]> git.kernelconcepts.de Git - karo-tx-redboot.git/blobdiff - tools/src/libcdl/expr.cxx
Initial revision
[karo-tx-redboot.git] / tools / src / libcdl / expr.cxx
diff --git a/tools/src/libcdl/expr.cxx b/tools/src/libcdl/expr.cxx
new file mode 100644 (file)
index 0000000..03d7d43
--- /dev/null
@@ -0,0 +1,3437 @@
+//{{{  Banner                           
+
+//============================================================================
+//
+//      expr.cxx
+//
+//      Implementation of the various CDL expression classes.
+//
+//============================================================================
+//####COPYRIGHTBEGIN####
+//                                                                          
+// ----------------------------------------------------------------------------
+// Copyright (C) 1999, 2000, 2001 Red Hat, Inc.
+//
+// This file is part of the eCos host tools.
+//
+// This program is free software; you can redistribute it and/or modify it 
+// under the terms of the GNU General Public License as published by the Free 
+// Software Foundation; either version 2 of the License, or (at your option) 
+// any later version.
+// 
+// This program is distributed in the hope that it will be useful, but WITHOUT 
+// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 
+// FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for 
+// more details.
+// 
+// You should have received a copy of the GNU General Public License along with
+// this program; if not, write to the Free Software Foundation, Inc., 
+// 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
+//
+// ----------------------------------------------------------------------------
+//                                                                          
+//####COPYRIGHTEND####
+//============================================================================
+//#####DESCRIPTIONBEGIN####
+//
+// Author(s):   bartv
+// Contact(s):  bartv
+// Date:        1999/02/02
+// Version:     0.02
+//
+//####DESCRIPTIONEND####
+//============================================================================
+
+//}}}
+//{{{  #include's                       
+
+// ----------------------------------------------------------------------------
+#include "cdlconfig.h"
+
+// Get the infrastructure types, assertions, tracing and similar
+// facilities.
+#include <cyg/infra/cyg_ass.h>
+#include <cyg/infra/cyg_trac.h>
+
+// <cdlcore.hxx> defines everything implemented in this module.
+// It implicitly supplies <string>, <vector> and <map> because
+// the class definitions rely on these headers.
+#include <cdlcore.hxx>
+
+//}}}
+
+//{{{  Statics                          
+
+// ----------------------------------------------------------------------------
+CYGDBG_DEFINE_MEMLEAK_COUNTER(CdlEvalContext);
+CYGDBG_DEFINE_MEMLEAK_COUNTER(CdlExpressionBody);
+CYGDBG_DEFINE_MEMLEAK_COUNTER(CdlListExpressionBody);
+CYGDBG_DEFINE_MEMLEAK_COUNTER(CdlGoalExpressionBody);
+
+//}}}
+//{{{  CdlEvalContext                   
+
+// ----------------------------------------------------------------------------
+// A utility class to keep track of the context in which expression
+// evaluation is happening.
+
+CdlEvalContext::CdlEvalContext(CdlTransaction transaction_arg, CdlNode node_arg, CdlProperty property_arg,
+                               CdlToplevel toplevel_arg)
+{
+    CYG_REPORT_FUNCNAME("CdlEvalContext::constructor");
+    CYG_REPORT_FUNCARG4XV(this, transaction_arg, node_arg, property_arg);
+
+    transaction = transaction_arg;
+    
+    if ((0 == property_arg) && (0 != transaction)) {
+        CdlConflict conflict = transaction->get_conflict();
+        if (0 != conflict) {
+            property_arg = conflict->get_property();
+        }
+    }
+    property    = property_arg;
+    
+    if ((0 == node_arg) && (0 != transaction)) {
+        CdlConflict conflict = transaction->get_conflict();
+        if (0 != conflict) {
+            node_arg = conflict->get_node();
+        }
+    }
+    node        = node_arg;
+    
+    if (0 == toplevel_arg) {
+        if (0 != transaction) {
+            toplevel_arg = transaction->get_toplevel();
+        } else if (0 != node) {
+            toplevel_arg = node->get_toplevel();
+        }
+    }
+    toplevel = toplevel_arg;
+    
+    cdlevalcontext_cookie = CdlEvalContext_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+CdlEvalContext::~CdlEvalContext()
+{
+    CYG_REPORT_FUNCNAME("CdlEvalContext::destructor");
+    CYG_PRECONDITION_THISC();
+
+    cdlevalcontext_cookie       = CdlEvalContext_Invalid;
+    transaction = 0;
+    node        = 0;
+    property    = 0;
+    toplevel    = 0;
+    CYGDBG_MEMLEAK_DESTRUCTOR();
+    
+    CYG_REPORT_RETURN();
+}
+
+// Given a context and a reference inside an expression, obtain the node
+// being referenced - if it is loaded.
+CdlNode
+CdlEvalContext::resolve_reference(CdlExpression expr, int index)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlEvalContext::resolve_reference", "result %");
+    CYG_REPORT_FUNCARG2XV(expr, index);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSC(expr);
+    CYG_PRECONDITIONC((0 <= index) && (index <= (int)expr->references.size()));
+
+    // This expression may be happening in the context of a particular
+    // property. If so then the destination may or may not be
+    // resolved, which will have been handled when the containing package
+    // was loaded. Alternatively this expression may be evaluated inside
+    // some arbitrary Tcl code, in which case references remain unbound
+    // and need to be resolved the hard way.
+    CdlNode result = 0;
+    if (0 != this->property) {
+        // There is a property, use the bound/unbound reference.
+        result = expr->references[index].get_destination();
+    } else {
+        // The destination name can be retrieved, but we still need some
+        // way of resolving it.
+        if (0 != this->toplevel) {
+            std::string destination_name = expr->references[index].get_destination_name();
+            result = this->toplevel->lookup(destination_name);
+        }
+    }
+    
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// Ditto, but also check that the result is a valuable.
+CdlValuable
+CdlEvalContext::resolve_valuable_reference(CdlExpression expr, int index)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlEvalContext::resolve_reference", "result %");
+    CYG_REPORT_FUNCARG2XV(expr, index);
+
+    CdlValuable result  = 0;
+    CdlNode     node = this->resolve_reference(expr, index);
+    if (0 != node) {
+        result = dynamic_cast<CdlValuable>(node);
+    }
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+bool
+CdlEvalContext::check_this(cyg_assert_class_zeal zeal) const
+{
+    if (CdlEvalContext_Magic != cdlevalcontext_cookie) {
+        return false;
+    }
+    CYGDBG_MEMLEAK_CHECKTHIS();
+    
+    if ((0 != transaction) && !transaction->check_this(zeal)) {
+        return false;
+    }
+    if ((0 != toplevel) && !toplevel->check_this(zeal)) {
+        return false;
+    }
+    if ((0 != node) && !node->check_this(zeal)) {
+        return false;
+    }
+    if ((0 != property) && !property->check_this(zeal)) {
+        return false;
+    }
+    return true;
+}
+
+//}}}
+//{{{  Expression parsing               
+
+//{{{  Description                      
+
+// ----------------------------------------------------------------------------
+// There are a number of different entry points related to expression parsing,
+// largely to support list and goal expressions. All of these eventually
+// end up calling the function
+//    continue_parse(expr, data, index, token, token_end)
+//
+// The expr argument holds an existing expression object that needs to be
+// updated. If token is Invalid then we are at the start of an expression
+// (but not necessarily at the start of the string).
+//
+// The data string holds all of the expression that should be parsed.
+// It is formed by concatenating all non-option arguments to the
+// appropriate property command, with spaces between them.
+//
+// index is an input/output variable. On input it indicates where in
+// the string parsing should continue. On output it indicates the
+// location within the string where the terminating token began.
+//
+// token is an input/output variable. On input it can have the values
+// Invalid or And. The former means that we are parsing a completely
+// new expression. The latter is used for goal expressions: it is
+// necessary to parse a new expression and then combine it with the
+// existing one.
+//
+// token_end is an output variable. It indicates the location within
+// the string where the terminating token ended. This is useful for
+// e.g. ranges in a list expression.
+//
+// A conventional recursive descent parser is used.
+
+//}}}
+//{{{  Tokenization                     
+
+// ----------------------------------------------------------------------------
+// Tokenization.
+
+//{{{  token enum                       
+
+// A separate token enum is necessary, rather than re-using the CdlExprOp
+// enum. Some tokens may correspond to several operators, and some tokens
+// such as close-bracket do not correspond directly to an operator at all.
+enum token {
+    T_Invalid           = -2,
+
+    T_EOD               = -1,
+    T_Reference         =  1,   // CYGPKG_HAL
+    T_String            =  2,   // "hello"
+    T_Integer           =  3,   // 123
+    T_Double            =  4,   // 3.1415
+    T_Range             =  5,   // to
+    T_OpenBracket       =  6,   // (
+    T_CloseBracket      =  7,   // )
+    T_Minus             =  8,   // -
+    T_Plus              =  9,   // +
+    T_Times             = 10,   // *
+    T_Divide            = 11,   // /
+    T_Exclamation       = 12,   // !
+    T_Tilde             = 13,   // ~
+    T_Questionmark      = 14,   // ?
+    T_Remainder         = 15,   // %
+    T_LeftShift         = 16,   // <<
+    T_RightShift        = 17,   // >>
+    T_LessThan          = 18,   // <
+    T_LessEqual         = 19,   // <=
+    T_GreaterThan       = 20,   // >
+    T_GreaterEqual      = 21,   // >=
+    T_Equal             = 22,   // ==
+    T_NotEqual          = 23,   // !=
+    T_BitAnd            = 24,   // &
+    T_BitXor            = 25,   // ^
+    T_BitOr             = 26,   // |
+    T_And               = 27,   // &&
+    T_Or                = 28,   // ||
+    T_Colon             = 29,   // : (in a conditional)
+    T_StringConcat      = 30,   // .
+    T_Function          = 31,   // is_substr etc.
+    T_Comma             = 32,   // , (inside a function)
+    T_Implies           = 33,   // implies
+    T_Xor               = 34,   // xor
+    T_Eqv               = 35    // eqv
+    
+};
+
+//}}}
+//{{{  Statics                          
+
+// Statics to keep track of the current state.
+static std::string      current_data            = "";
+static unsigned int     current_index           = 0;
+static unsigned int     token_start             = 0;
+static int              current_char            = EOF;
+static token            current_token           = T_Invalid;
+static std::string      current_string          = "";
+static std::string      current_reference       = "";
+static std::string      current_special         = "";
+static cdl_int          current_int             = 0;
+static double           current_double          = 0.0;
+static CdlValueFormat   current_format          = CdlValueFormat_Default;
+static int              current_function_id     = 0;
+
+//}}}
+//{{{  Character access                 
+
+// ----------------------------------------------------------------------------
+// Individual character access.
+// Note that current_index is one character past current_char.
+
+// Return the next character in the string, or EOF
+static void
+next_char()
+{
+    if (current_index >= current_data.size()) {
+        current_char = EOF;
+    } else {
+        current_char = current_data[current_index++];
+    }
+}
+
+// Go back a character. This is useful when parsing
+// strings. It is the responsibility of the calling code
+// to make sure that we are not at the start of the buffer.
+static void
+backup_char()
+{
+    CYG_ASSERTC(((EOF == current_char) && (0 < current_index)) || (1 < current_index));
+    if (EOF != current_char) {
+        current_index--;
+    }
+    current_char = current_data[current_index - 1];
+}
+
+//}}}
+//{{{  get_error_location()             
+
+// ----------------------------------------------------------------------------
+// Construct part of a diagnostic message, indicating the
+// area in the data where the error occurred. This string
+// is of the form {...data} ^char^ {data...}. Ideally
+// the ^ markers would be on a subsequent line, eliminating
+// the need for braces, but there is insufficient control
+// of how the message gets presented to the user.
+//
+// Care has to be taken with EOD.
+static std::string
+get_error_location()
+{
+    CYG_REPORT_FUNCNAME("get_error_location");
+    std::string result = "";
+
+    // token_start is probably the best place for centering the error.
+    // current_index is past the point where the error has occurred.
+    if (token_start > 1) {
+        if (token_start > 16) {
+            result = "{..." + current_data.substr(token_start - 13, 13) + "} ";
+        } else {
+            result = "{" + current_data.substr(0, token_start) + "}";
+        }
+    }
+
+    if (current_char == EOF) {
+        result += " <end of data>";
+    } else {
+        result += " ^" + std::string(1, current_data[token_start]) + "^ ";
+    }
+
+    if (token_start < current_data.size()) {
+        if ((token_start + 16) < current_data.size()) {
+            result += "{" + current_data.substr(token_start + 1, current_data.size() - (token_start+1)) + "}";
+        } else {
+            result += "{" + current_data.substr(token_start, 13) + "...}";
+        }
+    }
+
+    CYG_REPORT_RETURN();
+    return result;
+}
+
+// Export this functionality available to other modules, especially func.cxx and its
+// argument checking routines.
