[cfe-commits] r98608 - in /cfe/trunk: include/clang/Basic/SourceManager.h include/clang/Lex/Preprocessor.h lib/Basic/SourceManager.cpp lib/Lex/LiteralSupport.cpp lib/Lex/PPExpressions.cpp lib/Lex/Preprocessor.cpp test/PCH/changed-files.c
Douglas Gregor
dgregor at apple.com
Mon Mar 15 22:20:39 PDT 2010
Author: dgregor
Date: Tue Mar 16 00:20:39 2010
New Revision: 98608
URL: http://llvm.org/viewvc/llvm-project?rev=98608&view=rev
Log:
Introduce optional "Invalid" parameters to routines that invoke the
SourceManager's getBuffer() and, therefore, could fail, along with
Preprocessor::getSpelling(). Use the Invalid parameters in the literal
parsers (string, floating point, integral, character) to make them
robust against errors that stem from, e.g., PCH files that are not
consistent with the underlying file system.
I still need to audit every use caller to all of these routines, to
determine which ones need specific handling of error conditions.
Added:
cfe/trunk/test/PCH/changed-files.c (with props)
Modified:
cfe/trunk/include/clang/Basic/SourceManager.h
cfe/trunk/include/clang/Lex/Preprocessor.h
cfe/trunk/lib/Basic/SourceManager.cpp
cfe/trunk/lib/Lex/LiteralSupport.cpp
cfe/trunk/lib/Lex/PPExpressions.cpp
cfe/trunk/lib/Lex/Preprocessor.cpp
Modified: cfe/trunk/include/clang/Basic/SourceManager.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/SourceManager.h?rev=98608&r1=98607&r2=98608&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/SourceManager.h (original)
+++ cfe/trunk/include/clang/Basic/SourceManager.h Tue Mar 16 00:20:39 2010
@@ -417,7 +417,11 @@
unsigned Offset = 0);
/// \brief Retrieve the memory buffer associated with the given file.
- const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File);
+ ///
+ /// \param Invalid If non-NULL, will be set \c true if an error
+ /// occurs while retrieving the memory buffer.
+ const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
+ bool *Invalid = 0);
/// \brief Override the contents of the given source file by providing an
/// already-allocated buffer.
@@ -438,8 +442,9 @@
/// getBuffer - Return the buffer for the specified FileID. If there is an
/// error opening this buffer the first time, this manufactures a temporary
/// buffer and returns a non-empty error string.
- const llvm::MemoryBuffer *getBuffer(FileID FID) const {
- return getSLocEntry(FID).getFile().getContentCache()->getBuffer(Diag);
+ const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
+ return getSLocEntry(FID).getFile().getContentCache()->getBuffer(Diag,
+ Invalid);
}
/// getFileEntryForID - Returns the FileEntry record for the provided FileID.
@@ -571,31 +576,37 @@
/// getCharacterData - Return a pointer to the start of the specified location
/// in the appropriate spelling MemoryBuffer.
- const char *getCharacterData(SourceLocation SL) const;
+ ///
+ /// \param Invalid If non-NULL, will be set \c true if an error occurs.
+ const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
/// getColumnNumber - Return the column # for the specified file position.
/// This is significantly cheaper to compute than the line number. This
/// returns zero if the column number isn't known. This may only be called on
/// a file sloc, so you must choose a spelling or instantiation location
/// before calling this method.
- unsigned getColumnNumber(FileID FID, unsigned FilePos) const;
- unsigned getSpellingColumnNumber(SourceLocation Loc) const;
- unsigned getInstantiationColumnNumber(SourceLocation Loc) const;
+ unsigned getColumnNumber(FileID FID, unsigned FilePos,
+ bool *Invalid = 0) const;
+ unsigned getSpellingColumnNumber(SourceLocation Loc,
+ bool *Invalid = 0) const;
+ unsigned getInstantiationColumnNumber(SourceLocation Loc,
+ bool *Invalid = 0) const;
/// getLineNumber - Given a SourceLocation, return the spelling line number
/// for the position indicated. This requires building and caching a table of
/// line offsets for the MemoryBuffer, so this is not cheap: use only when
/// about to emit a diagnostic.
- unsigned getLineNumber(FileID FID, unsigned FilePos) const;
+ unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
- unsigned getInstantiationLineNumber(SourceLocation Loc) const;
- unsigned getSpellingLineNumber(SourceLocation Loc) const;
+ unsigned getInstantiationLineNumber(SourceLocation Loc,
+ bool *Invalid = 0) const;
+ unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
/// Return the filename or buffer identifier of the buffer the location is in.
