released a toolchain
> for android. And the one released at following link is not meant to
> compile android build.
>
> http://www.embinux.com/download_toolchain.php
ok, i stand corrected, but it's certainly confusing to have that
link at embinux's home page, immediately underneath a link labelled
"Android on Beagle".
Robert P. J. Day
> Robert,
>
> The toolchain you are mentioning is not from embinux. looking
> at source tar ball name (android-toolchain-20081019.tar.bz2 ), it
> seems that you have downloaded it from following link.
>
> http://android.git.kernel.org/pub/android-toolchain-20081019.tar.bz2
>
> I will like to clarify that embinux had never released a toolchain
> for android. And the one released at following link is not meant to
> compile android build.
>
> http://www.embinux.com/download_toolchain.php
ok, i stand corrected, but it's certainly confusing to have that
link at embinux's home page, immediately underneath a link labelled
"Android on Beagle".
rday
Rupesh Gujare
If you download toolchain from embinux it gives you following tar
ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
It is better to acknowledge your mistakes, rather than pointing at
others, and what others should do.
Regards,
Rupesh Gujare
Robert P. J. Day wrote:
>
>
> ok, i stand corrected, but it's certainly confusing to have that
> link at embinux's home page, immediately underneath a link labelled
> "Android on Beagle".
>
> rday
> --
>
>
>
--
Rupesh Gujare
Robert P. J. Day
>
> Robert,
>
> If you download toolchain from embinux it gives you following
> tar ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
>
> While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
>
> It is better to acknowledge your mistakes, rather than pointing at
> others, and what others should do.
ok, if i messed that up, i apologize, but i have no idea where that
other toolchain could have come from. i'll try again.
Robert P. J. Day
>
> Robert,
>
> If you download toolchain from embinux it gives you following
> tar ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
>
> While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
>
> It is better to acknowledge your mistakes, rather than pointing at
> others, and what others should do.
ok, if i messed that up, i apologize, but i have no idea where that
other toolchain could have come from. i'll try again.
rday
Robert P. J. Day
>
> Robert,
>
> If you download toolchain from embinux it gives you following tar
> ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
>
> While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
>
> It is better to acknowledge your mistakes, rather than pointing at
> others, and what others should do.
you're right, i apologize. i'm absolutely baffled as to where that
other toolchain came from. my mistake.
melwyn
I am trying porting Cortex patches from code sourcery to Android
toolchain. I did that and am now compiling with top Makefile.in
patches as follows:
-----------------------------------------------------------------------------
- GCC_CONFIG_ARGS += --with-float=soft --with-fpu=vfp --with-
arch=armv5te \
--enable-target-optspace --with-abi=aapcs
+ GCC_CONFIG_ARGS += --with-float=soft --with-fpu=vfp
+ --with-arch=armv7-a --with-mode=thumb --disable-multilib\
--enable-target-optspace --with-abi=aapcs
I am able to compile the toolchain successfully. Here is the verbose
"configured with" output of the compiled binary
root@melwyn-desktop:/home/melwyn/myandroid# prebuilt/linux-x86/
toolchain/arm-eabi-4.2.1/bin/arm-eabi-gcc -v
Using built-in specs.
Target: arm-eabi
Configured with: /home/melwyn/toolchain/android-ndk-toolchain-20090323/
gcc-4.3.2/configure --prefix=/home/melwyn/toolchain/
toolchain_install_ndk/ --target=arm-eabi --host=i686-pc-linux-gnu --
build=i686-pc-linux-gnu --enable-languages=c,c++ --with-gmp=/home/
melwyn/toolchain/toolchain_objs_ndk/temp-install --with-mpfr=/home/
melwyn/toolchain/toolchain_objs_ndk/temp-install --disable-libssp --
