bool cxxParserParseNextToken(void)
{
CXXToken * t = cxxTokenCreate();
// The token chain should not be allowed to grow arbitrairly large.
// The token structures are quite big and it's easy to grow up to
// 5-6GB or memory usage. However this limit should be large enough
// to accomodate all the reasonable statements that could have some
// information in them. This includes multiple function prototypes
// in a single statement (ImageMagick has some examples) but probably
// does NOT include large data tables.
if(g_cxx.pTokenChain->iCount > 16384)
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
cxxTokenChainAppend(g_cxx.pTokenChain,t);
g_cxx.pToken = t;
cxxParserSkipToNonWhiteSpace();
// FIXME: this cpp handling is kind of broken:
// it works only because the moon is in the correct phase.
cppBeginStatement();
// This must be done after getting char from input
t->iLineNumber = getInputLineNumber();
t->oFilePosition = getInputFilePosition();
if(g_cxx.iChar == EOF)
{
t->eType = CXXTokenTypeEOF;
t->bFollowedBySpace = false;
return false;
}
unsigned int uInfo = UINFO(g_cxx.iChar);
//fprintf(stderr,"Char %c %02x info %u\n",g_cxx.iChar,g_cxx.iChar,uInfo);
if(uInfo & CXXCharTypeStartOfIdentifier)
{
// word
t->eType = CXXTokenTypeIdentifier;
t->bFollowedBySpace = false;
vStringPut(t->pszWord,g_cxx.iChar);
// special case for tile, which may actually be an operator
if(g_cxx.iChar == '~')
{
// may be followed by space!
g_cxx.iChar = cppGetc();
if(isspace(g_cxx.iChar))
{
t->bFollowedBySpace = true;
g_cxx.iChar = cppGetc();
while(isspace(g_cxx.iChar))
g_cxx.iChar = cppGetc();
}
// non space
uInfo = UINFO(g_cxx.iChar);
if(!(uInfo & CXXCharTypeStartOfIdentifier))
{
// this is not an identifier after all
t->eType = CXXTokenTypeOperator;
if((!t->bFollowedBySpace) && g_cxx.iChar == '=')
{
// make ~= single token so it's not handled as
// a separate assignment
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
}
return true;
}
} else {
g_cxx.iChar = cppGetc();
}
for(;;)
{
uInfo = UINFO(g_cxx.iChar);
if(!(uInfo & CXXCharTypePartOfIdentifier))
break;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
}
int iCXXKeyword = lookupKeyword(t->pszWord->buffer,g_cxx.eLanguage);
if(iCXXKeyword >= 0)
{
if(
(
(iCXXKeyword == CXXKeywordFINAL) &&
(!g_cxx.bParsingClassStructOrUnionDeclaration)
) || (
(
(iCXXKeyword == CXXKeywordPUBLIC) ||
(iCXXKeyword == CXXKeywordPROTECTED) ||
(iCXXKeyword == CXXKeywordPRIVATE)
) &&
(!g_cxx.bEnablePublicProtectedPrivateKeywords)
)
)
{
t->eType = CXXTokenTypeIdentifier;
} else {
t->eType = CXXTokenTypeKeyword;
t->eKeyword = (enum CXXKeyword)iCXXKeyword;
if(iCXXKeyword == CXXKeyword__ATTRIBUTE__)
{
// special handling for __attribute__
return cxxParserParseNextTokenCondenseAttribute();
}
}
} else {
const cppMacroInfo * pMacro = cppFindMacro(vStringValue(t->pszWord));
if(pMacro)
{
CXX_DEBUG_PRINT("Macro %s",vStringValue(t->pszWord));
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
CXXToken * pParameterChain = NULL;
if(pMacro->hasParameterList)
{
CXX_DEBUG_PRINT("Macro has parameter list");
if(!cxxParserParseNextTokenSkipMacroParenthesis(&pParameterChain))
return false;
}
// This is used to avoid infinite recursion in substitution
// (things like -D foo=foo or similar)
static int iReplacementRecursionCount = 0;
if(pMacro->replacements)
{
CXX_DEBUG_PRINT("The token has replacements: applying");
if(iReplacementRecursionCount < 1024)
{
// unget last char
cppUngetc(g_cxx.iChar);
// unget the replacement
cxxParserParseNextTokenApplyReplacement(
pMacro,
pParameterChain
);
g_cxx.iChar = cppGetc();
}
}
if(pParameterChain)
cxxTokenDestroy(pParameterChain);
iReplacementRecursionCount++;
// Have no token to return: parse it
CXX_DEBUG_PRINT("Parse inner token");
bool bRet = cxxParserParseNextToken();
CXX_DEBUG_PRINT("Parsed inner token: %s type %d",g_cxx.pToken->pszWord->buffer,g_cxx.pToken->eType);
iReplacementRecursionCount--;
return bRet;
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
if(g_cxx.iChar == '-')
{
// special case for pointer
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == '>')
{
t->eType = CXXTokenTypePointerOperator;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
} else {
t->eType = CXXTokenTypeOperator;
if(g_cxx.iChar == '-')
{
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
#if 0
// As long as we use cppGetc() we don't need this
if(g_cxx.iChar == '"')
{
// special case for strings
t->eType = CXXTokenTypeStringConstant;
vStringPut(t->pszWord,g_cxx.iChar);
// We don't even care of storing the other chars: we don't need
// them for parsing
// FIXME: We might need them in signature:() tag.. maybe add
// them up to a certain length only?