+std::string
+CdlParse::get_expression_error_location(void)
+{
+    return get_error_location();
+}
+
+//}}}
+//{{{  Token translation                
+
+// ----------------------------------------------------------------------------
+
+// Convert a token into a binary expression operator
+static CdlExprOp
+token_to_binary_expr_op()
+{
+    CYG_REPORT_FUNCNAMETYPE("token_to_expr_op", "op %d");
+    CdlExprOp result = CdlExprOp_Invalid;
+
+    switch(current_token) {
+      case T_Minus:             result = CdlExprOp_Subtract; break;
+      case T_Plus:              result = CdlExprOp_Add; break;
+      case T_Times:             result = CdlExprOp_Multiply; break;
+      case T_Divide:            result = CdlExprOp_Divide; break;
+      case T_Remainder:         result = CdlExprOp_Remainder; break;
+      case T_LeftShift:         result = CdlExprOp_LeftShift; break;
+      case T_RightShift:        result = CdlExprOp_RightShift; break;
+      case T_LessThan:          result = CdlExprOp_LessThan; break;
+      case T_LessEqual:         result = CdlExprOp_LessEqual; break;
+      case T_GreaterThan:       result = CdlExprOp_GreaterThan; break;
+      case T_GreaterEqual:      result = CdlExprOp_GreaterEqual; break;
+      case T_Equal:             result = CdlExprOp_Equal; break;
+      case T_NotEqual:          result = CdlExprOp_NotEqual; break;
+      case T_BitAnd:            result = CdlExprOp_BitAnd; break;
+      case T_BitXor:            result = CdlExprOp_BitXor; break;
+      case T_BitOr:             result = CdlExprOp_BitOr; break;
+      case T_And:               result = CdlExprOp_And; break;
+      case T_Or:                result = CdlExprOp_Or; break;
+      case T_StringConcat:      result = CdlExprOp_StringConcat; break;
+      case T_Implies:           result = CdlExprOp_Implies; break;
+      case T_Xor:               result = CdlExprOp_Xor; break;
+      case T_Eqv:               result = CdlExprOp_Eqv; break;
+      default:                  result = CdlExprOp_Invalid; break;
+    }
+    
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// Convert a token into an ExprOp. This way the internal token enum does
+// not need to be exported in order to define the interface.
+//
+// In practice the higher level code will only look for a handful of
+// cases, mainly EOD and the range operator, but we might as well
+// do the job property.
+static CdlExprOp
+token_to_expr_op()
+{
+    CYG_REPORT_FUNCNAMETYPE("token_to_expr_op", "expr op %d");
+    CdlExprOp result;
+
+    // Many of the tokens are already handled for binary operators.
+    result = token_to_binary_expr_op();
+    if (CdlExprOp_Invalid == result) {
+        switch(current_token) {
+        case T_EOD:             result = CdlExprOp_EOD; break;
+        case T_Reference:       result = CdlExprOp_Reference; break;
+        case T_String:          result = CdlExprOp_StringConstant; break;
+        case T_Integer:         result = CdlExprOp_IntegerConstant; break;
+        case T_Double:          result = CdlExprOp_DoubleConstant; break;
+        case T_Range:           result = CdlExprOp_Range; break;
+        case T_Exclamation:     result = CdlExprOp_LogicalNot; break;
+        case T_Tilde:           result = CdlExprOp_BitNot; break;
+        case T_Questionmark:
+        case T_Colon:           result = CdlExprOp_Cond; break; // best guess
+        case T_Function:        result = CdlExprOp_Function; break;
+        case T_OpenBracket:
+        case T_CloseBracket:
+        case T_Invalid:
+        default:                result = CdlExprOp_Invalid; break;
+        }
+    }
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// A utility routine to turn the current token back into a string
+// This is used for diagnostics.
+static std::string
+token_to_string()
+{
+    CYG_REPORT_FUNCNAME("token_to_string");
+    std::string result = "";
+
+    switch(current_token) {
+      case T_EOD:               result = "<end of data>"; break;
+      case T_Reference:         result = "reference to " + current_reference; break;
+      case T_String:            result = "string \"" + current_string + "\""; break;
+      case T_Integer:
+      {
+          std::string tmp;
+          Cdl::integer_to_string(current_int, tmp, current_format);
+          result = "integer constant " + tmp;
+          break;
+      }
+      case T_Double:
+      {
+          std::string tmp;
+          Cdl::double_to_string(current_double, tmp, current_format);
+          result = "double constant " + tmp;
+          break;
+      }
+      case T_Range:             result = "range operator \"to\""; break;
+      case T_OpenBracket:       result = "open bracket ("; break;
+      case T_CloseBracket:      result = "close bracket )"; break;
+      case T_Minus:             result = "minus sign -"; break;
+      case T_Plus:              result = "plus sign +"; break;
+      case T_Times:             result = "multiply operator *"; break;
+      case T_Divide:            result = "divide operator /"; break;
+      case T_Exclamation:       result = "not operator !"; break;
+      case T_Tilde:             result = "bitwise not operator ~"; break;
+      case T_Questionmark:      result = "question mark ?"; break;
+      case T_Remainder:         result = "remainder operator %"; break;
+      case T_LeftShift:         result = "left shift operator <<"; break;
+      case T_RightShift:        result = "right shift operator >>"; break;
+      case T_LessThan:          result = "less-than operator <"; break;
+      case T_LessEqual:         result = "less-or-equal operator <="; break;
+      case T_GreaterThan:       result = "greater-than operator >"; break;
+      case T_GreaterEqual:      result = "greater-or-equal operator >="; break;
+      case T_Equal:             result = "equality operator =="; break;
+      case T_NotEqual:          result = "not-equal operator !="; break;
+      case T_BitAnd:            result = "bitwise and operator &"; break;
+      case T_BitXor:            result = "bitwise xor operator ^"; break;
+      case T_BitOr:             result = "bitwise or operator |"; break;
+      case T_And:               result = "and operator &&"; break;
+      case T_Or:                result = "or operator ||"; break;
+      case T_Colon:             result = "colon"; break;
+      case T_StringConcat:      result = "string concatenation operator ."; break;
+      case T_Implies:           result = "implies operator"; break;
+      case T_Xor:               result = "logical xor operator"; break;
+      case T_Eqv:               result = "logical equivalence operator eqv"; break;
+      case T_Function:          result = std::string("function call ") + CdlFunction::get_name(current_function_id); break;
+      case T_Invalid:
+      default:                  result = "<invalid token>"; break;
+    }
+
+    CYG_REPORT_RETURN();
+    return result;
+}
+
+//}}}
+//{{{  Literals                         
+
+// ----------------------------------------------------------------------------
+//{{{  process_string()                 
+
+// The start of a string has been detected. Work out the entire string,
+// allowing for backslash escapes.
+static void
+process_string()
+{
+    CYG_REPORT_FUNCNAME("process_string");
+    CYG_ASSERTC('"' == current_char);
+    CYG_ASSERTC("" == current_string);
+
+    std::string result = "";
+
+    // Move past the leading quote mark.
+    next_char();
+    while ('"' != current_char) {
+        if (EOF == current_char) {
+            throw CdlParseException("Premature end of data in string constant.\n" + get_error_location());
+        } else if ('\\' == current_char) {
+            // Allow \a, \b, \f, \n, \r, \t, \v, \ddd and \xhh.
+            // Also copy with \newline space.
+            // Any other character gets passed through unchanged.
+            next_char();
+            switch(current_char) {
+              case EOF:
+                throw CdlParseException("Premature end of data after backslash in string constant.\n" + get_error_location());
+              case 'a':
+                result += '\a';
+                break;
+              case 'b':
+                result += '\b';
+                break;
+              case 'f':
+                result += '\f';
+                break;
+              case 'n':
+                result += '\n';
+                break;
+              case 'r':
+                result += '\r';
+                break;
+              case 't':
+                result += '\t';
+                break;
+              case 'v':
+                result += '\v';
+                break;
+              case 'x':
+              {
+                cdl_int tmp = 0;
+                next_char();
+                if (!isxdigit(current_char)) {
+                    throw CdlParseException("Non-hexadecimal digit detected in string \\x escape sequence.\n" +
+                        get_error_location());
+                }
+                // NOTE: there is no overflow detection here.
+                do {
+                    tmp *= 16;
+                    if (('0' <= current_char) && (current_char <= '9')) {
+                        tmp += (current_char - '0');
+                    } else if (('a' <= current_char) && (current_char <= 'f')) {
+                        tmp += 10 + (current_char - 'a');
+                    } else if (('A' <= current_char) && (current_char <= 'F')) {
+                        tmp += 10 + (current_char - 'A');
+                    } else {
+                        CYG_FAIL("C library error, isxdigit() succeeded on non-hexadecimal character");
+                    }
+                    next_char();
+                } while(isxdigit(current_char));
+                backup_char();
+                result += (char) tmp;
+              }
+
+              case '\n':
+                next_char();
+                while ((EOF != current_char) && isspace(current_char)) {
+                    next_char();
+                }
+                // We have gone one too far, back up.
+                backup_char();
+                result += " ";
+                break;
+
+              default:
+                if (('0' <= current_char) && (current_char <= '7')) {
+                    // A sequence of octal digits.
+                    cdl_int tmp = 0;
+                    do {
+                        tmp = (8 * tmp) + (current_char - '0');
+                        next_char();
+                    } while (('0' <= current_char) && (current_char <= '7'));
+                    backup_char();
+                    result += (char) tmp;
+                } else {
+                    // For all other backslash sequences, just add the second character
+                    result += (char) current_char;
+                }
+            }
+        } else {
+            result += (char) current_char;
+        }
+        next_char();
+    }
+    // The closing quote has been reached, move past it.
+    next_char();
+
+    // And all done.
+    current_token  = T_String;
+    current_string = result;
+
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  process_number()                 
+
+// The start of a number has been detected. This number may be an
+// integer or a double. It is necessary to figure out where the number
+// ends and invoke the appropriate Cdl:: conversion utility.
+//
+// Care has to be taken with termination. Consider a token such as
+// 134_5. This is not a string because there are no quote marks, nor
+// is it a valid reference, and because it begins with a digit it
+// should be interpreted as a number. The 134 bit works fine, then
+// number processing stops leaving current_char as '_'. If we are
+// parsing a list expression then the following _5 will actually
+// be interpreted as a reference. To avoid this, here is a utility
+// which checks number completion and throws an exception if
+// necessary.
+static void check_number_termination()
+{
+    CYG_REPORT_FUNCNAME("check_number_termination");
+
+    // End-of-data or any whitespace is ok.
+    if ((EOF != current_char) && !isspace(current_char)) {
+        // Any valid operator is ok as well, or brackets for that matter.
+        if (('-' != current_char) && ('+' != current_char) && ('*' != current_char) &&
+            ('/' != current_char) && ('!' != current_char) && ('~' != current_char) &&
+            ('?' != current_char) && ('%' != current_char) && ('<' != current_char) &&
+            ('>' != current_char) && ('=' != current_char) && ('&' != current_char) &&
+            ('^' != current_char) && ('|' != current_char) && (':' != current_char) &&
+            ('(' != current_char) && (')' != current_char)) {
+
+            std::string tmp;
+            Cdl::integer_to_string(current_int, tmp);
+            throw CdlParseException("Invalid character detected after number " + tmp + "\n" + get_error_location());
+        }
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+process_number()
+{
+    CYG_REPORT_FUNCNAME("process_number");
+
+    std::string tmp      = "";
+    bool        is_float = false;
+
+    // Detect the special cases of 0x and octal numbers.
+    if ('0' == current_char) {
+        next_char();
+        if (('x' == current_char) || ('X' == current_char)) {
+            
+            next_char();
+            if (!isxdigit(current_char)) {
+                throw CdlParseException("Invalid hexadecimal number, expected at least one hexadecimal digit after 0x.\n"
+                                        + get_error_location());
+            }
+            current_int = 0;
+            do {
+                current_int *= 16;
+                if (('0' <= current_char) && (current_char <= '9')) {
+                    current_int += (current_char - '0');
+                } else if (('a' <= current_char) && (current_char <= 'f')) {
+                    current_int += 10 + (current_char - 'a');
+                } else {
+                    current_int += 10 + (current_char - 'A');
+                }
+                next_char();
+            } while(isxdigit(current_char));
+            current_token  = T_Integer;
+            current_format = CdlValueFormat_Hex;
+            check_number_termination();
+            CYG_REPORT_RETURN();
+            return;
+                
+        } else if (('0' <= current_char) && (current_char <= '7')) {
+
+            current_int = 0;
+            do {
+                current_int *= 8;
+                current_int += (current_char - '0');
+                next_char();
+            } while (('0' <= current_char) && (current_char <= '7'));
+            current_token  = T_Integer;
+            current_format = CdlValueFormat_Octal;
+            check_number_termination();
+            CYG_REPORT_RETURN();
+            return;
+            
+        } else if (('8' == current_char) || ('9' == current_char)) {
+            throw CdlParseException("08... and 09... are not valid  octal numbers.\n" + get_error_location());
+        } else {
+            // This could be plain 0, or 0.123
+            // Backup, and let the rest of the code take care of things
+            backup_char();
+        }
+    }
+    
+    do {
+        tmp += (char) current_char;
+        next_char();
+    } while(isdigit(current_char));
+
+    // If we have found a . then we have a floating point number with a fraction.