enable-threads --disable-nls --disable-libmudflap --disable-libgomp --
disable-libstdc__-v3 --disable-sjlj-exceptions --disable-shared --with-
float=soft --with-fpu=vfp --with-arch=armv7-a --with-mode=thumb --
enable-target-optspace --with-abi=aapcs --disable-multilib --with-gcc-
version=4.3.2 --with-binutils-version=2.18.50.0.6 --prefix=/home/
melwyn/toolchain/toolchain_install_ndk/ --program-transform-
name='s,^,arm-eabi-,'
Thread model: single
gcc version 4.3.2 (GCC)
However when I compile the Android package with the thus compiled
toolchain I get the following error:
**********************************************************************************
target thumb C++: libUAPI_jni <= external/srec/uapi/java/jniapi/jniapi/
JNIHelper.cpp
prebuilt/linux-x86/toolchain/arm-eabi-4.2.1/bin/arm-eabi-g++ -I
system/core/include -I hardware/libhardware/include -I hardware/
ril/include -I dalvik/libnativehelper/include -I frameworks/base/
include -I external/skia/include -I out/target/product/generic/obj/
include -I bionic/libc/arch-arm/include -I bionic/libc/include -
I bionic/libstdc++/include -I bionic/libc/kernel/common -I bionic/
libc/kernel/arch-arm -I bionic/libm/include -I bionic/libm/include/
arch/arm -I bionic/libthread_db/include -I external/srec/uapi/java/
jniapi/jniapi -I external/srec/uapi/java/jniapi/jniapi/../../../cpp/
api/include -I external/srec/uapi/java/jniapi/jniapi/../../../cpp/
grammar/include -I external/srec/uapi/java/jniapi/jniapi/../../../
cpp/recognizer/srec/include -I external/srec/uapi/java/jniapi/
jniapi/../../../cpp/utilities/include -I dalvik/libnativehelper/
include/nativehelper -I external/srec/uapi/java/jniapi/jniapi -I
out/target/product/generic/obj/SHARED_LIBRARIES/
libUAPI_jni_intermediates -c -fno-exceptions -Wno-multichar -msoft-
float -fpic -ffunction-sections -funwind-tables -fstack-protector -
march=armv7-a -mthumb -D__ARM_ARCH_7__ -include system/core/include/
arch/linux-arm/AndroidConfig.h -I system/core/include/arch/linux-arm/ -
mthumb-interwork -DANDROID -fmessage-length=0 -W -Wall -Wno-unused -
DSK_RELEASE -DNDEBUG -O2 -g -Wstrict-aliasing=2 -finline-functions -
fno-inline-functions-called-once -fgcse-after-reload -frerun-cse-after-
loop -frename-registers -DNDEBUG -UDEBUG -fvisibility-inlines-hidden
-mthumb -Os -fomit-frame-pointer -fno-strict-aliasing -finline-
limit=64 -DJNI_EXPORTS -DOS_UNIX=8 -DOS_EMBEDDED=16 -DOS_VXWORKS=32 -
DOS_PSOS=64 -DOS_QNX=1024 -DOS=8 -DHOST_OS=8 -DTARGET_OS=8 -
DCPU_PPC=16 -DCPU_SPARC=32 -DCPU_ARM=64 -DCPU_STRONGARM=128 -
DCPU_TMS320X=256 -DCPU_SH3=512 -DCPU_SH4=1024 -DCPU=64 -DHOST_CPU=1 -
DTARGET_CPU=64 -DC_GNU=64 -DC_SUNPRO=128 -DC_METROWERKS=512 -
DC_TICXC=2048 -DC_ARM=4096 -DCOMPILER=64 -DBUILD_SHIP=1 -
DBUILD_DEBUG=8 -DBUILD=1 -DUAPI_LINUX -DUAPI_MT -
DAUDIOIN_SUPPORT_CALLBACK -DUSE_CUSTOM_RECURSIVE_MUTEX -
DNO_ATTACHCURRENTTHREAD_CAST_VOIDPP -DPOSIX -D__linux -DNDEBUG -
D_SHIP -DFINAL_RELEASE -D_OPTIMIZE -fno-rtti -MD -o out/target/
product/generic/obj/SHARED_LIBRARIES/libUAPI_jni_intermediates/
JNIHelper.o external/srec/uapi/java/jniapi/jniapi/JNIHelper.cpp
external/srec/uapi/java/jniapi/jniapi/JNIHelper.cpp: In static member
function 'static android::speech::recognition::Logger::LogLevel
android::speech::recognition::jni::JNIHelper::toNativeLogLevel
(JNIEnv*, _jobject*)':
external/srec/uapi/java/jniapi/jniapi/JNIHelper.cpp:168: internal
compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make: *** [out/target/product/generic/obj/SHARED_LIBRARIES/
libUAPI_jni_intermediates/JNIHelper.o] Error 1
************************************************************************************
ie a segmentation fault. I am lost could anyone help me out.
Melwyn.
Rupesh Gujare
This might be of interest to you.
http://labs.embinux.org/index.php/Toolchain_For_Android
Regards,
-- Rupesh Gujare http://labs.embinux.org http://embinux.com
melwyn
I think I prefer to compile the toolchain rather than using a prebuilt
one.
