FAQ
After a full / clean / rebuild of everything, compile in thumb-mode armv7a
only, my app is now crashing with this message ( taken from ndk-stack )

********** Crash dump: **********
Build fingerprint:
'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
Stack frame #00 pc 00864f14
/mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
Unable to locate routine information for address 864f14 in module
obj/local/armeabi-v7a//libPinballMania.so


Any idea ?

--
You received this message because you are subscribed to the Google Groups "android-ndk" group.
To post to this group, send email to android-ndk@googlegroups.com.
To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.

Search Discussions

  • Olivier Chatry at Feb 17, 2012 at 5:39 pm
    Tried with another of my lib, baked on linux, clean build got that :
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8 /data/data/com.myVR.mMap.demo/lib/libmMap.so
    (powf): Unable to locate routine information for address 122de8 in module
    obj/local/armeabi-v7a//libmMap.so

    Guess the there is a problem with libmath :(
    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry wrote:

    After a full / clean / rebuild of everything, compile in thumb-mode armv7a
    only, my app is now crashing with this message ( taken from ndk-stack )

    ********** Crash dump: **********
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14
    /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
    Unable to locate routine information for address 864f14 in module
    obj/local/armeabi-v7a//libPinballMania.so


    Any idea ?
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Olivier Chatry at Feb 17, 2012 at 5:41 pm
    (very wild guess : I'm testing on a TEGRA chipset, is there any way that
    the libmath ( or whatever ) is using NEON ? It works find if I switch back
    to r7.
    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry wrote:

    Tried with another of my lib, baked on linux, clean build got that :
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8 /data/data/com.myVR.mMap.demo/lib/libmMap.so
    (powf): Unable to locate routine information for address 122de8 in module
    obj/local/armeabi-v7a//libmMap.so

    Guess the there is a problem with libmath :(
    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry wrote:

    After a full / clean / rebuild of everything, compile in thumb-mode
    armv7a only, my app is now crashing with this message ( taken from
    ndk-stack )

    ********** Crash dump: **********
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14
    /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
    Unable to locate routine information for address 864f14 in module
    obj/local/armeabi-v7a//libPinballMania.so


    Any idea ?
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • David Turner at Feb 17, 2012 at 6:55 pm
    The math library doesn't use NEON. What's surprising is that there seems to
    be a copy of cosf / powf inside your own shared library.
    How do you build this exactly?

    On Fri, Feb 17, 2012 at 9:41 AM, Olivier Chatry (very wild guess : I'm testing on a TEGRA chipset, is there any way that
    the libmath ( or whatever ) is using NEON ? It works find if I switch back
    to r7.

    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry wrote:

    Tried with another of my lib, baked on linux, clean build got that :
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8
    /data/data/com.myVR.mMap.demo/lib/libmMap.so (powf): Unable to locate
    routine information for address 122de8 in module
    obj/local/armeabi-v7a//libmMap.so

    Guess the there is a problem with libmath :(

    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry <olivier.chatry@gmail.com
    wrote:
    After a full / clean / rebuild of everything, compile in thumb-mode
    armv7a only, my app is now crashing with this message ( taken from
    ndk-stack )

    ********** Crash dump: **********
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14
    /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
    Unable to locate routine information for address 864f14 in module
    obj/local/armeabi-v7a//libPinballMania.so


    Any idea ?
    --
    You received this message because you are subscribed to the Google Groups
    "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at
    http://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Olivier Chatry at Feb 17, 2012 at 7:28 pm
    These are Android.mk and Application.mk. but these project are really
    different, using different rendering engine. Everything work very fine with
    r7, but crash with r7b.

    'm really not doing anything fancy.
    On Fri, Feb 17, 2012 at 7:55 PM, David Turner wrote:

    The math library doesn't use NEON. What's surprising is that there seems
    to be a copy of cosf / powf inside your own shared library.
    How do you build this exactly?
    On Fri, Feb 17, 2012 at 9:41 AM, Olivier Chatry wrote:

    (very wild guess : I'm testing on a TEGRA chipset, is there any way that
    the libmath ( or whatever ) is using NEON ? It works find if I switch back
    to r7.


