What I have tried
I am working sample application using Caffe model in MAC OSX. I downloaded Caffe source from https://github.com/BVLC/caffe.
Steps:
1.Installed dependencies packages for caffe as said in Caffe Link. Some packages is in /usr/local/lib and some packages /opt/local/lib
2.Builded shared library(SO) file using CMakeLists.txt in CMake GUI.
3.Created sample c++ application
4.Linked all dependencies and caffe library files with c++ application.
Built the application
After I executed the application, It needed library files of dependencies package from /usr/local/lib
What I want
After I installed caffe dependencies, I copied all library files in another location.
How to link specified location of the library files in Cmake.?
In Mac osx, when the application run, It takes the library form /usr/local/lib not in the current folder(application folder).
How do I set the application to take library files from current folder.?
Here is a command from a makefile that I use to make an OSX application (the names have been changed to protect the innocent):
install_name_tool -id "#executable_path/../Resources/libMyLib.dylib" ./libMyLib.dylib
You can also just run install_name_tool on the command line.
The first argument after -id is the destination you want (including the name of the library) and the second is the current path to the library -- ie the binary itself. In this example the library is in the same folder as the makefile.
Then, after I build the app (the linker will include this new path to the desired library in the binary you build) I copy the library to the Resources folder:
cp ./libMyLib.dylib App.app/Contents/Resources/
In OSX-land #executable_path is whatever directory your actual binary is in. The way apps are packaged is like this:
App.app/Contents/MacOS/App. There is also a folder in App.app/Contents called Resources, and this is generally where I stash dependencies if I don't want the user to have to install them him/herself. Thus, in this case, #executable_path/../Resources is a relative path from the app's binary to where the dylib is going to be.
You can used install_name_tool to put the library anywhere you want. I just feel like this is a good place to do it.
By the way, you can check what the current id of the library is with otool -L. Probably, if you run
> otool -L caffe.dylib
it will return something like /usr/local/lib, even if you move it somewhere else. Try to change the id with install_name_tool -id and then run otool -L again, and see what happens. It should make sense at that point.
Related
I have downloaded a ROS2 demo from the examples repository.
Specifically, I have used minimal_subscriber and minimal_publisher.
I have ROS2 installed in /opt/ros2 and when I build these two examples with colcon build, it generates an install/ directory with lib/, shared/ and the other usual directory structure.
I can execute and run these demos perfectly fine with my current setup in my machine, but these executables link to libraries present in /opt/ros2, so when I want to execute them in another machine without ROS2 installed or I move my ROS2 installation in my machine, the executables cannot find the shared objects.
I've added a really simple script that adds all dependencies to install/lib when building but the executables don't seem to care, they aren't looking for shared libraries in the generated lib directory, they keep trying to search in /opt/ros2.
I believe this is something I should solve in CMake and it's not ROS2 specific, so, is there any way I can tell my generated executables to search in a diferent directory? (../lib or ./lib in my case)
If you are building them yourself (assumed since you mention CMake), you can set the RPATH in CMake (docs here). Specifically set the CMAKE_INSTALL_RPATH something like:
set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib")
If you can't rebuild them, you can set LD_LIBRARY_PATH in your environment to where the libraries are located, or you can patch the executables themselves with an updated RPATH by using patchelf.
In order to get a relative RPATH rather than an absolute RPATH, use the $ORIGIN variable in your rpath spec. See "Recommendations" the the link above for more details.
I have a question related to how to install a built executable program with cmake when it relies on some external libraries. Suppose my executable is abc, and it relies on two external libraries: lib1.so and lib2.so. The structure of the codes are as follows:
-.........
|----bin (lib1.so lib2.so)
|----include(lib1.h lib2.h)
|----src(main.cpp)
When the executable program is installed using the following cmake commands:
INSTALL(TARGETS ${Exe_Name}
RUNTIME DESTINATION Path to bin
LIBRARY DESTINATION Path to bin)
I expect that the executable program will be in the same directory with lib1.so and lib2.so. However, when I execute the built program in the installation folder, I met the following error:
error while loading shared libraries: lib1 can not open shared object file No such file or directory
If I use ldd to check the executable, I found lib1.so and lib2.so not found. After searching for possible solutions, I found if I call the executable in this way, then it worked:
LD_LIBRARY_PATH=./ ./my_program_run
Then my question is how I can let my executable program knows the locations of the shared libraries with cmake when it is installed? Thanks.
