private Token stringify(Token pos, Argument arg) {
StringBuilder buf = new StringBuilder();
concat(buf, arg);
// System.out.println("Concat: " + arg + " -> " + buf);
StringBuilder str = new StringBuilder("\"");
escape(str, buf.ToString());
str.append("\"");
// System.out.println("Escape: " + buf + " -> " + str);
return new Token(Token.STRING,
pos.getLine(), pos.getColumn(),
str.toString(), buf.toString());
}
protected void pragma(Token name, List<Token> value)
{
warning(name, "Unknown #" + "pragma: " + name.getText());
}
/**
* Handles an error.
*
* If a PreprocessorListener is installed, it receives the
* error. Otherwise, an exception is thrown.
*
* @see #error(int, int, String)
*/
protected void error(Token tok, String msg)
{
error(tok.getLine(), tok.getColumn(), msg);
}
private Token _token()
{
SKIP_TOKEN:
for (;;) {
Token tok;
if (!isActive()) {
try {
/* XXX Tell lexer to ignore warnings. */
source.setActive(false);
tok = source_token();
}
finally {
/* XXX Tell lexer to stop ignoring warnings. */
source.setActive(true);
}
switch (tok.getType()) {
case Token.HASH:
case Token.NL:
case Token.EOF:
/* The preprocessor has to take action here. */
break;
case Token.WHITESPACE:
return tok;
case Token.CCOMMENT:
case Token.CPPCOMMENT:
// Patch up to preserve whitespace.
if (getFeature(Feature.KEEPALLCOMMENTS))
return tok;
if (!isActive())
return toWhitespace(tok);
if (getFeature(Feature.KEEPCOMMENTS))
return tok;
return toWhitespace(tok);
default:
// Return Token.NL to preserve whitespace.
/* XXX This might lose a comment. */
return source_skipline(false);
}
}
else {
tok = source_token();
}
LEX: switch (tok.getType()) {
case Token.EOF:
/* Pop the stacks. */
return tok;
case Token.WHITESPACE:
case Token.NL:
//goto SKIP_TOKEN;
return tok;
case Token.CCOMMENT:
case Token.CPPCOMMENT:
//if(!getFeature(Feature.KEEPALLCOMMENTS)) {
// goto SKIP_TOKEN;
//}
return tok;
case '!': case '%': case '&':
case '(': case ')': case '*':
case '+': case ',': case '-':
case '/': case ':': case ';':
case '<': case '=': case '>':
case '?': case '[': case ']':
case '^': case '{': case '|':
case '}': case '~': case '.':
/* From Olivier Chafik for Objective C? */
case '@':
/* The one remaining ASCII, might as well. */
case '`':
// case '#':
case Token.AND_EQ:
case Token.ARROW:
case Token.CHARACTER:
case Token.DEC:
case Token.DIV_EQ:
case Token.ELLIPSIS:
case Token.EQ:
case Token.GE:
case Token.HEADER: /* Should only arise from include() */
case Token.INC:
case Token.LAND:
case Token.LE:
case Token.LOR:
case Token.LSH:
case Token.LSH_EQ:
case Token.SUB_EQ:
case Token.MOD_EQ:
case Token.MULT_EQ:
case Token.NE:
case Token.OR_EQ:
case Token.PLUS_EQ:
case Token.RANGE:
case Token.RSH:
case Token.RSH_EQ:
case Token.STRING:
case Token.XOR_EQ:
return tok;
case Token.INTEGER:
return tok;
case Token.IDENTIFIER:
Macro m;
macros.TryGetValue(tok.getText(), out m);
if(tok.getText() == "__has_include_next") {
Console.WriteLine();
}
if (m == null)
return tok;
if (source.isExpanding(m))
return tok;
if (macro(m, tok))
break;
return tok;
case Token.P_LINE:
if (getFeature(Feature.LINEMARKERS))
return tok;
break;
case Token.INVALID:
if (getFeature(Feature.CSYNTAX))
error(tok, (String)tok.getValue());
return tok;
default:
throw new Exception("Bad token " + tok);
