170 lines
9.4 KiB
Ada
170 lines
9.4 KiB
Ada
------------------------------------------------------------------------------
|
|
-- --
|
|
-- GNAT COMPILER COMPONENTS --
|
|
-- --
|
|
-- E X P A N D E R --
|
|
-- --
|
|
-- S p e c --
|
|
-- --
|
|
-- Copyright (C) 1992-2013, Free Software Foundation, Inc. --
|
|
-- --
|
|
-- GNAT is free software; you can redistribute it and/or modify it under --
|
|
-- terms of the GNU General Public License as published by the Free Soft- --
|
|
-- ware Foundation; either version 3, or (at your option) any later ver- --
|
|
-- sion. GNAT is distributed in the hope that it will be useful, but WITH- --
|
|
-- OUT 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 distributed with GNAT; see file COPYING3. If not, go to --
|
|
-- http://www.gnu.org/licenses for a complete copy of the license. --
|
|
-- --
|
|
-- GNAT was originally developed by the GNAT team at New York University. --
|
|
-- Extensive contributions were provided by Ada Core Technologies Inc. --
|
|
-- --
|
|
------------------------------------------------------------------------------
|
|
|
|
-- This procedure performs any required expansion for the specified node.
|
|
-- The argument is the node that is a candidate for possible expansion.
|
|
-- If no expansion is required, then Expand returns without doing anything.
|
|
|
|
-- If the node does need expansion, then the subtree is replaced by the
|
|
-- tree corresponding to the required rewriting. This tree is a syntactic
|
|
-- tree, except that all Entity fields must be correctly set on all
|
|
-- direct names, since the expander presumably knows what it wants, and in
|
|
-- any case it doesn't work to have the semantic analyzer perform visibility
|
|
-- analysis on these trees (they may have references to non-visible runtime
|
|
-- routines etc.) There are a few exceptions to this rule in special cases,
|
|
-- but they must be documented clearly.
|
|
|
|
-- Expand is called in two different situations:
|
|
|
|
-- Nodes that are not subexpressions (Nkind not in N_Subexpr)
|
|
|
|
-- In this case, Expand is called from the body of Sem, immediately
|
|
-- after completing semantic analysis by calling the corresponding
|
|
-- Analyze_N_xxx procedure. If expansion occurs, the given node must
|
|
-- be replaced with another node that is also not a subexpression.
|
|
-- This seems naturally to be the case, since it is hard to imagine any
|
|
-- situation in which it would make sense to replace a non-expression
|
|
-- subtree with an expression. Once the substitution is completed, the
|
|
-- Expand routine must call Analyze on the resulting node to do any
|
|
-- required semantic analysis. Note that references to children copied
|
|
-- from the old tree won't be reanalyzed, since their Analyzed flag
|
|
-- is set.
|
|
|
|
-- Nodes that are subexpressions (Nkind in N_Subexpr)
|
|
|
|
-- In this case, Expand is called from Sem_Res.Resolve after completing
|
|
-- the resolution of the subexpression (this means that the expander sees
|
|
-- the fully typed subtree). If expansion occurs, the given node must be
|
|
-- replaced by a node that is also a subexpression. Again it is hard
|
|
-- to see how this restriction could possibly be violated. Once the
|
|
-- substitution is completed, the Expand routine must first call Analyze
|
|
-- on the resulting node to do any required semantic analysis, and then
|
|
-- call Resolve on the node to set the type (typically the type will be
|
|
-- the same as the original type of the input node, but this is not
|
|
-- always the case).
|
|
|
|
-- In both these cases, Replace or Rewrite must be used to achieve the
|
|
-- of the node, since the Expander routine is only passed the Node_Id
|
|
-- of the node to be expanded, and the resulting expanded Node_Id must
|
|
-- be the same (the parameter to Expand is mode in, not mode in-out).
|
|
|
|
-- For nodes other than subexpressions, it is not necessary to preserve the
|
|
-- original tree in the Expand routines, unlike the case for modifications
|
|
-- to the tree made in the semantic analyzer. This is because anyone who is
|
|
-- interested in working with the original tree (like ASIS) is required to
|
|
-- compile in semantics checks only mode. Thus Replace may be freely used
|
|
-- in such instances.
|
|
|
|
-- For subexpressions, preservation of the original tree is required because
|
|
-- of the need for conformance checking of default expressions, which occurs
|
|
-- on expanded trees. This means that Replace should not ever be used on
|
|
-- on subexpression nodes. Instead use Rewrite.