This is best solved this with the RPATH of the final executable. RPATH is a hardcoded search path for the executable itself, and allows the use of the string $ORIGIN, which expands to the location of the executable at runtime. See this reference: http://man7.org/linux/man-pages/man8/ld.so.8.html
CMake strips the rpath of a binary at installation time, to avoid the binary picking up libraries littered around your development tree. But it also provides a simple way to modify the installation rpath for exactly this reason. Here's the short answer:
IF(UNIX)
SET(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_RPATH}:\$ORIGIN/../bin:\$ORIGIN")
ENDIF()
This particular example appends to the existing rpath, and adds . and ../bin to the search path, all relative to the location of the binary.
Some developers claim that adjusting the RPATH of the binary is not a good idea. In the ideal world, all the libraries would live in the system library directories. But if you take this to the extreme, you end up with Windows (at least the older ones), where c:\windows\system32 is full of junk that came from who knows where, and may or may not conflict with other software, etc. Using rpath and installing everything in one place seems like a great solution.
If the application is to be cleanly installed to a standard linux distribution, then you should either install the supporting shared libraries into a standard location (/usr/lib), or you should add the libraries location to the ld.so config, by create an /etc/ld.so.conf.d/myprogram.conf file containing the name of the directory the libraries are in.
If the installation is temporary or more ad-hoc, then a script to set the LD_LIBRARY_PATH is suitable.
The libraries are searched in the predefined locations which includes standard library paths configured with ld.so.conf and LD_LIBRARY_PATH. You can also try to compile your app with -rpath, but it is not recommended by some developers. I suggest you to create a wrapper script which will set LD_LIBRARY_PATH and run the real application like that:
"theapp" script:
#!/bin/sh
dir="`dirname \"$0\"`"
export LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+$LD_LIBRARY_PATH:}"$dir"
exec "$dir/theapp.real" # your real application
The application, script and libraries should be in the same location (under bin/).
I'm trying to build wxWidgets library into a custom path on a Fedora 27 operative system.
I achieved the wx-config file path recognition and works with the cmake execution. Also, I load libraries and include dirs based on modified wxWidgets finder cmake file that sets thewx-config custom path successfully.
But cmake does not load my wxWidgets configuration. I mean, wx_gtk2u_jpeg-3.1 builded lib could not be founded (suposed to be /usr/lib/libwx_gtk2u_jpeg-3.1.so). I need jpeg dependency from wxWidgets for my project.
I'm sure that problem is not about cmake files. However, the problem is wxWidgets compilation because cmake can found the other builded dependencies into /usr/lib/
I actually installed the libjpeg-turbo-devel package that includes the libjpeg.h needed for wxWidgets building without success of libwx_gtk2u_jpeg-3.1.so creation.
The weirdest part is that $ wx-config --libs shows the wx_gtk2u_jpeg-3.1 lib to be linked and the hint paths that it should be founded.
wxWidgets commands for building:
$ ./configure --with-libjpeg=builtin --with-libpng=builtin --with-libtiff=builtin --with-zlib=builtin --with-expat=builtin --enable-webviewwebkit=no --prefix=/opt/cpp_dependencies/2018Q1/usr'
$ make -j 4
$ make install
You can check out my cmake files, the cmake output and wxWidgets building output in order to reproducing it: https://gist.github.com/jjalvarezl/b70accae269ef56c56010bedf157c27f
You can see line 1543 of wxWidgets building output file that jpeg library is buildin, and, 1564 of same file, the make install command that installs all libwx_<lib_name>.so libraries into final /usr/lib path. Anyway, no one contains the needed library.
Please show the exact error message, as it's not clear what the actual problem is. What I can say, is that the different built-in versions of 3rd party libraries, such as libjpeg, are always static libraries, even when wxWidgets themselves are shared. I.e. you're never going to have libwx_gtk2u_jpeg-3.1.so, only .a.
I'd also strongly recommend using system versions of the 3rd party libraries under Unix systems. This means that your wxWidgets applications will get security updates from your OS vendor and you don't risk running into any incompatibilities due to using 2 different versions of the same library in your application.
I have an application and a dylib that it links against. It is my understanding that the application bundle should contain the dylib. Is that correct?
Is there a way to copy the dylib to the application bundle when built?
Thanks,
Alan
Is there a way to copy the dylib to the application bundle when built?
Well, sure -- with a little shell scripting, anything is possible. :)
In your build script, after you've called macdeployqt, but before you've done any code-signing, you'd copy the .dylib file over into the Contents/Frameworks inside your .app folder.