for(;;)
{
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = false;
return true;
}
if(g_cxx.iChar == '\\')
{
// escape
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = false;
return true;
}
} else if(g_cxx.iChar == '"')
{
g_cxx.iChar = cppGetc();
break;
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
#else
if(g_cxx.iChar == STRING_SYMBOL)
{
t->eType = CXXTokenTypeStringConstant;
vStringPut(t->pszWord,'"');
vStringPut(t->pszWord,'"');
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
#endif
#if 0
// As long as we use cppGetc() we don't need this
if(g_cxx.iChar == '\'')
{
// special case for strings
t->eType = CXXTokenTypeCharacterConstant;
vStringPut(t->pszWord,g_cxx.iChar);
// We don't even care storing the other chars: we don't
// need them for parsing
for(;;)
{
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = false;
return true;
}
if(g_cxx.iChar == '\\')
{
// escape
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = false;
return true;
}
} else if(g_cxx.iChar == '\'')
{
g_cxx.iChar = cppGetc();
break;
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
#else
if(g_cxx.iChar == CHAR_SYMBOL)
{
t->eType = CXXTokenTypeCharacterConstant;
vStringPut(t->pszWord,'\'');
vStringPut(t->pszWord,'\'');
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
#endif
if(uInfo & CXXCharTypeDecimalDigit)
{
// number
t->eType = CXXTokenTypeNumber;
vStringPut(t->pszWord,g_cxx.iChar);
for(;;)
{
g_cxx.iChar = cppGetc();
uInfo = UINFO(g_cxx.iChar);
if(!(uInfo & CXXCharTypeValidInNumber))
break;
vStringPut(t->pszWord,g_cxx.iChar);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
if(uInfo & CXXCharTypeNamedSingleOrRepeatedCharToken)
{
t->eType = g_aCharTable[g_cxx.iChar].uSingleTokenType;
vStringPut(t->pszWord,g_cxx.iChar);
int iChar = g_cxx.iChar;
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == iChar)
{
t->eType = g_aCharTable[g_cxx.iChar].uMultiTokenType;
// We could signal a syntax error with more than two colons
// or equal signs...but we're tolerant
do {
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
} while(g_cxx.iChar == iChar);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
if(uInfo & CXXCharTypeNamedSingleOrOperatorToken)
{
t->eType = g_aCharTable[g_cxx.iChar].uSingleTokenType;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = UINFO(g_cxx.iChar);
if(uInfo & (CXXCharTypeOperator | CXXCharTypeNamedSingleOrOperatorToken))
{
t->eType = CXXTokenTypeOperator;
do {
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = UINFO(g_cxx.iChar);
} while(
uInfo &
(CXXCharTypeOperator | CXXCharTypeNamedSingleOrOperatorToken)
);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
if(uInfo & CXXCharTypeNamedSingleCharToken)
{
t->eType = g_aCharTable[g_cxx.iChar].uSingleTokenType;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
if(uInfo & CXXCharTypeOperator)
{
t->eType = CXXTokenTypeOperator;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = UINFO(g_cxx.iChar);
while(uInfo & CXXCharTypeOperator)
{
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = UINFO(g_cxx.iChar);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
t->eType = CXXTokenTypeUnknown;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return true;
}
static bool cxxParserParseBlockInternal(bool bExpectClosingBracket)
{
CXX_DEBUG_ENTER();
//char * szScopeName = cxxScopeGetFullName();
//CXX_DEBUG_PRINT("Scope name is '%s'",szScopeName ? szScopeName : "");
cxxParserNewStatement();
if(bExpectClosingBracket)
{
// FIXME: this cpp handling is kind of broken:
// it works only because the moon is in the correct phase.
cppBeginStatement();
}
for(;;)
{
if(!cxxParserParseNextToken())
{
found_eof:
if(bExpectClosingBracket)
{
CXX_DEBUG_LEAVE_TEXT(
"Syntax error: found EOF in block but a closing " \
"bracket was expected!"
