Home | Libraries | People | FAQ | More |
dlopen
like
functions. For those platforms Boost.DLL is safe in the manner as
all the C++ Standard Library containers are: it is safe to use different
instances of shared_library from different threads even if all the
instances loaded the same library. On other platforms it is not safe
to concurrently call any of the functions from Boost.DLL (even a
shared_library::location()
call triggers a race condition). See Limitations,
Thread safe library loading.
BOOST_SYMBOL_EXPORT
or BOOST_DLL_ALIAS
.
You can call shared_library(program_location())
to load yourself. Refer to the
Tutorial section for more info. You can also query executables, just
provide a path to the executable to library_info
class.
shared_library::get<T>
or import<T>
?
Segmentation Fault
.
However it is safe to make types more strict, for example making
const int
from an int
will not
harm.
BOOST_DLL_ALIAS
macros will not change with the change of compiler or platform. You
must take care of functions ABI and API stability by your own.
libsome_library
shared library. How to achieve that?
load_mode::rtld_deepbind
.
Workaround would be to write plugin into a temporary file in RAM and load plugin from it:
#include <boost/filesystem.hpp> #include <boost/dll.hpp> using namespace boost; dll:shared_library load_from_memory(unsigned char* data, std::size_t size, const filesystem::path& tmp_plugin_path = filesystem::unique_path() / "libplugin.so") { const filesystem::path plugin_location = filesystem::temp_directory_path() / tmp_plugin_path; filesystem::create_directories(plugin_location.parent_path()); filesystem::ofstream ofs(plugin_location, std::ios::out|std::ios::bin|std::ios::trunc); ofs.write(data, size); return dll::shared_library(plugin_location); }
But there's no guarantee that filesystem::temp_directory_path()
will actually write to RAM, that's very platform
dependent.