aboutsummaryrefslogtreecommitdiffstats
path: root/tests/manual/plain-cplusplus/Preprocessor.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'tests/manual/plain-cplusplus/Preprocessor.cpp')
-rw-r--r--tests/manual/plain-cplusplus/Preprocessor.cpp339
1 files changed, 0 insertions, 339 deletions
diff --git a/tests/manual/plain-cplusplus/Preprocessor.cpp b/tests/manual/plain-cplusplus/Preprocessor.cpp
deleted file mode 100644
index 876f18ae9b..0000000000
--- a/tests/manual/plain-cplusplus/Preprocessor.cpp
+++ /dev/null
@@ -1,339 +0,0 @@
-/****************************************************************************
-**
-** Copyright (C) 2014 Digia Plc and/or its subsidiary(-ies).
-** Contact: http://www.qt-project.org/legal
-**
-** This file is part of Qt Creator.
-**
-** Commercial License Usage
-** Licensees holding valid commercial Qt licenses may use this file in
-** accordance with the commercial license agreement provided with the
-** Software or, alternatively, in accordance with the terms contained in
-** a written agreement between you and Digia. For licensing terms and
-** conditions see http://qt.digia.com/licensing. For further information
-** use the contact form at http://qt.digia.com/contact-us.
-**
-** GNU Lesser General Public License Usage
-** Alternatively, this file may be used under the terms of the GNU Lesser
-** General Public License version 2.1 as published by the Free Software
-** Foundation and appearing in the file LICENSE.LGPL included in the
-** packaging of this file. Please review the following information to
-** ensure the GNU Lesser General Public License version 2.1 requirements
-** will be met: http://www.gnu.org/licenses/old-licenses/lgpl-2.1.html.
-**
-** In addition, as a special exception, Digia gives you certain additional
-** rights. These rights are described in the Digia Qt LGPL Exception
-** version 1.1, included in the file LGPL_EXCEPTION.txt in this package.
-**
-****************************************************************************/
-
-#include "Preprocessor.h"
-#include "Lexer.h"
-#include <list>
-#include <iostream>
-#include <cassert>
-
-using namespace CPlusPlus;
-
-std::ostream &operator << (std::ostream &out, const StringRef &s)
-{
- out.write(s.text(), s.size());
- return out;
-}
-
-struct Preprocessor::TokenBuffer
-{
- std::list<Token> tokens;
- const Macro *macro;
- TokenBuffer *next;
-
- template <typename _Iterator>
- TokenBuffer(_Iterator firstToken, _Iterator lastToken, const Macro *macro, TokenBuffer *next)
- : tokens(firstToken, lastToken), macro(macro), next(next) {}
-};
-
-Lexer *Preprocessor::switchLexer(Lexer *lex)
-{
- Lexer *previousLexer = _lexer;
- _lexer = lex;
- return previousLexer;
-}
-
-StringRef Preprocessor::switchSource(const StringRef &source)
-{
- StringRef previousSource = _source;
- _source = source;
- return previousSource;
-}
-
-const Preprocessor::Macro *Preprocessor::resolveMacro(const StringRef &name) const
-{
- std::map<StringRef, Macro>::const_iterator it = macros.find(name);
- if (it != macros.end()) {
- const Macro *m = &it->second;
- for (TokenBuffer *r = _tokenBuffer; r; r = r->next) {
- if (r->macro == m)
- return 0;
- }
- return m;
- }
-
- return 0;
-}
-
-void Preprocessor::collectActualArguments(Token *tk, std::vector<std::vector<Token> > *actuals)
-{
- lex(tk);
-
- assert(tk->is(T_LPAREN));
-
- lex(tk);
-
- std::vector<Token> tokens;
- scanActualArgument(tk, &tokens);
-
- actuals->push_back(tokens);
-
- while (tk->is(T_COMMA)) {
- lex(tk);
-
- std::vector<Token> tokens;
- scanActualArgument(tk, &tokens);
- actuals->push_back(tokens);
- }
-
- assert(tk->is(T_RPAREN));
- lex(tk);
-}
-
-void Preprocessor::scanActualArgument(Token *tk, std::vector<Token> *tokens)
-{
- int count = 0;
-
- while (tk->isNot(T_EOF_SYMBOL)) {
- if (tk->is(T_LPAREN)) {
- ++count;
- } else if (tk->is(T_RPAREN)) {
- if (! count)
- break;
-
- --count;
- } else if (! count && tk->is(T_COMMA)) {
- break;
- }
-
- tokens->push_back(*tk);
- lex(tk);
- }
-}
-
-void Preprocessor::lex(Token *tk)
-{
-_Lagain:
- if (_tokenBuffer) {
- if (_tokenBuffer->tokens.empty()) {
- TokenBuffer *r = _tokenBuffer;
- _tokenBuffer = _tokenBuffer->next;
- delete r;
- goto _Lagain;
- }
- *tk = _tokenBuffer->tokens.front();
