Stage two of getting CFE top correct.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@39734 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/Basic/Diagnostic.cpp b/Basic/Diagnostic.cpp
new file mode 100644
index 0000000..155b6fc
--- /dev/null
+++ b/Basic/Diagnostic.cpp
@@ -0,0 +1,147 @@
+//===--- Diagnostic.cpp - C Language Family Diagnostic Handling -----------===//
+//
+// The LLVM Compiler Infrastructure
+//
+// This file was developed by Chris Lattner and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the Diagnostic-related interfaces.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Basic/Diagnostic.h"
+#include "clang/Basic/SourceLocation.h"
+#include <cassert>
+using namespace clang;
+
+/// Flag values for diagnostics.
+enum {
+ // Diagnostic classes.
+ NOTE = 0x01,
+ WARNING = 0x02,
+ EXTENSION = 0x03,
+ ERROR = 0x04,
+ FATAL = 0x05,
+ class_mask = 0x07
+};
+
+/// DiagnosticFlags - A set of flags, or'd together, that describe the
+/// diagnostic.
+static unsigned char DiagnosticFlags[] = {
+#define DIAG(ENUM,FLAGS,DESC) FLAGS,
+#include "clang/Basic/DiagnosticKinds.def"
+ 0
+};
+
+/// getDiagClass - Return the class field of the diagnostic.
+///
+static unsigned getDiagClass(unsigned DiagID) {
+ assert(DiagID < diag::NUM_DIAGNOSTICS && "Diagnostic ID out of range!");
+ return DiagnosticFlags[DiagID] & class_mask;
+}
+
+/// DiagnosticText - An english message to print for the diagnostic. These
+/// should be localized.
+static const char * const DiagnosticText[] = {
+#define DIAG(ENUM,FLAGS,DESC) DESC,
+#include "clang/Basic/DiagnosticKinds.def"
+ 0
+};
+
+Diagnostic::Diagnostic(DiagnosticClient &client) : Client(client) {
+ WarningsAsErrors = false;
+ WarnOnExtensions = false;
+ ErrorOnExtensions = false;
+ // Clear all mappings, setting them to MAP_DEFAULT.
+ memset(DiagMappings, 0, sizeof(DiagMappings));
+
+ ErrorOccurred = false;
+ NumDiagnostics = 0;
+ NumErrors = 0;
+}
+
+/// isNoteWarningOrExtension - Return true if the unmapped diagnostic level of
+/// the specified diagnostic ID is a Note, Warning, or Extension.
+bool Diagnostic::isNoteWarningOrExtension(unsigned DiagID) {
+ return getDiagClass(DiagID) < ERROR;
+}
+
+
+/// getDescription - Given a diagnostic ID, return a description of the
+/// issue.
+const char *Diagnostic::getDescription(unsigned DiagID) {
+ assert(DiagID < diag::NUM_DIAGNOSTICS && "Diagnostic ID out of range!");
+ return DiagnosticText[DiagID];
+}
+
+/// getDiagnosticLevel - Based on the way the client configured the Diagnostic
+/// object, classify the specified diagnostic ID into a Level, consumable by
+/// the DiagnosticClient.
+Diagnostic::Level Diagnostic::getDiagnosticLevel(unsigned DiagID) const {
+ unsigned DiagClass = getDiagClass(DiagID);
+
+ // Specific non-error diagnostics may be mapped to various levels from ignored
+ // to error.
+ if (DiagClass < ERROR) {
+ switch (getDiagnosticMapping((diag::kind)DiagID)) {
+ case diag::MAP_DEFAULT: break;
+ case diag::MAP_IGNORE: return Ignored;
+ case diag::MAP_WARNING: DiagClass = WARNING; break;
+ case diag::MAP_ERROR: DiagClass = ERROR; break;
+ }
+ }
+
+ // Map diagnostic classes based on command line argument settings.
