/// decimal integer: [1-9] [0-9]* /// binary integer: "0" [bB] [01]+ /// octal integer: "0" [0-7]* /// hex integer: "0" [xX] [0-9a-fA-F]+ /// /// float: "0" [a-zA-Z except bB or xX] /// [-+]? [0-9]* ([.] [0-9]*)? ([eE] [-+]? [0-9]+)? /// GasNumericParser::GasNumericParser(llvm::StringRef str, SourceLocation loc, Preprocessor& pp, bool force_float) : NumericParser(str) { // This routine assumes that the range begin/end matches the regex for // integer and FP constants, and assumes that the byte at "*end" is both // valid and not part of the regex. Because of this, it doesn't have to // check for 'overscan' in various places. assert(!isalnum(*m_digits_end) && *m_digits_end != '.' && "Lexer didn't maximally munch?"); const char* s = str.begin(); // Look for key radix prefixes if (force_float) { // forced decimal float; skip the prefix if present m_radix = 10; m_is_float = true; if (*s == '0' && isalpha(s[1])) s += 2; } else if (*s == '0' && (s[1] == 'x' || s[1] == 'X')) { m_radix = 16; s += 2; } else if (*s == '0' && (s[1] == 'b' || s[1] == 'B')) { m_radix = 2; s += 2; } else if (*s == '0' && isalpha(s[1])) { // it's a decimal float; skip the prefix m_radix = 10; s += 2; m_is_float = true; } else if (*s == '0') { // It's an octal integer m_radix = 8; } else { // Otherwise it's a decimal m_radix = 10; } m_digits_begin = s; switch (m_radix) { case 2: s = SkipBinaryDigits(s); break; case 8: s = SkipOctalDigits(s); break; case 10: s = SkipDigits(s); break; case 16: s = SkipHexDigits(s); break; } if (s == m_digits_end) { // Done. } else if (isxdigit(*s) && (!m_is_float || (*s != 'e' && *s != 'E'))) { unsigned int err; switch (m_radix) { case 2: err = diag::err_invalid_binary_digit; break; case 8: err = diag::err_invalid_octal_digit; break; case 10: err = diag::err_invalid_decimal_digit; break; case 16: default: assert(false && "unexpected radix"); err = diag::err_invalid_decimal_digit; break; } pp.Diag(pp.AdvanceToTokenCharacter(loc, s-str.begin()), err) << std::string(s, s+1); m_had_error = true; return; } else if (m_is_float) { if (*s == '-' || *s == '+') { ++s; s = SkipDigits(s); } if (*s == '.') { ++s; s = SkipDigits(s); } if (*s == 'e' || *s == 'E') { // Float exponent const char* exponent = s; ++s; if (*s == '+' || *s == '-') // sign ++s; const char* first_non_digit = SkipDigits(s); if (first_non_digit == s) { pp.Diag(pp.AdvanceToTokenCharacter(loc, exponent-str.begin()), diag::err_exponent_has_no_digits); m_had_error = true; return; } s = first_non_digit; } } // Report an error if there are any. if (s != m_digits_end) { pp.Diag(pp.AdvanceToTokenCharacter(loc, s-str.begin()), m_is_float ? diag::err_invalid_suffix_float_constant : diag::err_invalid_suffix_integer_constant) << std::string(s, str.end()); m_had_error = true; return; } }
/// EvaluateHasIncludeCommon - Process a '__has_include("path")' /// or '__has_include_next("path")' expression. /// Returns true if successful. static bool EvaluateHasIncludeCommon(bool &Result, Token &Tok, IdentifierInfo *II, Preprocessor &PP, const DirectoryLookup *LookupFrom) { SourceLocation LParenLoc; // Get '('. PP.LexNonComment(Tok); // Ensure we have a '('. if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::err_pp_missing_lparen) << II->getName(); return false; } // Save '(' location for possible missing ')' message. LParenLoc = Tok.getLocation(); // Get the file name. PP.getCurrentLexer()->LexIncludeFilename(Tok); // Reserve a buffer to get the spelling. llvm::SmallString<128> FilenameBuffer; llvm::StringRef Filename; switch (Tok.getKind()) { case tok::eom: // If the token kind is EOM, the error has already been diagnosed. return false; case tok::angle_string_literal: case tok::string_literal: { FilenameBuffer.resize(Tok.getLength()); const char *FilenameStart = &FilenameBuffer[0]; unsigned Len = PP.getSpelling(Tok, FilenameStart); Filename = llvm::StringRef(FilenameStart, Len); break; } case tok::less: // This could be a <foo/bar.h> file coming from a macro expansion. In this // case, glue the tokens together into FilenameBuffer and interpret those. FilenameBuffer.push_back('<'); if (PP.ConcatenateIncludeName(FilenameBuffer)) return false; // Found <eom> but no ">"? Diagnostic already emitted. Filename = FilenameBuffer.str(); break; default: PP.Diag(Tok.getLocation(), diag::err_pp_expects_filename); return false; } bool isAngled = PP.GetIncludeFilenameSpelling(Tok.getLocation(), Filename); // If GetIncludeFilenameSpelling set the start ptr to null, there was an // error. if (Filename.empty()) return false; // Search include directories. const DirectoryLookup *CurDir; const FileEntry *File = PP.LookupFile(Filename, isAngled, LookupFrom, CurDir); // Get the result value. Result = true means the file exists. Result = File != 0; // Get ')'. PP.LexNonComment(Tok); // Ensure we have a trailing ). if (Tok.isNot(tok::r_paren)) { PP.Diag(Tok.getLocation(), diag::err_pp_missing_rparen) << II->getName(); PP.Diag(LParenLoc, diag::note_matching) << "("; return false; } return true; }
/// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of /// tokens into the literal string token that should be produced by the C # /// preprocessor operator. If Charify is true, then it should be turned into /// a character literal for the Microsoft charize (#@) extension. /// Token MacroArgs::StringifyArgument(const Token *ArgToks, Preprocessor &PP, bool Charify) { Token Tok; Tok.startToken(); Tok.setKind(Charify ? tok::char_constant : tok::string_literal); const Token *ArgTokStart = ArgToks; // Stringify all the tokens. llvm::SmallString<128> Result; Result += "\""; bool isFirst = true; for (; ArgToks->isNot(tok::eof); ++ArgToks) { const Token &Tok = *ArgToks; if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine())) Result += ' '; isFirst = false; // If this is a string or character constant, escape the token as specified // by 6.10.3.2p2. if (Tok.is(tok::string_literal) || // "foo" Tok.is(tok::wide_string_literal) || // L"foo" Tok.is(tok::char_constant)) { // 'x' and L'x'. std::string Str = Lexer::Stringify(PP.getSpelling(Tok)); Result.append(Str.begin(), Str.end()); } else { // Otherwise, just append the token. Do some gymnastics to get the token // in place and avoid copies where possible. unsigned CurStrLen = Result.size(); Result.resize(CurStrLen+Tok.getLength()); const char *BufPtr = &Result[CurStrLen]; unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr); // If getSpelling returned a pointer to an already uniqued version of the // string instead of filling in BufPtr, memcpy it onto our string. if (BufPtr != &Result[CurStrLen]) memcpy(&Result[CurStrLen], BufPtr, ActualTokLen); // If the token was dirty, the spelling may be shorter than the token. if (ActualTokLen != Tok.getLength()) Result.resize(CurStrLen+ActualTokLen); } } // If the last character of the string is a \, and if it isn't escaped, this // is an invalid string literal, diagnose it as specified in C99. if (Result.back() == '\\') { // Count the number of consequtive \ characters. If even, then they are // just escaped backslashes, otherwise it's an error. unsigned FirstNonSlash = Result.size()-2; // Guaranteed to find the starting " if nothing else. while (Result[FirstNonSlash] == '\\') --FirstNonSlash; if ((Result.size()-1-FirstNonSlash) & 1) { // Diagnose errors for things like: #define F(X) #X / F(\) PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal); Result.pop_back(); // remove one of the \'s. } } Result += '"'; // If this is the charify operation and the result is not a legal character // constant, diagnose it. if (Charify) { // First step, turn double quotes into single quotes: Result[0] = '\''; Result[Result.size()-1] = '\''; // Check for bogus character. bool isBad = false; if (Result.size() == 3) isBad = Result[1] == '\''; // ''' is not legal. '\' already fixed above. else isBad = (Result.size() != 4 || Result[1] != '\\'); // Not '\x' if (isBad) { PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify); Result = "' '"; // Use something arbitrary, but legal. } } PP.CreateString(&Result[0], Result.size(), Tok); return Tok; }
