fail to build Opencv in windows using code:block - opencv

Following this link. I am stuck in the last two steps during setting up my workstation.
As a compiler I use Code:Block MinGW; I have already generated the compiled opencv files, now I need to build the sln file in Windows. I use Code Block to build this OpenCV Project File in the path D:\OpenCV\Build\Opencv, where I put the generated bin file after using Cmake.
During the building, it stopped at 40%, saying;
Linking CXX executable ....\bin\opencv_perf_core.exe
c:/codeblock/codeblocks/mingw/bin/../lib/gcc/mingw32/4.4.1/../../../../mingw32/bin/ld.exe:
warning: auto-importing has been activated without
--enable-auto-import specified on the command line. This should work unless it involves constant data structures referencing symbols from
auto-imported DLLs. Cannot export _ZN12_GLOBAL__N_13ROp3allEv: symbol
not found Cannot export _ZN12_GLOBAL__N_17CmpType3allEv: symbol not
found collect2: ld returned 1 exit status mingw32-make.exe[2]: *
[bin/opencv_perf_core.exe] Error 1 mingw32-make.exe1: *
[modules/core/CMakeFiles/opencv_perf_core.dir/all] Error 2
mingw32-make.exe: * [all] Error 2 Info: resolving vtable for
cv::_OutputArray by linking to imp_ZTVN2cv12_OutputArrayE
(auto-import) Info: resolving vtable for cv::_InputArray by linking to
imp_ZTVN2cv11_InputArrayE (auto-import) Info: resolving vtable for cv::Exception by linking to imp_ZTVN2cv9ExceptionE (auto-import)
Creating library file: ....\bin\libopencv_perf_core.dll.a Process
terminated with status 2 (14 minutes, 29 seconds) 0 errors, 3 warnings
How can I solve this problem?

Unfortunately, there's not much you can do, according to http://code.opencv.org/issues/2523.
You will have to use a recent version of MinGW. It builds fine using the latest MinGW shipping with GCC 4.7.2.
This issue seems to have been introduced in OpenCV 2.4.3 as it is said version 2.4.2 builds fine.

Related

When trying to build Halide on Linux system with llvm-10.0 on release/10.x and master branches, facing the below issue

[ 53%] Linking CXX executable test_function_dag
/bin/ld: ../../libHalide.so.10.0.0: undefined reference to `typeinfo for llvm::SectionMemoryManager'
collect2: error: ld returned 1 exit status
make[2]: *** [src/autoschedulers/adams2019/CMakeFiles/test_function_dag.dir/build.make:134: src/autoschedulers/adams2019/test_function_dag] Error 1
make[1]: *** [CMakeFiles/Makefile2:2328: src/autoschedulers/adams2019/CMakeFiles/test_function_dag.dir/all] Error 2
make: *** [Makefile:182: all] Error 2
I'm using pre-built LLVM binaries and below cmake command to configure
cmake -DCMAKE_BUILD_TYPE=Release -DLLVM_DIR=/home/harry/llvm/lib/cmake/llvm ../
Any advice and suggestions will be greatly appreciated.
I cloned Halide from release/10.x, and compiled against the LLVM 10.0.0 binaries obtained here. On Ubuntu 20.04 LTS, I had to also apt install libtinfo5 because these binaries are built against old system libraries.
I noticed that the official binaries are not built with RTTI enabled (ugh) and was able to repro and fix. There's a bug in dependencies/llvm/CMakeLists.txt. Find the line that reads:
$<$<COMPILE_LANG_AND_ID:CXX,GCC,Clang,AppleClang>:-fno-rtti>)
and change GCC to GNU. We don't test with non-RTTI LLVM, which is why this wasn't caught.

Adding Library to CMake on CLI

I'm trying to build OpenCV with CUDA support using Arch Linux's AUR. In their build configuration some options are passed to CMake, I have modified those to add CUDA features. Upon attempting compilation I get the following error:
[ 10%] Built target pch_Generate_opencv_reg
In file included from /home/meurer/Downloads/opencv-git/src/modules/cudacodec/opencv_cudacodec_pch_dephelp.cxx:1:0:
/home/meurer/Downloads/opencv-git/src/opencv/modules/cudacodec/src/precomp.hpp:59:14: fatal error: dynlink_nvcuvid.h: No such file or directory
#include <dynlink_nvcuvid.h>
^~~~~~~~~~~~~~~~~~~
compilation terminated.
make[2]: *** [modules/cudacodec/CMakeFiles/opencv_cudacodec_pch_dephelp.dir/build.make:63: modules/cudacodec/CMakeFiles/opencv_cudacodec_pch_dephelp.dir/opencv_cudacodec_pch_dephelp.cxx.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:6834: modules/cudacodec/CMakeFiles/opencv_cudacodec_pch_dephelp.dir/all] Error 2
I searched my filesystem and found the rogue header file in /usr/include/nvidia-sdk/dynlink_nvcuvid.h. Due to how AUR works, I believe I can only configure CMake via command line options (See links 1, 2 under _cmakeopts=...).
What argument can I pass to CMake to make it search for the appropriate libraries under the aforementioned directory, as well as the ones it already uses?