+ if (DiagClass == EXTENSION) {
+ if (ErrorOnExtensions)
+ DiagClass = ERROR;
+ else if (WarnOnExtensions)
+ DiagClass = WARNING;
+ else
+ return Ignored;
+ }
+
+ // If warnings are to be treated as errors, indicate this as such.
+ if (DiagClass == WARNING && WarningsAsErrors)
+ DiagClass = ERROR;
+
+ switch (DiagClass) {
+ default: assert(0 && "Unknown diagnostic class!");
+ case NOTE: return Diagnostic::Note;
+ case WARNING: return Diagnostic::Warning;
+ case ERROR: return Diagnostic::Error;
+ case FATAL: return Diagnostic::Fatal;
+ }
+}
+
+/// Report - Issue the message to the client. If the client wants us to stop
+/// compilation, return true, otherwise return false. DiagID is a member of
+/// the diag::kind enum.
+void Diagnostic::Report(SourceLocation Pos, unsigned DiagID,
+ const std::string *Strs, unsigned NumStrs,
+ const SourceRange *Ranges, unsigned NumRanges) {
+ // Figure out the diagnostic level of this message.
+ Diagnostic::Level DiagLevel = getDiagnosticLevel(DiagID);
+
+ // If the client doesn't care about this message, don't issue it.
+ if (DiagLevel == Diagnostic::Ignored)
+ return;
+
+ if (DiagLevel >= Diagnostic::Error) {
+ ErrorOccurred = true;
+ ++NumErrors;
+ }
+
+ // Are we going to ignore this diagnosic?
+ if (Client.IgnoreDiagnostic(DiagLevel, Pos))
+ return;
+
+ // Finally, report it.
+ Client.HandleDiagnostic(DiagLevel, Pos, (diag::kind)DiagID, Strs, NumStrs,
+ Ranges, NumRanges);
+ ++NumDiagnostics;
+}
+
+DiagnosticClient::~DiagnosticClient() {}
diff --git a/Basic/FileManager.cpp b/Basic/FileManager.cpp
new file mode 100644
index 0000000..9886e03
--- /dev/null
+++ b/Basic/FileManager.cpp
@@ -0,0 +1,169 @@
+//===--- FileManager.cpp - File System Probing and Caching ----------------===//
+//
+// The LLVM Compiler Infrastructure
+//
+// This file was developed by Chris Lattner and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the FileManager interface.
+//
+//===----------------------------------------------------------------------===//
+//
+// TODO: This should index all interesting directories with dirent calls.
+// getdirentries ?
+// opendir/readdir_r/closedir ?
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Basic/FileManager.h"
+#include "llvm/ADT/SmallString.h"
+#include <iostream>
+using namespace clang;
+
+// FIXME: Enhance libsystem to support inode and other fields.
+#include <sys/stat.h>
+
+
+/// NON_EXISTANT_DIR - A special value distinct from null that is used to
+/// represent a dir name that doesn't exist on the disk.
+#define NON_EXISTANT_DIR reinterpret_cast<DirectoryEntry*>((intptr_t)-1)
+
+/// getDirectory - Lookup, cache, and verify the specified directory. This
+/// returns null if the directory doesn't exist.
+///
+const DirectoryEntry *FileManager::getDirectory(const char *NameStart,
+ const char *NameEnd) {
+ ++NumDirLookups;
+ llvm::StringMapEntry<DirectoryEntry *> &NamedDirEnt =
+ DirEntries.GetOrCreateValue(NameStart, NameEnd);
+
+ // See if there is already an entry in the map.
+ if (NamedDirEnt.getValue())
+ return NamedDirEnt.getValue() == NON_EXISTANT_DIR
+ ? 0 : NamedDirEnt.getValue();
+
+ ++NumDirCacheMisses;
+
+ // By default, initialize it to invalid.
+ NamedDirEnt.setValue(NON_EXISTANT_DIR);
+
+ // Get the null-terminated directory name as stored as the key of the
+ // DirEntries map.