/// Note that this name does not respect #line directives. Use getPresumedLoc
/// for normal clients.
- const char *getBufferName(SourceLocation Loc) const;
+ const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
/// getFileCharacteristic - return the file characteristic of the specified
/// source location, indicating whether this is a normal file, a system
Modified: cfe/trunk/include/clang/Lex/Preprocessor.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Lex/Preprocessor.h?rev=98608&r1=98607&r2=98608&view=diff
==============================================================================
--- cfe/trunk/include/clang/Lex/Preprocessor.h (original)
+++ cfe/trunk/include/clang/Lex/Preprocessor.h Tue Mar 16 00:20:39 2010
@@ -547,7 +547,9 @@
/// after trigraph expansion and escaped-newline folding. In particular, this
/// wants to get the true, uncanonicalized, spelling of things like digraphs
/// UCNs, etc.
- std::string getSpelling(const Token &Tok) const;
+ ///
+ /// \param Invalid If non-NULL, will be set \c true if an error occurs.
+ std::string getSpelling(const Token &Tok, bool *Invalid = 0) const;
/// getSpelling() - Return the 'spelling' of the Tok token. The spelling of a
/// token is the characters used to represent the token in the source file
@@ -556,7 +558,8 @@
/// UCNs, etc.
static std::string getSpelling(const Token &Tok,
const SourceManager &SourceMgr,
- const LangOptions &Features);
+ const LangOptions &Features,
+ bool *Invalid = 0);
/// getSpelling - This method is used to get the spelling of a token into a
/// preallocated buffer, instead of as an std::string. The caller is required
@@ -568,17 +571,20 @@
/// to point to a constant buffer with the data already in it (avoiding a
/// copy). The caller is not allowed to modify the returned buffer pointer
/// if an internal buffer is returned.
- unsigned getSpelling(const Token &Tok, const char *&Buffer) const;
+ unsigned getSpelling(const Token &Tok, const char *&Buffer,
+ bool *Invalid = 0) const;
/// getSpelling - This method is used to get the spelling of a token into a
/// SmallVector. Note that the returned StringRef may not point to the
/// supplied buffer if a copy can be avoided.
llvm::StringRef getSpelling(const Token &Tok,
- llvm::SmallVectorImpl<char> &Buffer) const;
+ llvm::SmallVectorImpl<char> &Buffer,
+ bool *Invalid = 0) const;
/// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant
/// with length 1, return the character.
- char getSpellingOfSingleCharacterNumericConstant(const Token &Tok) const {
+ char getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
+ bool *Invalid = 0) const {
assert(Tok.is(tok::numeric_constant) &&
Tok.getLength() == 1 && "Called on unsupported token");
assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
Modified: cfe/trunk/lib/Basic/SourceManager.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Basic/SourceManager.cpp?rev=98608&r1=98607&r2=98608&view=diff
==============================================================================
--- cfe/trunk/lib/Basic/SourceManager.cpp (original)
+++ cfe/trunk/lib/Basic/SourceManager.cpp Tue Mar 16 00:20:39 2010
@@ -457,10 +457,11 @@
}
const llvm::MemoryBuffer *
-SourceManager::getMemoryBufferForFile(const FileEntry *File) {
+SourceManager::getMemoryBufferForFile(const FileEntry *File,
+ bool *Invalid) {
const SrcMgr::ContentCache *IR = getOrCreateContentCache(File);
assert(IR && "getOrCreateContentCache() cannot return NULL");
- return IR->getBuffer(Diag);
+ return IR->getBuffer(Diag, Invalid);
}
bool SourceManager::overrideFileContents(const FileEntry *SourceFile,
@@ -701,21 +702,34 @@
/// getCharacterData - Return a pointer to the start of the specified location
/// in the appropriate MemoryBuffer.
-const char *SourceManager::getCharacterData(SourceLocation SL) const {
+const char *SourceManager::getCharacterData(SourceLocation SL,
+ bool *Invalid) const {
// Note that this is a hot function in the getSpelling() path, which is
// heavily used by -E mode.
std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(SL);
// Note that calling 'getBuffer()' may lazily page in a source file.
- return getSLocEntry(LocInfo.first).getFile().getContentCache()
- ->getBuffer(Diag)->getBufferStart() + LocInfo.second;
+ bool CharDataInvalid = false;
+ const llvm::MemoryBuffer *Buffer
+ = getSLocEntry(LocInfo.first).getFile().getContentCache()->getBuffer(Diag,
+ &CharDataInvalid);
+ if (Invalid)
+ *Invalid = CharDataInvalid;
+ return Buffer->getBufferStart() + (CharDataInvalid? 0 : LocInfo.second);
}
/// getColumnNumber - Return the column # for the specified file position.