Error Cross Compiling: hidden symbol ... final link failed

I am getting the following error when compiling opencv with ffmpeg.
..../arm-none-linux-gnueabi/bin/ld: ../../bin/opencv_test_core: hidden symbol `__sync_val_compare_and_swap_4' in ..../arm-none-linux-gnueabi/bin/ld: final link failed: Nonrepresentable section on output
collect2: ld returned 1 exit status
I have no idea what this means. How do I fix this?
Edit:
I am cross compiling for ARM using codesourcery's arm-2009q3 compiler.
I using cmake and make to compile.
Please check all the libraries that you use and go through this stackoverflow thread as well.
To me, I got the below error when I migrated my product to new GCC version (3.3.3 to 4.5.4)
hidden symbol `__clz_tab' in /../lib/gcc/mipsel-unknown-linux-uclibc/4.5.4/libgcc.a(_clz.o) is referenced by DSO
Later I found that one of the library that I tried to link was built using GCC 3.3.3. After building the other library under GCC 4.5.4, the 'hidden symbol' error disappeared.

Wireshark make error: undefined reference to symbol 'g_module_name'

I'm trying to compile wireshark-P2P-1.7.0 but it fails showing the following error:
/usr/bin/ld: gtk/libui.a(libui_a-plugins_dlg.o): undefined reference to symbol 'g_module_name'
//usr/lib/i386-linux-gnu/libgmodule-2.0.so.0: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
make[2]: *** [wireshark] Error 1
make[2]: Leaving directory `/home/sniffer/Desktop/P2P_install/Sigma_Sniffer-P2P-Ubuntu_v8.1.0/wireshark-P2P-1.7.0'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/sniffer/Desktop/P2P_install/Sigma_Sniffer-P2P-Ubuntu_v8.1.0/wireshark-P2P-1.7.0'
make: *** [all] Error 2
I'm not sure if it requires special libraries to run on a 64bit processor since this version of wireshark is 32 bit.
I cannot use a standard version of wireshark or download it from anywhere else. I have to use this one because is specially configure to make WFA tests.
Thank you!!
Oscar.
The only way a source code version of Wireshark would "[be] 32 bit" would be if somebody had added code to it that assumed that long or pointer data types were 32-bit; no code in the Wireshark source from wireshark.org is "32 bit" or "64 bit", it's all portable (and the Wireshark build infrastructure builds and tests both 32-bit and 64-bit versions).
What it does need is a version of GLib that includes g_module_name(), and you need to be linking with GLib (Wireshark uses it).

Crypto++ linker error using ANT

I am trying to reverse engineer a previous employee's build process for some custom software he wrote for us. I've seen to get everything to compile okay, but I get an error in the linking process that points to a problem with crypto++. Any clues?
compile.util:
compile:
[echo] Compiling: util
[cc] Starting dependency analysis for 55 files.
[cc] 55 files are up to date.
[cc] 0 files to be recompiled from dependency analysis.
[cc] 0 total files to be compiled.
link.util:
link:
[echo] Linking shared library: libutil
[cc] 0 total files to be compiled.
[cc] Starting link
[cc] /usr/bin/ld: /home/john/softwarename/thirdparty/crypto/5.6.1_64/libcryptopp.a(cryptlib.o): relocation R_X86_64_32 against `CryptoPP::DEFAULT_CHANNEL' can not be used when making a shared object; recompile with -fPIC
[cc] /home/john/softwarename/thirdparty/crypto/5.6.1_64/libcryptopp.a: could not read symbols: Bad value
[cc] collect2: ld returned 1 exit status
BUILD FAILED
/home/john/softwarename/build/build.xml:167: Following error occured while executing this line
/home/john/softwarename/build/link.xml:27: gcc failed with return code 1
Os is Fedora 11
Every object file you link into a shared library must be position independent, meaning the loader can move it wherever it wants in memory and it will still work. Your crypto library was not compiled that way, hence the admonition to recompile it with -fPIC added to your compile flags. Alternately, you can make libutil a static library instead of a shared one. Without more details on your build files, I can't provide more details on how to do that or which one to choose.

Resources