+    if ('.' == current_char) {
+        tmp += '.';
+        next_char();
+        if (!isdigit(current_char)) {
+            throw CdlParseException("Invalid floating point constant, expected a digit for the fractional part.\n" +
+                                    get_error_location());
+        }
+        is_float = true;
+        do {
+            tmp += (char) current_char;
+            next_char();
+        } while(isdigit(current_char));
+    }
+
+    // If we have found e or E then we have a floating point number with an exponent
+    if (('e' == current_char) || ('E' == current_char)) {
+        tmp += 'E';
+        next_char();
+        if (('+' == current_char) || ('-' == current_char)) {
+            tmp += current_char;
+            next_char();
+        }
+        if (!isdigit(current_char)) {
+            throw CdlParseException("Invalid floating point constant, expected a digit for the exponent.\n" +
+                                    get_error_location());
+        }
+        is_float = true;
+        do {
+            tmp += (char) current_char;
+            next_char();
+        } while(isdigit(current_char));
+    }
+
+    if (is_float) {
+        if (!Cdl::string_to_double(tmp, current_double)) {
+            throw CdlParseException("Invalid floating point constant `" + tmp + "'.\n" + get_error_location());
+        } else {
+            current_token = T_Double;
+        }
+    } else {
+        if (!Cdl::string_to_integer(tmp, current_int)) {
+            throw CdlParseException("Invalid integer constant `" + tmp + "'.\n" + get_error_location());
+        } else {
+            current_token = T_Integer;
+        }
+    }
+    
+    check_number_termination();
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  process_alphanumeric()           
+
+// The start of an alphanumeric sequence has been detected. This may
+// be a reference, a function call, or an operator like eq or to. All
+// such sequences must be a valid C preprocessor name, so the only
+// characters allowed are underscore, upper and lower case characters,
+// and digits. The first character cannot be a digit, but that has
+// been checked already.
+//
+// Some care has to be taken with locale's, the C library may decide
+// that a character is a letter even though the same character is not
+// valid as far as the preprocessor is concerned.
+static void
+process_alphanumeric()
+{
+    CYG_REPORT_FUNCNAME("process_alphanumeric");
+
+    do {
+       current_reference += (char) current_char;
+       next_char();
+    } while (('_' == current_char) || isdigit(current_char) ||
+             (('a' <= current_char) && (current_char <= 'z')) ||
+             (('A' <= current_char) && (current_char <= 'Z')));
+
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  process_special()                
+
+// Usually an alphanumeric sequence of characters is a reference, e.g.
+// CYGPKG_KERNEL. However there are only so many special characters
+// available so some operators are implemented as a sequence, e.g. 
+// "to". CDL also supports functions like is_substr().
+//
+// The data will have been collected into the current_reference string
+// by a call to process_alphanumeric().
+
+static bool
+process_special()
+{
+    CYG_REPORT_FUNCNAMETYPE("process_special", "special %d");
+    bool result = false;
+    
+    if ("to" == current_reference) {
+        current_token  = T_Range;
+        result = true;
+    } else if ("implies" == current_reference) {
+        current_token  = T_Implies;
+        result = true;
+    } else if ("xor" == current_reference) {
+        current_token  = T_Xor;
+        result = true;
+    } else if ("eqv" == current_reference) {
+        current_token  = T_Eqv;
+        result = true;
+    } else if (CdlFunction::is_function(current_reference.c_str(), current_function_id)) {
+        current_token  = T_Function;
+        result = true;
+    }
+
+    if (result) {
+        current_special     = current_reference;
+        current_reference   = "";
+    }
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+//}}}
+
+//}}}
+//{{{  next_token()                     
+
+// ----------------------------------------------------------------------------
+// Work out what the next token is. This includes the handling of
+// strings, integers, doubles, and references.
+static void
+next_token()
+{
+    CYG_REPORT_FUNCNAMETYPE("next_token", "token %d");
+
+    // Make sure there is no dross left lying around from the previous call.
+    current_token       = T_Invalid;
+    current_string      = "";
+    current_reference   = "";
+    current_special     = "";
+    current_int         = 0;
+    current_double      = 0.0;
+    current_format      = CdlValueFormat_Default;
+    current_function_id = 0;
+
+    // Skip leading white space. This includes newlines, tabs, etc,
+    // consider the case of:
+    //    ...
+    //    legal_values {
+    //        1
+    //        2
+    //        4
+    //        ..
+    //    }
+    //    ...
+    // which is perfectly legitimate. White space inside strings
+    // is handled by the string literal code, and does not get filtered
+    // out here.
+    //
+    // Exactly which characters are white-space is implementation-defined,
+    // so a special check for EOF is in order.
+    while ((EOF != current_char) && isspace(current_char)) {
+        next_char();
+    }
+
+    // Remember the token starting point. next_char() has actually moved
+    // the index on by one.
+    token_start = current_index - 1;
+
+    // The simple cases can be handled inline, the more complicated cases
+    // involve other functions
+    switch(current_char) {
+
+      case EOF:
+          current_token = T_EOD;
+          break;
+
+      case '"':
+          process_string();
+          break;
+
+      case '(':
+          current_token = T_OpenBracket;
+          next_char();
+          break;
+
+      case ')':
+          current_token = T_CloseBracket;
+          next_char();
+          break;
+
+          // At this level it is not possible to distinguish between
+          // unary and binary operators, so no attempt is made to
+          // turn - and + into part of a number.
+      case '-':
+          current_token = T_Minus;
+          next_char();
+          break;
+
+      case '+':
+          current_token = T_Plus;
+          next_char();
+          break;
+
+      case '*':
+          current_token = T_Times;
+          next_char();
+          break;
+
+      case '/':
+          current_token = T_Divide;
+          next_char();
+          break;
+
+      case '!':
+          next_char();
+          if ('=' == current_char) {
+              current_token = T_NotEqual;
+              next_char();
+          } else {
+              current_token = T_Exclamation;
+          }
+          break;
+
+      case '~':
+          current_token = T_Tilde;
+          next_char();
+          break;
+
+      case '?':
+          current_token = T_Questionmark;
+          next_char();
+          break;
+
+      case '%':
+          current_token = T_Remainder;
+          next_char();
+          break;
+
+      case '<':
+          next_char();
+          if ('<' == current_char) {
+              current_token = T_LeftShift;
+              next_char();
+          } else if ('=' == current_char) {
+              current_token = T_LessEqual;
+              next_char();
+          } else {
+              current_token = T_LessThan;
+          }
+          break;
+
+      case '>':
+          next_char();
+          if ('>' == current_char) {
+              current_token = T_RightShift;
+              next_char();
+          } else if ('=' == current_char) {
+              current_token = T_GreaterEqual;
+              next_char();
+          } else {
+              current_token = T_GreaterThan;
+          }
+          break;
+
+      case '=':
+          next_char();
+          if ('=' != current_char) {
+              throw CdlParseException(std::string("Incomplete == operator in expression.\n") + get_error_location());
+          } else {
+              current_token = T_Equal;
+              next_char();
+          }
+          break;
+
+      case '&':
+          next_char();
+          if ('&' == current_char) {
+              current_token = T_And;
+              next_char();
+          } else {
+              current_token = T_BitAnd;
+          }
+          break;
+
+      case '^':
+          current_token = T_BitXor;
+          next_char();
+          break;
+
+      case '|':
+          next_char();
+          if ('|' == current_char) {
+              current_token = T_Or;
+              next_char();
+          } else {
+              current_token = T_BitOr;
+          }
+          break;
+
+      case ':':
+          current_token = T_Colon;
+          next_char();
+          break;
+
+      case '.':
+          current_token = T_StringConcat;
+          next_char();
+          break;
+
+      case ',':
+          current_token = T_Comma;
+          next_char();
+          break;
+        
+      default:
+          // String constants have been handled already. The only
+          // valid tokens that are left are numbers, references,
+          // "specials" such as the range and string equality
+          // operators, and functions.
+          //
+          // Numbers should begin with a digit (plus and minus are
+          // tokenized separately).
+          //
+          // References must be valid C preprocessor symbols, i.e.
+          // they must begin with either a letter or an underscore.
+          // The range operator is handled most conveniently as
+          // a special case of a reference.
+          if (isdigit(current_char)) {
+              process_number();
+          } else if (('_' == current_char) ||
+                     (('a' <= current_char) && (current_char <= 'z')) ||
+                     (('A' <= current_char) && (current_char <= 'Z'))) {
+              process_alphanumeric();
+              if (!process_special()) {
+                  current_token = T_Reference;
+              }
+          } else {
+              std::string msg = "Unexpected character '";
+              msg += (char) current_char;
+              msg += "' in expression.\n";
+              msg += get_error_location();
+              throw CdlParseException(msg);
+          }
+          break;
+    }
+
+    CYG_REPORT_RETVAL(current_token);
+}
+
+//}}}
+//{{{  initialise_tokenisation()        
+
+// ----------------------------------------------------------------------------
+// This is called at the start of expression parsing. It
+// sets up the appropriate statics, and provides initial
+// values for current_char and current_token.
+static void
+initialise_tokenisation(std::string data, int index)
+{
+    CYG_REPORT_FUNCNAME("initialise_tokenization");
+
+    current_data        = data;
+    current_index       = static_cast<unsigned int>(index);
+    token_start         = current_index;
+    next_char();
+    next_token();
+
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+
+//}}}
+//{{{  Syntactic analysis               
+
+// ----------------------------------------------------------------------------
+// Syntactic analysis.
+//
+// The BNF of CDL expressions is something like this:
+//
+//   <expression>   ::= <conditional>
+//   <conditional>  ::= <implies> ? <conditional> : <conditional> | <implies>
+//   <implies>      ::= <eqv>    [<implies op>  <implies>]      implies
+//   <eqv>          ::= <or>     [<eqv op>      <eqv>]          xor, eqv        
+//   <or>           ::= <and>    [<or op>       <or>]           ||
+//   <and>          ::= <bitor>  [<and op>      <and>]          &&
+//   <bitor>        ::= <bitxor> [<bitor op>    <bitor>]        |
+//   <bitxor>       ::= <bitand> [<bitxor op>   <bitxor>]       ^
+//   <bitand>       ::= <eq>     [<bitand op>   <and>]          &
+//   <eq>           ::= <comp>   [<eq op>       <eq>]           == !=
+//   <comp>         ::= <shift>  [<comp op>     <comp>]         < <= > >=
+//   <shift>        ::= <add>    [<shift op>    <shift>]        << >>
+//   <add>          ::= <mult>   [<add op>      <add>]          + - .
+//   <mult>         ::= <unary>  [<mult op>     <mult>]         * / %
+//   <unary>        ::= -<unary> | +<unary> | !<unary> | *<unary> | ?<unary> |
+//                      ~<unary> |
+//                      <string constant> | <integer constant> |
+//                      <double constant> | <reference> |
+//                      ( <expression> ) | <function>
+//
+// There are separate functions for each of these terms.
+
+// A forward declaration, needed for bracketed subexpressions.
+static void parse_expression(CdlExpression);
+
+// A utility to add a reference to the current expression, returning
+// the index.
+static int
+push_reference(CdlExpression expr, const std::string& reference)
+{
+    CYG_REPORT_FUNCNAMETYPE("push_reference", "new index %d");
+    CYG_PRECONDITION_CLASSC(expr);
+
+    CdlReference ref(reference);
+    expr->references.push_back(ref);
+    int result = (int) expr->references.size() - 1;
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// A utility to add a subexpression, returning its index.