// break;
case Token.HASH:
tok = source_token_nonwhite();
// (new Exception("here")).printStackTrace();
switch (tok.getType()) {
case Token.NL:
goto BREAK_LEX; /* Some code has #\n */
case Token.IDENTIFIER:
break;
default:
error(tok,
"Preprocessor directive not a word " +
tok.getText());
return source_skipline(false);
}
int _ppcmd = ppcmds[tok.getText()];
if (_ppcmd == null) {
error(tok,
"Unknown preprocessor directive " +
tok.getText());
return source_skipline(false);
}
int ppcmd = _ppcmd;
PP: switch(ppcmd) {
case PP_DEFINE:
if(!isActive())
return source_skipline(false);
else
return define();
// break;
case PP_UNDEF:
if(!isActive())
return source_skipline(false);
else
return undef();
// break;
case PP_INCLUDE:
if(!isActive())
return source_skipline(false);
else
return include(false, false);
// break;
case PP_INCLUDE_NEXT:
if(!isActive())
return source_skipline(false);
if(!getFeature(Feature.INCLUDENEXT)) {
error(tok,
"Directive include_next not enabled"
);
return source_skipline(false);
}
return include(true, false);
// break;
case PP_WARNING:
case PP_ERROR:
if(!isActive())
return source_skipline(false);
else
error(tok, ppcmd == PP_ERROR);
break;
case PP_IF:
push_state();
if(!isActive()) {
return source_skipline(false);
}
_expr_token = null;
states.Peek().setActive(expr(0) != 0);
tok = expr_token(); /* unget */
if(tok.getType() == Token.NL)
return tok;
return source_skipline(true);
// break;
case PP_ELIF:
State state = states.Peek();
if(false) {
/* Check for 'if' */
;
} else if(state.sawElse()) {
error(tok,
"#elif after #" + "else");
return source_skipline(false);
} else if(!state.isParentActive()) {
/* Nested in skipped 'if' */
return source_skipline(false);
} else if(state.isActive()) {
/* The 'if' part got executed. */
state.setParentActive(false);
/* This is like # else # if but with
* only one # end. */
state.setActive(false);
return source_skipline(false);
} else {
_expr_token = null;
state.setActive(expr(0) != 0);
tok = expr_token(); /* unget */
if(tok.getType() == Token.NL)
return tok;
return source_skipline(true);
}
// break;
case PP_ELSE:
state = states.Peek();
if(false)
/* Check for 'if' */
;
else if(state.sawElse()) {
error(tok,
"#" + "else after #" + "else");
return source_skipline(false);
} else {
state.setSawElse();
state.setActive(!state.isActive());
return source_skipline(warnings.HasFlag(Warning.ENDIF_LABELS));
}
// break;
case PP_IFDEF:
push_state();
if(!isActive()) {
return source_skipline(false);
} else {
tok = source_token_nonwhite();
// System.out.println("ifdef " + tok);
if(tok.getType() != Token.IDENTIFIER) {
error(tok,
"Expected identifier, not " +
tok.getText());
return source_skipline(false);
} else {
String text = tok.getText();
bool exists =
macros.ContainsKey(text);
states.Peek().setActive(exists);
return source_skipline(true);
}
}
// break;
case PP_IFNDEF:
push_state();
if(!isActive()) {
return source_skipline(false);
} else {
tok = source_token_nonwhite();
if(tok.getType() != Token.IDENTIFIER) {
error(tok,
"Expected identifier, not " +
tok.getText());
return source_skipline(false);
} else {
String text = tok.getText();
bool exists =
macros.ContainsKey(text);
states.Peek().setActive(!exists);
return source_skipline(true);
}
}
// break;
case PP_ENDIF:
pop_state();
return source_skipline(warnings.HasFlag(Warning.ENDIF_LABELS));
// break;
case PP_LINE:
return source_skipline(false);
// break;
case PP_PRAGMA:
if(!isActive())
return source_skipline(false);
return pragma();
// break;
case PP_IMPORT:
if(!isActive())
return source_skipline(false);
else
return import();
default:
/* Actual unknown directives are
* processed above. If we get here,
* we succeeded the map lookup but
* failed to handle it. Therefore,
* this is (unconditionally?) fatal. */
// if (isActive()) /* XXX Could be warning. */
throw new Exception(
"Internal error: Unknown directive "
+ tok);
// return source_skipline(false);
}
BREAK_PP: ;
break;
}
BREAK_LEX: ;
}
}
private Token toWhitespace(Token tok)
{
String text = tok.getText();
int len = text.Length;
bool cr = false;
int nls = 0;
for (int i = 0; i < len; i++) {
char c = text[i];
switch (c) {
case '\r':
cr = true;
nls++;
break;
case '\n':
if (cr) {
cr = false;
break;
}
goto case '\u2028';
/* fallthrough */
case '\u2028':
case '\u2029':
case '\u000B':
case '\u000C':
case '\u0085':
cr = false;
nls++;
break;
}
}
char[] cbuf = new char[nls];
for(int i = 0; i < nls; i++) { cbuf[i] = '\n'; }
return new Token(Token.WHITESPACE,
tok.getLine(), tok.getColumn(),
new String(cbuf));
}
private void source_untoken(Token tok)
{
if (this._source_token != null)
throw new InvalidOperationException("Cannot return two tokens");
this._source_token = tok;
}
private Token source_token()
{
if(_source_token != null) {
Token tok = _source_token;
_source_token = null;
if (getFeature(Feature.DEBUG))
System.Console.Error.WriteLine("Returning unget token " + tok);
return tok;
}
for (;;) {
Source s = getSource();
if (s == null) {
if (inputs.Count == 0)
return new Token(Token.EOF);
Source t = inputs[0];
inputs.RemoveAt(0);
push_source(t, true);
if (getFeature(Feature.LINEMARKERS))
return line_token(t.getLine(), t.getName(), " 1");
continue;
}
Token tok = s.token();
/* XXX Refactor with skipline() */
if(tok.getType() == Token.EOF && s.isAutopop()) {
// System.out.println("Autopop " + s);
pop_source();
Source t = getSource();
if (getFeature(Feature.LINEMARKERS)
&& s.isNumbered()
&& t != null) {
/* We actually want 'did the nested source
* contain a newline token', which isNumbered()
* approximates. This is not perfect, but works. */
return line_token(t.getLine() + 1, t.getName(), " 2");
}
continue;
}
if (getFeature(Feature.DEBUG))
System.Console.Error.WriteLine("Returning fresh token " + tok);
return tok;
}
}
/**
* Adds a "paste" operator to the expansion of this macro.
*
* A paste operator causes the next token added to be pasted
* to the previous token when the macro is expanded.
* It is an error for a macro to end with a paste token.
*/
public void addPaste(Token tok)
{
/*
* Given: tok0 ## tok1
* We generate: M_PASTE, tok0, tok1
* This extends as per a stack language:
* tok0 ## tok1 ## tok2 ->
* M_PASTE, tok0, M_PASTE, tok1, tok2
*/
this.tokens.Insert(tokens.Count - 1, tok);
}
private bool isWhite(Token tok)
{
int type = tok.getType();
return (type == Token.WHITESPACE)
|| (type == Token.CCOMMENT)
|| (type == Token.CPPCOMMENT);
}
private void expr_untoken(Token tok)
{
if(_expr_token != null)
throw new Exception (
"Cannot unget two expression tokens."