);
return false;
}
CXX_DEBUG_LEAVE_TEXT("EOF in main block");
return true; // EOF
}
process_token:
CXX_DEBUG_PRINT(
"Token '%s' of type 0x%02x",
vStringValue(g_cxx.pToken->pszWord),
g_cxx.pToken->eType
);
switch(g_cxx.pToken->eType)
{
case CXXTokenTypeKeyword:
{
switch(g_cxx.pToken->eKeyword)
{
case CXXKeywordNAMESPACE:
{
enum CXXScopeType eScopeType = cxxScopeGetType();
if(
(
// toplevel or nested within a namespace
(eScopeType == CXXScopeTypeNamespace) ||
// namespace X = Y inside a function
(eScopeType == CXXScopeTypeFunction)
) && (
// either certainly C++
g_cxx.bConfirmedCPPLanguage ||
// or a "sane" namespace syntax
(
!cxxTokenChainPreviousTokenOfType(
g_cxx.pToken,
CXXTokenTypeStar |
CXXTokenTypeAnd |
CXXTokenTypeKeyword
)
)
)
)
{
if(!cxxParserParseNamespace())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse namespace");
return false;
}
} else {
// If we're pretty sure this is C++ then this is a syntax error.
// If we're not sure (namely when we're in a *.h file) then
// let's try to be flexible: treat the namespace keyword as an identifier.
if(!g_cxx.bConfirmedCPPLanguage)
{
CXX_DEBUG_LEAVE_TEXT(
"Found namespace in unexpected place, but we're not sure it's really C++ "
"so we'll treat it as an identifier instead"
);
g_cxx.pToken->eType = CXXTokenTypeIdentifier;
continue;
}
CXX_DEBUG_LEAVE_TEXT(
"Found namespace in a wrong place: we're probably out of sync"
);
return false;
}
cxxParserNewStatement();
}
break;
case CXXKeywordTEMPLATE:
if(!cxxParserParseTemplatePrefix())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse template");
return false;
}
// Here we are just after the "template<parameters>" prefix.
break;
case CXXKeywordTYPEDEF:
// Mark the next declaration as a typedef
g_cxx.uKeywordState |= CXXParserKeywordStateSeenTypedef;
cxxTokenChainClear(g_cxx.pTokenChain);
break;
case CXXKeywordENUM:
if(!cxxParserParseEnum())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse enum");
return false;
}
break;
case CXXKeywordCLASS:
if(!cxxParserParseClassStructOrUnion(CXXKeywordCLASS,CXXTagCPPKindCLASS,CXXScopeTypeClass))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse class/struct/union");
return false;
}
break;
case CXXKeywordSTRUCT:
if(!cxxParserParseClassStructOrUnion(CXXKeywordSTRUCT,CXXTagKindSTRUCT,CXXScopeTypeStruct))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse class/struct/union");
return false;
}
break;
case CXXKeywordUNION:
if(!cxxParserParseClassStructOrUnion(CXXKeywordUNION,CXXTagKindUNION,CXXScopeTypeUnion))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse class/struct/union");
return false;
}
break;
case CXXKeywordPUBLIC:
case CXXKeywordPROTECTED:
case CXXKeywordPRIVATE:
// Note that the class keyword has its own handler
// so the only possibility here is an access specifier
if(!cxxParserParseAccessSpecifier())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse access specifier");
return false;
}
break;
case CXXKeywordUSING:
if(!cxxParserParseUsingClause())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse using clause");
return false;
}
cxxParserNewStatement();
break;
case CXXKeywordIF:
case CXXKeywordFOR:
case CXXKeywordWHILE:
case CXXKeywordSWITCH:
if(!cxxParserParseIfForWhileSwitch())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse if/for/while/switch");
return false;
}
cxxParserNewStatement();
// Force the cpp preprocessor to think that we're in the middle of a statement.
cppBeginStatement();
break;
case CXXKeywordTRY:
case CXXKeywordELSE:
case CXXKeywordDO:
// parse as normal statement/block
cxxParserNewStatement();
// Force the cpp preprocessor to think that we're in the middle of a statement.
cppBeginStatement();
break;
case CXXKeywordRETURN:
if(cxxParserCurrentLanguageIsCPP())
{
// may be followed by a lambda, otherwise it's not interesting.
cxxParserNewStatement();
g_cxx.uKeywordState |= CXXParserKeywordStateSeenReturn;
} else {
// ignore
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF,
false))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse return");
return false;
}
cxxParserNewStatement();
}
break;
case CXXKeywordCONTINUE:
case CXXKeywordBREAK:
case CXXKeywordGOTO:
// ignore
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF,
false))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse continue/break/goto");
return false;
}
cxxParserNewStatement();
break;
case CXXKeywordTHROW:
// ignore when inside a function
if(cxxScopeGetType() == CXXScopeTypeFunction)
{
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF,
false))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse return/continue/break");
return false;
}
cxxParserNewStatement();
}
break;
case CXXKeywordCASE:
// ignore
if(!cxxParserParseUpToOneOf(
CXXTokenTypeSemicolon | CXXTokenTypeEOF | CXXTokenTypeSingleColon,
false
))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse case keyword");
return false;
}
cxxParserNewStatement();
break;
case CXXKeywordEXTERN:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenExtern;
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
if(!cxxParserParseNextToken())
goto found_eof;
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeStringConstant))
{
// assume extern "language"
// Strictly speaking this is a C++ only syntax.