+ const char *InterndDirName = NamedDirEnt.getKeyData();
+
+ // Check to see if the directory exists.
+ struct stat StatBuf;
+ if (stat(InterndDirName, &StatBuf) || // Error stat'ing.
+ !S_ISDIR(StatBuf.st_mode)) // Not a directory?
+ return 0;
+
+ // It exists. See if we have already opened a directory with the same inode.
+ // This occurs when one dir is symlinked to another, for example.
+ DirectoryEntry &UDE =
+ UniqueDirs[std::make_pair(StatBuf.st_dev, StatBuf.st_ino)];
+
+ NamedDirEnt.setValue(&UDE);
+ if (UDE.getName()) // Already have an entry with this inode, return it.
+ return &UDE;
+
+ // Otherwise, we don't have this directory yet, add it. We use the string
+ // key from the DirEntries map as the string.
+ UDE.Name = InterndDirName;
+ return &UDE;
+}
+
+/// NON_EXISTANT_FILE - A special value distinct from null that is used to
+/// represent a filename that doesn't exist on the disk.
+#define NON_EXISTANT_FILE reinterpret_cast<FileEntry*>((intptr_t)-1)
+
+/// getFile - Lookup, cache, and verify the specified file. This returns null
+/// if the file doesn't exist.
+///
+const FileEntry *FileManager::getFile(const char *NameStart,
+ const char *NameEnd) {
+ ++NumFileLookups;
+
+ // See if there is already an entry in the map.
+ llvm::StringMapEntry<FileEntry *> &NamedFileEnt =
+ FileEntries.GetOrCreateValue(NameStart, NameEnd);
+
+ // See if there is already an entry in the map.
+ if (NamedFileEnt.getValue())
+ return NamedFileEnt.getValue() == NON_EXISTANT_FILE
+ ? 0 : NamedFileEnt.getValue();
+
+ ++NumFileCacheMisses;
+
+ // By default, initialize it to invalid.
+ NamedFileEnt.setValue(NON_EXISTANT_FILE);
+
+ // Figure out what directory it is in. If the string contains a / in it,
+ // strip off everything after it.
+ // FIXME: this logic should be in sys::Path.
+ const char *SlashPos = NameEnd-1;
+ while (SlashPos >= NameStart && SlashPos[0] != '/')
+ --SlashPos;
+
+ const DirectoryEntry *DirInfo;
+ if (SlashPos < NameStart) {
+ // Use the current directory if file has no path component.
+ const char *Name = ".";
+ DirInfo = getDirectory(Name, Name+1);
+ } else if (SlashPos == NameEnd-1)
+ return 0; // If filename ends with a /, it's a directory.
+ else
+ DirInfo = getDirectory(NameStart, SlashPos);
+
+ if (DirInfo == 0) // Directory doesn't exist, file can't exist.
+ return 0;
+
+ // Get the null-terminated file name as stored as the key of the
+ // FileEntries map.
+ const char *InterndFileName = NamedFileEnt.getKeyData();
+
+ // FIXME: Use the directory info to prune this, before doing the stat syscall.
+ // FIXME: This will reduce the # syscalls.
+
+ // Nope, there isn't. Check to see if the file exists.
+ struct stat StatBuf;
+ //std::cerr << "STATING: " << Filename;
+ if (stat(InterndFileName, &StatBuf) || // Error stat'ing.
+ S_ISDIR(StatBuf.st_mode)) { // A directory?
+ // If this file doesn't exist, we leave a null in FileEntries for this path.
+ //std::cerr << ": Not existing\n";
+ return 0;
+ }
+ //std::cerr << ": exists\n";
+
+ // It exists. See if we have already opened a directory with the same inode.
+ // This occurs when one dir is symlinked to another, for example.