I got over the above issue by using --disable-multilib option
However when I use this compiler for compiling the Android package I
get the following error:
******************************************************************************************************************************************
root@melwyn-desktop:/home/melwyn/myandroid# prebuilt/linux-x86/
hardware/libhardware/include -I hardware/ril/include -I dalvik/
libnativehelper/include -I frameworks/base/include -I external/
skia/include -I out/target/product/generic/obj/include -I bionic/
libc/arch-arm/include -I bionic/libc/include -I bionic/libstdc++/
include -I bionic/libc/kernel/common -I bionic/libc/kernel/arch-
uapi/java/jniapi/jniapi/JNIHelper.cpp
function 'static android::speech::recognition::Logger::LogLevel
android::speech::recognition::jni::JNIHelper::toNativeLogLevel
(JNIEnv*, _jobject*)':
external/srec/uapi/java/jniapi/jniapi/JNIHelper.cpp:168: internal
compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
On Jun 11, 4:50 pm, Rupesh Gujare <rupesh.guj...@embinux.com> wrote:
> Melwyn,
>
> This might be of interest to you.
>
> http://labs.embinux.org/index.php/Toolchain_For_Android
>
> Regards,
>
> --
>
>
> melwyn wrote:
> > Hi all,
> > I am trying porting Cortex patches from code sourcery to Android
> > toolchain. I did that and am now compiling with top Makefile.in
> > patches as follows:
>
> > ie a segmentation fault. I am lost could anyone help me out.
>
> > Melwyn.
>
> > On May 20, 7:04 pm, "Robert P. J. Day" <rpj...@crashcourse.ca> wrote:
>
> >> On Wed, 20 May 2009, Rupesh Gujare wrote:
>
> >>> Robert,
>
> >>> If you download toolchain from embinux it gives you following tar
> >>> ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
>
> >>> While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
>
> >>> It is better to acknowledge your mistakes, rather than pointing at
> >>> others, and what others should do.
>
> >> you're right, i apologize. i'm absolutely baffled as to where that
> >> other toolchain came from. my mistake.
>
> >> rday
> >> --
>
> >> ========================================================================
> >> Robert P. J. Day Waterloo, Ontario, CANADA
>
> >> Linux Consulting, Training and Annoying Kernel Pedantry.
>
> >> Web page: http://crashcourse.ca
> >> Linked In: http://www.linkedin.com/in/rpjday
> >> Twitter: http://twitter.com/rpjday
>
> - Show quoted text -
melwyn
labs.embinux.org/index.php/Toolchain_For_Android"
I could use that as a reference.
Regards,
Melwyn
>
> This might be of interest to you.
>
> http://labs.embinux.org/index.php/Toolchain_For_Android
>
> Regards,
>
> --
>
>
> melwyn wrote:
> > Hi all,
> > I am trying porting Cortex patches from code sourcery to Android
> > toolchain. I did that and am now compiling with top Makefile.in
> > patches as follows:
>
> > ie a segmentation fault. I am lost could anyone help me out.
>
> > Melwyn.
>
> > On May 20, 7:04 pm, "Robert P. J. Day" <rpj...@crashcourse.ca> wrote:
>
> >> On Wed, 20 May 2009, Rupesh Gujare wrote:
>
> >>> Robert,
>
> >>> If you download toolchain from embinux it gives you following tar
> >>> ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
>
> >>> While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
>
> >>> It is better to acknowledge your mistakes, rather than pointing at
> >>> others, and what others should do.
>
> >> you're right, i apologize. i'm absolutely baffled as to where that
> >> other toolchain came from. my mistake.
>
> >> rday
> >> --
>
> >> ========================================================================
> >> Robert P. J. Day Waterloo, Ontario, CANADA
>
> >> Linux Consulting, Training and Annoying Kernel Pedantry.