// #pragma unused(identifier) void PragmaUnusedHandler::HandlePragma(Preprocessor &PP, PragmaIntroducerKind Introducer, Token &UnusedTok) { // FIXME: Should we be expanding macros here? My guess is no. SourceLocation UnusedLoc = UnusedTok.getLocation(); // Lex the left '('. Token Tok; PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_lparen) << "unused"; return; } // Lex the declaration reference(s). SmallVector<Token, 5> Identifiers; SourceLocation RParenLoc; bool LexID = true; while (true) { PP.Lex(Tok); if (LexID) { if (Tok.is(tok::identifier)) { Identifiers.push_back(Tok); LexID = false; continue; } // Illegal token! PP.Diag(Tok.getLocation(), diag::warn_pragma_unused_expected_var); return; } // We are execting a ')' or a ','. if (Tok.is(tok::comma)) { LexID = true; continue; } if (Tok.is(tok::r_paren)) { RParenLoc = Tok.getLocation(); break; } // Illegal token! PP.Diag(Tok.getLocation(), diag::warn_pragma_unused_expected_punc); return; } PP.Lex(Tok); if (Tok.isNot(tok::eod)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_extra_tokens_at_eol) << "unused"; return; } // Verify that we have a location for the right parenthesis. assert(RParenLoc.isValid() && "Valid '#pragma unused' must have ')'"); assert(!Identifiers.empty() && "Valid '#pragma unused' must have arguments"); // For each identifier token, insert into the token stream a // annot_pragma_unused token followed by the identifier token. // This allows us to cache a "#pragma unused" that occurs inside an inline // C++ member function. Token *Toks = new Token[2*Identifiers.size()]; for (unsigned i=0; i != Identifiers.size(); i++) { Token &pragmaUnusedTok = Toks[2*i], &idTok = Toks[2*i+1]; pragmaUnusedTok.startToken(); pragmaUnusedTok.setKind(tok::annot_pragma_unused); pragmaUnusedTok.setLocation(UnusedLoc); idTok = Identifiers[i]; } PP.EnterTokenStream(Toks, 2*Identifiers.size(), /*DisableMacroExpansion=*/true, /*OwnsTokens=*/true); }
// #pragma pack(...) comes in the following delicious flavors: // pack '(' [integer] ')' // pack '(' 'show' ')' // pack '(' ('push' | 'pop') [',' identifier] [, integer] ')' void PragmaPackHandler::HandlePragma(Preprocessor &PP, PragmaIntroducerKind Introducer, Token &PackTok) { SourceLocation PackLoc = PackTok.getLocation(); Token Tok; PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_lparen) << "pack"; return; } Sema::PragmaPackKind Kind = Sema::PPK_Default; IdentifierInfo *Name = 0; ExprResult Alignment; SourceLocation LParenLoc = Tok.getLocation(); PP.Lex(Tok); if (Tok.is(tok::numeric_constant)) { Alignment = Actions.ActOnNumericConstant(Tok); if (Alignment.isInvalid()) return; PP.Lex(Tok); // In MSVC/gcc, #pragma pack(4) sets the alignment without affecting // the push/pop stack. // In Apple gcc, #pragma pack(4) is equivalent to #pragma pack(push, 4) if (PP.getLangOptions().ApplePragmaPack) Kind = Sema::PPK_Push; } else if (Tok.is(tok::identifier)) { const IdentifierInfo *II = Tok.getIdentifierInfo(); if (II->isStr("show")) { Kind = Sema::PPK_Show; PP.Lex(Tok); } else { if (II->isStr("push")) { Kind = Sema::PPK_Push; } else if (II->isStr("pop")) { Kind = Sema::PPK_Pop; } else { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_invalid_action); return; } PP.Lex(Tok); if (Tok.is(tok::comma)) { PP.Lex(Tok); if (Tok.is(tok::numeric_constant)) { Alignment = Actions.ActOnNumericConstant(Tok); if (Alignment.isInvalid()) return; PP.Lex(Tok); } else if (Tok.is(tok::identifier)) { Name = Tok.getIdentifierInfo(); PP.Lex(Tok); if (Tok.is(tok::comma)) { PP.Lex(Tok); if (Tok.isNot(tok::numeric_constant)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_malformed); return; } Alignment = Actions.ActOnNumericConstant(Tok); if (Alignment.isInvalid()) return; PP.Lex(Tok); } } else { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_malformed); return; } } } } else if (PP.getLangOptions().ApplePragmaPack) { // In MSVC/gcc, #pragma pack() resets the alignment without affecting // the push/pop stack. // In Apple gcc #pragma pack() is equivalent to #pragma pack(pop). Kind = Sema::PPK_Pop; } if (Tok.isNot(tok::r_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_rparen) << "pack"; return; } SourceLocation RParenLoc = Tok.getLocation(); PP.Lex(Tok); if (Tok.isNot(tok::eod)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_extra_tokens_at_eol) << "pack"; return; } Actions.ActOnPragmaPack(Kind, Name, Alignment.release(), PackLoc, LParenLoc, RParenLoc); }
/// EvaluateDefined - Process a 'defined(sym)' expression. static bool EvaluateDefined(PPValue &Result, Token &PeekTok, DefinedTracker &DT, bool ValueLive, Preprocessor &PP) { IdentifierInfo *II; Result.setBegin(PeekTok.getLocation()); // Get the next token, don't expand it. PP.LexUnexpandedNonComment(PeekTok); // Two options, it can either be a pp-identifier or a (. SourceLocation LParenLoc; if (PeekTok.is(tok::l_paren)) { // Found a paren, remember we saw it and skip it. LParenLoc = PeekTok.getLocation(); PP.LexUnexpandedNonComment(PeekTok); } if (PeekTok.is(tok::code_completion)) { if (PP.getCodeCompletionHandler()) PP.getCodeCompletionHandler()->CodeCompleteMacroName(false); PP.setCodeCompletionReached(); PP.LexUnexpandedNonComment(PeekTok); } // If we don't have a pp-identifier now, this is an error. if ((II = PeekTok.getIdentifierInfo()) == 0) { PP.Diag(PeekTok, diag::err_pp_defined_requires_identifier); return true; } // Otherwise, we got an identifier, is it defined to something? Result.Val = II->hasMacroDefinition(); Result.Val.setIsUnsigned(false); // Result is signed intmax_t. MacroDirective *Macro = 0; // If there is a macro, mark it used. if (Result.Val != 0 && ValueLive) { Macro = PP.getMacroDirective(II); PP.markMacroAsUsed(Macro->getMacroInfo()); } // Invoke the 'defined' callback. if (PPCallbacks *Callbacks = PP.getPPCallbacks()) { MacroDirective *MD = Macro; // Pass the MacroInfo for the macro name even if the value is dead. if (!MD && Result.Val != 0) MD = PP.getMacroDirective(II); Callbacks->Defined(PeekTok, MD); } // If we are in parens, ensure we have a trailing ). if (LParenLoc.isValid()) { // Consume identifier. Result.setEnd(PeekTok.getLocation()); PP.LexUnexpandedNonComment(PeekTok); if (PeekTok.isNot(tok::r_paren)) { PP.Diag(PeekTok.getLocation(), diag::err_pp_missing_rparen) << "defined"; PP.Diag(LParenLoc, diag::note_matching) << "("; return true; } // Consume the ). Result.setEnd(PeekTok.getLocation()); PP.LexNonComment(PeekTok); } else { // Consume identifier. Result.setEnd(PeekTok.getLocation()); PP.LexNonComment(PeekTok); } // Success, remember that we saw defined(X). DT.State = DefinedTracker::DefinedMacro; DT.TheMacro = II; return false; }