);
_expr_token = tok;
}
private Token expr_token()
{
Token tok = _expr_token;
if (tok != null) {
// System.out.println("ungetting");
_expr_token = null;
}
else {
tok = expanded_token_nonwhite();
// System.out.println("expt is " + tok);
if (tok.getType() == Token.IDENTIFIER &&
tok.getText() == "defined") {
Token la = source_token_nonwhite();
bool paren = false;
if (la.getType() == '(') {
paren = true;
la = source_token_nonwhite();
}
// System.out.println("Core token is " + la);
if (la.getType() != Token.IDENTIFIER) {
error(la,
"defined() needs identifier, not " +
la.getText());
tok = new Token(Token.INTEGER,
la.getLine(), la.getColumn(),
"0", 0);
}
else if (macros.ContainsKey(la.getText())) {
// System.out.println("Found macro");
tok = new Token(Token.INTEGER,
la.getLine(), la.getColumn(),
"1", 1);
} else if(la.getText() == "__has_include_next" || la.getText() == "__has_include" || la.getText() == "__has_feature") {
tok = new Token(Token.INTEGER,
la.getLine(), la.getColumn(),
"1", 1);
} else {
// System.out.println("Not found macro");
tok = new Token(Token.INTEGER,
la.getLine(), la.getColumn(),
"0", 0);
}
if (paren) {
la = source_token_nonwhite();
if (la.getType() != ')') {
expr_untoken(la);
error(la, "Missing ) in defined()");
}
}
}
}
// System.out.println("expr_token returns " + tok);
return tok;
}
private int expr_priority(Token op)
{
switch (op.getType()) {
case '/': return 11;
case '%': return 11;
case '*': return 11;
case '+': return 10;
case '-': return 10;
case Token.LSH: return 9;
case Token.RSH: return 9;
case '<': return 8;
case '>': return 8;
case Token.LE: return 8;
case Token.GE: return 8;
case Token.EQ: return 7;
case Token.NE: return 7;
case '&': return 6;
case '^': return 5;
case '|': return 4;
case Token.LAND: return 3;
case Token.LOR: return 2;
case '?': return 1;
default:
// System.out.println("Unrecognised operator " + op);
return 0;
}
}
public override Token token() {
Token tok = null;
int _l = line;
int _c = column;
int c = read();
int d;
switch (c) {
case '\n':
if (ppvalid) {
bol = true;
if (include) {
tok = new Token(Token.NL, _l, _c, "\n");
}
else {
int nls = 0;
do {
nls++;
d = read();
} while (d == '\n');
unread(d);
char[] text = new char[nls];
for (int i = 0; i < text.Length; i++)
text[i] = '\n';
// Skip the bol = false below.
tok = new Token(Token.NL, _l, _c, new String(text));
}
if (DEBUG)
System.Console.Error.WriteLine("lx: Returning NL: " + tok);
return tok;
}
/* Let it be handled as whitespace. */
break;
case '!':
tok = cond('=', Token.NE, '!');
break;
case '#':
if (bol)
tok = new Token(Token.HASH);
else
tok = cond('#', Token.PASTE, '#');
break;
case '+':
d = read();
if (d == '+')
tok = new Token(Token.INC);
else if (d == '=')
tok = new Token(Token.PLUS_EQ);
else
unread(d);
break;
case '-':
d = read();
if (d == '-')
tok = new Token(Token.DEC);
else if (d == '=')
tok = new Token(Token.SUB_EQ);
else if (d == '>')
tok = new Token(Token.ARROW);
else
unread(d);
break;
case '*':
tok = cond('=', Token.MULT_EQ, '*');
break;
case '/':
d = read();
if (d == '*')
tok = ccomment();
else if (d == '/')
tok = cppcomment();
else if (d == '=')
tok = new Token(Token.DIV_EQ);
else
unread(d);
break;
case '%':
d = read();
if (d == '=')
tok = new Token(Token.MOD_EQ);
else if (digraphs && d == '>')
tok = new Token('}'); // digraph
else if (digraphs && d == ':') {
bool paste = true;
d = read();
if (d != '%') {
unread(d);
tok = new Token('#'); // digraph
paste = false;
}
d = read();
if (d != ':') {
unread(d); // Unread 2 chars here.
unread('%');
tok = new Token('#'); // digraph
paste = false;
}
if(paste) {
tok = new Token(Token.PASTE); // digraph
}
}
else
unread(d);
break;
case ':':
/* :: */
d = read();
if (digraphs && d == '>')
tok = new Token(']'); // digraph
else
unread(d);
break;
case '<':
if (include) {
tok = String('<', '>');
}
else {
d = read();
if (d == '=')
tok = new Token(Token.LE);
else if (d == '<')
tok = cond('=', Token.LSH_EQ, Token.LSH);
else if (digraphs && d == ':')
tok = new Token('['); // digraph
else if (digraphs && d == '%')
tok = new Token('{'); // digraph
else
unread(d);
}
break;
case '=':
tok = cond('=', Token.EQ, '=');
break;
case '>':
d = read();
if (d == '=')
tok = new Token(Token.GE);
else if (d == '>')
tok = cond('=', Token.RSH_EQ, Token.RSH);
else
unread(d);
break;
case '^':
tok = cond('=', Token.XOR_EQ, '^');
break;
case '|':
d = read();
if (d == '=')
tok = new Token(Token.OR_EQ);
else if (d == '|')
tok = cond('=', Token.LOR_EQ, Token.LOR);
else
unread(d);
break;
case '&':
d = read();
if (d == '&')
tok = cond('=', Token.LAND_EQ, Token.LAND);
else if (d == '=')
tok = new Token(Token.AND_EQ);
else
unread(d);
break;
case '.':
d = read();
if (d == '.')