|
|
|
|
-- Note: the front end avoids calls to any of the expand routines if code
|
|
-- is not being generated. This is done for three reasons:
|
|
|
|
-- 1. Make sure tree does not get mucked up by the expander if no
|
|
-- code is being generated, and is thus usable by ASIS etc.
|
|
|
|
-- 2. Save time, since expansion is not needed if a compilation is
|
|
-- being done only to check the semantics, or if code generation
|
|
-- has been canceled due to previously detected errors.
|
|
|
|
-- 3. Allow the expand routines to assume that the tree is error free.
|
|
-- This results from the fact that code generation mode is always
|
|
-- cancelled when any error occurs.
|
|
|
|
-- If we ever decide to implement a feature allowing object modules to be
|
|
-- generated even if errors have been detected, then point 3 will no longer
|
|
-- hold, and the expand routines will have to be modified to operate properly
|
|
-- in the presence of errors (for many reasons this is not currently true).
|
|
|
|
-- Note: a consequence of this approach is that error messages must never
|
|
-- be generated in the expander, since this would mean that such error
|
|
-- messages are not generated when the expander is not being called.
|
|
|
|
-- Expansion is the last stage of analyzing a node, so Expand sets the
|
|
-- Analyzed flag of the node being analyzed as its last action. This is
|
|
-- done even if expansion is off (in this case, the only effect of the
|
|
-- call to Expand is to set the Analyzed flag to True).
|
|
|
|
with Types; use Types;
|
|
|
|
package Expander is
|
|
|
|
-- The flag Opt.Expander_Active controls whether expansion is active
|
|
-- (True) or deactivated (False). When expansion is deactivated all
|
|
-- calls to expander routines have no effect. To temporarily disable
|
|
-- expansion, always call the routines defined below, do NOT change
|
|
-- Expander_Active directly.
|
|
--
|
|
-- You should not use this flag to test if you are currently processing
|
|
-- a generic spec or body. Use the flag Inside_A_Generic instead (see
|
|
-- the spec of package Sem).
|
|
--
|
|
-- There is no good reason for permanently changing the value of this flag
|
|
-- except after detecting a syntactic or semantic error. In this event
|
|
-- this flag is set to False to disable all subsequent expansion activity.
|
|
--
|
|
-- In general this flag should be used as a read only value. The only
|
|
-- exceptions where it makes sense to temporarily change its value are:
|
|
--
|
|
-- (a) when starting/completing the processing of a generic definition
|
|
-- or declaration (see routines Start_Generic_Processing and
|
|
-- End_Generic_Processing in Sem_Ch12)
|
|
--
|
|
-- (b) when starting/completing the pre-analysis of an expression
|
|
-- (see the spec of package Sem for more info on pre-analysis.)
|
|
--
|
|
-- Note that when processing a spec expression (In_Spec_Expression
|
|
-- is True) or performing semantic analysis of a generic spec or body
|
|
-- (Inside_A_Generic) or when performing pre-analysis (Full_Analysis is
|
|
-- False) the Expander_Active flag is False.
|
|
|
|
procedure Expand (N : Node_Id);
|
|
-- Expand node N, as described above
|
|
|
|
procedure Expander_Mode_Save_And_Set (Status : Boolean);
|
|
-- Saves the current setting of the Expander_Active flag on an internal
|
|
-- stack and then sets the flag to the given value.
|
|
--
|
|
-- Note: this routine has no effect in ASIS and GNATprove modes. In ASIS
|
|
-- mode, all expansion activity is always off, since we want the original
|
|
-- semantic tree for ASIS purposes without any expansion. In GNATprove
|
|
-- mode, a very light expansion is performed on specific nodes. Both are
|
|
-- achieved by setting Expander_Active False in ASIS and GNATprove modes.
|
|
-- In situations such as the call to Instantiate_Bodies in Frontend,
|
|
-- Expander_Mode_Save_And_Set may be called to temporarily turn the
|
|
-- expander on, but this will have no effect in ASIS and GNATprove modes.
|
|
|
|
procedure Expander_Mode_Restore;
|
|
-- Restores the setting of the Expander_Active flag using the top entry
|
|
-- pushed onto the stack by Expander_Mode_Save_And_Reset, popping the
|
|
-- stack, except that if any errors have been detected, then the state of
|
|
-- the flag is left set to False. Disabled for ASIS and GNATprove modes
|
|
-- (see above).
|
|
|
|
end Expander;
|