/// EvaluateDefined - Process a 'defined(sym)' expression. static bool EvaluateDefined(PPValue &Result, Token &PeekTok, DefinedTracker &DT, bool ValueLive, Preprocessor &PP) { SourceLocation beginLoc(PeekTok.getLocation()); Result.setBegin(beginLoc); // Get the next token, don't expand it. PP.LexUnexpandedNonComment(PeekTok); // Two options, it can either be a pp-identifier or a (. SourceLocation LParenLoc; if (PeekTok.is(tok::l_paren)) { // Found a paren, remember we saw it and skip it. LParenLoc = PeekTok.getLocation(); PP.LexUnexpandedNonComment(PeekTok); } if (PeekTok.is(tok::code_completion)) { if (PP.getCodeCompletionHandler()) PP.getCodeCompletionHandler()->CodeCompleteMacroName(false); PP.setCodeCompletionReached(); PP.LexUnexpandedNonComment(PeekTok); } // If we don't have a pp-identifier now, this is an error. if (PP.CheckMacroName(PeekTok, MU_Other)) return true; // Otherwise, we got an identifier, is it defined to something? IdentifierInfo *II = PeekTok.getIdentifierInfo(); MacroDefinition Macro = PP.getMacroDefinition(II); Result.Val = !!Macro; Result.Val.setIsUnsigned(false); // Result is signed intmax_t. // If there is a macro, mark it used. if (Result.Val != 0 && ValueLive) PP.markMacroAsUsed(Macro.getMacroInfo()); // Save macro token for callback. Token macroToken(PeekTok); // If we are in parens, ensure we have a trailing ). if (LParenLoc.isValid()) { // Consume identifier. Result.setEnd(PeekTok.getLocation()); PP.LexUnexpandedNonComment(PeekTok); if (PeekTok.isNot(tok::r_paren)) { PP.Diag(PeekTok.getLocation(), diag::err_pp_expected_after) << "'defined'" << tok::r_paren; PP.Diag(LParenLoc, diag::note_matching) << tok::l_paren; return true; } // Consume the ). Result.setEnd(PeekTok.getLocation()); PP.LexNonComment(PeekTok); } else { // Consume identifier. Result.setEnd(PeekTok.getLocation()); PP.LexNonComment(PeekTok); } // Invoke the 'defined' callback. if (PPCallbacks *Callbacks = PP.getPPCallbacks()) { Callbacks->Defined(macroToken, Macro, SourceRange(beginLoc, PeekTok.getLocation())); } // Success, remember that we saw defined(X). DT.State = DefinedTracker::DefinedMacro; DT.TheMacro = II; return false; }
/// ParseNumberStartingWithZero - This method is called when the first character /// of the number is found to be a zero. This means it is either an octal /// number (like '04') or a hex number ('0x123a') a binary number ('0b1010') or /// a floating point number (01239.123e4). Eat the prefix, determining the /// radix etc. void NumericLiteralParser::ParseNumberStartingWithZero(SourceLocation TokLoc) { assert(s[0] == '0' && "Invalid method call"); s++; int c1 = s[0]; // Handle a hex number like 0x1234. if ((c1 == 'x' || c1 == 'X') && (isHexDigit(s[1]) || s[1] == '.')) { s++; assert(s < ThisTokEnd && "didn't maximally munch?"); radix = 16; DigitsBegin = s; s = SkipHexDigits(s); bool noSignificand = (s == DigitsBegin); if (s == ThisTokEnd) { // Done. } else if (*s == '.') { s++; saw_period = true; const char *floatDigitsBegin = s; checkSeparator(TokLoc, s, CSK_BeforeDigits); s = SkipHexDigits(s); noSignificand &= (floatDigitsBegin == s); } if (noSignificand) { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin), diag::err_hexconstant_requires_digits); hadError = true; return; } // A binary exponent can appear with or with a '.'. If dotted, the // binary exponent is required. if (*s == 'p' || *s == 'P') { checkSeparator(TokLoc, s, CSK_AfterDigits); const char *Exponent = s; s++; saw_exponent = true; if (*s == '+' || *s == '-') s++; // sign const char *first_non_digit = SkipDigits(s); if (first_non_digit == s) { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin), diag::err_exponent_has_no_digits); hadError = true; return; } checkSeparator(TokLoc, s, CSK_BeforeDigits); s = first_non_digit; if (!PP.getLangOpts().HexFloats) PP.Diag(TokLoc, diag::ext_hexconstant_invalid); } else if (saw_period) { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin), diag::err_hexconstant_requires_exponent); hadError = true; } return; } // Handle simple binary numbers 0b01010 if ((c1 == 'b' || c1 == 'B') && (s[1] == '0' || s[1] == '1')) { // 0b101010 is a C++1y / GCC extension. PP.Diag(TokLoc, PP.getLangOpts().CPlusPlus14 ? diag::warn_cxx11_compat_binary_literal : PP.getLangOpts().CPlusPlus ? diag::ext_binary_literal_cxx14 : diag::ext_binary_literal); ++s; assert(s < ThisTokEnd && "didn't maximally munch?"); radix = 2; DigitsBegin = s; s = SkipBinaryDigits(s); if (s == ThisTokEnd) { // Done. } else if (isHexDigit(*s)) { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin), diag::err_invalid_binary_digit) << StringRef(s, 1); hadError = true; } // Other suffixes will be diagnosed by the caller. return; } // For now, the radix is set to 8. If we discover that we have a // floating point constant, the radix will change to 10. Octal floating // point constants are not permitted (only decimal and hexadecimal). radix = 8; DigitsBegin = s; s = SkipOctalDigits(s); if (s == ThisTokEnd) return; // Done, simple octal number like 01234 // If we have some other non-octal digit that *is* a decimal digit, see if // this is part of a floating point number like 094.123 or 09e1. if (isDigit(*s)) { const char *EndDecimal = SkipDigits(s); if (EndDecimal[0] == '.' || EndDecimal[0] == 'e' || EndDecimal[0] == 'E') { s = EndDecimal; radix = 10; } } // If we have a hex digit other than 'e' (which denotes a FP exponent) then // the code is using an incorrect base. if (isHexDigit(*s) && *s != 'e' && *s != 'E') { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s-ThisTokBegin), diag::err_invalid_octal_digit) << StringRef(s, 1); hadError = true; return; } if (*s == '.') { s++; radix = 10; saw_period = true; checkSeparator(TokLoc, s, CSK_BeforeDigits); s = SkipDigits(s); // Skip suffix. } if (*s == 'e' || *s == 'E') { // exponent checkSeparator(TokLoc, s, CSK_AfterDigits); const char *Exponent = s; s++; radix = 10; saw_exponent = true; if (*s == '+' || *s == '-') s++; // sign const char *first_non_digit = SkipDigits(s); if (first_non_digit != s) { checkSeparator(TokLoc, s, CSK_BeforeDigits); s = first_non_digit; } else { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent-ThisTokBegin), diag::err_exponent_has_no_digits); hadError = true; return; } } }
/// EvaluateValue - Evaluate the token PeekTok (and any others needed) and /// return the computed value in Result. Return true if there was an error /// parsing. This function also returns information about the form of the /// expression in DT. See above for information on what DT means. /// /// If ValueLive is false, then this value is being evaluated in a context where /// the result is not used. As such, avoid diagnostics that relate to /// evaluation. static bool EvaluateValue(PPValue &Result, Token &PeekTok, DefinedTracker &DT, bool ValueLive, Preprocessor &PP) { DT.State = DefinedTracker::Unknown; if (PeekTok.is(tok::code_completion)) { if (PP.getCodeCompletionHandler()) PP.getCodeCompletionHandler()->CodeCompletePreprocessorExpression(); PP.setCodeCompletionReached(); PP.LexNonComment(PeekTok); } // If this token's spelling is a pp-identifier, check to see if it is // 'defined' or if it is a macro. Note that we check here because many // keywords are pp-identifiers, so we can't check the kind. if (IdentifierInfo *II = PeekTok.getIdentifierInfo()) { // Handle "defined X" and "defined(X)". if (II->isStr("defined")) return(EvaluateDefined(Result, PeekTok, DT, ValueLive, PP)); // If this identifier isn't 'defined' or one of the special // preprocessor keywords and it wasn't macro expanded, it turns // into a simple 0, unless it is the C++ keyword "true", in which case it // turns into "1". if (ValueLive && II->getTokenID() != tok::kw_true && II->getTokenID() != tok::kw_false) PP.Diag(PeekTok, diag::warn_pp_undef_identifier) << II; Result.Val = II->getTokenID() == tok::kw_true; Result.Val.setIsUnsigned(false); // "0" is signed intmax_t 0. Result.setRange(PeekTok.getLocation()); PP.LexNonComment(PeekTok); return false; } switch (PeekTok.getKind()) { default: // Non-value