don't use singleton for symbol_table This is the first of a series of changes to make the symbol table a part of the interpreter instead of a global object. These changes also aim to simplify the implementation of symbol table so that it is easier to understand and modify. * Functions now own their scope (workspace) data. * The list of subfunctions is contained in the scope rather than a global list. * symtab.h, symtab.cc (class symbol_table): Don't use singleton pattern. * interpreter.h, interpreter.cc (interpreter::m_symbol_table): New data member. (interpreter::~interpreter): Don't set instance to 0. * interpreter-private.h, interpreter-private.cc (__get_symbol_table__): New function. Change all uses of call_stack to access call_stack object from the interpreter.

--- a/libinterp/parse-tree/pt-id.cc	Fri Jun 09 11:36:34 2017 -0400
+++ b/libinterp/parse-tree/pt-id.cc	Fri Jun 09 02:21:28 2017 -0400
@@ -25,6 +25,7 @@
 #endif
 
 #include "error.h"
+#include "interpreter-private.h"
 #include "ovl.h"
 #include "oct-lvalue.h"
 #include "pager.h"
@@ -75,9 +76,10 @@
     // The new tree_identifier object contains a symbol_record
     // entry from the duplicated scope.
 
+    symbol_table& symtab = octave::__get_symbol_table__ ("tree_identifier::dup");
+
     // FIXME: is this the best way?
-    symbol_table::symbol_record new_sym
-      = symbol_table::find_symbol (name (), sc);
+    symbol_table::symbol_record new_sym = symtab.find_symbol (name (), sc);
 
     tree_identifier *new_id
       = new tree_identifier (new_sym, line (), column ());