wrong `empty` handling in RPCs

[jktjkt]

As reported in #27, trying to set a leaf within a prepare context fails:

/> prepare /ietf-netconf:get-config
(prepare: /ietf-netconf:get-config) > set source/candidate 
(prepare: /ietf-netconf:get-config) > exec

Here's the BT which eventually throws:

(gdb) bt
#0  lyplg_type_store_empty (ctx=ctx@entry=0x513000017180, type=<optimized out>, value=0x7ffff477aac0, value_len=7, options=0, UNUSED_format=<optimized out>, UNUSED_prefix_data=<optimized out>, hints=1011, UNUSED_ctx_node=<optimized out>, storage=0x50800007c8d8, UNUSED_unres=<optimized out>, err=0x7ffff44c9420)
    at /home/jkt/work/cesnet/gerrit/github/CESNET/libyang/src/plugins_types/empty.c:54
#1  0x00007ffff690f46c in lyd_value_store (ctx=<optimized out>, val=<optimized out>, type=<optimized out>, value=<optimized out>, value_len=<optimized out>, is_utf8=<optimized out>, store_only=<optimized out>, dynamic=<optimized out>, format=<optimized out>, prefix_data=<optimized out>, hints=<optimized out>, ctx_node=<optimized out>, 
    incomplete=<optimized out>) at /home/jkt/work/cesnet/gerrit/github/CESNET/libyang/src/tree_data_common.c:535
#2  0x00007ffff6926c89 in lyd_create_term (schema=schema@entry=0x50c0000a4b00, value=value@entry=0x7ffff477aac0 "[empty]", value_len=value_len@entry=7, is_utf8=is_utf8@entry=0 '\000', store_only=store_only@entry=0 '\000', dynamic=dynamic@entry=0x0, format=LY_VALUE_JSON, prefix_data=0x0, hints=1011, incomplete=0x0, node=0x7ffff477b510)
    at /home/jkt/work/cesnet/gerrit/github/CESNET/libyang/src/tree_data_new.c:71
#3  0x00007ffff69442ef in lyd_new_path_ (parent=0x5070000ab200, ctx=0x513000017180, ext=<optimized out>, path=<optimized out>, value=<optimized out>, value_len=7, value_type=<optimized out>, options=<optimized out>, new_parent=<optimized out>, new_node=<optimized out>) at /home/jkt/work/cesnet/gerrit/github/CESNET/libyang/src/tree_data_new.c:1857
#4  0x00007ffff6940583 in lyd_new_path (parent=<optimized out>, ctx=<optimized out>, path=<optimized out>, value=<optimized out>, options=32, node=<optimized out>) at /home/jkt/work/cesnet/gerrit/github/CESNET/libyang/src/tree_data_new.c:1919
#5  0x00007ffff7eb7a7a in libyang::impl::newPath (node=<optimized out>, ctx=<optimized out>, refs=..., path="/ietf-netconf:get-config/source/candidate", value=..., options=std::optional = {...}) at /home/jkt/work/cesnet/gerrit/CzechLight/libyang-cpp/src/utils/newPath.cpp:20
#6  0x00007ffff7e37087 in libyang::DataNode::newPath (this=<optimized out>, path=..., value=..., options=...) at /home/jkt/work/cesnet/gerrit/CzechLight/libyang-cpp/src/DataNode.cpp:264
#7  0x000055555599a899 in treeToRpcInput (ctx=..., path=..., in=...) at /home/jkt/work/cesnet/gerrit/CzechLight/netconf-cli/src/libyang_utils.cpp:130
#8  0x000055555578d427 in SysrepoAccess::execute (this=0x50b000007e20, path=..., input=...) at /home/jkt/work/cesnet/gerrit/CzechLight/netconf-cli/src/sysrepo_access.cpp:175
#9  0x000055555579cd58 in ProxyDatastore::execute[abi:cxx11]() (this=0x7ffff4a00350) at /home/jkt/work/cesnet/gerrit/CzechLight/netconf-cli/src/proxy_datastore.cpp:81
#10 0x00005555558da0d0 in Interpreter::operator() (this=0x7ffff4a00a90, exec=...) at /home/jkt/work/cesnet/gerrit/CzechLight/netconf-cli/src/interpreter.cpp:279