token. PP.Diag(PeekTok, diag::err_pp_expr_bad_token_start_expr); return true; case tok::eod: case tok::r_paren: // If there is no expression, report and exit. PP.Diag(PeekTok, diag::err_pp_expected_value_in_expr); return true; case tok::numeric_constant: { SmallString<64> IntegerBuffer; bool NumberInvalid = false; StringRef Spelling = PP.getSpelling(PeekTok, IntegerBuffer, &NumberInvalid); if (NumberInvalid) return true; // a diagnostic was already reported NumericLiteralParser Literal(Spelling, PeekTok.getLocation(), PP); if (Literal.hadError) return true; // a diagnostic was already reported. if (Literal.isFloatingLiteral() || Literal.isImaginary) { PP.Diag(PeekTok, diag::err_pp_illegal_floating_literal); return true; } assert(Literal.isIntegerLiteral() && "Unknown ppnumber"); // Complain about, and drop, any ud-suffix. if (Literal.hasUDSuffix()) PP.Diag(PeekTok, diag::err_pp_invalid_udl) << /*integer*/1; // 'long long' is a C99 or C++11 feature. if (!PP.getLangOpts().C99 && Literal.isLongLong) { if (PP.getLangOpts().CPlusPlus) PP.Diag(PeekTok, PP.getLangOpts().CPlusPlus11 ? diag::warn_cxx98_compat_longlong : diag::ext_cxx11_longlong); else PP.Diag(PeekTok, diag::ext_c99_longlong); } // Parse the integer literal into Result. if (Literal.GetIntegerValue(Result.Val)) { // Overflow parsing integer literal. if (ValueLive) PP.Diag(PeekTok, diag::err_integer_literal_too_large) << /* Unsigned */ 1; Result.Val.setIsUnsigned(true); } else { // Set the signedness of the result to match whether there was a U suffix // or not. Result.Val.setIsUnsigned(Literal.isUnsigned); // Detect overflow based on whether the value is signed. If signed // and if the value is too large, emit a warning "integer constant is so // large that it is unsigned" e.g. on 12345678901234567890 where intmax_t // is 64-bits. if (!Literal.isUnsigned && Result.Val.isNegative()) { // Octal, hexadecimal, and binary literals are implicitly unsigned if // the value does not fit into a signed integer type. if (ValueLive && Literal.getRadix() == 10) PP.Diag(PeekTok, diag::ext_integer_literal_too_large_for_signed); Result.Val.setIsUnsigned(true); } } // Consume the token. Result.setRange(PeekTok.getLocation()); PP.LexNonComment(PeekTok); return false; } case tok::char_constant: // 'x' case tok::wide_char_constant: // L'x' case tok::utf8_char_constant: // u8'x' case tok::utf16_char_constant: // u'x' case tok::utf32_char_constant: { // U'x' // Complain about, and drop, any ud-suffix. if (PeekTok.hasUDSuffix()) PP.Diag(PeekTok, diag::err_pp_invalid_udl) << /*character*/0; SmallString<32> CharBuffer; bool CharInvalid = false; StringRef ThisTok = PP.getSpelling(PeekTok, CharBuffer, &CharInvalid); if (CharInvalid) return true; CharLiteralParser Literal(ThisTok.begin(), ThisTok.end(), PeekTok.getLocation(), PP, PeekTok.getKind()); if (Literal.hadError()) return true; // A diagnostic was already emitted. // Character literals are always int or wchar_t, expand to intmax_t. const TargetInfo &TI = PP.getTargetInfo(); unsigned NumBits; if (Literal.isMultiChar()) NumBits = TI.getIntWidth(); else if (Literal.isWide()) NumBits = TI.getWCharWidth(); else if (Literal.isUTF16()) NumBits = TI.getChar16Width(); else if (Literal.isUTF32()) NumBits = TI.getChar32Width(); else NumBits = TI.getCharWidth(); // Set the width. llvm::APSInt Val(NumBits); // Set the value. Val = Literal.getValue(); // Set the signedness. UTF-16 and UTF-32 are always unsigned if (Literal.isWide()) Val.setIsUnsigned(!TargetInfo::isTypeSigned(TI.getWCharType())); else if (!Literal.isUTF16() && !Literal.isUTF32()) Val.setIsUnsigned(!PP.getLangOpts().CharIsSigned); if (Result.Val.getBitWidth() > Val.getBitWidth()) { Result.Val = Val.extend(Result.Val.getBitWidth()); } else { assert(Result.Val.getBitWidth() == Val.getBitWidth() && "intmax_t smaller than char/wchar_t?"); Result.Val = Val; } // Consume the token. Result.setRange(PeekTok.getLocation()); PP.LexNonComment(PeekTok); return false; } case tok::l_paren: { SourceLocation Start = PeekTok.getLocation(); PP.LexNonComment(PeekTok); // Eat the (. // Parse the value and if there are any binary operators involved, parse // them. if (EvaluateValue(Result, PeekTok, DT, ValueLive, PP)) return true; // If this is a silly value like (X), which doesn't need parens, check for // !(defined X). if (PeekTok.is(tok::r_paren)) { // Just use DT unmodified as our result. } else { // Otherwise, we have something like (x+y), and we consumed '(x'. if (EvaluateDirectiveSubExpr(Result, 1, PeekTok, ValueLive, PP)) return true; if (PeekTok.isNot(tok::r_paren)) { PP.Diag(PeekTok.getLocation(), diag::err_pp_expected_rparen) << Result.getRange(); PP.Diag(Start, diag::note_matching) << tok::l_paren; return true; } DT.State = DefinedTracker::Unknown; } Result.setRange(Start, PeekTok.getLocation()); PP.LexNonComment(PeekTok); // Eat the ). return false; } case tok::plus: { SourceLocation Start = PeekTok.getLocation(); // Unary plus doesn't modify the value. PP.LexNonComment(PeekTok); if (EvaluateValue(Result, PeekTok, DT, ValueLive, PP)) return true; Result.setBegin(Start); return false; } case tok::minus: { SourceLocation Loc = PeekTok.getLocation(); PP.LexNonComment(PeekTok); if (EvaluateValue(Result, PeekTok, DT, ValueLive, PP)) return true; Result.setBegin(Loc); // C99 6.5.3.3p3: The sign of the result matches the sign of the operand. Result.Val = -Result.Val; // -MININT is the only thing that overflows. Unsigned never overflows. bool Overflow = !Result.isUnsigned() && Result.Val.isMinSignedValue(); // If this operator is live and overflowed, report the issue. if (Overflow && ValueLive) PP.Diag(Loc, diag::warn_pp_expr_overflow) << Result.getRange(); DT.State = DefinedTracker::Unknown; return false; } case tok::tilde: { SourceLocation Start = PeekTok.getLocation(); PP.LexNonComment(PeekTok); if (EvaluateValue(Result, PeekTok, DT, ValueLive, PP)) return true; Result.setBegin(Start); // C99 6.5.3.3p4: The sign of the result matches the sign of the operand. Result.Val = ~Result.Val; DT.State = DefinedTracker::Unknown; return false; } case tok::exclaim: { SourceLocation Start = PeekTok.getLocation(); PP.LexNonComment(PeekTok); if (EvaluateValue(Result, PeekTok, DT, ValueLive, PP)) return true; Result.setBegin(Start); Result.Val = !Result.Val; // C99 6.5.3.3p5: The sign of the result is 'int', aka it is signed. Result.Val.setIsUnsigned(false); if (DT.State == DefinedTracker::DefinedMacro) DT.State = DefinedTracker::NotDefinedMacro; else if (DT.State == DefinedTracker::NotDefinedMacro) DT.State = DefinedTracker::DefinedMacro; return false; } // FIXME: Handle #assert } }
// #pragma unused(identifier) void PragmaUnusedHandler::HandlePragma(Preprocessor &PP, Token &UnusedTok) { // FIXME: Should we be expanding macros here? My guess is no. SourceLocation UnusedLoc = UnusedTok.getLocation(); // Lex the left '('. Token Tok; PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_lparen) << "unused"; return; } SourceLocation LParenLoc = Tok.getLocation(); // Lex the declaration reference(s). llvm::SmallVector<Action::ExprTy*, 5> Ex; SourceLocation RParenLoc; bool LexID = true; while (true) { PP.Lex(Tok); if (LexID) { if (Tok.is(tok::identifier)) { Action::OwningExprResult Name = Actions.ActOnIdentifierExpr(parser.CurScope, Tok.getLocation(), *Tok.getIdentifierInfo(), false); if (Name.isInvalid()) { if (!Ex.empty()) Action::MultiExprArg Release(Actions, &Ex[0], Ex.size()); return; } Ex.push_back(Name.release()); LexID = false; continue; } // Illegal token! Release the parsed expressions (if any) and emit // a warning. if (!Ex.empty()) Action::MultiExprArg Release(Actions, &Ex[0], Ex.size()); PP.Diag(Tok.getLocation(), diag::warn_pragma_unused_expected_var); return; } // We are execting a ')' or a ','. if (Tok.is(tok::comma)) { LexID = true; continue; } if (Tok.is(tok::r_paren)) { RParenLoc = Tok.getLocation(); break; } // Illegal token! Release the parsed expressions (if any) and emit // a warning. if (!Ex.empty()) Action::MultiExprArg Release(Actions, &Ex[0], Ex.size()); PP.Diag(Tok.getLocation(), diag::warn_pragma_unused_expected_punc); return; } // Verify that we have a location for the right parenthesis. assert(RParenLoc.isValid() && "Valid '#pragma unused' must have ')'"); assert(!Ex.empty() && "Valid '#pragma unused' must have arguments"); // Perform the action to handle the pragma. Actions.ActOnPragmaUnused(&Ex[0], Ex.size(), UnusedLoc, LParenLoc, RParenLoc); }