- _tokenBuffer->tokens.pop_front();
- } else {
- _lexer->scan(tk);
- }
-
-_Lclassify:
- if (! inPreprocessorDirective) {
- if (tk->newline() && tk->is(T_POUND)) {
- handlePreprocessorDirective(tk);
- goto _Lclassify;
-
- } else if (tk->is(T_IDENTIFIER)) {
- const StringRef id = asStringRef(*tk);
-
- if (const Macro *macro = resolveMacro(id)) {
- std::vector<Token> body = macro->body;
-
- if (macro->isFunctionLike) {
- std::vector<std::vector<Token> > actuals;
- collectActualArguments(tk, &actuals);
-
- std::vector<Token> expanded;
- for (size_t i = 0; i < body.size(); ++i) {
- const Token &token = body[i];
-
- if (token.isNot(T_IDENTIFIER)) {
- expanded.push_back(token);
- } else {
- const StringRef id = asStringRef(token);
- size_t j = 0;
- for (; j < macro->formals.size(); ++j) {
- if (macro->formals[j] == id) {
- expanded.insert(expanded.end(), actuals[j].begin(), actuals[j].end());
- break;
- }
- }
-
- if (j == macro->formals.size())
- expanded.push_back(token);
- }
- }
-
- const Token currentTokenBuffer[] = { *tk };
- _tokenBuffer = new TokenBuffer(currentTokenBuffer, currentTokenBuffer + 1,
- /*macro */ 0, _tokenBuffer);
-
- body = expanded;
- }
-
- _tokenBuffer = new TokenBuffer(body.begin(), body.end(),
- macro, _tokenBuffer);
- goto _Lagain;
- }
- }
- }
-}
-
-void Preprocessor::handlePreprocessorDirective(Token *tk)
-{
- inPreprocessorDirective = true;
-
- lex(tk); // scan the directive
-
- if (tk->newline() && ! tk->joined())
- return; // nothing to do.
-
- const StringRef ppDefine("define", 6);
-
- if (tk->is(T_IDENTIFIER)) {
- const StringRef directive = asStringRef(*tk);
-
- if (directive == ppDefine)
- handleDefineDirective(tk);
- else
- skipPreprocesorDirective(tk);
- }
-
- inPreprocessorDirective = false;
-}
-
-bool Preprocessor::isValidToken(const Token &tk) const
-{
- if (tk.isNot(T_EOF_SYMBOL) && (! tk.newline() || tk.joined()))
- return true;
-
- return false;
-}
-
-void Preprocessor::handleDefineDirective(Token *tk)
-{
- lex(tk);
-
- if (tk->is(T_IDENTIFIER)) {
- const StringRef macroName = asStringRef(*tk);
- Macro macro;
-
- lex(tk);
-
- if (isValidToken(*tk) && tk->is(T_LPAREN) && ! tk->whitespace()) {
- macro.isFunctionLike = true;
-
- lex(tk); // skip `('
-
- if (isValidToken(*tk) && tk->is(T_IDENTIFIER)) {
- macro.formals.push_back(asStringRef(*tk));
-
- lex(tk);
-
- while (isValidToken(*tk) && tk->is(T_COMMA)) {
- lex(tk);
-
- if (isValidToken(*tk) && tk->is(T_IDENTIFIER)) {
- macro.formals.push_back(asStringRef(*tk));
- lex(tk);
- }
- }
- }
-
- if (isValidToken(*tk) && tk->is(T_RPAREN))
- lex(tk); // skip `)'
- }
-
- while (isValidToken(*tk)) {
- macro.body.push_back(*tk);
- lex(tk);
- }
-
- macros.insert(std::make_pair(macroName, macro));
- } else {
- skipPreprocesorDirective(tk);
- }
-}
-
-void Preprocessor::skipPreprocesorDirective(Token *tk)
-{
- do {
- lex(tk);
- } while (isValidToken(*tk));
-}
-
-StringRef Preprocessor::asStringRef(const Token &tk) const
-{ return StringRef(_source.begin() + tk.bytesBegin(), tk.bytes()); }
-
-Preprocessor::Preprocessor(std::ostream &out)
- : out(out), _lexer(0), inPreprocessorDirective(false)
-{ }
-
-void Preprocessor::operator()(const char *source, unsigned size, const StringRef &currentFileName)
-{
- _currentFileName = currentFileName;
- run(source, size);
-}
-
-void Preprocessor::run(const char *source, unsigned size)
-{
- _tokenBuffer = 0;
-
- const StringRef previousSource = switchSource(StringRef(source, size));
- Lexer thisLexer(source, source + size);
- thisLexer.setScanKeywords(false);
- Lexer *previousLexer = switchLexer(&thisLexer);
- inPreprocessorDirective = false;
-
- Token tk;
- unsigned lineno = 0;
- do {
- lex(&tk);
-
- if (lineno != tk.lineno) {
- if (lineno > tk.lineno || tk.lineno - lineno > 3) {
- out << std::endl << "#line " << tk.lineno << " \"" << _currentFileName << "\"" << std::endl;
- } else {
- for (unsigned i = lineno; i < tk.lineno; ++i)
- out << std::endl;
- }
-
- lineno = tk.lineno;
-
- } else {
- if (tk.newline())
- out << std::endl;
-
- if (tk.whitespace())
- out << ' ';
- }
-
- out << asStringRef(tk);
- lineno = tk.lineno;
- } while (tk.isNot(T_EOF_SYMBOL));
-
- out << std::endl;
-
- switchLexer(previousLexer);
- switchSource(previousSource);
-}
-
-
-
-