+ FileEntry &UFE = UniqueFiles[std::make_pair(StatBuf.st_dev, StatBuf.st_ino)];
+
+ NamedFileEnt.setValue(&UFE);
+ if (UFE.getName()) // Already have an entry with this inode, return it.
+ return &UFE;
+
+ // Otherwise, we don't have this directory yet, add it.
+ // FIXME: Change the name to be a char* that points back to the 'FileEntries'
+ // key.
+ UFE.Name = InterndFileName;
+ UFE.Size = StatBuf.st_size;
+ UFE.ModTime = StatBuf.st_mtime;
+ UFE.Dir = DirInfo;
+ UFE.UID = NextFileUID++;
+ return &UFE;
+}
+
+void FileManager::PrintStats() const {
+ std::cerr << "\n*** File Manager Stats:\n";
+ std::cerr << UniqueFiles.size() << " files found, "
+ << UniqueDirs.size() << " dirs found.\n";
+ std::cerr << NumDirLookups << " dir lookups, "
+ << NumDirCacheMisses << " dir cache misses.\n";
+ std::cerr << NumFileLookups << " file lookups, "
+ << NumFileCacheMisses << " file cache misses.\n";
+
+ //std::cerr << PagesMapped << BytesOfPagesMapped << FSLookups;
+}
diff --git a/Basic/Makefile b/Basic/Makefile
new file mode 100644
index 0000000..1db0a7f
--- /dev/null
+++ b/Basic/Makefile
@@ -0,0 +1,22 @@
+##===- clang/Basic/Makefile --------------------------------*- Makefile -*-===##
+#
+# The LLVM Compiler Infrastructure
+#
+# This file was developed by Chris Lattner and is distributed under
+# the University of Illinois Open Source License. See LICENSE.TXT for details.
+#
+##===----------------------------------------------------------------------===##
+#
+# This implements the Basic library for the C-Language front-end.
+#
+##===----------------------------------------------------------------------===##
+
+LEVEL = ../../..
+LIBRARYNAME := clangBasic
+BUILD_ARCHIVE = 1
+CXXFLAGS = -fno-rtti
+
+CPPFLAGS += -I$(PROJ_SRC_DIR)/../include
+
+include $(LEVEL)/Makefile.common
+
diff --git a/Basic/SourceManager.cpp b/Basic/SourceManager.cpp
new file mode 100644
index 0000000..f6148c16
--- /dev/null
+++ b/Basic/SourceManager.cpp
@@ -0,0 +1,370 @@
+//===--- SourceManager.cpp - Track and cache source files -----------------===//
+//
+// The LLVM Compiler Infrastructure
+//
+// This file was developed by Chris Lattner and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the SourceManager interface.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Basic/SourceManager.h"
+#include "clang/Basic/FileManager.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/System/Path.h"
+#include <algorithm>
+#include <iostream>
+using namespace clang;
+using namespace SrcMgr;
+using llvm::MemoryBuffer;
+
+SourceManager::~SourceManager() {
+ for (std::map<const FileEntry *, FileInfo>::iterator I = FileInfos.begin(),
+ E = FileInfos.end(); I != E; ++I) {
+ delete I->second.Buffer;
+ delete[] I->second.SourceLineCache;
+ }
+
+ for (std::list<InfoRec>::iterator I = MemBufferInfos.begin(),
+ E = MemBufferInfos.end(); I != E; ++I) {
+ delete I->second.Buffer;
+ delete[] I->second.SourceLineCache;
+ }
+}
+
+
+// FIXME: REMOVE THESE
+#include <unistd.h>
+#include <sys/types.h>
+#include <sys/uio.h>
+#include <sys/fcntl.h>
+#include <cerrno>
+
+static const MemoryBuffer *ReadFileFast(const FileEntry *FileEnt) {
+#if 0
+ // FIXME: Reintroduce this and zap this function once the common llvm stuff
+ // is fast for the small case.