    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry <olivier.chatry@gmail.com
    wrote:
    Tried with another of my lib, baked on linux, clean build got that :
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8
    /data/data/com.myVR.mMap.demo/lib/libmMap.so (powf): Unable to locate
    routine information for address 122de8 in module
    obj/local/armeabi-v7a//libmMap.so

    Guess the there is a problem with libmath :(

    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry <
    olivier.chatry@gmail.com> wrote:
    After a full / clean / rebuild of everything, compile in thumb-mode
    armv7a only, my app is now crashing with this message ( taken from
    ndk-stack )

    ********** Crash dump: **********
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14
    /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
    Unable to locate routine information for address 864f14 in module
    obj/local/armeabi-v7a//libPinballMania.so


    Any idea ?
    --
    You received this message because you are subscribed to the Google Groups
    "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at
    http://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups
    "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at
    http://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Olivier Chatry at Feb 18, 2012 at 8:15 pm
    Still crashing on TEGRA, but working on a HTC desire :(

    --
    Olivier Chatry
    Sent with Sparrow (http://www.sparrowmailapp.com)

    On Friday, February 17, 2012 at 8:28 PM, Olivier Chatry wrote:

    These are Android.mk and Application.mk. but these project are really different, using different rendering engine. Everything work very fine with r7, but crash with r7b.

    'm really not doing anything fancy.
    On Fri, Feb 17, 2012 at 7:55 PM, David Turner (mailto:digit@android.com)> wrote:
    The math library doesn't use NEON. What's surprising is that there seems to be a copy of cosf / powf inside your own shared library.
    How do you build this exactly?
    On Fri, Feb 17, 2012 at 9:41 AM, Olivier Chatry (mailto:olivier.chatry@gmail.com)> wrote:
    (very wild guess : I'm testing on a TEGRA chipset, is there any way that the libmath ( or whatever ) is using NEON ? It works find if I switch back to r7.

    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry (mailto:olivier.chatry@gmail.com)> wrote:
    Tried with another of my lib, baked on linux, clean build got that :
    Build fingerprint: 'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'

    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8 /data/data/com.myVR.mMap.demo/lib/libmMap.so (powf): Unable to locate routine information for address 122de8 in module obj/local/armeabi-v7a//libmMap.so

    Guess the there is a problem with libmath :(
    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry (mailto:olivier.chatry@gmail.com)> wrote:
    After a full / clean / rebuild of everything, compile in thumb-mode armv7a only, my app is now crashing with this message ( taken from ndk-stack )

    ********** Crash dump: **********
    Build fingerprint: 'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14 /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf): Unable to locate routine information for address 864f14 in module obj/local/armeabi-v7a//libPinballMania.so


    Any idea ?
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com (mailto:android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com (mailto:android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com (mailto:android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com (mailto:android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.

    Attachments:
    - Android.mk

    - Application.mk

    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Michael at Feb 27, 2012 at 4:33 pm
    We are experiencing the same issue and were able to build a minimal
    example (based on hello-jni) to reproduce the problem.
    Our example project can be downloaded here: http://bit.ly/yaHBEL and
    then installed with, e.g.
    android update project -p .
    ndk-build -B NDK_DEBUG=1
    ant all clean debug install
    The relevant content of the project is listed below.
    Note that the crash only occurs on TEGRA devices and when the native
    code is compiled with NDK r7b (everything works with r7).

    The immediate trigger of the SIGILL crash seems to be the use of a
    cmath function on a volatile floating point variable. Making the
    variables non-volatile or substituting <cmath> by <math.h> will fix
    the issue in this particular example.

    The below code will produce a crash, but there seem to be other not-so-
    clear ways to make it crash, too. In our production code, for example,
    we are not using 'volatile', yet it crashes with
    Stack frame #00 pc 00002db4 /data/data/com.X/lib/libX.so (sinf):
    Unable to locate routine information for address 2db4 in module ./obj/
    local/armeabi-v7a/libX.so
    As a result, we currently don't have a workaround for the issue,
    besides reverting to use of NDK r7.