>
> >> Web page: http://crashcourse.ca
> >> Linked In: http://www.linkedin.com/in/rpjday
> >> Twitter: http://twitter.com/rpjday
melwyn
While using your toolchain binaries (after following the README) I get
the following error:
prebuilt/linux-x86/toolchain/arm-eabi-4.4.0/bin/arm-eabi-gcc -I
ril/include -I dalvik/libnativehelper/include -I frameworks/base/
include -I external/skia/include -I out/target/product/generic/obj/
include -I bionic/libc/arch-arm/include -I bionic/libc/include -
I bionic/libstdc++/include -I bionic/libc/kernel/common -I bionic/
libc/kernel/arch-arm -I bionic/libm/include -I bionic/libm/include/
test/AudioHardwareRecordLoop/include -I external/srec/audio/AudioIn/
UNIX/include -I external/srec/audio/test/AudioHardwareRecordLoop -
I out/target/product/generic/obj/EXECUTABLES/
AudioHardwareRecordLoop_intermediates -c -fno-exceptions -Wno-
include/arch/linux-arm/AndroidConfig.h -I system/core/include/arch/
linux-arm/ -mthumb-interwork -DANDROID -fmessage-length=0 -W -Wall -
Wno-unused -DSK_RELEASE -DNDEBUG -O2 -g -Wstrict-aliasing=2 -finline-
functions -fno-inline-functions-called-once -fgcse-after-reload -
frerun-cse-after-loop -frename-registers -DNDEBUG -UDEBUG -mthumb -Os -
fomit-frame-pointer -fno-strict-aliasing -finline-limit=64 -MD -o
out/target/product/generic/obj/EXECUTABLES/
AudioHardwareRecordLoop_intermediates/src/AudioHardwareRecordLoop.o
external/srec/audio/test/AudioHardwareRecordLoop/src/
AudioHardwareRecordLoop.c
/home/melwyn/myandroid/prebuilt/linux-x86/toolchain/arm-eabi-4.4.0/
bin/../libexec/gcc/arm-eabi/4.4.0/cc1: error while loading shared
libraries: libmpfr.so.1: cannot open shared object file: No such file
or directory
>
> This might be of interest to you.
>
> http://labs.embinux.org/index.php/Toolchain_For_Android
>
> Regards,
>
> --
>
>
> melwyn wrote:
> > Hi all,
> > I am trying porting Cortex patches from code sourcery to Android
> > toolchain. I did that and am now compiling with top Makefile.in
> > patches as follows:
>
> > ie a segmentation fault. I am lost could anyone help me out.
>
> > Melwyn.
>
> > On May 20, 7:04 pm, "Robert P. J. Day" <rpj...@crashcourse.ca> wrote:
>
> >> On Wed, 20 May 2009, Rupesh Gujare wrote:
>
> >>> Robert,
>
> >>> If you download toolchain from embinux it gives you following tar
> >>> ball. "arm-none-linux-gnueabi-cortex-core-v2.tar.bz2"
>
> >>> While the one you have mentioned is "android-toolchain-20081019.tar.bz2"
>
> >>> It is better to acknowledge your mistakes, rather than pointing at
> >>> others, and what others should do.
>
> >> you're right, i apologize. i'm absolutely baffled as to where that
> >> other toolchain came from. my mistake.
>
> >> rday
> >> --
>
> >> ========================================================================
> >> Robert P. J. Day Waterloo, Ontario, CANADA
>
> >> Linux Consulting, Training and Annoying Kernel Pedantry.
>
> >> Web page: http://crashcourse.ca
> >> Linked In: http://www.linkedin.com/in/rpjday
> >> Twitter: http://twitter.com/rpjday
Anish N
>
> Rupesh,
> While using your toolchain binaries (after following the README) I get
> the following error:
>
> /home/melwyn/myandroid/prebuilt/linux-x86/toolchain/arm-eabi-4.4.0/
> bin/../libexec/gcc/arm-eabi/4.4.0/cc1: error while loading shared
> libraries: libmpfr.so.1: cannot open shared object file: No such file
> or directory
Seems like you don't have MPFR library(http://www.mpfr.org) on your
host machine, please install it and give a try.
On ubuntu you could do,
$ sudo apt-get install libmpfr1ldbl
-Thanks,
Anish
melwyn
I get this error on a newly built toolchain. Could you tell what this
could be?
external/srec/uapi/java/jniapi/jniapi/JNIHelper.cpp:168: internal
compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make: *** [out/target/product/generic/obj/SHARED_LIBRARIES/
libUAPI_jni_intermediates/JNIHelper.o] Error 1
Anish N
>
> I get this error on a newly built toolchain. Could you tell what this
> could be?
>
> external/srec/uapi/java/jniapi/jniapi/JNIHelper.cpp:168: internal
> compiler error: Segmentation fault
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <http://gcc.gnu.org/bugs.html> for instructions.
> make: *** [out/target/product/generic/obj/SHARED_LIBRARIES/
> libUAPI_jni_intermediates/JNIHelper.o] Error 1
Hi,
Think these kinds of errors are encountered when you mix the
optimization flags passed to the compiler in a way that is not
supported.
You could try building in the verbose mode with
#make showcommands
and send the line where the fault occurs.
-Thanks,
Anish
Partha
It seems to be new error from the gcc , can you please send some more
detail log of the error.
Regards
Partha