// However we allow it also in C as it doesn't really hurt.
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
// Note that extern "C" may be followed by a block with declarations
//
// extern "C" { ... }
//
// However in this case the declarations are ALSO definitions
// and extern "C" is used only to specify the name mangling mode.
//
// extern "C" int x; <-- a declaration and not a definition
// extern "C" { int x; } <-- a declaration and definition: x IS defined
// here and is NOT extern.
//
// A variable in an extern "C" block has to be re-declared extern again
// to be really treated as declaration only.
//
// extern "C" { extern int x; }
//
// So in this case we do NOT treat the inner declarations as extern
// and we don't need specific handling code for this case.
} else {
// something else: handle it the normal way
goto process_token;
}
break;
case CXXKeywordSTATIC:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenStatic;
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
break;
case CXXKeywordINLINE:
case CXXKeyword__INLINE:
case CXXKeyword__INLINE__:
case CXXKeyword__FORCEINLINE:
case CXXKeyword__FORCEINLINE__:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenInline;
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
break;
case CXXKeywordEXPLICIT:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenExplicit;
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
break;
case CXXKeywordOPERATOR:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenOperator;
break;
case CXXKeywordVIRTUAL:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenVirtual;
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
break;
case CXXKeywordMUTABLE:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenMutable;
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
break;
// "const" and "volatile" are part of the type. Don't treat them specially
// and don't attempt to extract an eventual typedef yet,
// as there might be a struct/class/union keyword following.
case CXXKeywordVOLATILE:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenVolatile;
break;
case CXXKeywordCONST:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenConst;
break;
default:
if(g_cxx.uKeywordState & CXXParserKeywordStateSeenTypedef)
{
g_cxx.uKeywordState &= ~CXXParserKeywordStateSeenTypedef;
if(!cxxParserParseGenericTypedef())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse generic typedef");
return false;
}
cxxParserNewStatement();
}
break;
}
}
break;
case CXXTokenTypeSemicolon:
{
if(
(cxxParserCurrentLanguageIsC()) &&
cxxScopeIsGlobal() &&
(!(g_cxx.uKeywordState & CXXParserKeywordStateSeenExtern)) &&
(!(g_cxx.uKeywordState & CXXParserKeywordStateSeenTypedef))
)
{
// Special handling of K&R style function declarations.
// We might be in the following situation:
//
// type whatever fname(par1,par2) int par1; int par2; {
// ^
//
switch(cxxParserMaybeParseKnRStyleFunctionDefinition())
{
case 1:
// K&R parser did the job and started a new statement
break;
case 0:
// something else
cxxParserAnalyzeOtherStatement();
break;
default:
CXX_DEBUG_LEAVE_TEXT("Failed to check for K&R style function definition");
return false;
break;
}
} else {
// K&R style function declarations not allowed here.
cxxParserAnalyzeOtherStatement();
}
cxxParserNewStatement();
}
break;
case CXXTokenTypeSingleColon:
{
// label ?
if(
(g_cxx.pTokenChain->iCount == 2) &&
cxxTokenTypeIs(
cxxTokenChainFirst(g_cxx.pTokenChain),
CXXTokenTypeIdentifier
)
)
{
CXXToken * pFirst = cxxTokenChainFirst(g_cxx.pTokenChain);
// assume it's label
tagEntryInfo * tag = cxxTagBegin(CXXTagKindLABEL,pFirst);
if(tag)
{
tag->isFileScope = true;
cxxTagCommit();
}
} else {
// what is this? (default: and similar things have been handled at keyword level)
}
}
break;
case CXXTokenTypeOpeningBracket:
if(!cxxParserParseBlockHandleOpeningBracket())
{
CXX_DEBUG_LEAVE_TEXT("Failed to handle opening bracket");
return false;
}
break;
case CXXTokenTypeClosingBracket:
// scope finished
if(!bExpectClosingBracket)
{
CXX_DEBUG_LEAVE_TEXT(
"Found unexpected closing bracket: probably preprocessing problem"
);
return false;
}
CXX_DEBUG_LEAVE_TEXT("Closing bracket!");
cxxParserNewStatement();
return true;
break;
case CXXTokenTypeOpeningParenthesis:
case CXXTokenTypeOpeningSquareParenthesis:
if(!cxxParserParseAndCondenseCurrentSubchain(
CXXTokenTypeOpeningBracket | CXXTokenTypeOpeningParenthesis |
CXXTokenTypeOpeningSquareParenthesis,
true,
false
))
{
CXX_DEBUG_LEAVE_TEXT("Parsing the parenthesis failed");
return false;
}
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeEOF))
{
if(bExpectClosingBracket)
{
CXX_DEBUG_LEAVE_TEXT(
"Syntax error: found EOF in block but a closing bracket was expected!"