    I hope the provided example will help to identify and fix the problem.

    Best,
    Michael

    -----------------

    Android.mk
    ----------

    LOCAL_PATH := $(call my-dir)

    include $(CLEAR_VARS)
    LOCAL_MODULE := hello-jni
    LOCAL_SRC_FILES := hello-jni.cpp
    include $(BUILD_SHARED_LIBRARY)

    Application.mk
    --------------

    APP_ABI := armeabi armeabi-v7a
    APP_OPTIM := release
    APP_STL := gnustl_static

    hello-jni.cpp
    -------------

    #include <string.h>
    #include <jni.h>
    #include <cmath>

    extern "C"
    {
    jstring
    Java_com_example_hellojni_HelloJni_stringFromJNI( JNIEnv* env,
    jobject thiz )
    {
    volatile float sinresult = std::sin(1.0f);
    volatile float cosresult = std::sin(sinresult);
    volatile float powresult = std::pow(cosresult, 2);

    if (powresult < 0)
    return env->NewStringUTF("RESULT 0");
    else
    return env->NewStringUTF("RESULT 1");
    }
    }

    On Feb 18, 9:14 pm, Olivier Chatry wrote:
    Still crashing on TEGRA, but working on a HTC desire :(

    --
    Olivier Chatry
    Sent with Sparrow (http://www.sparrowmailapp.com)






    On Friday, February 17, 2012 at 8:28 PM, Olivier Chatry wrote:
    These are  Android.mk and Application.mk. but these project are really different, using different rendering engine. Everything work very fine with r7, but crash with r7b.
    'm really not doing anything fancy.
    On Fri, Feb 17, 2012 at 7:55 PM, David Turner (mailto:di...@android.com)> wrote:
    The math library doesn't use NEON. What's surprising is that there seems to be a copy of cosf / powf inside your own shared library.
    How do you build this exactly?
    On Fri, Feb 17, 2012 at 9:41 AM, Olivier Chatry (mailto:olivier.cha...@gmail.com)> wrote:
    (very wild guess : I'm testing on a TEGRA chipset, is there any way that the libmath ( or whatever ) is using NEON ? It works find if I switch back to r7.
    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry (mailto:olivier.cha...@gmail.com)> wrote:
    Tried with another of my lib, baked on linux, clean build got that :
    Build fingerprint: 'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146  >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00  pc 00122de8  /data/data/com.myVR.mMap.demo/lib/libmMap.so (powf): Unable to locate routine information for address 122de8 in module obj/local/armeabi-v7a//libmMap.so
    Guess the there is a problem with libmath :(
    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry (mailto:olivier.cha...@gmail.com)> wrote:
    After a full / clean / rebuild of everything, compile in thumb-mode armv7a only, my app is now crashing with this message ( taken from ndk-stack )
    ********** Crash dump: **********
    Build fingerprint: 'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321  >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00  pc 00864f14  /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf): Unable to locate routine information for address 864f14 in module obj/local/armeabi-v7a//libPinballMania.so
    Any idea ?
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com (mailto:android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com (mailto:android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group athttp://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com (mailto:android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com (mailto:android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group athttp://groups.google.com/group/android-ndk?hl=en.
    Attachments:
    - Android.mk
    - Application.mk
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Olivier Chatry at Feb 27, 2012 at 9:17 pm
    I'm happy ( and sad for you ) to not be the only one having the problem !
    On Mon, Feb 27, 2012 at 5:33 PM, Michael wrote:

    We are experiencing the same issue and were able to build a minimal
    example (based on hello-jni) to reproduce the problem.
    Our example project can be downloaded here: http://bit.ly/yaHBEL and
    then installed with, e.g.
    android update project -p .
    ndk-build -B NDK_DEBUG=1
    ant all clean debug install
    The relevant content of the project is listed below.
    Note that the crash only occurs on TEGRA devices and when the native
    code is compiled with NDK r7b (everything works with r7).