+ return MemoryBuffer::getFile(FileEnt->getName(), strlen(FileEnt->getName()),
+ FileEnt->getSize());
+#endif
+
+ // If the file is larger than some threshold, use 'read', otherwise use mmap.
+ if (FileEnt->getSize() >= 4096*4)
+ return MemoryBuffer::getFile(FileEnt->getName(), strlen(FileEnt->getName()),
+ 0, FileEnt->getSize());
+
+ MemoryBuffer *SB = MemoryBuffer::getNewUninitMemBuffer(FileEnt->getSize(),
+ FileEnt->getName());
+ char *BufPtr = const_cast<char*>(SB->getBufferStart());
+
+ int FD = ::open(FileEnt->getName(), O_RDONLY);
+ if (FD == -1) {
+ delete SB;
+ return 0;
+ }
+
+ unsigned BytesLeft = FileEnt->getSize();
+ while (BytesLeft) {
+ ssize_t NumRead = ::read(FD, BufPtr, BytesLeft);
+ if (NumRead != -1) {
+ BytesLeft -= NumRead;
+ BufPtr += NumRead;
+ } else if (errno == EINTR) {
+ // try again
+ } else {
+ // error reading.
+ close(FD);
+ delete SB;
+ return 0;
+ }
+ }
+ close(FD);
+
+ return SB;
+}
+
+
+/// getFileInfo - Create or return a cached FileInfo for the specified file.
+///
+const InfoRec *
+SourceManager::getInfoRec(const FileEntry *FileEnt) {
+ assert(FileEnt && "Didn't specify a file entry to use?");
+ // Do we already have information about this file?
+ std::map<const FileEntry *, FileInfo>::iterator I =
+ FileInfos.lower_bound(FileEnt);
+ if (I != FileInfos.end() && I->first == FileEnt)
+ return &*I;
+
+ // Nope, get information.
+ const MemoryBuffer *File = ReadFileFast(FileEnt);
+ if (File == 0)
+ return 0;
+
+ const InfoRec &Entry =
+ *FileInfos.insert(I, std::make_pair(FileEnt, FileInfo()));
+ FileInfo &Info = const_cast<FileInfo &>(Entry.second);
+
+ Info.Buffer = File;
+ Info.SourceLineCache = 0;
+ Info.NumLines = 0;
+ return &Entry;
+}
+
+
+/// createMemBufferInfoRec - Create a new info record for the specified memory
+/// buffer. This does no caching.
+const InfoRec *
+SourceManager::createMemBufferInfoRec(const MemoryBuffer *Buffer) {
+ // Add a new info record to the MemBufferInfos list and return it.
+ FileInfo FI;
+ FI.Buffer = Buffer;
+ FI.SourceLineCache = 0;
+ FI.NumLines = 0;
+ MemBufferInfos.push_back(InfoRec(0, FI));
+ return &MemBufferInfos.back();
+}
+
+
+/// createFileID - Create a new fileID for the specified InfoRec and include
+/// position. This works regardless of whether the InfoRec corresponds to a
+/// file or some other input source.
+unsigned SourceManager::createFileID(const InfoRec *File,
+ SourceLocation IncludePos) {
+ // If FileEnt is really large (e.g. it's a large .i file), we may not be able
+ // to fit an arbitrary position in the file in the FilePos field. To handle
+ // this, we create one FileID for each chunk of the file that fits in a
+ // FilePos field.
+ unsigned FileSize = File->second.Buffer->getBufferSize();
+ if (FileSize+1 < (1 << SourceLocation::FilePosBits)) {
+ FileIDs.push_back(FileIDInfo::getNormalBuffer(IncludePos, 0, File));
+ assert(FileIDs.size() < (1 << SourceLocation::FileIDBits) &&
+ "Ran out of file ID's!");
+ return FileIDs.size();
+ }
+
+ // Create one FileID for each chunk of the file.