+static void
+push_subexpression(CdlExpression expr, const CdlSubexpression& subexpr)
+{
+    CYG_REPORT_FUNCNAME("push_subexpression");
+    CYG_PRECONDITION_CLASSC(expr);
+
+    expr->sub_expressions.push_back(subexpr);
+    expr->first_subexpression = ((int) expr->sub_expressions.size()) - 1;
+
+    CYG_REPORT_RETURN();
+}
+
+// Another utility to hold of the most recent subexpression
+static CdlSubexpression&
+current_subexpression(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("current_subexpression");
+
+    CdlSubexpression& result = expr->sub_expressions[expr->first_subexpression];
+
+    CYG_REPORT_RETURN();
+    return result;
+}
+
+static void
+parse_function(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_function");
+    CYG_REPORT_FUNCARG1XV(expr);
+    CYG_PRECONDITION_CLASSC(expr);
+
+    CdlSubexpression subexpr;
+    subexpr.op          = CdlExprOp_Function;
+    subexpr.func        = current_function_id;
+
+    int number_of_args  = CdlFunction::get_args_count(current_function_id);
+    CYG_ASSERTC((0 < number_of_args) && (number_of_args <= CdlFunction_MaxArgs));
+    std::string name    = current_special;
+
+    // check for the opening bracket: xyzzy(arg1, arg2)
+    next_token();
+    if (T_OpenBracket != current_token) {
+        throw CdlParseException(std::string("Expected opening bracket after function ") + name + "\n" + get_error_location());
+    }
+    next_token();
+
+    int i;
+    for (i = 0; i < number_of_args; i++) {
+        parse_expression(expr);
+        subexpr.args[i] = expr->first_subexpression;
+        if (i < (number_of_args - 1)) {
+            if (T_Comma != current_token) {
+                throw CdlParseException(std::string("Expected comma between arguments in function ") +
+                                        name + "\n" + get_error_location());
+            }
+            next_token();
+        }
+    }
+    if (T_Comma == current_token) {
+        throw CdlParseException(std::string("Too many arguments passed to function ") + name + "\n" + get_error_location());
+    }
+    if (T_CloseBracket != current_token) {
+        throw CdlParseException(std::string("Expected closing bracket after function ") + name + "\n" + get_error_location());
+    }
+    next_token();
+    
+    // Allow the function implementation to check its arguments if it is so inclined.
+    CdlFunction::check(expr, subexpr);
+    
+    push_subexpression(expr, subexpr);
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_unary(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_operand");
+    CYG_REPORT_FUNCARG1XV(expr);
+    CYG_PRECONDITION_CLASSC(expr);
+
+    CdlSubexpression subexpr;
+
+    switch(current_token) {
+      case T_EOD :
+      {
+        // This warrants a special case
+        throw CdlParseException("End of expression reached when expecting an operand.\n" + get_error_location());
+      }
+
+      case T_Function :
+      {
+          parse_function(expr);
+          break;
+      }
+      
+      case T_Reference :
+      {
+        subexpr.op              = CdlExprOp_Reference;
+        subexpr.reference_index = push_reference(expr, current_reference);
+        push_subexpression(expr, subexpr);
+        next_token();
+        break;
+      }
+      
+      case T_String :
+      {
+        subexpr.op              = CdlExprOp_StringConstant;
+        subexpr.constants       = current_string;
+        push_subexpression(expr, subexpr);
+        next_token();
+        break;
+      }
+      
+      case T_Integer :
+      {
+        subexpr.op               = CdlExprOp_IntegerConstant;
+        subexpr.constants.set_integer_value(current_int, current_format);
+        push_subexpression(expr, subexpr);
+        next_token();
+        break;
+      }
+      
+      case T_Double :
+      {
+        subexpr.op              = CdlExprOp_DoubleConstant;
+        subexpr.constants.set_double_value(current_double, current_format);
+        push_subexpression(expr, subexpr);
+        next_token();
+        break;
+      }
+      
+      case T_OpenBracket :
+      {
+        next_token();
+        parse_expression(expr);
+        if (T_CloseBracket != current_token) {
+            throw CdlParseException("Missing close bracket after subexpression.\n" + get_error_location());
+        }
+        next_token();
+        break;
+      }
+      
+      case T_Minus :
+      {
+        next_token();
+        parse_unary(expr);
+        CdlSubexpression& last_sub      = current_subexpression(expr);
+        if (CdlExprOp_IntegerConstant == last_sub.op) {
+            // Do the negating inline, no need for another subexpression.
+            last_sub.constants = last_sub.constants.get_integer_value() * -1;
+        } else if (CdlExprOp_DoubleConstant == last_sub.op) {
+            last_sub.constants = last_sub.constants.get_double_value() * -1;
+        } else {
+            // We could detect certain cases such as string constants etc.
+            // For now don't bother.
+            subexpr.op          = CdlExprOp_Negate;
+            subexpr.lhs_index   = expr->first_subexpression;
+            push_subexpression(expr, subexpr);
+        }
+        break;
+      }
+      
+      case T_Plus :
+      {
+        next_token();
+        parse_unary(expr);
+        CdlSubexpression& last_sub      = current_subexpression(expr);
+        if ((CdlExprOp_IntegerConstant == last_sub.op) || (CdlExprOp_DoubleConstant == last_sub.op)) {
+            // No need to do anything here.
+        } else {
+            subexpr.op          = CdlExprOp_Plus;
+            subexpr.lhs_index   = expr->first_subexpression;
+            push_subexpression(expr, subexpr);
+        }
+        break;
+      }
+
+      case T_Times :
+      {
+          next_token();
+          parse_unary(expr);
+          subexpr.op            = CdlExprOp_Indirect;
+          subexpr.lhs_index     = expr->first_subexpression;
+          push_subexpression(expr, subexpr);
+          break;
+      }
+      
+      case T_Exclamation :
+      {
+          next_token();
+          parse_unary(expr);
+          subexpr.op            = CdlExprOp_LogicalNot;
+          subexpr.lhs_index     = expr->first_subexpression;
+          push_subexpression(expr, subexpr);
+          break;
+      }
+
+      case T_Tilde :
+      {
+          next_token();
+          parse_unary(expr);
+          subexpr.op            = CdlExprOp_BitNot;
+          subexpr.lhs_index     = expr->first_subexpression;
+          push_subexpression(expr, subexpr);
+          break;
+      }
+
+      case T_Questionmark:
+      {
+          // This is the `active' operator, it can only be applied directly to a reference.
+          next_token();
+          parse_unary(expr);
+          CdlSubexpression& last_sub = current_subexpression(expr);
+          if (CdlExprOp_Reference != last_sub.op) {
+              throw CdlParseException("The active operator ? can only be applied directly to a reference.\n" +
+                                      get_error_location());
+          }
+          // There is no point in creating a new subexpression object, just modify
+          // the existing one. This has the useful side effect of avoiding
+          // reference substitution in the eval code.
+          last_sub.op           = CdlExprOp_Active;
+          break;
+      }
+      default:
+      {
+        throw CdlParseException("Unexpected token `" + token_to_string() + "', expecting an operand.\n" +
+                                get_error_location());
+      }
+    }
+
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_multiply(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_multiply");
+    
+    parse_unary(expr);
+    while ((T_Times == current_token) || (T_Divide == current_token) || (T_Remainder == current_token)) {
+
+        CdlSubexpression subexpr;
+        subexpr.op      =
+            (T_Times  == current_token) ? CdlExprOp_Multiply :
+            (T_Divide == current_token) ? CdlExprOp_Divide : CdlExprOp_Remainder;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_unary(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_add(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_add");
+    
+    parse_multiply(expr);
+    while ((T_Plus == current_token)  ||
+           (T_Minus == current_token) ||
+           (T_StringConcat == current_token)) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = (T_Plus == current_token) ? CdlExprOp_Add :
+                     (T_Minus == current_token) ? CdlExprOp_Subtract :
+                     CdlExprOp_StringConcat;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_multiply(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_shift(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_shift");
+    
+    parse_add(expr);
+    while ((T_LeftShift == current_token) || (T_RightShift == current_token)) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = (T_LeftShift == current_token) ? CdlExprOp_LeftShift : CdlExprOp_RightShift;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_add(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_comparison(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_comparison");
+    
+    parse_shift(expr);
+    while ((T_LessThan == current_token)    || (T_LessEqual    == current_token) ||
+           (T_GreaterThan == current_token) || (T_GreaterEqual == current_token))  {
+
+        CdlSubexpression subexpr;
+        subexpr.op =
+            (T_LessThan    == current_token) ? CdlExprOp_LessThan : 
+            (T_LessEqual   == current_token) ? CdlExprOp_LessEqual :
+            (T_GreaterThan == current_token) ? CdlExprOp_GreaterThan : CdlExprOp_GreaterEqual;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_shift(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_equals(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_equals");
+    
+    parse_comparison(expr);
+    while ((T_Equal == current_token) ||
+           (T_NotEqual == current_token)) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = (T_Equal == current_token) ? CdlExprOp_Equal : CdlExprOp_NotEqual;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_comparison(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_bitand(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_bitand");
+    
+    parse_equals(expr);
+    while (T_BitAnd == current_token) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_BitAnd;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_equals(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_bitxor(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_bitxor");
+    
+    parse_bitand(expr);
+    while (T_BitXor == current_token) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_BitXor;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_bitand(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_bitor(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_bitor");
+    
+    parse_bitxor(expr);
+    while (T_BitOr == current_token) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_BitOr;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_bitxor(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_and(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_and");
+    parse_bitor(expr);
+    while (T_And == current_token) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_And;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_bitor(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_or(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_or");
+
+    parse_and(expr);
+    while (T_Or == current_token) {
+
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_Or;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_and(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_eqv(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_eqv");
+
+    parse_or(expr);
+    while ((T_Xor == current_token) || (T_Eqv == current_token)) {
+        
+        CdlSubexpression subexpr;
+        subexpr.op = (T_Xor == current_token) ? CdlExprOp_Xor : CdlExprOp_Eqv;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_or(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_implies(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_implies");
+
+    parse_eqv(expr);
+    while (T_Implies == current_token) {
+        
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_Implies;
+        subexpr.lhs_index = expr->first_subexpression;
+        
+        next_token();
+        parse_eqv(expr);
+
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_conditional(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_conditional");
+
+    parse_implies(expr);
+    if (T_Questionmark == current_token) {
+        CdlSubexpression subexpr;
+        subexpr.op = CdlExprOp_Cond;
+        subexpr.lhs_index = expr->first_subexpression;
+
+        next_token();
+        parse_conditional(expr);
+        subexpr.rhs_index = expr->first_subexpression;
+
+        if (T_Colon != current_token) {
+            throw CdlParseException("Expected colon in conditional expression.\n" + get_error_location());
+        }
+
+        next_token();
+        parse_conditional(expr);
+        subexpr.rrhs_index = expr->first_subexpression;
+
+        push_subexpression(expr, subexpr);
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+static void
+parse_expression(CdlExpression expr)
+{
+    CYG_REPORT_FUNCNAME("parse_expression");
+
+    parse_conditional(expr);
+    
+    CYG_REPORT_RETURN();
+}
+
+// ----------------------------------------------------------------------------
+// The entry point.
+void
+CdlExpressionBody::continue_parse(CdlExpression expr, std::string data, int& index, CdlExprOp& token, int& token_end)
+{
+    CYG_REPORT_FUNCNAME("CdlExpression::continue_parse");
+    CYG_REPORT_FUNCARG1XV(expr);
+    CYG_PRECONDITION_CLASSC(expr);
+    CYG_PRECONDITIONC((CdlExprOp_Invalid == token) || (CdlExprOp_And == token));
+
+    int current_subexpr = expr->first_subexpression;
+    initialise_tokenisation(data, index);
+    parse_expression(expr);
+    if (CdlExprOp_And == token) {
+        CdlSubexpression subexpr;
+        subexpr.op        = CdlExprOp_And;
+        subexpr.lhs_index = current_subexpr;
+        subexpr.rhs_index = expr->first_subexpression;
+        push_subexpression(expr, subexpr);
+    }
+    token       = token_to_expr_op();
+    index       = token_start;
+    token_end   = current_index;
+
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+
+//}}}
+//{{{  Expression Evaluation            
+
+// ----------------------------------------------------------------------------
+// Expression evaluation. This always happens in the context of a
+// particular toplevel. The parsed expression is held in what amounts
+// to a simple tree, so evaluation involves some recursion and a big
+// switch statement.
+
+static void
+evaluate_subexpr(CdlEvalContext& context, CdlExpression expr, int subexpr_index, CdlSimpleValue& result)
+{
+    CYG_REPORT_FUNCNAME("evaluate_subexpr");
+    CYG_REPORT_FUNCARG2XV(expr, subexpr_index);
+    CYG_ASSERTC((subexpr_index >= 0) && ((unsigned int)subexpr_index < expr->sub_expressions.size()));
+
+    const CdlSubexpression& subexpr = expr->sub_expressions[subexpr_index];
+    switch(subexpr.op) {
+    case CdlExprOp_StringConstant :
+    case CdlExprOp_IntegerConstant :
+    case CdlExprOp_DoubleConstant :
+    {
+        result = subexpr.constants;
+        break;
+    }
+    case CdlExprOp_Function :
+    {
+        CdlFunction::eval(context, expr, subexpr, result);
+        break;
+    }
+    case CdlExprOp_Reference :
+    {
+        // This expression may be happening in the context of a particular
+        // property. If so then the destination may or may not be resolved,
+        // and this is significant in the context of loading and unloading.