    The immediate trigger of the SIGILL crash seems to be the use of a
    cmath function on a volatile floating point variable. Making the
    variables non-volatile or substituting <cmath> by <math.h> will fix
    the issue in this particular example.

    The below code will produce a crash, but there seem to be other not-so-
    clear ways to make it crash, too. In our production code, for example,
    we are not using 'volatile', yet it crashes with
    Stack frame #00 pc 00002db4 /data/data/com.X/lib/libX.so (sinf):
    Unable to locate routine information for address 2db4 in module ./obj/
    local/armeabi-v7a/libX.so
    As a result, we currently don't have a workaround for the issue,
    besides reverting to use of NDK r7.

    I hope the provided example will help to identify and fix the problem.

    Best,
    Michael

    -----------------

    Android.mk
    ----------

    LOCAL_PATH := $(call my-dir)

    include $(CLEAR_VARS)
    LOCAL_MODULE := hello-jni
    LOCAL_SRC_FILES := hello-jni.cpp
    include $(BUILD_SHARED_LIBRARY)

    Application.mk
    --------------

    APP_ABI := armeabi armeabi-v7a
    APP_OPTIM := release
    APP_STL := gnustl_static

    hello-jni.cpp
    -------------

    #include <string.h>
    #include <jni.h>
    #include <cmath>

    extern "C"
    {
    jstring
    Java_com_example_hellojni_HelloJni_stringFromJNI( JNIEnv* env,
    jobject thiz )
    {
    volatile float sinresult = std::sin(1.0f);
    volatile float cosresult = std::sin(sinresult);
    volatile float powresult = std::pow(cosresult, 2);

    if (powresult < 0)
    return env->NewStringUTF("RESULT 0");
    else
    return env->NewStringUTF("RESULT 1");
    }
    }

    On Feb 18, 9:14 pm, Olivier Chatry wrote:
    Still crashing on TEGRA, but working on a HTC desire :(

    --
    Olivier Chatry
    Sent with Sparrow (http://www.sparrowmailapp.com)






    On Friday, February 17, 2012 at 8:28 PM, Olivier Chatry wrote:
    These are Android.mk and Application.mk. but these project are really
    different, using different rendering engine. Everything work very fine with
    r7, but crash with r7b.
    'm really not doing anything fancy.
    On Fri, Feb 17, 2012 at 7:55 PM, David Turner (mailto:
    di...@android.com)> wrote:
    The math library doesn't use NEON. What's surprising is that there
    seems to be a copy of cosf / powf inside your own shared library.
    How do you build this exactly?
    On Fri, Feb 17, 2012 at 9:41 AM, Olivier Chatry <
    olivier.cha...@gmail.com (mailto:olivier.cha...@gmail.com)> wrote:
    (very wild guess : I'm testing on a TEGRA chipset, is there any
    way that the libmath ( or whatever ) is using NEON ? It works find if I
    switch back to r7.
    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry <
    olivier.cha...@gmail.com (mailto:olivier.cha...@gmail.com)> wrote:
    Tried with another of my lib, baked on linux, clean build got
    that :
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8
    /data/data/com.myVR.mMap.demo/lib/libmMap.so (powf): Unable to locate
    routine information for address 122de8 in module
    obj/local/armeabi-v7a//libmMap.so
    Guess the there is a problem with libmath :(
    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry <
    olivier.cha...@gmail.com (mailto:olivier.cha...@gmail.com)> wrote:
    After a full / clean / rebuild of everything, compile in
    thumb-mode armv7a only, my app is now crashing with this message ( taken
    from ndk-stack )
    ********** Crash dump: **********
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14
    /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
    Unable to locate routine information for address 864f14 in module
    obj/local/armeabi-v7a//libPinballMania.so
    Any idea ?
    --
    You received this message because you are subscribed to the Google
    Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com(mailto:
    android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com (mailto:
    android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group athttp://
    groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google
    Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com(mailto:
    android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com (mailto:
    android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group athttp://
    groups.google.com/group/android-ndk?hl=en.
    Attachments:
    - Android.mk
    - Application.mk
    --
    You received this message because you are subscribed to the Google Groups
    "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at
    http://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • David Turner at Feb 28, 2012 at 10:39 pm
    thanks, having a reproducible test case is _very_ helpful. Can I ask you to
    upload it to b.android.com as a new entry? That will help tremendously for
    proper tracking.

