OpenCV 2.4.8: java ant example causes SIGSEGV on CentOS 6.5 (Bug #3489)
Description
When I try to run the OpenCV 2.4.8 java ant example (SimpleSample.jar) on CentOS 6.5 64-bit, I get a SIGSEGV:
[java] #
[java] # A fatal error has been detected by the Java Runtime Environment:
[java] #
[java] # SIGSEGV (0xb) at pc=0x000000393380c380, pid=30957, tid=140227056252672
[java] #
[java] # JRE version: Java(TM) SE Runtime Environment (7.0_51-b13) (build 1.7.0_51-b13)
[java] # Java VM: Java HotSpot(TM) 64-Bit Server VM (24.51-b03 mixed mode linux-amd64 compressed oops)
[java] # Problematic frame:
[java] # C [libpthread.so.0+0xc380] pthread_spin_lock+0x0
[java] #
[java] # Failed to write core dump. Core dumps have been disabled. To enable core dumping, try "ulimit -c unlimited" before starting Java again
[java] #
I used cmake-2.8.12.1, python-2.7.3, and jdk-7u51-linux-x64. If you only have a Windows machine, you can reproduce this problem by loading a centos 6.5 iso into an Oracle VM VirtualBox. Here are the steps to reproduce the problem (centos requires a little extra work to get python 2.7 installed, because you can't uninstall python 2.6):
1. Download the opencv zip: http://sourceforge.net/projects/opencvlibrary/files/opencv-unix/2.4.8/opencv-2.4.8.zip/download
2. Run the following commands:
sudo yum groupinstall "Development tools"
sudo yum install zlib-devel bzip2-devel openssl-devel ncurses-devel sqlite-devel readline-devel tk-devel gtk2-devel ant
wget http://www.cmake.org/files/v2.8/cmake-2.8.12.1.tar.gz
tar xf cmake-2.8.12.1.tar.gz
cd cmake-2.8.12.1
./bootstrap
make
sudo make install
cd ..
wget http://download.oracle.com/otn-pub/java/jdk/7u51-b13/jdk-7u51-linux-x64.rpm
sudo rpm -ihv jdk-7u51-linux-x64.rpm
wget http://python.org/ftp/python/2.7.3/Python-2.7.3.tar.bz2
tar xf Python-2.7.3.tar.bz2
cd Python-2.7.3
./configure --prefix=/usr/local --enable-shared
make
sudo make altinstall
sudo -s
echo "/usr/local/lib" > /etc/ld.so.conf.d/libpython2.7.conf
ldconfig
exit
cd ..
wget http://pypi.python.org/packages/source/d/distribute/distribute-0.6.35.tar.gz --no-check-certificate
tar xf distribute-0.6.35.tar.gz
cd distribute-0.6.35
sudo /usr/local/bin/python2.7 setup.py install
cd ..
sudo /usr/local/bin/easy_install-2.7 numpy
unzip opencv-2.4.8.zip
cd opencv-2.4.8
mkdir release
cd release
export JAVA_HOME=/usr/java/latest
cmake .. -DBUILD_SHARED_LIBS=OFF -DCMAKE_BUILD_TYPE=RELEASE -DCMAKE_INSTALL_PREFIX=/usr/local -DPYTHON_EXECUTABLE=/usr/local/bin/python2.7 -DPYTHON_INCLUDE_DIR=/usr/local/include/python2.7/ -DPYTHON_LIBRARY=/usr/local/lib/libpython2.7.so.1.0 -DPYTHON_NUMPY_INCLUDE_DIR=/usr/local/lib/python2.7/site-packages/numpy-1.8.0-py2.7-linux-x86_64.egg/numpy/core/include/ -DPYTHON_PACKAGES_PATH=/usr/local/lib/python2.7/site-packages/
make
sudo make install
cd ../samples/java/ant
ant -DocvJarDir=/usr/local/share/OpenCV/java/ -DocvLibDir=/usr/local/share/OpenCV/java/
Associated revisions
Fixed bug #3489: The code assumed that two global variables would be constructed in a particular order, but global variable initialization order is compiler-dependent.
Fixed bug #3489: The code assumed that two global variables would be constructed in a particular order, but global variable initialization order is compiler-dependent.