);
return false;
}
return true; // EOF
}
break;
case CXXTokenTypeIdentifier:
if(g_cxx.uKeywordState & CXXParserKeywordStateSeenTypedef)
{
g_cxx.uKeywordState &= ~CXXParserKeywordStateSeenTypedef;
if(!cxxParserParseGenericTypedef())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse generic typedef");
return false;
}
cxxParserNewStatement();
}
break;
default:
// something else we didn't handle
break;
}
}
CXX_DEBUG_LEAVE_TEXT("WARNING: Not reached");
return true;
}
boolean cxxParserParseNextToken(void)
{
CXXToken * t = cxxTokenCreate();
// The token chain should not be allowed to grow arbitrairly large.
// The token structures are quite big and it's easy to grow up to
// 5-6GB or memory usage. However this limit should be large enough
// to accomodate all the reasonable statements that could have some
// information in them. This includes multiple function prototypes
// in a single statement (ImageMagick has some examples) but probably
// does NOT include large data tables.
if(g_cxx.pTokenChain->iCount > 16384)
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
cxxTokenChainAppend(g_cxx.pTokenChain,t);
g_cxx.pToken = t;
cxxParserSkipToNonWhiteSpace();
// FIXME: this cpp handling is kind of broken:
// it works only because the moon is in the correct phase.
cppBeginStatement();
// This must be done after getting char from input
t->iLineNumber = getInputLineNumber();
t->oFilePosition = getInputFilePosition();
if(g_cxx.iChar == EOF)
{
t->eType = CXXTokenTypeEOF;
t->bFollowedBySpace = FALSE;
return FALSE;
}
unsigned int uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
//printf("Char %c %02x info %u\n",g_cxx.iChar,g_cxx.iChar,uInfo);
if(uInfo & CXXCharTypeStartOfIdentifier)
{
// word
t->eType = CXXTokenTypeIdentifier;
t->bFollowedBySpace = FALSE;
vStringPut(t->pszWord,g_cxx.iChar);
// special case for tile, which may actually be an operator
if(g_cxx.iChar == '~')
{
// may be followed by space!
g_cxx.iChar = cppGetc();
if(isspace(g_cxx.iChar))
{
t->bFollowedBySpace = TRUE;
g_cxx.iChar = cppGetc();
while(isspace(g_cxx.iChar))
g_cxx.iChar = cppGetc();
}
// non space
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
if(!(uInfo & CXXCharTypeStartOfIdentifier))
{
// this is not an identifier after all
t->eType = CXXTokenTypeOperator;
if((!t->bFollowedBySpace) && g_cxx.iChar == '=')
{
// make ~= single token so it's not handled as
// a separate assignment
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
}
return TRUE;
}
} else {
g_cxx.iChar = cppGetc();
}
for(;;)
{
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
if(!(uInfo & CXXCharTypePartOfIdentifier))
break;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
}
int iCXXKeyword = lookupKeyword(t->pszWord->buffer,g_cxx.eLanguage);
if(iCXXKeyword >= 0)
{
if(
(
(iCXXKeyword == CXXKeywordFINAL) &&
(!g_cxx.bParsingClassStructOrUnionDeclaration)
) || (
(
(iCXXKeyword == CXXKeywordPUBLIC) ||
(iCXXKeyword == CXXKeywordPROTECTED) ||
(iCXXKeyword == CXXKeywordPRIVATE)
) &&
(!g_cxx.bEnablePublicProtectedPrivateKeywords)
)
)
{
t->eType = CXXTokenTypeIdentifier;
} else {
t->eType = CXXTokenTypeKeyword;
t->eKeyword = (enum CXXKeyword)iCXXKeyword;
if(iCXXKeyword == CXXKeyword__ATTRIBUTE__)
{
// special handling for __attribute__
return cxxParserParseNextTokenCondenseAttribute();
}
}
} else {
boolean bIgnoreParens = FALSE;
const char * szReplacement = NULL;
if(isIgnoreToken(
vStringValue(t->pszWord),
&bIgnoreParens,
&szReplacement
))
{
CXX_DEBUG_PRINT("Ignore token %s",vStringValue(t->pszWord));
// FIXME: Handle ignore parens!