+        // Alternatively this expression may be being evaluated inside
+        // some Tcl code, with no particular context.
+        CdlNode destination = 0;
+        if (0 != context.property) {
+            // There is a property, use the bound/unbound reference.
+            destination = expr->references[subexpr.reference_index].get_destination();
+        } else {
+            // The destination name can be retrieved, but we still need some
+            // way of resolving it.
+            if (0 != context.toplevel) {
+                std::string destination_name = expr->references[subexpr.reference_index].get_destination_name();
+                destination = context.toplevel->lookup(destination_name);
+            }
+        }
+        if (0 == destination) {
+            // There are two ways of handling this.
+            //   1) throw an eval exception, which will usually result
+            //      in a new conflict object
+            //   2) substitute a value of 0.
+            // There should already be a conflict object for an
+            // unresolved reference, and having two conflicts for
+            // essentially the same error is not useful. Using a value
+            // of 0 allows things to continue for a bit longer. It is
+            // consistent with active vs. inactive values, gives
+            // basically the right result for "requires" properties,
+            // and so on.
+            //
+            // For now option (2) has it, but this decision may be
+            // reversed in future.
+            result = false;
+        } else {
+            CdlValuable valuable = dynamic_cast<CdlValuable>(destination);
+            if (0 == valuable) {
+                // This is a serious problem, an exception is warranted.
+                throw CdlEvalException("The expression references `" + destination->get_class_name() + " " +
+                                       destination->get_name() + "' which does not have a value.");
+            } else {
+                CdlSimpleValue::eval_valuable(context, valuable, result);
+            }
+        }
+        break;
+    }
+    case CdlExprOp_Negate :
+    {
+        // Unary -. Evaluate the target. If it is numeric, fine. Otherwise
+        // an error is warranted.
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if (result.has_integer_value()) {
+            result.set_integer_value(-1 * result.get_integer_value());
+        } else if (result.has_double_value()) {
+            result.set_double_value(-1.0 * result.get_double_value());
+        } else {
+            throw CdlEvalException("Attempt to negate non-numeric value `" + result.get_value() + "'.");
+        }
+        break;
+    }
+    case CdlExprOp_Plus :
+    {
+        // Unary +. Essentially this just checks that the current value is numeric.
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if ((!result.has_integer_value()) && (!result.has_double_value())) {
+            throw CdlEvalException("Attempt to apply unary + operator to non-numeric value `" + result.get_value() + "'.");
+        }
+        break;
+    }
+    case CdlExprOp_LogicalNot :
+    {
+        // !x
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if (result.get_bool_value()) {
+            result = false;;
+        } else {
+            result = true;
+        }
+        result.set_value_format(CdlValueFormat_Default);
+        break;
+    }
+    case CdlExprOp_BitNot :
+    {
+        // ~x. The operand must be an integer value.
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if (result.has_integer_value()) {
+            cdl_int tmp = result.get_integer_value();
+            result = ~tmp;
+        } else {
+            throw CdlEvalException("Attempt to apply unary ~ operator to non-integer value `" + result.get_value() + "'.");
+        }
+        break;
+    }
+    case CdlExprOp_Indirect :
+    {
+        // *x. The operand must evaluate to a string, and that string should be
+        // the name of a CdlValuable object.
+        CdlNode destination = 0;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        std::string name = result.get_value();
+        
+        if (0 != context.toplevel) {
+            destination = context.toplevel->lookup(name);
+        } else {
+            CYG_FAIL("This situation should probably never happen.");
+        }
+        
+        if (0 == destination) {
+            throw CdlEvalException("Attempt to apply unary indirection operator * to `" + name +
+                                   "', which is not the name of a known CDL entity.");
+        } else {
+            CdlValuable valuable = dynamic_cast<CdlValuable>(destination);
+            if (0 == valuable) {
+                throw CdlEvalException("Attempt to apply unary indirection operator * to `" + name +
+                                       "', which does not have a value.");
+            } else {
+                CdlSimpleValue::eval_valuable(context, valuable, result);
+            }
+        }
+        break;
+    }
+    case CdlExprOp_Active :
+    {
+        // ?x. If x is currently unresolved then default to 0.
+        // See the CdlExprOp_Reference code above for a similar case.
+        CdlNode destination = 0;
+        if (0 != context.property) {
+            destination =  expr->references[subexpr.reference_index].get_destination();
+        } else {
+            if (0 != context.toplevel) {
+                std::string destination_name = expr->references[subexpr.reference_index].get_destination_name();
+                destination = context.toplevel->lookup(destination_name);
+            }
+        }
+
+        bool active = false;
+        if ((0 != destination) && context.transaction->is_active(destination)) {
+            active = true;
+        }
+        if (active) {
+            result = true;
+        } else {
+            result = false;
+        }
+        break;
+    }
+    case CdlExprOp_Multiply :
+    {
+        // x * y. For now this only makes sense for numerical data,
+        // but it is possible to mix and match integer and double
+        // precision data.
+        //
+        // Strictly speaking the rhs need only be evaluated if it
+        // is known that the lhs is numeric.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((!(lhs.has_integer_value() || lhs.has_double_value())) ||
+            (!(rhs.has_integer_value() || rhs.has_double_value()))) {
+            throw CdlEvalException("Attempt to multiply non-numerical values: `" + lhs.get_value() + "' * `" +
+                                   rhs.get_value() + "'.");
+        }
+        if (lhs.has_integer_value() && rhs.has_integer_value()) {
+            result = lhs.get_integer_value() * rhs.get_integer_value();
+        } else {
+            result = lhs.get_double_value() * rhs.get_double_value();
+        }
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_Divide :
+    {
+        // x / y. Basically the same as multiplication, apart from a check for
+        // division by zero.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((!(lhs.has_integer_value() || lhs.has_double_value())) ||
+            (!(rhs.has_integer_value() || rhs.has_double_value()))) {
+            throw CdlEvalException("Attempt to divide non-numerical values: `" + lhs.get_value() + "' / `" +
+                                   rhs.get_value() + "'.");
+        }
+        if (lhs.has_integer_value() && rhs.has_integer_value()) {
+            cdl_int rhs_val = rhs.get_integer_value();
+            if (0 == rhs_val) {
+                throw CdlEvalException("Division by zero error: `" + lhs.get_value() + "' / `" + rhs.get_value() + "'.");
+            } else {
+                result = lhs.get_integer_value() / rhs_val;
+            }
+        } else {
+            double rhs_val = rhs.get_double_value();
+            if (0.0 == rhs_val) {
+                throw CdlEvalException("Division by zero error: `" + lhs.get_value() + "' / `" + rhs.get_value() + "'.");
+            }
+            result = lhs.get_double_value() / rhs_val;
+        }
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_Remainder :
+    {
+        // x % y. Both operands must be integral.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if (!(lhs.has_integer_value() && rhs.has_integer_value())) {
+            throw CdlEvalException("Attempt to use the remainder operator on non integral data: `" +
+                                   lhs.get_value() + "' % `" + rhs.get_value() + "'.");
+        }
+        cdl_int rhs_val = rhs.get_integer_value();
+        if (0 == rhs_val) {
+            throw CdlEvalException("Division by zero error: `" + lhs.get_value() + "' % `" + rhs.get_value() + "'.");
+        }
+        result = lhs.get_integer_value() % rhs_val;
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_Add :
+    {
+        // x + y. For now this only makes sense for numerical data,
+        // but it is possible to mix and match integer and double
+        // precision data. Arguably for string data this operator
+        // should mean concatenation, but it would probably be
+        // safer to have a separate operator for that.
+        //
+        // Strictly speaking the rhs need only be evaluated if it
+        // is known that the lhs is numeric.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((!(lhs.has_integer_value() || lhs.has_double_value())) ||
+            (!(rhs.has_integer_value() || rhs.has_double_value()))) {
+            throw CdlEvalException("Attempt to add non-numerical values: `" + lhs.get_value() + "' + `" +
+                                   rhs.get_value() + "'.");
+        }
+        if (lhs.has_integer_value() && rhs.has_integer_value()) {
+            result = lhs.get_integer_value() + rhs.get_integer_value();
+        } else {
+            result = lhs.get_double_value() + rhs.get_double_value();
+        }
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_Subtract :
+    {
+        // x - y. Again only numerical data is supported for now.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((!(lhs.has_integer_value() || lhs.has_double_value())) ||
+            (!(rhs.has_integer_value() || rhs.has_double_value()))) {
+            throw CdlEvalException("Attempt to subtract non-numerical values: `" + lhs.get_value() + "' - `" +
+                                   rhs.get_value() + "'.");
+        }
+        if (lhs.has_integer_value() && rhs.has_integer_value()) {
+            result = lhs.get_integer_value() - rhs.get_integer_value();
+        } else {
+            result = lhs.get_double_value() - rhs.get_double_value();
+        }
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_LeftShift :
+    {
+        // x << y. Both operands must be integral. For now there is no
+        // check on the value of y.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if (!(lhs.has_integer_value() && rhs.has_integer_value())) {
+            throw CdlEvalException("Attempt to use the left-shift operator on non integral data: `" +
+                                   lhs.get_value() + "' << `" + rhs.get_value() + "'.");
+        }
+        result = lhs.get_integer_value() << rhs.get_integer_value();
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_RightShift :
+    {
+        // x >> y. Both operands must be integral. For now there is no
+        // check on the value of y.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if (!(lhs.has_integer_value() && rhs.has_integer_value())) {
+            throw CdlEvalException("Attempt to use the right-shift operator on non integral data: `" +
+                                   lhs.get_value() + "' >> `" + rhs.get_value() + "'.");
+        }
+        result = lhs.get_integer_value() >> rhs.get_integer_value();
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_LessThan :
+    case CdlExprOp_LessEqual :
+    case CdlExprOp_GreaterThan :
+    case CdlExprOp_GreaterEqual :
+    {
+        // x < y, and similar comparison operators. These share
+        // sufficient code to warrant a common implementation. Only
+        // numerical data is supported for now. These operator could
+        // be interpreted as e.g. substring operations, but arguably
+        // separate operators would be better for that.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((!(lhs.has_integer_value() || lhs.has_double_value())) ||
+            (!(rhs.has_integer_value() || rhs.has_double_value()))) {
+            
+            std::string op_str =
+                (CdlExprOp_LessThan    == subexpr.op) ? "<" :
+                (CdlExprOp_LessEqual   == subexpr.op) ? "<=" :
+                (CdlExprOp_GreaterThan == subexpr.op) ? ">" : ">=";
+
+            throw CdlEvalException("Attempt to compare non-numerical values: `" + lhs.get_value() +
+                                   "' " + op_str + " `" + rhs.get_value() + "'.");
+        }
+        bool val = false;
+        if (lhs.has_integer_value() && rhs.has_integer_value()) {
+            cdl_int lhs_val = lhs.get_integer_value();
+            cdl_int rhs_val = rhs.get_integer_value();
+            val =
+                (CdlExprOp_LessThan    == subexpr.op) ? (lhs_val <  rhs_val) :
+                (CdlExprOp_LessEqual   == subexpr.op) ? (lhs_val <= rhs_val) :
+                (CdlExprOp_GreaterThan == subexpr.op) ? (lhs_val >  rhs_val) : (lhs_val >= rhs_val);
+        } else {
+            double lhs_val = lhs.get_double_value();
+            double rhs_val = rhs.get_double_value();
+            val =
+                (CdlExprOp_LessThan    == subexpr.op) ? (lhs_val <  rhs_val) :
+                (CdlExprOp_LessEqual   == subexpr.op) ? (lhs_val <= rhs_val) :
+                (CdlExprOp_GreaterThan == subexpr.op) ? (lhs_val >  rhs_val) : (lhs_val >= rhs_val);
+        }
+        result = val;
+        break;
+    }
+    case CdlExprOp_Equal :
+    {
+        // x == y. For numerical data this should be a numerical comparison.
+        // Otherwise a string comparison has to be used.
+        bool val = false;
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((lhs.has_integer_value() || lhs.has_double_value()) &&
+            (rhs.has_integer_value() || rhs.has_double_value())) {
+
+            if (lhs.has_integer_value() && rhs.has_integer_value()) {
+                if (lhs.get_integer_value() == rhs.get_integer_value()) {
+                    val = true;
+                } else {
+                    val = false;
+                }
+            } else {
+                if (lhs.get_double_value() == rhs.get_double_value()) {
+                    val = true;
+                } else {
+                    val = false;
+                }
+                  
+            }
+        } else {
+            // At least one of the two sides is non-numerical. Do a string comparison.
+            if (lhs.get_value() == rhs.get_value()) {
+                val = true;
+            } else {
+                val = false;
+            }
+        }
+        result = val;
+        break;
+    }
+    case CdlExprOp_NotEqual :
+    {
+        // x != y. For numerical data this should be a numerical comparison.
+        // Otherwise a string comparison has to be used.