(cherry picked from commit 6bf599b1bca8a58c7a656ddc169f7be0fc3094c6)
History
Updated by Drew Jetter about 11 years ago
Here is a backtrace from gdb:
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f4736b19700 (LWP 14218)]
0x000000393380c380 in pthread_spin_lock () from /lib64/libpthread.so.0
(gdb) backtrace
#0 0x000000393380c380 in pthread_spin_lock () from /lib64/libpthread.so.0
#1 0x00007f4719053e0c in cv::Mutex::Impl::lock (this=0x0)
at /home/djetter/Downloads/opencv-2.4.8/modules/core/src/system.cpp:901
#2 0x00007f4719053f93 in cv::Mutex::lock (this=0x7f471a0ce080)
at /home/djetter/Downloads/opencv-2.4.8/modules/core/src/system.cpp:953
#3 0x00007f4719051c4b in cv::AutoLock::AutoLock (this=0x7f4736b17560, m=...)
at /home/djetter/Downloads/opencv-2.4.8/modules/core/include/opencv2/core/core.hpp:4815
#4 0x00007f4719054a94 in cv::TLSContainerStorage::allocateKey (this=
0x7f471a0ce080, pContainer=0x7f471a1668e0)
at /home/djetter/Downloads/opencv-2.4.8/modules/core/src/system.cpp:1104
#5 0x00007f47190550b2 in cv::TLSDataContainer::TLSDataContainer (this=
0x7f471a1668e0)
at /home/djetter/Downloads/opencv-2.4.8/modules/core/src/system.cpp:1137
#6 0x00007f4719758b2e in cv::TLSData<cv::ocl::CommandQueue>::TLSData (this=
0x7f471a1668e0)
at /home/djetter/Downloads/opencv-2.4.8/modules/core/include/opencv2/core/core.hpp:4842
#7 0x00007f471975f6a7 in __static_initialization_and_destruction_0 (
__initialize_p=1, __priority=65535)
at /home/djetter/Downloads/opencv-2.4.8/modules/ocl/src/cl_context.cpp:560
#8 0x00007f471975f843 in global constructors keyed to cl_context.cpp(void) ()
at /home/djetter/Downloads/opencv-2.4.8/modules/ocl/src/cl_context.cpp:924
#9 0x00007f4719a54276 in __do_global_ctors_aux ()
from /home/djetter/Downloads/opencv-2.4.8/samples/java/ant/build/jar/libopencv_java248.so
#10 0x00007f4718d7a1bb in _init ()
from /home/djetter/Downloads/opencv-2.4.8/samples/java/ant/build/jar/libopencv_java248.so
#11 0x0000003900000000 in ?? ()
#12 0x0000003932c0e555 in _dl_init_internal () from /lib64/ld-linux-x86-64.so.2
#13 0x0000003932c12cb5 in dl_open_worker () from /lib64/ld-linux-x86-64.so.2
#14 0x0000003932c0e1b6 in _dl_catch_error () from /lib64/ld-linux-x86-64.so.2
#15 0x0000003932c124fa in _dl_open () from /lib64/ld-linux-x86-64.so.2
#16 0x0000003933000f66 in dlopen_doit () from /lib64/libdl.so.2
#17 0x0000003932c0e1b6 in _dl_catch_error () from /lib64/ld-linux-x86-64.so.2
#18 0x000000393300129c in _dlerror_run () from /lib64/libdl.so.2
#19 0x0000003933000ee1 in dlopen@@GLIBC_2.2.5 () from /lib64/libdl.so.2
#20 0x00007f47372d262e in ?? ()
from /usr/lib/jvm/java-1.7.0-openjdk-1.7.0.51.x86_64/jre/lib/amd64/server/libjvm.so
#21 0x00007f473713c21d in JVM_LoadLibrary ()
from /usr/lib/jvm/java-1.7.0-openjdk-1.7.0.51.x86_64/jre/lib/amd64/server/libjvm.so
#22 0x00007f47365ed8b7 in Java_java_lang_ClassLoader_00024NativeLibrary_load ()
from /usr/lib/jvm/java-1.7.0-openjdk-1.7.0.51.x86_64/jre/lib/amd64/libjava.so
Updated by Drew Jetter about 11 years ago
After inspecting the backtrace, I believe this is a case of "static initialization order fiasco". The two problem static variables are:
static TLSContainerStorage tlsContainerStorage; (at modules/core/src/system.cpp:1132) cv::TLSData<CommandQueue> commandQueueTLSData; (at modules/ocl/src/cl_context.cpp:560)
It looks like the code is assuming that tlsContainerStorage will be constructed before commandQueueTLSData, but in this case, commandQueueTLSData is constructed before tlsContainerStorage. The commandQueueTLSData constructor attempts to call tlsContainerStorage.allocateKey, and that function in turn tries to create an AutoLock by passing in the mutex_ member variable.
However, since tlsContainerStorage has not been constructed, then mutex_ has not been constructed either, and so its impl variable is just pointing to random memory. When impl->lock() gets called, pthread_spin_lock is called on the sl structure. However, since the impl pointer is invalid, then the sl structure must be invalid as well, and a segmentation fault occurs.
So, I believe I have identified the problem. Now I just need to figure out how to fix it...
Updated by Drew Jetter about 11 years ago
Submitted pull request 2171.
Updated by Alexander Smorkalov about 11 years ago
- Category set to core
- Target version set to 2.4.9
- Pull request set to https://github.com/Itseez/opencv/pull/2171
Updated by Andrew Senin about 11 years ago
Thank you for submitting the pull request!
- Status changed from New to Open
Updated by Andrey Pavlenko about 11 years ago
- Status changed from Open to Done