if(szReplacement && *szReplacement)
{
vStringClear(t->pszWord);
vStringCatS(t->pszWord,szReplacement);
} else {
// skip
cxxTokenChainDestroyLast(g_cxx.pTokenChain);
return cxxParserParseNextToken();
}
}
}
t->bFollowedBySpace = t->bFollowedBySpace | isspace(g_cxx.iChar);
return TRUE;
}
if(g_cxx.iChar == '-')
{
// special case for pointer
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == '>')
{
t->eType = CXXTokenTypePointerOperator;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
} else {
t->eType = CXXTokenTypeOperator;
if(g_cxx.iChar == '-')
{
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
#if 0
// As long as we use cppGetc() we don't need this
if(g_cxx.iChar == '"')
{
// special case for strings
t->eType = CXXTokenTypeStringConstant;
vStringPut(t->pszWord,g_cxx.iChar);
// We don't even care of storing the other chars: we don't need
// them for parsing
// FIXME: We might need them in signature:() tag.. maybe add
// them up to a certain length only?
for(;;)
{
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = FALSE;
return TRUE;
}
if(g_cxx.iChar == '\\')
{
// escape
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = FALSE;
return TRUE;
}
} else if(g_cxx.iChar == '"')
{
g_cxx.iChar = cppGetc();
break;
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
#else
if(g_cxx.iChar == STRING_SYMBOL)
{
t->eType = CXXTokenTypeStringConstant;
vStringPut(t->pszWord,'"');
vStringPut(t->pszWord,'"');
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
#endif
#if 0
// As long as we use cppGetc() we don't need this
if(g_cxx.iChar == '\'')
{
// special case for strings
t->eType = CXXTokenTypeCharacterConstant;
vStringPut(t->pszWord,g_cxx.iChar);
// We don't even care storing the other chars: we don't
// need them for parsing
for(;;)
{
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = FALSE;
return TRUE;
}
if(g_cxx.iChar == '\\')
{
// escape
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == EOF)
{
t->bFollowedBySpace = FALSE;
return TRUE;
}
} else if(g_cxx.iChar == '\'')
{
g_cxx.iChar = cppGetc();
break;
}
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
#else
if(g_cxx.iChar == CHAR_SYMBOL)
{
t->eType = CXXTokenTypeCharacterConstant;
vStringPut(t->pszWord,'\'');
vStringPut(t->pszWord,'\'');
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
#endif
if(uInfo & CXXCharTypeDecimalDigit)
{
// number
t->eType = CXXTokenTypeNumber;
vStringPut(t->pszWord,g_cxx.iChar);
for(;;)
{
g_cxx.iChar = cppGetc();
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
if(!(uInfo & CXXCharTypeValidInNumber))
break;
vStringPut(t->pszWord,g_cxx.iChar);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
if(uInfo & CXXCharTypeNamedSingleOrRepeatedCharToken)
{
t->eType = g_aCharTable[g_cxx.iChar].uSingleTokenType;
vStringPut(t->pszWord,g_cxx.iChar);
int iChar = g_cxx.iChar;
g_cxx.iChar = cppGetc();
if(g_cxx.iChar == iChar)
{
t->eType = g_aCharTable[g_cxx.iChar].uMultiTokenType;
// We could signal a syntax error with more than two colons
// or equal signs...but we're tolerant
do {
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
} while(g_cxx.iChar == iChar);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
if(uInfo & CXXCharTypeNamedSingleOrOperatorToken)
{
t->eType = g_aCharTable[g_cxx.iChar].uSingleTokenType;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
if(uInfo & (CXXCharTypeOperator | CXXCharTypeNamedSingleOrOperatorToken))
{
t->eType = CXXTokenTypeOperator;
do {
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
} while(
uInfo &
(CXXCharTypeOperator | CXXCharTypeNamedSingleOrOperatorToken)
);
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
if(uInfo & CXXCharTypeNamedSingleCharToken)
{
t->eType = g_aCharTable[g_cxx.iChar].uSingleTokenType;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
if(uInfo & CXXCharTypeOperator)
{
t->eType = CXXTokenTypeOperator;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
while(uInfo & CXXCharTypeOperator)
{
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
uInfo = (g_cxx.iChar < 0x80) ? g_aCharTable[g_cxx.iChar].uType : 0;
}
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
t->eType = CXXTokenTypeUnknown;
vStringPut(t->pszWord,g_cxx.iChar);
g_cxx.iChar = cppGetc();
t->bFollowedBySpace = isspace(g_cxx.iChar);
return TRUE;
}
//
// This is the toplevel scanning function. It's a forward-only scanner that keeps
// accumulating tokens in the chain until either a characteristic token is found
// or the statement ends. When a characteristic token is found it usually enters
// a specialized scanning routine (e.g for classes, namespaces, structs...).