+ unsigned Result = FileIDs.size()+1;
+
+ unsigned ChunkNo = 0;
+ while (1) {
+ FileIDs.push_back(FileIDInfo::getNormalBuffer(IncludePos, ChunkNo++, File));
+
+ if (FileSize+1 < (1 << SourceLocation::FilePosBits)) break;
+ FileSize -= (1 << SourceLocation::FilePosBits);
+ }
+
+ assert(FileIDs.size() < (1 << SourceLocation::FileIDBits) &&
+ "Ran out of file ID's!");
+ return Result;
+}
+
+/// getInstantiationLoc - Return a new SourceLocation that encodes the fact
+/// that a token from physloc PhysLoc should actually be referenced from
+/// InstantiationLoc.
+SourceLocation SourceManager::getInstantiationLoc(SourceLocation PhysLoc,
+ SourceLocation InstantLoc) {
+ assert(getFIDInfo(PhysLoc.getFileID())->IDType !=
+ SrcMgr::FileIDInfo::MacroExpansion &&
+ "Location instantiated in a macro?");
+
+ // Resolve InstantLoc down to a real logical location.
+ InstantLoc = getLogicalLoc(InstantLoc);
+
+ unsigned InstantiationFileID;
+ // If this is the same instantiation as was requested last time, return this
+ // immediately.
+ if (PhysLoc.getFileID() == LastInstantiationLoc_MacroFID &&
+ InstantLoc == LastInstantiationLoc_InstantLoc) {
+ InstantiationFileID = LastInstantiationLoc_Result;
+ } else {
+ // Add a FileID for this. FIXME: should cache these!
+ FileIDs.push_back(FileIDInfo::getMacroExpansion(InstantLoc,
+ PhysLoc.getFileID()));
+ InstantiationFileID = FileIDs.size();
+
+ // Remember this in the single-entry cache for next time.
+ LastInstantiationLoc_MacroFID = PhysLoc.getFileID();
+ LastInstantiationLoc_InstantLoc = InstantLoc;
+ LastInstantiationLoc_Result = InstantiationFileID;
+ }
+ return SourceLocation(InstantiationFileID, PhysLoc.getRawFilePos());
+}
+
+
+
+/// getCharacterData - Return a pointer to the start of the specified location
+/// in the appropriate MemoryBuffer.
+const char *SourceManager::getCharacterData(SourceLocation SL) const {
+ // Note that this is a hot function in the getSpelling() path, which is
+ // heavily used by -E mode.
+ unsigned FileID = SL.getFileID();
+ assert(FileID && "Invalid source location!");
+
+ return getFileInfo(FileID)->Buffer->getBufferStart() + getFilePos(SL);
+}
+
+/// getIncludeLoc - Return the location of the #include for the specified
+/// FileID.
+SourceLocation SourceManager::getIncludeLoc(unsigned FileID) const {
+ const SrcMgr::FileIDInfo *FIDInfo = getFIDInfo(FileID);
+
+ // For Macros, the physical loc is specified by the MacroTokenFileID.
+ if (FIDInfo->IDType == SrcMgr::FileIDInfo::MacroExpansion)
+ FIDInfo = &FileIDs[FIDInfo->u.MacroTokenFileID-1];
+
+ return FIDInfo->IncludeLoc;
+}
+
+
+/// getColumnNumber - Return the column # for the specified include position.
+/// this is significantly cheaper to compute than the line number. This returns
+/// zero if the column number isn't known.
+unsigned SourceManager::getColumnNumber(SourceLocation Loc) const {
+ Loc = getLogicalLoc(Loc);
+ unsigned FileID = Loc.getFileID();
+ if (FileID == 0) return 0;
+
+ unsigned FilePos = getFilePos(Loc);
+ const MemoryBuffer *Buffer = getBuffer(FileID);
+ const char *Buf = Buffer->getBufferStart();
+
+ unsigned LineStart = FilePos;
+ while (LineStart && Buf[LineStart-1] != '\n' && Buf[LineStart-1] != '\r')
+ --LineStart;
+ return FilePos-LineStart+1;
+}
+
+/// getSourceName - This method returns the name of the file or buffer that
+/// the SourceLocation specifies. This can be modified with #line directives,
+/// etc.