/// integer-constant: [C99 6.4.4.1] /// decimal-constant integer-suffix /// octal-constant integer-suffix /// hexadecimal-constant integer-suffix /// binary-literal integer-suffix [GNU, C++1y] /// user-defined-integer-literal: [C++11 lex.ext] /// decimal-literal ud-suffix /// octal-literal ud-suffix /// hexadecimal-literal ud-suffix /// binary-literal ud-suffix [GNU, C++1y] /// decimal-constant: /// nonzero-digit /// decimal-constant digit /// octal-constant: /// 0 /// octal-constant octal-digit /// hexadecimal-constant: /// hexadecimal-prefix hexadecimal-digit /// hexadecimal-constant hexadecimal-digit /// hexadecimal-prefix: one of /// 0x 0X /// binary-literal: /// 0b binary-digit /// 0B binary-digit /// binary-literal binary-digit /// integer-suffix: /// unsigned-suffix [long-suffix] /// unsigned-suffix [long-long-suffix] /// long-suffix [unsigned-suffix] /// long-long-suffix [unsigned-sufix] /// nonzero-digit: /// 1 2 3 4 5 6 7 8 9 /// octal-digit: /// 0 1 2 3 4 5 6 7 /// hexadecimal-digit: /// 0 1 2 3 4 5 6 7 8 9 /// a b c d e f /// A B C D E F /// binary-digit: /// 0 /// 1 /// unsigned-suffix: one of /// u U /// long-suffix: one of /// l L /// long-long-suffix: one of /// ll LL /// /// floating-constant: [C99 6.4.4.2] /// TODO: add rules... /// NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling) : ThisTokBegin(TokSpelling.begin()), ThisTokEnd(TokSpelling.end()) { // This routine assumes that the range begin/end matches the regex for integer // and FP constants (specifically, the 'pp-number' regex), and assumes that // the byte at "*end" is both valid and not part of the regex. Because of // this, it doesn't have to check for 'overscan' in various places. assert(!isPreprocessingNumberBody(*ThisTokEnd) && "didn't maximally munch?"); s = DigitsBegin = ThisTokBegin; saw_exponent = false; saw_period = false; saw_ud_suffix = false; isLong = false; isUnsigned = false; isLongLong = false; isFloat = false; isImaginary = false; MicrosoftInteger = 0; hadError = false; SourceLocation TokLoc; if (*s == '0') { // parse radix ParseNumberStartingWithZero(TokLoc); if (hadError) return; } else { // the first digit is non-zero radix = 10; s = SkipDigits(s); if (s == ThisTokEnd) { // Done. } else if (isHexDigit(*s) && !(*s == 'e' || *s == 'E')) { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, s - ThisTokBegin), diag::err_invalid_decimal_digit) << StringRef(s, 1); hadError = true; return; } else if (*s == '.') { checkSeparator(TokLoc, s, CSK_AfterDigits); s++; saw_period = true; checkSeparator(TokLoc, s, CSK_BeforeDigits); s = SkipDigits(s); } if ((*s == 'e' || *s == 'E')) { // exponent checkSeparator(TokLoc, s, CSK_AfterDigits); const char *Exponent = s; s++; saw_exponent = true; if (*s == '+' || *s == '-') s++; // sign checkSeparator(TokLoc, s, CSK_BeforeDigits); const char *first_non_digit = SkipDigits(s); if (first_non_digit != s) { s = first_non_digit; } else { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, Exponent - ThisTokBegin), diag::err_exponent_has_no_digits); hadError = true; return; } } } SuffixBegin = s; checkSeparator(TokLoc, s, CSK_AfterDigits); // Parse the suffix. At this point we can classify whether we have an FP or // integer constant. bool isFPConstant = isFloatingLiteral(); const char *ImaginarySuffixLoc = nullptr; // Loop over all of the characters of the suffix. If we see something bad, // we break out of the loop. for (; s != ThisTokEnd; ++s) { switch (*s) { case 'f': // FP Suffix for "float" case 'F': if (!isFPConstant) break; // Error for integer constant. if (isFloat || isLong) break; // FF, LF invalid. isFloat = true; continue; // Success. case 'u': case 'U': if (isFPConstant) break; // Error for floating constant. if (isUnsigned) break; // Cannot be repeated. isUnsigned = true; continue; // Success. case 'l': case 'L': if (isLong || isLongLong) break; // Cannot be repeated. if (isFloat) break; // LF invalid. // Check for long long. The L's need to be adjacent and the same case. if (s[1] == s[0]) { assert(s + 1 < ThisTokEnd && "didn't maximally munch?"); if (isFPConstant) break; // long long invalid for floats. isLongLong = true; ++s; // Eat both of them. } else { isLong = true; } continue; // Success. case 'i': case 'I': if (PP.getLangOpts().MicrosoftExt) { if (isLong || isLongLong || MicrosoftInteger) break; if (!isFPConstant) { // Allow i8, i16, i32, i64, and i128. switch (s[1]) { case '8': s += 2; // i8 suffix MicrosoftInteger = 8; break; case '1': if (s[2] == '6') { s += 3; // i16 suffix MicrosoftInteger = 16; } else if (s[2] == '2' && s[3] == '8') { s += 4; // i128 suffix MicrosoftInteger = 128; } break; case '3': if (s[2] == '2') { s += 3; // i32 suffix MicrosoftInteger = 32; } break; case '6': if (s[2] == '4') { s += 3; // i64 suffix MicrosoftInteger = 64; } break; default: break; } } if (MicrosoftInteger) { assert(s <= ThisTokEnd && "didn't maximally munch?"); break; } } // "i", "if", and "il" are user-defined suffixes in C++1y. if (*s == 'i' && PP.getLangOpts().CPlusPlus14) break; // fall through. case 'j': case 'J': if (isImaginary) break; // Cannot be repeated. isImaginary = true; ImaginarySuffixLoc = s; continue; // Success. } // If we reached here, there was an error or a ud-suffix. break; } if (s != ThisTokEnd) { // FIXME: Don't bother expanding UCNs if !tok.hasUCN(). expandUCNs(UDSuffixBuf, StringRef(SuffixBegin, ThisTokEnd - SuffixBegin)); if (isValidUDSuffix(PP.getLangOpts(), UDSuffixBuf)) { // Any suffix pieces we might have parsed are actually part of the // ud-suffix. isLong = false; isUnsigned = false; isLongLong = false; isFloat = false; isImaginary = false; MicrosoftInteger = 0; saw_ud_suffix = true; return; } // Report an error if there are any. PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, SuffixBegin - ThisTokBegin), isFPConstant ? diag::err_invalid_suffix_float_constant : diag::err_invalid_suffix_integer_constant) << StringRef(SuffixBegin, ThisTokEnd-SuffixBegin); hadError = true; return; } if (isImaginary) { PP.Diag(PP.AdvanceToTokenCharacter(TokLoc, ImaginarySuffixLoc - ThisTokBegin), diag::ext_imaginary_constant); } }