/// this is significantly cheaper to compute than the line number.
-unsigned SourceManager::getColumnNumber(FileID FID, unsigned FilePos) const {
- const char *Buf = getBuffer(FID)->getBufferStart();
+unsigned SourceManager::getColumnNumber(FileID FID, unsigned FilePos,
+ bool *Invalid) const {
+ bool MyInvalid = false;
+ const char *Buf = getBuffer(FID, &MyInvalid)->getBufferStart();
+ if (Invalid)
+ *Invalid = MyInvalid;
+
+ if (MyInvalid)
+ return 1;
unsigned LineStart = FilePos;
while (LineStart && Buf[LineStart-1] != '\n' && Buf[LineStart-1] != '\r')
@@ -723,27 +737,30 @@
return FilePos-LineStart+1;
}
-unsigned SourceManager::getSpellingColumnNumber(SourceLocation Loc) const {
+unsigned SourceManager::getSpellingColumnNumber(SourceLocation Loc,
+ bool *Invalid) const {
if (Loc.isInvalid()) return 0;
std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc);
- return getColumnNumber(LocInfo.first, LocInfo.second);
+ return getColumnNumber(LocInfo.first, LocInfo.second, Invalid);
}
-unsigned SourceManager::getInstantiationColumnNumber(SourceLocation Loc) const {
+unsigned SourceManager::getInstantiationColumnNumber(SourceLocation Loc,
+ bool *Invalid) const {
if (Loc.isInvalid()) return 0;
std::pair<FileID, unsigned> LocInfo = getDecomposedInstantiationLoc(Loc);
- return getColumnNumber(LocInfo.first, LocInfo.second);
+ return getColumnNumber(LocInfo.first, LocInfo.second, Invalid);
}
-
-
static DISABLE_INLINE void ComputeLineNumbers(Diagnostic &Diag,
ContentCache* FI,
- llvm::BumpPtrAllocator &Alloc);
+ llvm::BumpPtrAllocator &Alloc,
+ bool &Invalid);
static void ComputeLineNumbers(Diagnostic &Diag, ContentCache* FI,
- llvm::BumpPtrAllocator &Alloc){
+ llvm::BumpPtrAllocator &Alloc, bool &Invalid) {
// Note that calling 'getBuffer()' may lazily page in the file.
- const MemoryBuffer *Buffer = FI->getBuffer(Diag);
+ const MemoryBuffer *Buffer = FI->getBuffer(Diag, &Invalid);
+ if (Invalid)
+ return;
// Find the file offsets of all of the *physical* source lines. This does
// not look at trigraphs, escaped newlines, or anything else tricky.
@@ -789,7 +806,8 @@
/// for the position indicated. This requires building and caching a table of
/// line offsets for the MemoryBuffer, so this is not cheap: use only when
/// about to emit a diagnostic.
-unsigned SourceManager::getLineNumber(FileID FID, unsigned FilePos) const {
+unsigned SourceManager::getLineNumber(FileID FID, unsigned FilePos,
+ bool *Invalid) const {
ContentCache *Content;
if (LastLineNoFileIDQuery == FID)
Content = LastLineNoContentCache;
@@ -799,8 +817,15 @@
// If this is the first use of line information for this buffer, compute the
/// SourceLineCache for it on demand.
- if (Content->SourceLineCache == 0)
- ComputeLineNumbers(Diag, Content, ContentCacheAlloc);
+ if (Content->SourceLineCache == 0) {
+ bool MyInvalid = false;
+ ComputeLineNumbers(Diag, Content, ContentCacheAlloc, MyInvalid);
+ if (Invalid)
+ *Invalid = MyInvalid;
+ if (MyInvalid)
+ return 1;
+ } else if (Invalid)
+ *Invalid = false;
// Okay, we know we have a line number table. Do a binary search to find the
// line number that this character position lands on.
@@ -886,12 +911,14 @@
return LineNo;
}
-unsigned SourceManager::getInstantiationLineNumber(SourceLocation Loc) const {
+unsigned SourceManager::getInstantiationLineNumber(SourceLocation Loc,
+ bool *Invalid) const {
if (Loc.isInvalid()) return 0;
std::pair<FileID, unsigned> LocInfo = getDecomposedInstantiationLoc(Loc);
return getLineNumber(LocInfo.first, LocInfo.second);
}
-unsigned SourceManager::getSpellingLineNumber(SourceLocation Loc) const {
+unsigned SourceManager::getSpellingLineNumber(SourceLocation Loc,
+ bool *Invalid) const {
if (Loc.isInvalid()) return 0;
std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc);
return getLineNumber(LocInfo.first, LocInfo.second);
@@ -931,10 +958,11 @@
/// Return the filename or buffer identifier of the buffer the location is in.