// When the statement ends without finding any characteristic token the chain
// is passed to an analysis routine which does a second scan pass.
//
boolean cxxParserParseBlock(boolean bExpectClosingBracket)
{
CXX_DEBUG_ENTER();
//char * szScopeName = cxxScopeGetFullName();
//CXX_DEBUG_PRINT("Scope name is '%s'",szScopeName ? szScopeName : "");
cxxParserNewStatement();
if(bExpectClosingBracket)
cppBeginStatement(); // FIXME: this cpp handling is broken: it works only because the moon is in the correct phase.
for(;;)
{
if(!cxxParserParseNextToken())
{
if(bExpectClosingBracket)
{
CXX_DEBUG_LEAVE_TEXT("Syntax error: found EOF in block but a closing bracket was expected!");
return FALSE;
}
CXX_DEBUG_LEAVE_TEXT("EOF in main block");
return TRUE; // EOF
}
CXX_DEBUG_PRINT("Token '%s' of type 0x%02x",vStringValue(g_cxx.pToken->pszWord),g_cxx.pToken->eType);
switch(g_cxx.pToken->eType)
{
case CXXTokenTypeKeyword:
{
switch(g_cxx.pToken->eKeyword)
{
case CXXKeywordNAMESPACE:
{
int iCurrentScopeKind = cxxScopeGetKind();
if(iCurrentScopeKind == CXXTagKindNAMESPACE)
{
// namespaces can be nested only within themselves
if(!cxxParserParseNamespace())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse namespace");
return FALSE;
}
} else {
// hm... syntax error?
CXX_DEBUG_LEAVE_TEXT("Found namespace in a wrong place: we're probably out of sync");
return FALSE;
}
//cxxParserNewStatement(); <-- already called by cxxParserParseNamespace()
}
break;
case CXXKeywordTEMPLATE:
if(!cxxParserParseTemplatePrefix())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse template");
return FALSE;
}
// Here we are just after the "template<parameters>" prefix.
break;
case CXXKeywordTYPEDEF:
// Mark the next declaration as a typedef
g_cxx.uKeywordState |= CXXParserKeywordStateSeenTypedef;
cxxTokenChainClear(g_cxx.pTokenChain);
break;
case CXXKeywordENUM:
if(!cxxParserParseEnum())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse enum");
return FALSE;
}
break;
case CXXKeywordCLASS:
if(!cxxParserParseClassStructOrUnion(CXXKeywordCLASS,CXXTagKindCLASS))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse class/struct/union");
return FALSE;
}
break;
case CXXKeywordSTRUCT:
if(!cxxParserParseClassStructOrUnion(CXXKeywordSTRUCT,CXXTagKindSTRUCT))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse class/struct/union");
return FALSE;
}
break;
case CXXKeywordUNION:
if(!cxxParserParseClassStructOrUnion(CXXKeywordUNION,CXXTagKindUNION))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse class/struct/union");
return FALSE;
}
break;
case CXXKeywordPUBLIC:
case CXXKeywordPROTECTED:
case CXXKeywordPRIVATE:
// Note that the class keyword has its own handler so the only possibility here is an access specifier
if(!cxxParserParseAccessSpecifier())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse access specifier");
return FALSE;
}
break;
case CXXKeywordUSING:
if(!cxxParserParseUsingClause())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse using clause");
return FALSE;
}
cxxParserNewStatement();
break;
case CXXKeywordIF:
case CXXKeywordFOR:
case CXXKeywordWHILE:
case CXXKeywordSWITCH:
if(!cxxParserParseIfForWhileSwitch())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse if/for/while/switch");
return FALSE;
}
cxxParserNewStatement();
break;
case CXXKeywordTRY:
case CXXKeywordELSE:
case CXXKeywordDO:
// parse as normal statement/block
cxxParserNewStatement();
break;
case CXXKeywordRETURN:
if(cxxParserCurrentLanguageIsCPP())
{
// may be followed by a lambda, otherwise it's not interesting.
