It's not a test as such, since it's got its own main() that takes
arguments, doesn't return a result that can be directly compared
against expected results, and doesn't make use of the test_common
infrastructure.
The CMake build attempts to build all files named test_*.cpp with
test_common, whose main() and load_file() symbols clash with those
that the benchmark provides. Instead of making an exception, let's
rename it to something that doesn't start with "test_". Its usage
message even claims it's called bdecode_benchmark, so we use that.
Also, complete the file listing for test_common in CMake and use
an OBJECT pseudo-library there instead of STATIC. It's just as good
and jives better with having symbols provided by the executable.
(Follow-up patch in the works to provide Asio symbols that way.)
test_natpmp is a proper test but doesn't currently build,
so exclude it from the build for now.
This is necessary because the client_test example spans multiple
files, so creating an executable for each one separately doesn't
make so much sense anymore.
As _BSD_SOURCE has been deprecated since glibc 2.20, also add
_DEFAULT_SOURCE to the definitions. Both can be any value as long
as they're defined, so in order to save two lines, just #undef
them instead of checking with #ifdef/#endif.
Refactored dht_put_item to dht_put_immutable_item in session_impl since the get operations are not overloaded (consistency).
Added allow_threading_guard to add_dht_node python binding.
Making dht_put_item in session returns a zeroed hash if TORRENT_DISABLE_DHT (avoid unnecessary calculation and it's more consistent).
Added missing DHT operations to the python binding.
Added missing allow_threading_guard to add_dht_node in the python binding.
The session_handle class exposes the same public interface to session_impl as
the session class. The difference being that session_handle does not own the
underlying session_impl so multiple session_handle instances can be created
and copied about as needed.
The intent is to chenge the plugin API to pass an instance of session_handle
instead of a pointer to session_impl. This change will be made in a separate
patch.
To maintain ABI compatibilty, the type signature of session was not changed.
The relevent functions have been modified to forward to session_handle and
all enums have been left in session.