+std::string SourceManager::getSourceName(SourceLocation Loc) {
+ Loc = getLogicalLoc(Loc);
+ unsigned FileID = Loc.getFileID();
+ if (FileID == 0) return "";
+ return getFileInfo(FileID)->Buffer->getBufferIdentifier();
+}
+
+
+/// getLineNumber - Given a SourceLocation, return the physical 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 SourceManager::getLineNumber(SourceLocation Loc) {
+ Loc = getLogicalLoc(Loc);
+ unsigned FileID = Loc.getFileID();
+ if (FileID == 0) return 0;
+ FileInfo *FileInfo = getFileInfo(FileID);
+
+ // If this is the first use of line information for this buffer, compute the
+ /// SourceLineCache for it on demand.
+ if (FileInfo->SourceLineCache == 0) {
+ const MemoryBuffer *Buffer = FileInfo->Buffer;
+
+ // Find the file offsets of all of the *physical* source lines. This does
+ // not look at trigraphs, escaped newlines, or anything else tricky.
+ std::vector<unsigned> LineOffsets;
+
+ // Line #1 starts at char 0.
+ LineOffsets.push_back(0);
+
+ const unsigned char *Buf = (const unsigned char *)Buffer->getBufferStart();
+ const unsigned char *End = (const unsigned char *)Buffer->getBufferEnd();
+ unsigned Offs = 0;
+ while (1) {
+ // Skip over the contents of the line.
+ // TODO: Vectorize this? This is very performance sensitive for programs
+ // with lots of diagnostics and in -E mode.
+ const unsigned char *NextBuf = (const unsigned char *)Buf;
+ while (*NextBuf != '\n' && *NextBuf != '\r' && *NextBuf != '\0')
+ ++NextBuf;
+ Offs += NextBuf-Buf;
+ Buf = NextBuf;
+
+ if (Buf[0] == '\n' || Buf[0] == '\r') {
+ // If this is \n\r or \r\n, skip both characters.
+ if ((Buf[1] == '\n' || Buf[1] == '\r') && Buf[0] != Buf[1])
+ ++Offs, ++Buf;
+ ++Offs, ++Buf;
+ LineOffsets.push_back(Offs);
+ } else {
+ // Otherwise, this is a null. If end of file, exit.
+ if (Buf == End) break;
+ // Otherwise, skip the null.
+ ++Offs, ++Buf;
+ }
+ }
+ LineOffsets.push_back(Offs);
+
+ // Copy the offsets into the FileInfo structure.
+ FileInfo->NumLines = LineOffsets.size();
+ FileInfo->SourceLineCache = new unsigned[LineOffsets.size()];
+ std::copy(LineOffsets.begin(), LineOffsets.end(),
+ FileInfo->SourceLineCache);
+ }
+
+ // Okay, we know we have a line number table. Do a binary search to find the
+ // line number that this character position lands on.
+ unsigned NumLines = FileInfo->NumLines;
+ unsigned *SourceLineCache = FileInfo->SourceLineCache;
+
+ // TODO: If this is performance sensitive, we could try doing simple radix
+ // type approaches to make good (tight?) initial guesses based on the
+ // assumption that all lines are the same average size.
+ unsigned *Pos = std::lower_bound(SourceLineCache, SourceLineCache+NumLines,
+ getFilePos(Loc)+1);
+ return Pos-SourceLineCache;
+}
+
+/// getSourceFilePos - This method returns the *logical* offset from the start
+/// of the file that the specified SourceLocation represents. This returns
+/// the location of the *logical* character data, not the physical file
+/// position. In the case of macros, for example, this returns where the
+/// macro was instantiated, not where the characters for the macro can be
+/// found.