/// EvaluateDefined - Process a 'defined(sym)' expression. static bool EvaluateDefined(PPValue &Result, Token &PeekTok, DefinedTracker &DT, bool ValueLive, Preprocessor &PP) { SourceLocation beginLoc(PeekTok.getLocation()); Result.setBegin(beginLoc); // Get the next token, don't expand it. PP.LexUnexpandedNonComment(PeekTok); // Two options, it can either be a pp-identifier or a (. SourceLocation LParenLoc; if (PeekTok.is(tok::l_paren)) { // Found a paren, remember we saw it and skip it. LParenLoc = PeekTok.getLocation(); PP.LexUnexpandedNonComment(PeekTok); } if (PeekTok.is(tok::code_completion)) { if (PP.getCodeCompletionHandler()) PP.getCodeCompletionHandler()->CodeCompleteMacroName(false); PP.setCodeCompletionReached(); PP.LexUnexpandedNonComment(PeekTok); } // If we don't have a pp-identifier now, this is an error. if (PP.CheckMacroName(PeekTok, MU_Other)) return true; // Otherwise, we got an identifier, is it defined to something? IdentifierInfo *II = PeekTok.getIdentifierInfo(); MacroDefinition Macro = PP.getMacroDefinition(II); Result.Val = !!Macro; Result.Val.setIsUnsigned(false); // Result is signed intmax_t. // If there is a macro, mark it used. if (Result.Val != 0 && ValueLive) PP.markMacroAsUsed(Macro.getMacroInfo()); // Save macro token for callback. Token macroToken(PeekTok); // If we are in parens, ensure we have a trailing ). if (LParenLoc.isValid()) { // Consume identifier. Result.setEnd(PeekTok.getLocation()); PP.LexUnexpandedNonComment(PeekTok); if (PeekTok.isNot(tok::r_paren)) { PP.Diag(PeekTok.getLocation(), diag::err_pp_expected_after) << "'defined'" << tok::r_paren; PP.Diag(LParenLoc, diag::note_matching) << tok::l_paren; return true; } // Consume the ). Result.setEnd(PeekTok.getLocation()); PP.LexNonComment(PeekTok); } else { // Consume identifier. Result.setEnd(PeekTok.getLocation()); PP.LexNonComment(PeekTok); } // [cpp.cond]p4: // Prior to evaluation, macro invocations in the list of preprocessing // tokens that will become the controlling constant expression are replaced // (except for those macro names modified by the 'defined' unary operator), // just as in normal text. If the token 'defined' is generated as a result // of this replacement process or use of the 'defined' unary operator does // not match one of the two specified forms prior to macro replacement, the // behavior is undefined. // This isn't an idle threat, consider this program: // #define FOO // #define BAR defined(FOO) // #if BAR // ... // #else // ... // #endif // clang and gcc will pick the #if branch while Visual Studio will take the // #else branch. Emit a warning about this undefined behavior. if (beginLoc.isMacroID()) { bool IsFunctionTypeMacro = PP.getSourceManager() .getSLocEntry(PP.getSourceManager().getFileID(beginLoc)) .getExpansion() .isFunctionMacroExpansion(); // For object-type macros, it's easy to replace // #define FOO defined(BAR) // with // #if defined(BAR) // #define FOO 1 // #else // #define FOO 0 // #endif // and doing so makes sense since compilers handle this differently in // practice (see example further up). But for function-type macros, // there is no good way to write // # define FOO(x) (defined(M_ ## x) && M_ ## x) // in a different way, and compilers seem to agree on how to behave here. // So warn by default on object-type macros, but only warn in -pedantic // mode on function-type macros. if (IsFunctionTypeMacro) PP.Diag(beginLoc, diag::warn_defined_in_function_type_macro); else PP.Diag(beginLoc, diag::warn_defined_in_object_type_macro); } // Invoke the 'defined' callback. if (PPCallbacks *Callbacks = PP.getPPCallbacks()) { Callbacks->Defined(macroToken, Macro, SourceRange(beginLoc, PeekTok.getLocation())); } // Success, remember that we saw defined(X). DT.State = DefinedTracker::DefinedMacro; DT.TheMacro = II; return false; }
// #pragma 'align' '=' {'native','natural','mac68k','power','reset'} // #pragma 'options 'align' '=' {'native','natural','mac68k','power','reset'} static void ParseAlignPragma(Preprocessor &PP, Token &FirstTok, bool IsOptions) { Token Tok; if (IsOptions) { PP.Lex(Tok); if (Tok.isNot(tok::identifier) || !Tok.getIdentifierInfo()->isStr("align")) { PP.Diag(Tok.getLocation(), diag::warn_pragma_options_expected_align); return; } } PP.Lex(Tok); if (Tok.isNot(tok::equal)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_align_expected_equal) << IsOptions; return; } PP.Lex(Tok); if (Tok.isNot(tok::identifier)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_identifier) << (IsOptions ? "options" : "align"); return; } Sema::PragmaOptionsAlignKind Kind = Sema::POAK_Natural; const IdentifierInfo *II = Tok.getIdentifierInfo(); if (II->isStr("native")) Kind = Sema::POAK_Native; else if (II->isStr("natural")) Kind = Sema::POAK_Natural; else if (II->isStr("packed")) Kind = Sema::POAK_Packed; else if (II->isStr("power")) Kind = Sema::POAK_Power; else if (II->isStr("mac68k")) Kind = Sema::POAK_Mac68k; else if (II->isStr("reset")) Kind = Sema::POAK_Reset; else { PP.Diag(Tok.getLocation(), diag::warn_pragma_align_invalid_option) << IsOptions; return; } PP.Lex(Tok); if (Tok.isNot(tok::eod)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_extra_tokens_at_eol) << (IsOptions ? "options" : "align"); return; } Token *Toks = (Token*) PP.getPreprocessorAllocator().Allocate( sizeof(Token) * 1, llvm::alignOf<Token>()); new (Toks) Token(); Toks[0].startToken(); Toks[0].setKind(tok::annot_pragma_align); Toks[0].setLocation(FirstTok.getLocation()); Toks[0].setAnnotationValue(reinterpret_cast<void*>( static_cast<uintptr_t>(Kind))); PP.EnterTokenStream(Toks, 1, /*DisableMacroExpansion=*/true, /*OwnsTokens=*/false); }
// #pragma pack(...) comes in the following delicious flavors: // pack '(' [integer] ')' // pack '(' 'show' ')' // pack '(' ('push' | 'pop') [',' identifier] [, integer] ')' void PragmaPackHandler::HandlePragma(Preprocessor &PP, PragmaIntroducerKind Introducer, Token &PackTok) { SourceLocation PackLoc = PackTok.getLocation(); Token Tok; PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_lparen) << "pack"; return; } Sema::PragmaPackKind Kind = Sema::PPK_Default; IdentifierInfo *Name = 0; Token Alignment; Alignment.startToken(); SourceLocation LParenLoc = Tok.getLocation(); PP.Lex(Tok); if (Tok.is(tok::numeric_constant)) { Alignment = Tok; PP.Lex(Tok); // In MSVC/gcc, #pragma pack(4) sets the alignment without affecting // the push/pop stack. // In Apple gcc, #pragma pack(4) is equivalent to #pragma pack(push, 4) if (PP.getLangOpts().ApplePragmaPack) Kind = Sema::PPK_Push; } else if (Tok.is(tok::identifier)) { const IdentifierInfo *II = Tok.getIdentifierInfo(); if (II->isStr("show")) { Kind = Sema::PPK_Show; PP.Lex(Tok); } else { if (II->isStr("push")) { Kind = Sema::PPK_Push; } else if (II->isStr("pop")) { Kind = Sema::PPK_Pop; } else { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_invalid_action); return; } PP.Lex(Tok); if (Tok.is(tok::comma)) { PP.Lex(Tok); if (Tok.is(tok::numeric_constant)) { Alignment = Tok; PP.Lex(Tok); } else if (Tok.is(tok::identifier)) { Name = Tok.getIdentifierInfo(); PP.Lex(Tok); if (Tok.is(tok::comma)) { PP.Lex(Tok); if (Tok.isNot(tok::numeric_constant)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_malformed); return; } Alignment = Tok; PP.Lex(Tok); } } else { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_malformed); return; } } } } else if (PP.getLangOpts().ApplePragmaPack) { // In MSVC/gcc, #pragma pack() resets the alignment without affecting // the push/pop stack. // In Apple gcc #pragma pack() is equivalent to #pragma pack(pop). Kind = Sema::PPK_Pop; } if (Tok.isNot(tok::r_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_rparen) << "pack"; return; } SourceLocation RParenLoc = Tok.getLocation(); PP.Lex(Tok); if (Tok.isNot(tok::eod)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_extra_tokens_at_eol) << "pack"; return; } PragmaPackInfo *Info = (PragmaPackInfo*) PP.getPreprocessorAllocator().Allocate( sizeof(PragmaPackInfo), llvm::alignOf<PragmaPackInfo>()); new (Info) PragmaPackInfo(); Info->Kind = Kind; Info->Name = Name; Info->Alignment = Alignment; Info->LParenLoc = LParenLoc; Info->RParenLoc = RParenLoc; Token *Toks = (Token*) PP.getPreprocessorAllocator().Allocate( sizeof(Token) * 1, llvm::alignOf<Token>()); new (Toks) Token(); Toks[0].startToken(); Toks[0].setKind(tok::annot_pragma_pack); Toks[0].setLocation(PackLoc); Toks[0].setAnnotationValue(static_cast<void*>(Info)); PP.EnterTokenStream(Toks, 1, /*DisableMacroExpansion=*/true, /*OwnsTokens=*/false); }