    Thanks in advance

    - David

    On Mon, Feb 27, 2012 at 5:33 PM, Michael wrote:

    We are experiencing the same issue and were able to build a minimal
    example (based on hello-jni) to reproduce the problem.
    Our example project can be downloaded here: http://bit.ly/yaHBEL and
    then installed with, e.g.
    android update project -p .
    ndk-build -B NDK_DEBUG=1
    ant all clean debug install
    The relevant content of the project is listed below.
    Note that the crash only occurs on TEGRA devices and when the native
    code is compiled with NDK r7b (everything works with r7).

    The immediate trigger of the SIGILL crash seems to be the use of a
    cmath function on a volatile floating point variable. Making the
    variables non-volatile or substituting <cmath> by <math.h> will fix
    the issue in this particular example.

    The below code will produce a crash, but there seem to be other not-so-
    clear ways to make it crash, too. In our production code, for example,
    we are not using 'volatile', yet it crashes with
    Stack frame #00 pc 00002db4 /data/data/com.X/lib/libX.so (sinf):
    Unable to locate routine information for address 2db4 in module ./obj/
    local/armeabi-v7a/libX.so
    As a result, we currently don't have a workaround for the issue,
    besides reverting to use of NDK r7.

    I hope the provided example will help to identify and fix the problem.

    Best,
    Michael

    -----------------

    Android.mk
    ----------

    LOCAL_PATH := $(call my-dir)

    include $(CLEAR_VARS)
    LOCAL_MODULE := hello-jni
    LOCAL_SRC_FILES := hello-jni.cpp
    include $(BUILD_SHARED_LIBRARY)

    Application.mk
    --------------

    APP_ABI := armeabi armeabi-v7a
    APP_OPTIM := release
    APP_STL := gnustl_static

    hello-jni.cpp
    -------------

    #include <string.h>
    #include <jni.h>
    #include <cmath>

    extern "C"
    {
    jstring
    Java_com_example_hellojni_HelloJni_stringFromJNI( JNIEnv* env,
    jobject thiz )
    {
    volatile float sinresult = std::sin(1.0f);
    volatile float cosresult = std::sin(sinresult);
    volatile float powresult = std::pow(cosresult, 2);

    if (powresult < 0)
    return env->NewStringUTF("RESULT 0");
    else
    return env->NewStringUTF("RESULT 1");
    }
    }

    On Feb 18, 9:14 pm, Olivier Chatry wrote:
    Still crashing on TEGRA, but working on a HTC desire :(

    --
    Olivier Chatry
    Sent with Sparrow (http://www.sparrowmailapp.com)