tok = cond('.', Token.ELLIPSIS, Token.RANGE);
else
unread(d);
/* XXX decimal fraction */
break;
case '0':
/* octal or hex */
d = read();
if (d == 'x' || d == 'X')
tok = number_hex((char)d);
else {
unread(d);
tok = number_octal();
}
break;
case '\'':
tok = character();
break;
case '"':
tok = String('"', '"');
break;
case -1:
close();
tok = new Token(Token.EOF, _l, _c, "<eof>");
break;
}
if (tok == null) {
if (char.IsWhiteSpace((char)c)) {
tok = whitespace(c);
}
else if (char.IsDigit((char)c)) {
tok = number_decimal(c);
}
else if (isJavaIdentifierStart(c)) {
tok = identifier(c);
}
else {
tok = new Token(c);
}
}
if (bol) {
switch (tok.getType()) {
case Token.WHITESPACE:
case Token.CCOMMENT:
break;
default:
bol = false;
break;
}
}
tok.setLocation(_l, _c);
if (DEBUG)
System.Console.WriteLine("lx: Returning " + tok);
// (new Exception("here")).printStackTrace(System.out);
return tok;
}
/* At this point, we have consumed the first M_PASTE.
* @see Macro#addPaste(Token) */
private void paste(Token ptok) {
StringBuilder buf = new StringBuilder();
Token err = null;
/* We know here that arg is null or expired,
* since we cannot paste an expanded arg. */
int count = 2;
for (int i = 0; i < count; i++) {
if (!tokens.hasNext()) {
/* XXX This one really should throw. */
error(ptok.getLine(), ptok.getColumn(),
"Paste at end of expansion");
buf.append(' ').append(ptok.getText());
break;
}
Token tok = tokens.next();
// System.out.println("Paste " + tok);
switch (tok.getType()) {
case Token.M_PASTE:
/* One extra to paste, plus one because the
* paste token didn't count. */
count += 2;
ptok = tok;
break;
case Token.M_ARG:
int idx = (int)tok.getValue();
concat(buf, args.get(idx));
break;
/* XXX Test this. */
case Token.CCOMMENT:
case Token.CPPCOMMENT:
break;
default:
buf.append(tok.getText());
break;
}
}
/* Push and re-lex. */
/*
StringBuilder src = new StringBuilder();
escape(src, buf);
StringLexerSource sl = new StringLexerSource(src.toString());
*/
StringLexerSource sl = new StringLexerSource(buf.toString());
/* XXX Check that concatenation produces a valid token. */
arg = new SourceIterator(sl);
}
/**
* Handles a warning.
*
* If a PreprocessorListener is installed, it receives the
* warning. Otherwise, an exception is thrown.
*
* @see #warning(int, int, String)
*/
protected void warning(Token tok, String msg)
{
warning(tok.getLine(), tok.getColumn(), msg);
}
/* processes and expands a macro. */
private bool macro(Macro m, Token orig)
{
Token tok;
List<Argument> args;
// System.out.println("pp: expanding " + m);
if (m.isFunctionLike()) {
for (;;) {
tok = source_token();
// System.out.println("pp: open: token is " + tok);
switch (tok.getType()) {
case Token.WHITESPACE: /* XXX Really? */
case Token.CCOMMENT:
case Token.CPPCOMMENT:
case Token.NL:
break; /* continue */
case '(':
goto BREAK_OPEN;
default:
source_untoken(tok);
return false;
}
}
BREAK_OPEN:
// tok = expanded_token_nonwhite();
tok = source_token_nonwhite();
/* We either have, or we should have args.