g_cxx.uKeywordState |= CXXParserKeywordStateSeenReturn;
cxxParserNewStatement();
} else {
// ignore
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse return");
return FALSE;
}
cxxParserNewStatement();
}
break;
case CXXKeywordCONTINUE:
case CXXKeywordBREAK:
case CXXKeywordGOTO:
// ignore
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse continue/break/goto");
return FALSE;
}
cxxParserNewStatement();
break;
case CXXKeywordTHROW:
// ignore when inside a function
if(cxxScopeGetKind() == CXXTagKindFUNCTION)
{
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse return/continue/break");
return FALSE;
}
cxxParserNewStatement();
}
break;
break;
case CXXKeywordCASE:
// ignore
if(!cxxParserParseUpToOneOf(CXXTokenTypeSemicolon | CXXTokenTypeEOF | CXXTokenTypeSingleColon))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse case keyword");
return FALSE;
}
cxxParserNewStatement();
break;
case CXXKeywordEXTERN:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenExtern;
cxxTokenChainClear(g_cxx.pTokenChain);
break;
case CXXKeywordSTATIC:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenStatic;
cxxTokenChainClear(g_cxx.pTokenChain);
break;
case CXXKeywordINLINE:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenInline;
cxxTokenChainClear(g_cxx.pTokenChain);
break;
case CXXKeywordEXPLICIT:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenExplicit;
cxxTokenChainClear(g_cxx.pTokenChain);
break;
case CXXKeywordOPERATOR:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenOperator;
break;
case CXXKeywordVIRTUAL:
g_cxx.uKeywordState |= CXXParserKeywordStateSeenVirtual;
break;
default:
if(g_cxx.uKeywordState & CXXParserKeywordStateSeenTypedef)
{
g_cxx.uKeywordState &= ~CXXParserKeywordStateSeenTypedef;
if(!cxxParserParseGenericTypedef())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse generic typedef");
return FALSE;
}
cxxParserNewStatement();
}
break;
}
}
break;
case CXXTokenTypeSemicolon:
{
if(
(g_cxx.eLanguage == g_cxx.eCLanguage) &&
cxxScopeIsGlobal() &&
(!(g_cxx.uKeywordState & CXXParserKeywordStateSeenExtern)) &&
(!(g_cxx.uKeywordState & CXXParserKeywordStateSeenTypedef))
)
{
// Special handling of K&R style function declarations.
// We might be in the following situation:
//
// type whatever fname(par1,par2) int par1; int par2; {
// ^
//
switch(cxxParserMaybeExtractKnRStyleFunctionDefinition())
{
case 1:
// got K&R style function definition, one scope was pushed.
cxxParserNewStatement();
if(!cxxParserParseBlock(TRUE))
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse nested block");
return FALSE;
}
cxxScopePop();
break;
case 0:
// something else
cxxParserAnalyzeOtherStatement();
break;
default:
CXX_DEBUG_LEAVE_TEXT("Failed to check for K&R style function definition");
return FALSE;
break;
}
} else {
// K&R style function declarations not allowed here.
cxxParserAnalyzeOtherStatement();
}
cxxParserNewStatement();
}
break;
case CXXTokenTypeSingleColon:
{
// label ?
if((g_cxx.pTokenChain->iCount == 2) && cxxTokenTypeIs(cxxTokenChainFirst(g_cxx.pTokenChain),CXXTokenTypeIdentifier))
{
CXXToken * pFirst = cxxTokenChainFirst(g_cxx.pTokenChain);
// assume it's label
tagEntryInfo * tag = cxxTagBegin(CXXTagKindLABEL,pFirst);
if(tag)
{
tag->isFileScope = TRUE;
cxxTagCommit();
}
} else {
// what is this? (default: and similar things have been handled at keyword level)
}
}
break;
case CXXTokenTypeOpeningBracket:
if(!cxxParserParseBlockHandleOpeningBracket())
{
CXX_DEBUG_LEAVE_TEXT("Failed to handle opening bracket");
return FALSE;
}
break;
case CXXTokenTypeClosingBracket:
// scope finished
CXX_DEBUG_LEAVE_TEXT("Closing bracket!");
cxxParserNewStatement();
return TRUE;
break;
case CXXTokenTypeOpeningParenthesis:
case CXXTokenTypeOpeningSquareParenthesis:
if(!cxxParserParseAndCondenseCurrentSubchain(
CXXTokenTypeOpeningBracket | CXXTokenTypeOpeningParenthesis | CXXTokenTypeOpeningSquareParenthesis,
TRUE
))
{
CXX_DEBUG_LEAVE_TEXT("Parsing the parenthesis failed");
return FALSE;
}
if(cxxTokenTypeIs(g_cxx.pToken,CXXTokenTypeEOF))
{
if(bExpectClosingBracket)
{
CXX_DEBUG_LEAVE_TEXT("Syntax error: found EOF in block but a closing bracket was expected!");
return FALSE;
}
return TRUE; // EOF
}
break;
case CXXTokenTypeIdentifier:
if(g_cxx.uKeywordState & CXXParserKeywordStateSeenTypedef)
{
g_cxx.uKeywordState &= ~CXXParserKeywordStateSeenTypedef;
if(!cxxParserParseGenericTypedef())
{
CXX_DEBUG_LEAVE_TEXT("Failed to parse generic typedef");
return FALSE;
}
cxxParserNewStatement();
}
break;
default:
// something else we didn't handle
break;
}
}
CXX_DEBUG_LEAVE_TEXT("WARNING: Not reached");
return TRUE;
}