/// Note that this name does not respect #line directives. Use getPresumedLoc
/// for normal clients.
-const char *SourceManager::getBufferName(SourceLocation Loc) const {
+const char *SourceManager::getBufferName(SourceLocation Loc,
+ bool *Invalid) const {
if (Loc.isInvalid()) return "<invalid loc>";
- return getBuffer(getFileID(Loc))->getBufferIdentifier();
+ return getBuffer(getFileID(Loc), Invalid)->getBufferIdentifier();
}
@@ -1014,8 +1042,12 @@
// If this is the first use of line information for this buffer, compute the
/// SourceLineCache for it on demand.
- if (Content->SourceLineCache == 0)
- ComputeLineNumbers(Diag, Content, ContentCacheAlloc);
+ if (Content->SourceLineCache == 0) {
+ bool MyInvalid = false;
+ ComputeLineNumbers(Diag, Content, ContentCacheAlloc, MyInvalid);
+ if (MyInvalid)
+ return SourceLocation();
+ }
// Find the first file ID that corresponds to the given file.
FileID FirstFID;
Modified: cfe/trunk/lib/Lex/LiteralSupport.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Lex/LiteralSupport.cpp?rev=98608&r1=98607&r2=98608&view=diff
==============================================================================
--- cfe/trunk/lib/Lex/LiteralSupport.cpp (original)
+++ cfe/trunk/lib/Lex/LiteralSupport.cpp Tue Mar 16 00:20:39 2010
@@ -806,7 +806,14 @@
// Get the spelling of the token, which eliminates trigraphs, etc. We know
// that ThisTokBuf points to a buffer that is big enough for the whole token
// and 'spelled' tokens can only shrink.
- unsigned ThisTokLen = PP.getSpelling(StringToks[i], ThisTokBuf);
+ bool StringInvalid = false;
+ unsigned ThisTokLen = PP.getSpelling(StringToks[i], ThisTokBuf,
+ &StringInvalid);
+ if (StringInvalid) {
+ hadError = 1;
+ continue;
+ }
+
const char *ThisTokEnd = ThisTokBuf+ThisTokLen-1; // Skip end quote.
// TODO: Input character set mapping support.
@@ -904,8 +911,12 @@
llvm::SmallString<16> SpellingBuffer;
SpellingBuffer.resize(Tok.getLength());
+ bool StringInvalid = false;
const char *SpellingPtr = &SpellingBuffer[0];
- unsigned TokLen = PP.getSpelling(Tok, SpellingPtr);
+ unsigned TokLen = PP.getSpelling(Tok, SpellingPtr, &StringInvalid);
+ if (StringInvalid) {
+ return 0;
+ }
assert(SpellingPtr[0] != 'L' && "Doesn't handle wide strings yet");
Modified: cfe/trunk/lib/Lex/PPExpressions.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Lex/PPExpressions.cpp?rev=98608&r1=98607&r2=98608&view=diff
==============================================================================
--- cfe/trunk/lib/Lex/PPExpressions.cpp (original)
+++ cfe/trunk/lib/Lex/PPExpressions.cpp Tue Mar 16 00:20:39 2010
@@ -170,7 +170,12 @@
return true;
case tok::numeric_constant: {
llvm::SmallString<64> IntegerBuffer;
- llvm::StringRef Spelling = PP.getSpelling(PeekTok, IntegerBuffer);
+ bool NumberInvalid = false;
+ llvm::StringRef Spelling = PP.getSpelling(PeekTok, IntegerBuffer,
+ &NumberInvalid);
+ if (NumberInvalid)
+ return true; // a diagnostic was already reported
+
NumericLiteralParser Literal(Spelling.begin(), Spelling.end(),
PeekTok.getLocation(), PP);
if (Literal.hadError)
@@ -216,7 +221,10 @@
}
case tok::char_constant: { // 'x'
llvm::SmallString<32> CharBuffer;
- llvm::StringRef ThisTok = PP.getSpelling(PeekTok, CharBuffer);
+ bool CharInvalid = false;
+ llvm::StringRef ThisTok = PP.getSpelling(PeekTok, CharBuffer, &CharInvalid);
+ if (CharInvalid)
+ return true;
CharLiteralParser Literal(ThisTok.begin(), ThisTok.end(),
PeekTok.getLocation(), PP);
Modified: cfe/trunk/lib/Lex/Preprocessor.cpp
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Lex/Preprocessor.cpp?rev=98608&r1=98607&r2=98608&view=diff
==============================================================================
--- cfe/trunk/lib/Lex/Preprocessor.cpp (original)
+++ cfe/trunk/lib/Lex/Preprocessor.cpp Tue Mar 16 00:20:39 2010
@@ -282,11 +282,19 @@
/// UCNs, etc.
std::string Preprocessor::getSpelling(const Token &Tok,
const SourceManager &SourceMgr,
- const LangOptions &Features) {
+ const LangOptions &Features,
+ bool *Invalid) {
assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
// If this token contains nothing interesting, return it directly.