#11 0x000055555577a82e in boost::apply_visitor<Interpreter, boost::variant<boost::detail::variant::over_sequence<boost::mpl::l_item<mpl_::long_<18l>, cancel_, boost::mpl::l_item<mpl_::long_<17l>, cd_, boost::mpl::l_item<mpl_::long_<16l>, commit_, boost::mpl::l_item<mpl_::long_<15l>, copy_, boost::mpl::l_item<mpl_::long_<14l>, create_, boost::mpl::l_item<mpl_::long_<13l>, delete_, boost::mpl::l_item<mpl_::long_<12l>, describe_, boost::mpl::l_item<mpl_::long_<11l>, discard_, boost::mpl::l_item<mpl_::long_<10l>, dump_, boost::mpl::l_item<mpl_::long_<9l>, exec_, boost::mpl::l_item<mpl_::long_<8l>, get_, boost::mpl::l_item<mpl_::long_<7l>, help_, boost::mpl::l_item<mpl_::long_<6l>, ls_, boost::mpl::l_item<mpl_::long_<5l>, move_, boost::mpl::l_item<mpl_::long_<4l>, prepare_, boost::mpl::l_item<mpl_::long_<3l>, quit_, boost::mpl::l_item<mpl_::long_<2l>, set_, boost::mpl::l_item<mpl_::long_<1l>, switch_, boost::mpl::l_end> > > > > > > > > > > > > > > > > > >>&>(Interpreter const&, boost::variant<boost::detail::variant::over_sequence<boost::mpl::l_item<mpl_::long_<18l>, cancel_, boost::mpl::l_item<mpl_::long_<17l>, cd_, boost::mpl::l_item<mpl_::long_<16l>, commit_, boost::mpl::l_item<mpl_::long_<15l>, copy_, boost::mpl::l_item<mpl_::long_<14l>, create_, boost::mpl::l_item<mpl_::long_<13l>, delete_, boost::mpl::l_item<mpl_::long_<12l>, describe_, boost::mpl::l_item<mpl_::long_<11l>, discard_, boost::mpl::l_item<mpl_::long_<10l>, dump_, boost::mpl::l_item<mpl_::long_<9l>, exec_, boost::mpl::l_item<mpl_::long_<8l>, get_, boost::mpl::l_item<mpl_::long_<7l>, help_, boost::mpl::l_item<mpl_::long_<6l>, ls_, boost::mpl::l_item<mpl_::long_<5l>, move_, boost::mpl::l_item<mpl_::long_<4l>, prepare_, boost::mpl::l_item<mpl_::long_<3l>, quit_, boost::mpl::l_item<mpl_::long_<2l>, set_, boost::mpl::l_item<mpl_::long_<1l>, switch_, boost::mpl::l_end> > > > > > > > > > > > > > > > > > >>&) (visitor=..., visitable=...) at include/boost/variant/detail/apply_visitor_unary.hpp:68
#12 main (argc=<optimized out>, argv=<optimized out>) at /home/jkt/work/cesnet/gerrit/CzechLight/netconf-cli/src/cli.cpp:299

The root cause is in the implementation of leafDataToStringVisitor which encodes an instance of YANG's empty type into a string [empty]. This was probably done so in an attempt at being user-friendly, but this thing is used not just in the get command for printing retrieved leaf values, but also in other contexts:

• As reported in Unable to display candidate datastore content #27, it's used in the ProxyDatastore::execute to build a temporary DataNode tree.
• It's used in ast_path.cpp's pathToDataString(), and therefore for (e.g.) XPaths which contain list key values (right? and is it used for XPaths in general?)
• When moving data, it's used as an anchor to specify where to move the value relative in a list or a list-instance.
• As a result, the code in sysrepo_access.cpp, netconf_access.cpp and yang_access.cpp has some special handling of their setLeaf methods for instances of the empty_ class.

Maybe we should get rid of that special [empty] pretty-printing? Or should we be more careful and clearly separate "pretty printing" from "YANG canonical value printing"?