The trick after that is convincing the executable to look for the library in its new location rather than the location it was in when you linked the executable. (If you do a "otool -L ./MyProgram.app/Contents/MacOS/MyProgram" you will see the places where the program is looking for shared libraries, and you will see that it is not looking for your .dylib file in the Frameworks folder, yet)
To do that you use the install_name_tool command, e.g.:
install_name_tool -change /the/old/path/to/the_library_name.dylib "#executable_path/../Frameworks/the_library_name.dylib" ./MyProgram.app/Contents/MacOS/MyProgram
After doing that you can run "otool -L" on your executable again to verify to your satisfaction that the executable is now set to look for the .dylib file inside the Frameworks folder.
TL;DR
Objective-C app linked with static library that dynamic links Boost Filesystem. App can be run from output directory using Terminal, but trying to run from Xcode debugger or Finder gives error dyld: Library not loaded: libboost_filesystem.dylib <snip> Reason: image not found.
Problem
In my Xcode project I have a structure that looks like this:
MainProject (Objective-C)
- static_lib_that_uses_filesystem (C++)
To get everything to link, I added libboost_system and libboost_filesystem dylibs to the "Link Binary with Libraries" build phase in MainProject.
When I try to run the app from either Xcode or Finder I get:
sharedlibrary apply-load-rules all
warning: Unable to read symbols for libboost_filesystem.dylib (file not found).
warning: Unable to read symbols from "libboost_filesystem.dylib" (not yet mapped into memory).
warning: Unable to read symbols for libboost_system.dylib (file not found).
warning: Unable to read symbols from "libboost_system.dylib" (not yet mapped into memory).
[Switching to process 43957 thread 0x0]
dyld: Library not loaded: libboost_filesystem.dylib
Referenced from: /Users/ssteele/Library/Developer/Xcode/DerivedData/MainProject-dqrhyuarllykslftblocjdzxlran/Build/Products/Debug/MainProject.app/Contents/MacOS/MainProject
Reason: image not found
I added a build stage to copy the dylibs to the Frameworks directory in the bundle, this doesn't help. I changed this to copy them to the Executables directory which also didn't help.
Having them in the Executables directory does allow me to run the app from Terminal.
How can I get the app to find the dylibs when run from Finder/Xcode?
Background Info
I'm using Xcode 4.2 on Lion and currently targeting Lion only. I built my shared libraries for filesystem like this:
./b2 threading=multi macosx-version=10.7 --with-filesystem stage
This creates libboost_system.dylib, libboost_filesystem.dylib, and also .a equivalents in the stage/lib directory, I'm referencing them in the project directly from there.
Dynamic libs (dylibs) on OSX bake in the path that they should be loaded from. For example...
/usr/lib/some_awesome.dylib.
When you link to a dylib, the linker embeds this path in your executable as the place to look for it at runtime. This is fine and easy with installed libs, but for relative path linking it's more complicated.
When you build the boost libs, they just get their names embedded rather than a full or relative path (i.e. libboost_system.dylib rather than /usr/lib/libboost_filesystem.dylib). You should be able to change this with the dll-path option, but that seems broken currently.
To fix your problem you either need to get the correct path relative to your application embedded (e.g. #executable_path/libwhatever.dylib) into the dylibs somehow, which would probably require the bjam dll-path option to work, or instead you can fix your executable to look in a different location.
To do this, use something like the following as a script step in your build:
install_name_tool -change libboost_filesystem.dylib #executable_path/libboost_filesystem.dylib$BUILT_PRODUCTS_DIR/$EXECUTABLE_PATH
Note that if you have multiple dylibs that reference each other with broken paths, you'll need to fix the paths between them too, e.g.
install_name_tool -change libboost_system.dylib #executable_path/libboost_system.dylib$BUILT_PRODUCTS_DIR/$EXECUTABLE_FOLDER_PATH/libboost_filesystem.dylib
The following is a good article on this: Creating working dylibs
The issue is that boost needs to be installed e.g. b2 ..... install. This copies the libraries and headers into /usr/local/lib and /usr/local/include.
OSX dynamic libraries only run from the directory that they are built for.
You can change the install directory by using the -prefix argument to boost build. However the libraries would still need to be in the same directory for all users.
There should be a way of building boost as a framework and/or embedding #executable_path in the library.
An alternative is to use boost's static libraries - build the static only or delete the dynamic ones, Xcode looks for dynamic before static. If using static then the path to the libraries does not matter at run time as all the code is now in the executable.
I think you need to add the path to the directory where you saved libboost_filesystem.a in the "library search paths"
Click on your project profile -> build settings -> expand "Search path" -> "library search paths"