- const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation());
+ bool CharDataInvalid = false;
+ const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
+ &CharDataInvalid);
+ if (Invalid)
+ *Invalid = CharDataInvalid;
+ if (CharDataInvalid)
+ return std::string();
+
if (!Tok.needsCleaning())
return std::string(TokStart, TokStart+Tok.getLength());
@@ -310,8 +318,8 @@
/// after trigraph expansion and escaped-newline folding. In particular, this
/// wants to get the true, uncanonicalized, spelling of things like digraphs
/// UCNs, etc.
-std::string Preprocessor::getSpelling(const Token &Tok) const {
- return getSpelling(Tok, SourceMgr, Features);
+std::string Preprocessor::getSpelling(const Token &Tok, bool *Invalid) const {
+ return getSpelling(Tok, SourceMgr, Features, Invalid);
}
/// getSpelling - This method is used to get the spelling of a token into a
@@ -325,7 +333,7 @@
/// copy). The caller is not allowed to modify the returned buffer pointer
/// if an internal buffer is returned.
unsigned Preprocessor::getSpelling(const Token &Tok,
- const char *&Buffer) const {
+ const char *&Buffer, bool *Invalid) const {
assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
// If this token is an identifier, just return the string from the identifier
@@ -341,8 +349,16 @@
if (Tok.isLiteral())
TokStart = Tok.getLiteralData();
- if (TokStart == 0)
- TokStart = SourceMgr.getCharacterData(Tok.getLocation());
+ if (TokStart == 0) {
+ bool CharDataInvalid = false;
+ TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
+ if (Invalid)
+ *Invalid = CharDataInvalid;
+ if (CharDataInvalid) {
+ Buffer = "";
+ return 0;
+ }
+ }
// If this token contains nothing interesting, return it directly.
if (!Tok.needsCleaning()) {
@@ -368,7 +384,8 @@
/// SmallVector. Note that the returned StringRef may not point to the
/// supplied buffer if a copy can be avoided.
llvm::StringRef Preprocessor::getSpelling(const Token &Tok,
- llvm::SmallVectorImpl<char> &Buffer) const {
+ llvm::SmallVectorImpl<char> &Buffer,
+ bool *Invalid) const {
// Try the fast path.
if (const IdentifierInfo *II = Tok.getIdentifierInfo())
return II->getName();
@@ -378,7 +395,7 @@
Buffer.resize(Tok.getLength());
const char *Ptr = Buffer.data();
- unsigned Len = getSpelling(Tok, Ptr);
+ unsigned Len = getSpelling(Tok, Ptr, Invalid);
return llvm::StringRef(Ptr, Len);
}
Added: cfe/trunk/test/PCH/changed-files.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/PCH/changed-files.c?rev=98608&view=auto
==============================================================================
--- cfe/trunk/test/PCH/changed-files.c (added)
+++ cfe/trunk/test/PCH/changed-files.c Tue Mar 16 00:20:39 2010
@@ -0,0 +1,11 @@
+const char *s0 = m0;
+
+// RUN: echo '#define m0 ""' > %t.h
+// RUN: %clang_cc1 -emit-pch -o %t.h.pch %t.h
+// RUN: echo '' > %t.h
+// RUN: not %clang_cc1 -include-pch %t.h.pch %s 2>&1 | grep "size of file"
+
+// RUN: echo '#define m0 000' > %t.h
+// RUN: %clang_cc1 -emit-pch -o %t.h.pch %t.h
+// RUN: echo '' > %t.h
+// RUN: not %clang_cc1 -include-pch %t.h.pch %s 2>&1 | grep "size of file"
Propchange: cfe/trunk/test/PCH/changed-files.c
------------------------------------------------------------------------------
svn:eol-style = native
Propchange: cfe/trunk/test/PCH/changed-files.c
------------------------------------------------------------------------------
svn:keywords = Id
Propchange: cfe/trunk/test/PCH/changed-files.c
------------------------------------------------------------------------------
svn:mime-type = text/plain
More information about the cfe-commits
mailing list