// #pragma unused(identifier) void PragmaUnusedHandler::HandlePragma(Preprocessor &PP, Token &UnusedTok) { // FIXME: Should we be expanding macros here? My guess is no. SourceLocation UnusedLoc = UnusedTok.getLocation(); // Lex the left '('. Token Tok; PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_lparen) << "unused"; return; } SourceLocation LParenLoc = Tok.getLocation(); // Lex the declaration reference(s). llvm::SmallVector<Token, 5> Identifiers; SourceLocation RParenLoc; bool LexID = true; while (true) { PP.Lex(Tok); if (LexID) { if (Tok.is(tok::identifier)) { Identifiers.push_back(Tok); LexID = false; continue; } // Illegal token! PP.Diag(Tok.getLocation(), diag::warn_pragma_unused_expected_var); return; } // We are execting a ')' or a ','. if (Tok.is(tok::comma)) { LexID = true; continue; } if (Tok.is(tok::r_paren)) { RParenLoc = Tok.getLocation(); break; } // Illegal token! PP.Diag(Tok.getLocation(), diag::warn_pragma_unused_expected_punc); return; } PP.Lex(Tok); if (Tok.isNot(tok::eom)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_extra_tokens_at_eol) << "unused"; return; } // Verify that we have a location for the right parenthesis. assert(RParenLoc.isValid() && "Valid '#pragma unused' must have ')'"); assert(!Identifiers.empty() && "Valid '#pragma unused' must have arguments"); // Perform the action to handle the pragma. Actions.ActOnPragmaUnused(Identifiers.data(), Identifiers.size(), parser.CurScope, UnusedLoc, LParenLoc, RParenLoc); }
/// EvaluateDirectiveSubExpr - Evaluate the subexpression whose first token is /// PeekTok, and whose precedence is PeekPrec. This returns the result in LHS. /// /// If ValueLive is false, then this value is being evaluated in a context where /// the result is not used. As such, avoid diagnostics that relate to /// evaluation, such as division by zero warnings. static bool EvaluateDirectiveSubExpr(PPValue &LHS, unsigned MinPrec, Token &PeekTok, bool ValueLive, Preprocessor &PP) { unsigned PeekPrec = getPrecedence(PeekTok.getKind()); // If this token isn't valid, report the error. if (PeekPrec == ~0U) { PP.Diag(PeekTok.getLocation(), diag::err_pp_expr_bad_token_binop) << LHS.getRange(); return true; } while (1) { // If this token has a lower precedence than we are allowed to parse, return // it so that higher levels of the recursion can parse it. if (PeekPrec < MinPrec) return false; tok::TokenKind Operator = PeekTok.getKind(); // If this is a short-circuiting operator, see if the RHS of the operator is // dead. Note that this cannot just clobber ValueLive. Consider // "0 && 1 ? 4 : 1 / 0", which is parsed as "(0 && 1) ? 4 : (1 / 0)". In // this example, the RHS of the && being dead does not make the rest of the // expr dead. bool RHSIsLive; if (Operator == tok::ampamp && LHS.Val == 0) RHSIsLive = false; // RHS of "0 && x" is dead. else if (Operator == tok::pipepipe && LHS.Val != 0) RHSIsLive = false; // RHS of "1 || x" is dead. else if (Operator == tok::question && LHS.Val == 0) RHSIsLive = false; // RHS (x) of "0 ? x : y" is dead. else RHSIsLive = ValueLive; // Consume the operator, remembering the operator's location for reporting. SourceLocation OpLoc = PeekTok.getLocation(); PP.LexNonComment(PeekTok); PPValue RHS(LHS.getBitWidth()); // Parse the RHS of the operator. DefinedTracker DT; if (EvaluateValue(RHS, PeekTok, DT, RHSIsLive, PP)) return true; // Remember the precedence of this operator and get the precedence of the // operator immediately to the right of the RHS. unsigned ThisPrec = PeekPrec; PeekPrec = getPrecedence(PeekTok.getKind()); // If this token isn't valid, report the error. if (PeekPrec == ~0U) { PP.Diag(PeekTok.getLocation(), diag::err_pp_expr_bad_token_binop) << RHS.getRange(); return true; } // Decide whether to include the next binop in this subexpression. For // example, when parsing x+y*z and looking at '*', we want to recursively // handle y*z as a single subexpression. We do this because the precedence // of * is higher than that of +. The only strange case we have to handle // here is for the ?: operator, where the precedence is actually lower than // the LHS of the '?'. The grammar rule is: // // conditional-expression ::= // logical-OR-expression ? expression : conditional-expression // where 'expression' is actually comma-expression. unsigned RHSPrec; if (Operator == tok::question) // The RHS of "?" should be maximally consumed as an expression. RHSPrec = getPrecedence(tok::comma); else // All others should munch while higher precedence. RHSPrec = ThisPrec+1; if (PeekPrec >= RHSPrec) { if (EvaluateDirectiveSubExpr(RHS, RHSPrec, PeekTok, RHSIsLive, PP)) return true; PeekPrec = getPrecedence(PeekTok.getKind()); } assert(PeekPrec <= ThisPrec && "Recursion didn't work!"); // Usual arithmetic conversions (C99 6.3.1.8p1): result is unsigned if // either operand is unsigned. llvm::APSInt Res(LHS.getBitWidth()); switch (Operator) { case tok::question: // No UAC for x and y in "x ? y : z". case tok::lessless: // Shift amount doesn't UAC with shift value. case tok::greatergreater: // Shift amount doesn't UAC with shift value. case tok::comma: // Comma operands are not subject to UACs. case tok::pipepipe: // Logical || does not do UACs. case tok::ampamp: // Logical && does not do UACs. break; // No UAC default: Res.setIsUnsigned(LHS.isUnsigned()|RHS.isUnsigned()); // If this just promoted something from signed to unsigned, and if the // value was negative, warn about it. if (ValueLive && Res.isUnsigned()) { if (!LHS.isUnsigned() && LHS.Val.isNegative()) PP.Diag(OpLoc, diag::warn_pp_convert_lhs_to_positive) << LHS.Val.toString(10, true) + " to " + LHS.Val.toString(10, false) << LHS.getRange() << RHS.getRange(); if (!RHS.isUnsigned() && RHS.Val.isNegative()) PP.Diag(OpLoc, diag::warn_pp_convert_rhs_to_positive) << RHS.Val.toString(10, true) + " to " + RHS.Val.toString(10, false) << LHS.getRange() << RHS.getRange(); } LHS.Val.setIsUnsigned(Res.isUnsigned()); RHS.Val.setIsUnsigned(Res.isUnsigned()); } bool Overflow = false; switch (Operator) { default: llvm_unreachable("Unknown operator token!"); case tok::percent: if (RHS.Val != 0) Res = LHS.Val % RHS.Val; else if (ValueLive) { PP.Diag(OpLoc, diag::err_pp_remainder_by_zero) << LHS.getRange() << RHS.getRange(); return true; } break; case tok::slash: if (RHS.Val != 0) { if (LHS.Val.isSigned()) Res = llvm::APSInt(LHS.Val.sdiv_ov(RHS.Val, Overflow), false); else Res = LHS.Val / RHS.Val; } else if (ValueLive) { PP.Diag(OpLoc, diag::err_pp_division_by_zero) << LHS.getRange() << RHS.getRange(); return true; } break; case tok::star: if (Res.isSigned()) Res = llvm::APSInt(LHS.Val.smul_ov(RHS.Val, Overflow), false); else Res = LHS.Val * RHS.Val; break; case tok::lessless: { // Determine whether overflow is about to happen. if (LHS.isUnsigned()) Res = LHS.Val.ushl_ov(RHS.Val, Overflow); else Res = llvm::APSInt(LHS.Val.sshl_ov(RHS.Val, Overflow), false); break; } case tok::greatergreater: { // Determine whether overflow is about to happen. unsigned ShAmt = static_cast<unsigned>(RHS.Val.getLimitedValue()); if (ShAmt >= LHS.getBitWidth()) Overflow = true, ShAmt = LHS.getBitWidth()-1; Res = LHS.Val >> ShAmt; break; } case tok::plus: if (LHS.isUnsigned()) Res = LHS.Val + RHS.Val; else Res = llvm::APSInt(LHS.Val.sadd_ov(RHS.Val, Overflow), false); break; case tok::minus: if (LHS.isUnsigned()) Res = LHS.Val - RHS.Val; else Res = llvm::APSInt(LHS.Val.ssub_ov(RHS.Val, Overflow), false); break; case tok::lessequal: Res = LHS.Val <= RHS.Val; Res.setIsUnsigned(false); // C99 6.5.8p6, result is always int (signed) break; case tok::less: Res = LHS.Val < RHS.Val; Res.setIsUnsigned(false); // C99 6.5.8p6, result is always int (signed) break; case tok::greaterequal: Res = LHS.Val >= RHS.Val; Res.setIsUnsigned(false); // C99 6.5.8p6, result is always int (signed) break; case tok::greater: Res = LHS.Val > RHS.Val; Res.setIsUnsigned(false); // C99 6.5.8p6, result is always int (signed) break; case tok::exclaimequal: Res = LHS.Val != RHS.Val; Res.setIsUnsigned(false); // C99 6.5.9p3, result is always int (signed) break; case tok::equalequal: Res = LHS.Val == RHS.Val; Res.setIsUnsigned(false); // C99 6.5.9p3, result is always int (signed) break; case tok::amp: Res = LHS.Val & RHS.Val; break; case tok::caret: Res = LHS.Val ^ RHS.Val; break; case tok::pipe: Res = LHS.Val | RHS.Val; break; case tok::ampamp: Res = (LHS.Val != 0 && RHS.Val != 0); Res.setIsUnsigned(false); // C99 6.5.13p3, result is always int (signed) break; case tok::pipepipe: Res = (LHS.Val != 0 || RHS.Val != 0); Res.setIsUnsigned(false); // C99 6.5.14p3, result is always int (signed) break; case tok::comma: // Comma is invalid in pp expressions in c89/c++ mode, but is valid in C99 // if not being evaluated. if (!PP.getLangOpts().C99 || ValueLive) PP.Diag(OpLoc, diag::ext_pp_comma_expr) << LHS.getRange() << RHS.getRange(); Res = RHS.Val; // LHS = LHS,RHS -> RHS. break; case tok::question: { // Parse the : part of the expression. if (PeekTok.isNot(tok::colon)) { PP.Diag(PeekTok.getLocation(), diag::err_expected) << tok::colon << LHS.getRange() << RHS.getRange(); PP.Diag(OpLoc, diag::note_matching) << tok::question; return true; } // Consume the :. PP.LexNonComment(PeekTok); // Evaluate the value after the :. bool AfterColonLive = ValueLive && LHS.Val == 0; PPValue AfterColonVal(LHS.getBitWidth()); DefinedTracker DT; if (EvaluateValue(AfterColonVal, PeekTok, DT, AfterColonLive, PP)) return true; // Parse anything after the : with the same precedence as ?. We allow // things of equal precedence because ?: is right associative. if (EvaluateDirectiveSubExpr(AfterColonVal, ThisPrec, PeekTok, AfterColonLive, PP)) return true; // Now that we have the condition, the LHS and the RHS of the :, evaluate. Res = LHS.Val != 0 ? RHS.Val : AfterColonVal.Val; RHS.setEnd(AfterColonVal.getRange().getEnd()); // Usual arithmetic conversions (C99 6.3.1.8p1): result is unsigned if // either operand is unsigned. Res.setIsUnsigned(RHS.isUnsigned() | AfterColonVal.isUnsigned()); // Figure out the precedence of the token after the : part. PeekPrec = getPrecedence(PeekTok.getKind()); break; } case tok::colon: // Don't allow :'s to float around without being part of ?: exprs. PP.Diag(OpLoc, diag::err_pp_colon_without_question) << LHS.getRange() << RHS.getRange(); return true; } // If this operator is live and overflowed, report the issue. if (Overflow && ValueLive) PP.Diag(OpLoc, diag::warn_pp_expr_overflow) << LHS.getRange() << RHS.getRange(); // Put the result back into 'LHS' for our next iteration. LHS.Val = Res; LHS.setEnd(RHS.getRange().getEnd()); } }
// #pragma pack(...) comes in the following delicious flavors: // pack '(' [integer] ')' // pack '(' 'show' ')' // pack '(' ('push' | 'pop') [',' identifier] [, integer] ')' void PragmaPackHandler::HandlePragma(Preprocessor &PP, Token &PackTok) { // FIXME: Should we be expanding macros here? My guess is no. SourceLocation PackLoc = PackTok.getLocation(); Token Tok; PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_lparen) << "pack"; return; } Action::PragmaPackKind Kind = Action::PPK_Default; IdentifierInfo *Name = 0; Action::OwningExprResult Alignment(Actions); SourceLocation LParenLoc = Tok.getLocation(); PP.Lex(Tok); if (Tok.is(tok::numeric_constant)) { Alignment = Actions.ActOnNumericConstant(Tok); if (Alignment.isInvalid()) return; PP.Lex(Tok); } else if (Tok.is(tok::identifier)) { const IdentifierInfo *II = Tok.getIdentifierInfo(); if (II->isStr("show")) { Kind = Action::PPK_Show; PP.Lex(Tok); } else { if (II->isStr("push")) { Kind = Action::PPK_Push; } else if (II->isStr("pop")) { Kind = Action::PPK_Pop; } else { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_invalid_action); return; } PP.Lex(Tok); if (Tok.is(tok::comma)) { PP.Lex(Tok); if (Tok.is(tok::numeric_constant)) { Alignment = Actions.ActOnNumericConstant(Tok); if (Alignment.isInvalid()) return; PP.Lex(Tok); } else if (Tok.is(tok::identifier)) { Name = Tok.getIdentifierInfo(); PP.Lex(Tok); if (Tok.is(tok::comma)) { PP.Lex(Tok); if (Tok.isNot(tok::numeric_constant)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_malformed); return; } Alignment = Actions.ActOnNumericConstant(Tok); if (Alignment.isInvalid()) return; PP.Lex(Tok); } } else { PP.Diag(Tok.getLocation(), diag::warn_pragma_pack_malformed); return; } } } } if (Tok.isNot(tok::r_paren)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_expected_rparen) << "pack"; return; } PP.Lex(Tok); if (Tok.isNot(tok::eom)) { PP.Diag(Tok.getLocation(), diag::warn_pragma_extra_tokens_at_eol) << "pack"; return; } SourceLocation RParenLoc = Tok.getLocation(); Actions.ActOnPragmaPack(Kind, Name, Alignment.release(), PackLoc, LParenLoc, RParenLoc); }
/// \brief Handle the microsoft \#pragma comment extension. /// /// The syntax is: /// \code /// #pragma comment(linker, "foo") /// \endcode /// 'linker' is one of five identifiers: compiler, exestr, lib, linker, user. /// "foo" is a string, which is fully macro expanded, and permits string /// concatenation, embedded escape characters etc. See MSDN for more details. void PragmaCommentHandler::HandlePragma(Preprocessor &PP, PragmaIntroducerKind Introducer, Token &Tok) { SourceLocation CommentLoc = Tok.getLocation(); PP.Lex(Tok); if (Tok.isNot(tok::l_paren)) { PP.Diag(CommentLoc, diag::err_pragma_comment_malformed); return; } // Read the identifier. PP.Lex(Tok); if (Tok.isNot(tok::identifier)) { PP.Diag(CommentLoc, diag::err_pragma_comment_malformed); return; } // Verify that this is one of the 5 whitelisted options. IdentifierInfo *II = Tok.getIdentifierInfo(); Sema::PragmaMSCommentKind Kind = llvm::StringSwitch<Sema::PragmaMSCommentKind>(II->getName()) .Case("linker", Sema::PCK_Linker) .Case("lib", Sema::PCK_Lib) .Case("compiler", Sema::PCK_Compiler) .Case("exestr", Sema::PCK_ExeStr) .Case("user", Sema::PCK_User) .Default(Sema::PCK_Unknown); if (Kind == Sema::PCK_Unknown) { PP.Diag(Tok.getLocation(), diag::err_pragma_comment_unknown_kind); return; } // Read the optional string if present. PP.Lex(Tok); std::string ArgumentString; if (Tok.is(tok::comma) && !PP.LexStringLiteral(Tok, ArgumentString, "pragma comment", /*MacroExpansion=*/true)) return; // FIXME: warn that 'exestr' is deprecated. // FIXME: If the kind is "compiler" warn if the string is present (it is // ignored). // The MSDN docs say that "lib" and "linker" require a string and have a short // whitelist of linker options they support, but in practice MSVC doesn't // issue a diagnostic. Therefore neither does clang. if (Tok.isNot(tok::r_paren)) { PP.Diag(Tok.getLocation(), diag::err_pragma_comment_malformed); return; } PP.Lex(Tok); // eat the r_paren. if (Tok.isNot(tok::eod)) { PP.Diag(Tok.getLocation(), diag::err_pragma_comment_malformed); return; } // If the pragma is lexically sound, notify any interested PPCallbacks. if (PP.getPPCallbacks()) PP.getPPCallbacks()->PragmaComment(CommentLoc, II, ArgumentString); Actions.ActOnPragmaMSComment(Kind, ArgumentString); }