+        bool val = false;
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if ((lhs.has_integer_value() || lhs.has_double_value()) &&
+            (rhs.has_integer_value() || rhs.has_double_value())) {
+
+            if (lhs.has_integer_value() && rhs.has_integer_value()) {
+                if (lhs.get_integer_value() != rhs.get_integer_value()) {
+                    val = true;
+                } else {
+                    val = false;
+                }
+            } else {
+                if (lhs.get_double_value() != rhs.get_double_value()) {
+                    val = true;
+                } else {
+                    val = false;
+                }
+                  
+            }
+        } else {
+            // At least one of the two sides is non-numerical. Do a string comparison.
+            if (lhs.get_value() != rhs.get_value()) {
+                val = true;
+            } else {
+                val = false;
+            }
+        }
+        result = val;
+        break;
+    }
+    case CdlExprOp_BitAnd :
+    {
+        // x & y. Only integer data is supported.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if (!(lhs.has_integer_value() && rhs.has_integer_value())) {
+            throw CdlEvalException("Attempt to use the bitwise and operator on non integral data: `" +
+                                   lhs.get_value() + "' & `" + rhs.get_value() + "'.");
+        }
+        result = lhs.get_integer_value() & rhs.get_integer_value();
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_BitXor :
+    {
+        // x ^ y. Only integer data is supported.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if (!(lhs.has_integer_value() && rhs.has_integer_value())) {
+            throw CdlEvalException("Attempt to use the bitwise xor operator on non integral data: `" +
+                                   lhs.get_value() + "' ^ `" + rhs.get_value() + "'.");
+        }
+        result = lhs.get_integer_value() ^ rhs.get_integer_value();
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_BitOr :
+    {
+        // x | y. Only integer data is supported.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        if (!(lhs.has_integer_value() && rhs.has_integer_value())) {
+            throw CdlEvalException("Attempt to use the bitwise or operator on non integral data: `" +
+                                   lhs.get_value() + "' | `" + rhs.get_value() + "'.");
+        }
+        result = lhs.get_integer_value() | rhs.get_integer_value();
+        result.set_value_format(lhs, rhs);
+        break;
+    }
+    case CdlExprOp_And :
+    {
+        // x && y. Both sides should be interpreted as boolean values,
+        // and "y" should only be evaluated if necessary.
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if (!result.get_bool_value()) {
+            result = false;
+        } else {
+            evaluate_subexpr(context, expr, subexpr.rhs_index, result);
+            if (result.get_bool_value()) {
+                result = true;
+            } else {
+                result = false;
+            }
+        }
+        break;
+    }
+    case CdlExprOp_Or :
+    {
+        // x || y. Both sides should be interpreted as boolean values,
+        // and "y" should only be evaluated if necessary.
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if (result.get_bool_value()) {
+            result = true;
+        } else {
+            evaluate_subexpr(context, expr, subexpr.rhs_index, result);
+            if (result.get_bool_value()) {
+                result = true;
+            } else {
+                result = false;
+            }
+        }
+        break;
+    }
+    case CdlExprOp_Xor :
+    {
+        // x xor y. Both sides should be interpreted as boolean values.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+
+        bool lhs_bool = lhs.get_bool_value();
+        bool rhs_bool = rhs.get_bool_value();
+        if ((lhs_bool && !rhs_bool) || (!lhs_bool && rhs_bool)) {
+            result = true;
+        } else {
+            result = false;
+        }
+        
+        break;
+    }
+    case CdlExprOp_Eqv :
+    {
+        // x eqv y. Both sides should be interpreted as boolean values.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+
+        bool lhs_bool = lhs.get_bool_value();
+        bool rhs_bool = rhs.get_bool_value();
+        if ((!lhs_bool && !rhs_bool) || (lhs_bool && rhs_bool)) {
+            result = true;
+        } else {
+            result = false;
+        }
+        
+        break;
+    }
+    case CdlExprOp_Implies :
+    {
+        // x implies y. Both sides should be interpreted as boolean values.
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+
+        bool lhs_bool = lhs.get_bool_value();
+        bool rhs_bool = rhs.get_bool_value();
+        if (!lhs_bool || rhs_bool) {
+            result = true;
+        } else {
+            result = false;
+        }
+        
+        break;
+    }
+    case CdlExprOp_Cond :
+    {
+        // x ? a : b.
+        // First evaluate the condition. Then evaluate either the second
+        // or third argument, as appropriate.
+        evaluate_subexpr(context, expr, subexpr.lhs_index, result);
+        if (result.get_bool_value()) {
+            evaluate_subexpr(context, expr, subexpr.rhs_index, result);
+        } else {
+            evaluate_subexpr(context, expr, subexpr.rrhs_index, result);
+        }
+        break;
+    }
+    case CdlExprOp_StringConcat :
+    {
+        // a . b
+        CdlSimpleValue lhs;
+        CdlSimpleValue rhs;
+        evaluate_subexpr(context, expr, subexpr.lhs_index, lhs);
+        evaluate_subexpr(context, expr, subexpr.rhs_index, rhs);
+        result = lhs.get_value() + rhs.get_value();
+        break;
+    }
+
+    default:
+        break;
+    }
+
+    CYG_REPORT_RETURN();
+}
+
+// ----------------------------------------------------------------------------
+void
+CdlExpressionBody::eval_internal(CdlEvalContext& context, CdlSimpleValue& result)
+{
+    CYG_REPORT_FUNCNAME("CdlExpression::eval_internal)");
+    CYG_REPORT_FUNCARG3XV(this, &context, &result);
+    CYG_INVARIANT_THISC(CdlExpressionBody);
+    CYG_PRECONDITION_CLASSOC(context);
+
+    evaluate_subexpr(context, this, first_subexpression, result);
+    
+    CYG_REPORT_RETURN();
+}
+
+void
+CdlExpressionBody::eval_subexpression(CdlEvalContext& context, int index, CdlSimpleValue& result)
+{
+    CYG_REPORT_FUNCNAME("CdlExpression::eval_subexpression)");
+    CYG_REPORT_FUNCARG4XV(this, &context, index, &result);
+    CYG_INVARIANT_THISC(CdlExpressionBody);
+    CYG_PRECONDITION_CLASSOC(context);
+
+    evaluate_subexpr(context, this, index, result);
+    
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+
+//{{{  CdlExpression                    
+
+//{{{  Construction                                     
+
+// ----------------------------------------------------------------------------
+// Ordinary expressions.
+//
+// The default constructor is private and does very little. Expressions
+// are created primarily by means of the parse() member function. There
+// is an argument for having constructors that take the same arguments
+// as the parse() member functions and relying on exception handling,
+// but that gets tricky for goal expressions and continue_parse().
+//
+// The copy constructor is protected and is used when creating e.g.
+// a default_value property object, which inherits from the ordinary
+// expression class. Again it might be better to do the parsing in
+// the constructor itself.
+//
+// The assignment operator is private and illegal.
+
+CdlExpressionBody::CdlExpressionBody()
+{
+    CYG_REPORT_FUNCNAME("CdlExpression:: default constructor");
+    CYG_REPORT_FUNCARG1XV(this);
+
+    expression_string           = "";
+    first_subexpression         = -1;
+
+    cdlexpressionbody_cookie    = CdlExpressionBody_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+    
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+CdlExpressionBody::CdlExpressionBody(const CdlExpressionBody& original)
+{
+    CYG_REPORT_FUNCNAME("CdlExpression:: copy constructor");
+    CYG_REPORT_FUNCARG2XV(this, &original);
+    CYG_INVARIANT_CLASSOC(CdlExpressionBody, original);
+
+    // Sub-expressions are simple structs, so this should result in a bit-wise
+    // copy of each vector element
+    sub_expressions     = original.sub_expressions;
+
+    // Simple scalar
+    first_subexpression = original.first_subexpression;
+
+    // The CdlReference class has a valid copy constructor and assignment
+    // operator, provided that the reference is not yet bound. This should
+    // be true when this copy constructor gets invoked, after parsing
+    // and during the construction of a derived property object.
+    references          = original.references;
+    expression_string   = original.expression_string;
+
+    cdlexpressionbody_cookie    = CdlExpressionBody_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+    
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  check_this()                                     
+
+// ----------------------------------------------------------------------------
+// check_this(). Expression objects can exist before any parsing has
+// happened, not to mention in the middle of parsing. The
+// first_subexpression field can be used to detect this.
+
+bool
+CdlExpressionBody::check_this(cyg_assert_class_zeal zeal) const
+{
+    if (CdlExpressionBody_Magic != cdlexpressionbody_cookie) {
+        return false;
+    }
+    CYGDBG_MEMLEAK_CHECKTHIS();
+
+    if (-1 == first_subexpression) {
+        return true;
+    }
+
+    switch(zeal) {
+      case cyg_system_test    :
+      case cyg_extreme        :
+      case cyg_thorough       :
+      {
+          for (std::vector<CdlReference>::const_iterator i = references.begin(); i != references.end(); i++) {
+              if (!i->check_this(cyg_quick)) {
+                  return false;
+              }
+          }
+      }
+      case cyg_quick          :
+          if ((unsigned)first_subexpression >= sub_expressions.size()) {
+              return false;
+          }
+      case cyg_trivial        :
+      case cyg_none           :
+        break;
+    }
+
+    return true;
+}
+
+//}}}
+//{{{  Destruction                                      
+
+CdlExpressionBody::~CdlExpressionBody()
+{
+    CYG_REPORT_FUNCNAME("CdlExpression::destructor");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    cdlexpressionbody_cookie    = CdlExpressionBody_Invalid;
+    first_subexpression         = -1;
+    sub_expressions.clear();
+    expression_string           = "";
+
+    // This assumes that all references have been unbound already by
+    // higher-level destructors.
+    references.clear();
+
+    CYGDBG_MEMLEAK_DESTRUCTOR();
+
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  Parsing - exported interface                     
+
+// ----------------------------------------------------------------------------
+// parse(string) invokes parse(string, ...) and checks that the expression
+// has terminated with EOD. Parsing of list expressions etc. can terminate
+// with some other token.
+//
+// parse(string, ...) allocates the expression object and invokes
+// continue_parse().
+//
+// continue_parse() is supposed to do all the hard work.
+
+CdlExpression
+CdlExpressionBody::parse(std::string data)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlExpression::parse", "result %p");
+
+    CdlExpression       result  = 0;
+    int                 index   = 0;
+    CdlExprOp           next_op = CdlExprOp_Invalid;
+    int                 end_index;
+
+    result = parse(data, index, next_op, end_index);
+    
+    // Either there has already been a parsing or out-of-memory
+    // exception, or we should be at the end of the expression string.
+    if (CdlExprOp_EOD != next_op) {
+        delete result;
+        throw CdlParseException("Unexpected data at end of expression.\n" + get_error_location());
+    }
+
+    // Keep a copy of the original string for diagnostics purposes.
+    result->expression_string = data;
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+CdlExpression
+CdlExpressionBody::parse(std::string data, int& index, CdlExprOp& next_token, int& token_end)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlExpression::parse", "result %d");
+
+    CdlExpression result = new CdlExpressionBody;
+
+    try {
+        continue_parse(result, data, index, next_token, token_end);
+    }
+    catch (...) {
+        delete result;
+        throw;
+    }
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+//}}}
+//{{{  update()                                         
+
+// ----------------------------------------------------------------------------
+// There has been a change in the toplevel which involves entities being
+// created or destroyed, and reference resolution is required.
+
+bool
+CdlExpressionBody::update(CdlTransaction transaction, CdlNode source, CdlProperty source_prop, CdlNode dest, CdlUpdate change)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlExpression::update", "result %d");
+    CYG_REPORT_FUNCARG6XV(this, transaction, source, source_prop, dest, change);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSC(source);
+    CYG_PRECONDITION_CLASSC(source_prop);
+
+    CdlToplevel toplevel        = source->get_toplevel();
+    bool        result          = false;
+    std::vector<CdlReference>::iterator ref_i;
+
+    switch(change) {
+      case CdlUpdate_Loaded:
+      {
+        // The source package has just been loaded. Try to resolve every
+        // reference, creating CdlConflict objects where necessary.
+        CYG_ASSERTC(0 == dest);
+        for (ref_i = references.begin(); ref_i != references.end(); ref_i++) {
+            dest = toplevel->lookup(ref_i->get_destination_name());
+            if (0 == dest) {
+                CdlConflict_UnresolvedBody::make(transaction, source, source_prop, ref_i->get_destination_name());
+            } else {
+                ref_i->bind(source, source_prop, dest);
+            }
+        }
+        result = true;
+        break;
+      }
+
+      case CdlUpdate_Unloading:
+      {
+        // The source package is being unloaded. Unbind all currently bound references.
+        // Also destroy any unresolved conflicts.