    On Friday, February 17, 2012 at 8:28 PM, Olivier Chatry wrote:
    These are Android.mk and Application.mk. but these project are really
    different, using different rendering engine. Everything work very fine with
    r7, but crash with r7b.
    'm really not doing anything fancy.
    On Fri, Feb 17, 2012 at 7:55 PM, David Turner (mailto:
    di...@android.com)> wrote:
    The math library doesn't use NEON. What's surprising is that there
    seems to be a copy of cosf / powf inside your own shared library.
    How do you build this exactly?
    On Fri, Feb 17, 2012 at 9:41 AM, Olivier Chatry <
    olivier.cha...@gmail.com (mailto:olivier.cha...@gmail.com)> wrote:
    (very wild guess : I'm testing on a TEGRA chipset, is there any
    way that the libmath ( or whatever ) is using NEON ? It works find if I
    switch back to r7.
    On Fri, Feb 17, 2012 at 6:39 PM, Olivier Chatry <
    olivier.cha...@gmail.com (mailto:olivier.cha...@gmail.com)> wrote:
    Tried with another of my lib, baked on linux, clean build got
    that :
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 17138, tid: 17146 >>> com.myVR.mMap.demo <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 80522de8
    Stack frame #00 pc 00122de8
    /data/data/com.myVR.mMap.demo/lib/libmMap.so (powf): Unable to locate
    routine information for address 122de8 in module
    obj/local/armeabi-v7a//libmMap.so
    Guess the there is a problem with libmath :(
    On Fri, Feb 17, 2012 at 2:47 PM, Olivier Chatry <
    olivier.cha...@gmail.com (mailto:olivier.cha...@gmail.com)> wrote:
    After a full / clean / rebuild of everything, compile in
    thumb-mode armv7a only, my app is now crashing with this message ( taken
    from ndk-stack )
    ********** Crash dump: **********
    Build fingerprint:
    'lge/lge_star/p990_EUR-xx:2.3.4/GRJ22/lgp990-V20l.422C563E:user/release-keys'
    pid: 309, tid: 321 >>> com.massivefinger.PinballMania <<<
    signal 4 (SIGILL), code 1 (ILL_ILLOPC), fault addr 81864f14
    Stack frame #00 pc 00864f14
    /mnt/asec/com.massivefinger.PinballMania-2/lib/libPinballMania.so (cosf):
    Unable to locate routine information for address 864f14 in module
    obj/local/armeabi-v7a//libPinballMania.so
    Any idea ?
    --
    You received this message because you are subscribed to the Google
    Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com(mailto:
    android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com (mailto:
    android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group athttp://
    groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google
    Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com(mailto:
    android-ndk@googlegroups.com).
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com (mailto:
    android-ndk%2Bunsubscribe@googlegroups.com).
    For more options, visit this group athttp://
    groups.google.com/group/android-ndk?hl=en.
    Attachments:
    - Android.mk
    - Application.mk
    --
    You received this message because you are subscribed to the Google Groups
    "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to
    android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at
    http://groups.google.com/group/android-ndk?hl=en.
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Michael at Feb 29, 2012 at 9:55 am
    Thanks, I have just done that:
    https://code.google.com/p/android/issues/detail?id=26199

    Michael

    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To view this discussion on the web visit https://groups.google.com/d/msg/android-ndk/-/MHO0T7-mwF8J.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.
  • Daniel Smith at Mar 5, 2012 at 10:23 pm
    Just adding a "me too" to this.
    Under ndk-r7b, using a toshiba thrive (tegra2), i get a crash calling sinf()
    When i switch back to r7, the crash vanishes.
    On Wednesday, February 29, 2012 1:55:03 AM UTC-8, Michael wrote:

    Thanks, I have just done that:
    https://code.google.com/p/android/issues/detail?id=26199

    Michael


    On Wednesday, February 29, 2012 1:55:03 AM UTC-8, Michael wrote:

    Thanks, I have just done that:
    https://code.google.com/p/android/issues/detail?id=26199

    Michael


    On Wednesday, February 29, 2012 1:55:03 AM UTC-8, Michael wrote:

    Thanks, I have just done that:
    https://code.google.com/p/android/issues/detail?id=26199

    Michael
    --
    You received this message because you are subscribed to the Google Groups "android-ndk" group.
    To view this discussion on the web visit https://groups.google.com/d/msg/android-ndk/-/Fn8VzGH6XB4J.
    To post to this group, send email to android-ndk@googlegroups.com.
    To unsubscribe from this group, send email to android-ndk+unsubscribe@googlegroups.com.
    For more options, visit this group at http://groups.google.com/group/android-ndk?hl=en.

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupandroid-ndk @
categoriesandroid
postedFeb 17, '12 at 1:47p
activeMar 5, '12 at 10:23p
posts11
users4
websitedeveloper.android.com...

People

Translate

site design / logo © 2021 Grokbase