* This deals elegantly with the case that we have
* one empty arg. */
if (tok.getType() != ')' || m.getArgs() > 0) {
args = new List<Argument>();
Argument arg = new Argument();
int depth = 0;
bool space = false;
ARGS: for (;;) {
// System.out.println("pp: arg: token is " + tok);
switch (tok.getType()) {
case Token.EOF:
error(tok, "EOF in macro args");
return false;
case ',':
if (depth == 0) {
if (m.isVariadic() &&
/* We are building the last arg. */
args.Count == m.getArgs() - 1) {
/* Just add the comma. */
arg.addToken(tok);
}
else {
args.Add(arg);
arg = new Argument();
}
}
else {
arg.addToken(tok);
}
space = false;
break;
case ')':
if (depth == 0) {
args.Add(arg);
goto BREAK_ARGS;
}
else {
depth--;
arg.addToken(tok);
}
space = false;
break;
case '(':
depth++;
arg.addToken(tok);
space = false;
break;
case Token.WHITESPACE:
case Token.CCOMMENT:
case Token.CPPCOMMENT:
/* Avoid duplicating spaces. */
space = true;
break;
default:
/* Do not put space on the beginning of
* an argument token. */
if (space && arg.Count != 0)
arg.addToken(Token.space);
arg.addToken(tok);
space = false;
break;
}
// tok = expanded_token();
tok = source_token();
}
BREAK_ARGS:
if(m.isVariadic() && args.Count < m.getArgs()) {
args.Add(new Argument());
}
/* space may still be true here, thus trailing space
* is stripped from arguments. */
if (args.Count != m.getArgs()) {
error(tok,
"macro " + m.getName() +
" has " + m.getArgs() + " parameters " +
"but given " + args.Count + " args");
/* We could replay the arg tokens, but I
* note that GNU cpp does exactly what we do,
* i.e. output the macro name and chew the args.
*/
return false;
}
/*
for (Argument a : args)
a.expand(this);
*/
for (int i = 0; i < args.Count; i++) {
args[i].expand(this);
}
// System.out.println("Macro " + m + " args " + args);
}
else {
/* nargs == 0 and we (correctly) got () */
args = null;
}
}
else {
/* Macro without args. */
args = null;
}
if (m == __LINE__) {
push_source(new FixedTokenSource(
new Token[] { new Token(Token.INTEGER,
orig.getLine(), orig.getColumn(),
orig.getLine().ToString(),
orig.getLine()) }
), true);
}
else if (m == __FILE__) {
StringBuilder buf = new StringBuilder("\"");
String name = getSource().getName();
if (name == null)
name = "<no file>";
for (int i = 0; i < name.Length; i++) {
char c = name[i];
switch (c) {
case '\\':
buf.Append("\\\\");
break;
case '"':
buf.Append("\\\"");
break;
default:
buf.Append(c);
break;
}
}
buf.Append("\"");
String text = buf.ToString();
push_source(new FixedTokenSource(
new Token[] { new Token(Token.STRING,
orig.getLine(), orig.getColumn(),
text, text) }
), true);
}
else if (m == __COUNTER__) {
/* This could equivalently have been done by adding
* a special Macro subclass which overrides getTokens(). */
int value = this.counter++;
push_source(new FixedTokenSource(
new Token[] { new Token(Token.INTEGER,
orig.getLine(), orig.getColumn(),
value.ToString(),
value) }
), true);
}
else {
push_source(new MacroTokenSource(m, args), true);
}
return true;
}
/* For #error and #warning. */
private void error(Token pptok, bool is_error)
{
StringBuilder buf = new StringBuilder();
buf.Append('#').Append(pptok.getText()).Append(' ');
/* Peculiar construction to ditch first whitespace. */
Token tok = source_token_nonwhite();
for (;;) {
switch (tok.getType()) {
case Token.NL:
case Token.EOF:
goto BREAK_ERROR;
default:
buf.Append(tok.getText());
break;
}
tok = source_token();
}
BREAK_ERROR:
if (is_error)
error(pptok, buf.ToString());
else
warning(pptok, buf.ToString());
}
/**
* Adds a token to the expansion of this macro.
*/
public void addToken(Token tok)
{
this.tokens.Add(tok);
}