+        CYG_ASSERTC(0 == dest);
+        for (ref_i = references.begin(); ref_i != references.end(); ref_i++) {
+            dest = ref_i->get_destination();
+            if (0 != dest) {
+                ref_i->unbind(source, source_prop);
+            }
+        }
+        result = true;
+        break;
+      }
+
+      case CdlUpdate_Created :
+      {
+        
+        // A previously unresolved reference can now be resolved.
+        // It is necessary to search the vector for an unresolved
+        // reference with the desired name, and do the binding.
+        // This search may fail in the case of list expressions.
+        CYG_ASSERT_CLASSC(dest);
+        std::string dest_name = dest->get_name();
+        for (ref_i = references.begin(); !result && (ref_i != references.end()); ref_i++) {
+            if ((dest_name == ref_i->get_destination_name()) && (0 == ref_i->get_destination())) {
+                ref_i->bind(source, source_prop, dest);
+                result = true;
+                
+                std::vector<CdlConflict> conflicts;
+                std::vector<CdlConflict>::iterator conf_i;
+                transaction->get_structural_conflicts(source, source_prop, &CdlConflict_UnresolvedBody::test, conflicts);
+                for (conf_i = conflicts.begin(); conf_i != conflicts.end(); conf_i++) {
+                    CdlConflict_Unresolved real_conf = dynamic_cast<CdlConflict_Unresolved>(*conf_i);
+                    CYG_ASSERTC(0 != real_conf);
+                    if (dest_name == real_conf->get_target_name()) {
+                        transaction->clear_conflict(real_conf);
+                        break;
+                    }
+                }
+                CYG_ASSERTC(conf_i != conflicts.end());
+            }
+        }
+        break;
+      }
+
+      case CdlUpdate_Destroyed :
+      {
+        // A previously resolved reference is about to become illegal.
+        // Search the vector for a resolved reference object matching
+        // the destination, and unbind it. Also create a conflict
+        // object. The search can fail in the case of list expressions
+        CYG_ASSERT_CLASSC(dest);
+        for (ref_i = references.begin(); !result && (ref_i != references.end()); ref_i++) {
+            if (dest == ref_i->get_destination()) {
+                ref_i->unbind(source, source_prop);
+                CdlConflict_UnresolvedBody::make(transaction, source, source_prop, ref_i->get_destination_name());
+                result = true;
+            }
+        }
+        break;
+      }
+
+      default :
+          CYG_FAIL("Illegal change type passed to CdlExpression::update");
+          break;
+    }
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+//}}}
+//{{{  Evaluation                                       
+
+// ----------------------------------------------------------------------------
+// Expression evaluation. At the end of the day everything filters through
+// to eval_internal() which should all the hard work.
+//
+// The eval() member function handles EvalException conflicts. The
+// eval_internal() member function does not, and is used for list
+// and goal expressions as well.
+
+void
+CdlExpressionBody::eval(CdlEvalContext& context, CdlSimpleValue& result)
+{
+    CYG_REPORT_FUNCNAME("CdlExpression::eval");
+
+    try {
+        
+        eval_internal(context, result);
+        
+        // Evaluation has succeeded, so if there was an EvalException
+        // conflict get rid of it. This can only happen in the context
+        // of a transaction.
+        if ((0 != context.transaction) && (0 != context.node) && (0 != context.property)) {
+            context.transaction->clear_conflicts(context.node, context.property, &CdlConflict_EvalExceptionBody::test);
+        }
+        
+    } catch(CdlEvalException e) {
+
+        if ((0 != context.transaction) && (0 != context.node) && (0 != context.property)) {
+
+            CdlConflict conflict = context.transaction->get_conflict(context.node, context.property,
+                                                                              &CdlConflict_EvalExceptionBody::test);
+            if (0 == conflict) {
+                CdlConflict_EvalExceptionBody::make(context.transaction, context.node, context.property, e.get_message());
+            } else {
+                
+                CdlConflict_EvalException eval_conf = dynamic_cast<CdlConflict_EvalException>(conflict);
+                CYG_ASSERTC(0 != eval_conf);
+                if (eval_conf->get_explanation() != e.get_message()) {
+                    
+                    // Replace the conflict object. That way higher level code gets informed
+                    // there has been a change.
+                    context.transaction->clear_conflict(conflict);
+                    CdlConflict_EvalExceptionBody::make(context.transaction, context.node, context.property, e.get_message());
+                }
+            }
+        }
+
+        throw;
+    }
+}
+
+//}}}
+//{{{  Misc                                             
+
+// ----------------------------------------------------------------------------
+
+std::string
+CdlExpressionBody::get_original_string() const
+{
+    CYG_REPORT_FUNCNAME("CdlExpression::get_original_string");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    CYG_REPORT_RETURN();
+    return expression_string;
+}
+
+//}}}
+
+//}}}
+//{{{  CdlListExpression                
+
+//{{{  Construction                             
+
+// ----------------------------------------------------------------------------
+// The normal sequence of events is:
+//
+// 1) higher level code calls CdlListExpressionbody::parse()
+// 2) this static member creates a new and empty list expression object.
+//    The constructor need not do very much.
+// 3) the parse() member then fills in the newly created object
+// 4) the object is returned to higher-level code
+// 5) usually the list expression will now become part of
+//    a property object by means of a copy constructor.
+//
+// The only complication is that a list expression contains a vector
+// of CdlExpression pointers which must be freed during the destructor.
+// The copy constructor does not make duplicates of the individual
+// expression objects, instead ownership is transferred.
+
+CdlListExpressionBody::CdlListExpressionBody()
+{
+    CYG_REPORT_FUNCNAME("CdlListExpression:: default constructor");
+    CYG_REPORT_FUNCARG1XV(this);
+
+    expression_string           = "";
+
+    cdllistexpressionbody_cookie = CdlListExpressionBody_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+    
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+CdlListExpressionBody::CdlListExpressionBody(const CdlListExpressionBody& original)
+{
+    CYG_REPORT_FUNCNAME("CdlListExpression:: copy constructor");
+    CYG_REPORT_FUNCARG2XV(this, &original);
+    CYG_INVARIANT_CLASSOC(CdlListExpressionBody, original);
+
+    expression_string           = original.expression_string;
+
+    // These copy across the pointers
+    data        = original.data;
+    ranges      = original.ranges;
+
+    // And this clears out the pointers, but leaves the expression objects lying around
+    CdlListExpression tmp = const_cast<CdlListExpression>(&original);
+    tmp->data.clear();
+    tmp->ranges.clear();
+
+    cdllistexpressionbody_cookie = CdlListExpressionBody_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+    
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  Destruction                              
+
+CdlListExpressionBody::~CdlListExpressionBody()
+{
+    CYG_REPORT_FUNCNAME("CdlListExpression:: destructor");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    cdllistexpressionbody_cookie        = CdlListExpressionBody_Invalid;
+    expression_string                   = "";
+
+    for (std::vector<CdlExpression>::iterator i = data.begin(); i != data.end(); i++) {
+        delete *i;
+        *i = 0;
+    }
+    for (std::vector<std::pair<CdlExpression, CdlExpression> >::iterator j = ranges.begin(); j != ranges.end(); j++) {
+        delete j->first;
+        delete j->second;
+        j->first = 0;
+        j->second = 0;
+    }
+    data.clear();
+    ranges.clear();
+    CYGDBG_MEMLEAK_DESTRUCTOR();
+    
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  check_this()                             
+
+// ----------------------------------------------------------------------------
+bool
+CdlListExpressionBody::check_this(cyg_assert_class_zeal zeal) const
+{
+    if (CdlListExpressionBody_Magic != cdllistexpressionbody_cookie) {
+        return false;
+    }
+    CYGDBG_MEMLEAK_CHECKTHIS();
+    switch(zeal) {
+      case cyg_system_test      :
+      case cyg_extreme          :
+      case cyg_thorough         :
+      {
+          for (std::vector<CdlExpression>::const_iterator i = data.begin(); i != data.end(); i++) {
+              if (!(*i)->check_this(cyg_quick)) {
+                  return false;
+              }
+          }
+          for (std::vector<std::pair<CdlExpression,CdlExpression> >::const_iterator j = ranges.begin();
+               j != ranges.end();
+               j++) {
+              if (!(j->first->check_this(cyg_quick)) || !(j->second->check_this(cyg_quick))) {
+                  return false;
+              }
+          }
+      }
+      case cyg_quick            :
+      case cyg_trivial          :
+      case cyg_none             :
+      default                   :
+          break;
+    }
+
+    return true;
+}
+
+//}}}
+//{{{  Parsing                                  
+
+// ----------------------------------------------------------------------------
+// Parsing a list expression involves repeated parsing of ordinary
+// expressions until an EOD token is reached.
+
+CdlListExpression
+CdlListExpressionBody::parse(std::string data)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlListExpression::parse", "result %p");
+
+    // Allocate an expression object that can then be filled in.
+    CdlListExpression result = new CdlListExpressionBody;
+
+    // Do the parsing in a try/catch statement to make sure the
+    // allocated expression gets freed on a parse error.
+    try {
+        int             index           = 0;
+        int             end_index       = 0;
+        CdlExprOp       op              = CdlExprOp_Invalid;
+        CdlExpression   expr1           = 0;
+
+        do {
+            // Try to parse the next expression in the list
+            op    = CdlExprOp_Invalid;
+            expr1 = CdlExpressionBody::parse(data, index, op, end_index);
+
+            // There should now be a valid expression, failure would have
+            // resulted in an exception.
+            CYG_ASSERT_CLASSC(expr1);
+
+            // Allow for ranges.
+            if (CdlExprOp_Range != op) {
+                // A simple expression, just add it to the current data vector
+                // "index" will contain the appropriate value.
+                result->data.push_back(expr1);
+            } else {
+                // A range expression. Get the other end of the range.
+                // This requires manipulating index a bit.
+                CdlExpression expr2 = 0;
+                index = end_index;
+                op    = CdlExprOp_Invalid;
+                try {
+                    expr2 = CdlExpressionBody::parse(data, index, op, end_index);
+                }
+                catch (...) {
+                    delete expr1;
+                    throw;
+                }
+                result->ranges.push_back(std::make_pair(expr1, expr2));
+            }
+        } while (CdlExprOp_EOD != op);
+    }
+    catch (...) {
+        delete result;
+        throw;
+    }
+
+    // Keep track of the original string for diagnostics purposes
+    result->expression_string = data;
+    
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+//}}}
+//{{{  update()                                 
+
+// ----------------------------------------------------------------------------
+// This code is invoked when it is necessary to update the references
+// for the list expression. There are four situations in which this
+// can happen: the package has just been loaded; the package is being
+// unloaded; a referenced target is being created; a referenced target is
+// being destroyed.
+//
+// The first two cases simply involve processing every expression that
+// makes up the overall list expression. The last two cases involve
+// searching through the expressions until an applicable one is found.
+// Note that an expression may contain multiple references to another
+// object, resulting in multiple calls to this function.
+
+bool
+CdlListExpressionBody::update(CdlTransaction transact, CdlNode source, CdlProperty source_prop, CdlNode dest, CdlUpdate change)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlListExpression::update", "result %d");
+    CYG_REPORT_FUNCARG6XV(this, transact, source, source_prop, dest, change);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSC(source);
+    CYG_PRECONDITION_CLASSC(source_prop);
+
+    bool result = false;
+
+    if ((CdlUpdate_Loaded == change) || (CdlUpdate_Unloading == change)) {
+
+        std::vector<CdlExpression>::const_iterator expr_i;
+        std::vector<std::pair<CdlExpression, CdlExpression> >::const_iterator pair_i;
+
+        for (expr_i = data.begin(); expr_i != data.end(); expr_i++) {
+            bool handled = (*expr_i)->update(transact, source, source_prop, dest, change);
+            CYG_ASSERTC(handled);
+            CYG_UNUSED_PARAM(bool, handled);
+        }
+        for (pair_i = ranges.begin(); pair_i != ranges.end(); pair_i++) {
+            bool handled = pair_i->first->update(transact, source, source_prop, dest, change);
+            CYG_ASSERTC(handled);
+            handled = pair_i->second->update(transact, source, source_prop, dest, change);
+            CYG_ASSERTC(handled);
+        }
+        
+        result = true;
+        
+    } else {
+        CYG_ASSERTC((CdlUpdate_Created == change) || (CdlUpdate_Destroyed == change));
+
+        std::vector<CdlExpression>::const_iterator expr_i;
+        std::vector<std::pair<CdlExpression, CdlExpression> >::const_iterator pair_i;
+
+        for (expr_i = data.begin(); !result && (expr_i != data.end()); expr_i++) {
+            result = (*expr_i)->update(transact, source, source_prop, dest, change);
+        }
+        for (pair_i = ranges.begin(); !result && (pair_i != ranges.end()); pair_i++) {
+            result = pair_i->first->update(transact, source, source_prop, dest, change);
+            if (!result) {
+                result = pair_i->second->update(transact, source, source_prop, dest, change);
+            }
+        }
+    }
+    
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+//}}}
+//{{{  Evaluation                               
+
+// ----------------------------------------------------------------------------
+// Evaluation. The hard work is actually done in eval_internal()
+
+void
+CdlListExpressionBody::eval(CdlEvalContext& context, CdlListValue& result)
+{
+    CYG_REPORT_FUNCNAME("CdlListExpression::eval");
+    CYG_REPORT_FUNCARG3XV(this, &context, &result);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    this->eval_internal(context, result);
+
+    CYG_REPORT_RETURN();
+}
+
+// ----------------------------------------------------------------------------
+// This requires evaluating each expression in the data and ranges
+// vectors and adding the result to the appropriate vector in result.
+// Various error conditions are possible.
+
+void
+CdlListExpressionBody::eval_internal(CdlEvalContext& context, CdlListValue& result)
+{
+    CYG_REPORT_FUNCNAME("CdlListExpression::eval_internal");
+    CYG_REPORT_FUNCARG2XV(this, &context);
+
+    result.table.clear();
+    result.integer_ranges.clear();
+    result.double_ranges.clear();
+
+    CdlSimpleValue val1;
+    CdlSimpleValue val2;
+
+    try {
+        for (std::vector<CdlExpression>::const_iterator i = data.begin(); i != data.end(); i++) {
+            (*i)->eval_internal(context, val1);
+            if ("" != val1.get_value()) {
+              result.table.push_back(val1);
+            }
+        }
+        for (std::vector<std::pair<CdlExpression,CdlExpression> >::const_iterator j = ranges.begin(); j != ranges.end(); j++) {
+            j->first->eval_internal(context, val1);
+            j->second->eval_internal(context, val2);
+
+            if (val1.has_integer_value() && val2.has_integer_value()) {
+                cdl_int x1 = val1.get_integer_value();
+                cdl_int x2 = val2.get_integer_value();
+                if (x1 > x2) {
+                    cdl_int tmp = x1;
+                    x1 = x2;
+                    x2 = tmp;
+                }
+                result.integer_ranges.push_back(std::make_pair(x1, x2));
+            } else if (val1.has_double_value() && val2.has_double_value()) {
+                double x1 = val1.get_double_value();
+                double x2 = val2.get_double_value();
+                if (x1 > x2) {
+                    double tmp = x1;
+                    x1 = x2;
+                    x2 = tmp;
+                }
+                result.double_ranges.push_back(std::make_pair(x1, x2));
+            } else {
+                throw CdlEvalException("Range expression involves non-numerical limits");
+            }
+        }
+        
+        // Any problems would have resulted in an exception. If there
+        // was a previous EvalExeption for this property, it is no
+        // longer applicable
+        if ((0 != context.transaction) && (0 != context.node) && (0 != context.property)) {
+            context.transaction->clear_conflicts(context.node, context.property, &CdlConflict_EvalExceptionBody::test);
+        }
+        
+    } catch(CdlEvalException e) {
+
+        if ((0 != context.transaction) && (0 != context.node) && (0 != context.property)) {
+            
+            CdlConflict conflict = context.transaction->get_conflict(context.node, context.property,
+                                                                              &CdlConflict_EvalExceptionBody::test);
+            if (0 == conflict) {
+                CdlConflict_EvalExceptionBody::make(context.transaction, context.node, context.property, e.get_message());
+            } else {
+                CdlConflict_EvalException eval_conf = dynamic_cast<CdlConflict_EvalException>(conflict);
+                CYG_ASSERTC(0 != eval_conf);
+                if (eval_conf->get_explanation() != e.get_message()) {
+                
+                    // Replace the conflict object. Higher level will be informed about this.
+                    context.transaction->clear_conflict(conflict);
+                    CdlConflict_EvalExceptionBody::make(context.transaction, context.node, context.property, e.get_message());
+                }
+            }
+        }
+
+        throw;
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+//}}}
+//{{{  is_member()                              
+
+// ----------------------------------------------------------------------------
+
+bool
+CdlListExpressionBody::is_member(CdlEvalContext& context, CdlSimpleValue& val)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlListExpression::is_member (value)", "result %d");
+    CYG_REPORT_FUNCARG3XV(this, &context, &val);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    bool result = false;
+    CdlListValue list_val;
+    eval_internal(context, list_val);
+    result = list_val.is_member(val);
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+bool
+CdlListExpressionBody::is_member(CdlEvalContext& context, std::string val)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlListExpression::is_member (string)", "result %d");
+    CYG_REPORT_FUNCARG2XV(this, &context);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    bool result = false;
+    CdlListValue list_val;
+    eval_internal(context, list_val);
+    result = list_val.is_member(val);
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+bool
+CdlListExpressionBody::is_member(CdlEvalContext& context, cdl_int val)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlListExpression::is_member (int)", "result %d");
+    CYG_REPORT_FUNCARG3XV(this, &context, (int) val);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    bool result = false;
+    CdlListValue list_val;
+    eval_internal(context, list_val);
+    result = list_val.is_member(val);
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+bool
+CdlListExpressionBody::is_member(CdlEvalContext& context, double val)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlListExpression::is_member (double)", "result %d");
+    CYG_REPORT_FUNCARG2XV(this, &context);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    bool result = false;
+    CdlListValue list_val;
+    eval_internal(context, list_val);
+    result = list_val.is_member(val);
+
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+//}}}
+//{{{  Misc                                     
+
+// ----------------------------------------------------------------------------
+
+std::string
+CdlListExpressionBody::get_original_string() const
+{
+    CYG_REPORT_FUNCNAME("CdlListExpression::get_original_string");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    CYG_REPORT_RETURN();
+    return expression_string;
+}
+
+//}}}
+
+//}}}
+//{{{  CdlGoalExpression                
+
+// ----------------------------------------------------------------------------
+// Constructors etc. are pretty much as per ordinary and list
+// expressions. Most of the work is done in the private base class.
+
+CdlGoalExpressionBody::CdlGoalExpressionBody()
+    : CdlExpressionBody()
+{
+    CYG_REPORT_FUNCNAME("CdlGoalExpression::default_constructor");
+    CYG_REPORT_FUNCARG1XV(this);
+
+    expression_string           = "";
+    cdlgoalexpressionbody_cookie = CdlGoalExpressionBody_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+    
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+CdlGoalExpressionBody::CdlGoalExpressionBody(const CdlGoalExpressionBody& original)
+    : CdlExpressionBody(original)
+{
+    CYG_REPORT_FUNCNAME("CdlGoalExpression:: copy constructor");
+    CYG_REPORT_FUNCARG2XV(this, &original);
+    CYG_INVARIANT_CLASSOC(CdlGoalExpressionBody, original);
+
+    expression_string           = original.expression_string;
+    cdlgoalexpressionbody_cookie = CdlGoalExpressionBody_Magic;
+    CYGDBG_MEMLEAK_CONSTRUCTOR();
+    
+    CYG_POSTCONDITION_THISC();
+    CYG_REPORT_RETURN();
+}
+
+CdlGoalExpressionBody::~CdlGoalExpressionBody()
+{
+    CYG_REPORT_FUNCNAME("CdlGoalExpression:: destructor");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    cdlgoalexpressionbody_cookie = CdlGoalExpressionBody_Invalid;
+    expression_string            = "";
+    CYGDBG_MEMLEAK_DESTRUCTOR();
+
+    CYG_REPORT_RETURN();
+}
+
+// ----------------------------------------------------------------------------
+// Parsing. A goal expression acts a bit like a list expression with
+// implicit && operators between the various expressions. It could be
+// implemented as a vector of expressions (which might make diagnostics
+// easier) but it is almost as easy to derive a goal expression from
+// an ordinary one.
+
+CdlGoalExpression
+CdlGoalExpressionBody::parse(std::string data)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlGoalExpression::parse", "result %p");
+
+    CdlGoalExpression result = new CdlGoalExpressionBody;
+
+    try {
+        int       index         = 0;
+        CdlExprOp op            = CdlExprOp_Invalid;
+        int       end_index     = 0;
+
+        // Parse the first expression in the data.
+        CdlExpressionBody::continue_parse(result, data, index, op, end_index);
+
+        // At this stage we have reached end-of-data or we should be
+        // at the start of another expression - any binary or ternary
+        // operands would have been subsumed in the previous expression.
+        // We need to keep adding && operators and new expressions until
+        // end-of-data.
+        while (CdlExprOp_EOD != op) {
+            op = CdlExprOp_And;
+            CdlExpressionBody::continue_parse(result, data, index, op, end_index);
+        }
+    }
+    catch(...) {
+        delete result;
+        throw;
+    }
+
+    // Keep track of the original expression string for diagnostics purposes
+    result->expression_string = data;
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// ----------------------------------------------------------------------------
+void
+CdlGoalExpressionBody::eval(CdlEvalContext& context, bool& result)
+{
+    CYG_REPORT_FUNCNAME("CdlGoalExpression::eval");
+    CYG_REPORT_FUNCARG2XV(this, &context);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    eval_internal(context, result);
+
+    CYG_REPORT_RETURN();
+}
+
+bool
+CdlGoalExpressionBody::eval(CdlEvalContext& context)
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlGoalExpression::eval", "result %d");
+    CYG_REPORT_FUNCARG2XV(this, &context);
+    CYG_PRECONDITION_THISC();
+    CYG_PRECONDITION_CLASSOC(context);
+
+    bool result;
+    eval_internal(context, result);
+    
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// ----------------------------------------------------------------------------
+// Provide access to the underlying CdlExpression object. This allows the
+// inference engine etc. to work out why a goal expression is failing
+
+CdlExpression
+CdlGoalExpressionBody::get_expression()
+{
+    CYG_REPORT_FUNCNAMETYPE("CdlGoalExpression::get_expression", "result %p");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    CdlExpression result = this;
+    CYG_REPORT_RETVAL(result);
+    return result;
+}
+
+// ----------------------------------------------------------------------------
+
+bool
+CdlGoalExpressionBody::check_this(cyg_assert_class_zeal zeal) const
+{
+    if (CdlGoalExpressionBody_Magic != cdlgoalexpressionbody_cookie) {
+        return false;
+    }
+    CYGDBG_MEMLEAK_CHECKTHIS();
+
+    // There is no data specific to a goal expression, just let the
+    // underlying check_this() member do its stuff.
+    
+    return inherited::check_this(zeal);
+}
+
+// ----------------------------------------------------------------------------
+
+std::string
+CdlGoalExpressionBody::get_original_string() const
+{
+    CYG_REPORT_FUNCNAME("CdlGoalExpression::get_original_string");
+    CYG_REPORT_FUNCARG1XV(this);
+    CYG_PRECONDITION_THISC();
+
+    CYG_REPORT_RETURN();
+    return expression_string;
+}
+
+// ----------------------------------------------------------------------------
+
+void
+CdlGoalExpressionBody::eval_internal(CdlEvalContext& context, bool& result)
+{
+    CYG_REPORT_FUNCNAME("CdlGoalExpression::eval_internal");
+    CYG_REPORT_FUNCARG2XV(this, &context);
+    // The assertions are all done in the calling code
+
+    // Start by evaluating the underlying expression
+    CdlSimpleValue      val;
+    try {
+        inherited::eval_internal(context, val);
+
+        // The evaluation succeeded. Do we have an integer, a string, ...?
+        if (val.has_integer_value()) {
+            result = (0 != val.get_integer_value());
+        } else if (val.has_double_value()) {
+            result = (0.0 != val.get_double_value());
+        } else {
+            result = ("" != val.get_value());
+        }
+
+        // If there is an EvalException conflict for this property, it is no longer applicable
+        if ((0 != context.transaction) && (0 != context.node) && (0 != context.property)) {
+            context.transaction->clear_conflicts(context.node, context.property,
+                                                          &CdlConflict_EvalExceptionBody::test);
+        }
+        
+    } catch(CdlEvalException e) {
+        if ((0 != context.transaction) && (0 != context.node) && (0 != context.property)) {
+            CdlConflict conflict = context.transaction->get_conflict(context.node, context.property,
+                                                                          &CdlConflict_EvalExceptionBody::test);
+            if (0 == conflict) {
+                CdlConflict_EvalExceptionBody::make(context.transaction, context.node, context.property, e.get_message());
+            } else {
+                CdlConflict_EvalException eval_conf = dynamic_cast<CdlConflict_EvalException>(conflict);
+                CYG_ASSERTC(0 != eval_conf);
+                if (eval_conf->get_explanation() != e.get_message()) {
+                    // Replace the conflict object. Higher level can detect this.
+                    context.transaction->clear_conflict(conflict);
+                    CdlConflict_EvalExceptionBody::make(context.transaction, context.node, context.property, e.get_message());
+                }
+            }
+            throw;
+        }
+    }
+    
+    CYG_REPORT_RETURN();
+}
+
+//}}}