+unsigned SourceManager::getSourceFilePos(SourceLocation Loc) const {
+
+ // If this is a macro, we need to get the instantiation location.
+ const SrcMgr::FileIDInfo *FIDInfo = getFIDInfo(Loc.getFileID());
+ while (FIDInfo->IDType == SrcMgr::FileIDInfo::MacroExpansion) {
+ Loc = FIDInfo->IncludeLoc;
+ FIDInfo = getFIDInfo(Loc.getFileID());
+ }
+
+ return getFilePos(Loc);
+}
+
+
+/// PrintStats - Print statistics to stderr.
+///
+void SourceManager::PrintStats() const {
+ std::cerr << "\n*** Source Manager Stats:\n";
+ std::cerr << FileInfos.size() << " files mapped, " << MemBufferInfos.size()
+ << " mem buffers mapped, " << FileIDs.size()
+ << " file ID's allocated.\n";
+ unsigned NumBuffers = 0, NumMacros = 0;
+ for (unsigned i = 0, e = FileIDs.size(); i != e; ++i) {
+ if (FileIDs[i].IDType == FileIDInfo::NormalBuffer)
+ ++NumBuffers;
+ else if (FileIDs[i].IDType == FileIDInfo::MacroExpansion)
+ ++NumMacros;
+ else
+ assert(0 && "Unknown FileID!");
+ }
+ std::cerr << " " << NumBuffers << " normal buffer FileID's, "
+ << NumMacros << " macro expansion FileID's.\n";
+
+
+
+ unsigned NumLineNumsComputed = 0;
+ unsigned NumFileBytesMapped = 0;
+ for (std::map<const FileEntry *, FileInfo>::const_iterator I =
+ FileInfos.begin(), E = FileInfos.end(); I != E; ++I) {
+ NumLineNumsComputed += I->second.SourceLineCache != 0;
+ NumFileBytesMapped += I->second.Buffer->getBufferSize();
+ }
+ std::cerr << NumFileBytesMapped << " bytes of files mapped, "
+ << NumLineNumsComputed << " files with line #'s computed.\n";
+}
diff --git a/Basic/TargetInfo.cpp b/Basic/TargetInfo.cpp
new file mode 100644
index 0000000..008e99b
--- /dev/null
+++ b/Basic/TargetInfo.cpp
@@ -0,0 +1,223 @@
+//===--- TargetInfo.cpp - Information about Target machine ----------------===//
+//
+// The LLVM Compiler Infrastructure
+//
+// This file was developed by Chris Lattner and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the TargetInfo and TargetInfoImpl interfaces.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Basic/TargetInfo.h"
+#include "clang/Basic/Diagnostic.h"
+#include "clang/AST/Builtins.h"
+#include <map>
+#include <set>
+using namespace clang;
+
+void TargetInfoImpl::ANCHOR() {} // out-of-line virtual method for class.
+
+
+/// DiagnoseNonPortability - When a use of a non-portable target feature is
+/// used, this method emits the diagnostic and marks the translation unit as
+/// non-portable.
+void TargetInfo::DiagnoseNonPortability(SourceLocation Loc, unsigned DiagKind) {
+ NonPortable = true;
+ if (Diag && Loc.isValid()) Diag->Report(Loc, DiagKind);
+}
+
+/// GetTargetDefineMap - Get the set of target #defines in an associative
+/// collection for easy lookup.
+static void GetTargetDefineMap(const TargetInfoImpl *Target,
+ std::map<std::string, std::string> &Map) {
+ std::vector<std::string> PrimaryDefines;
+ Target->getTargetDefines(PrimaryDefines);
+
+ while (!PrimaryDefines.empty()) {
+ const char *Str = PrimaryDefines.back().c_str();
+ if (const char *Equal = strchr(Str, '=')) {
+ // Split at the '='.
+ Map.insert(std::make_pair(std::string(Str, Equal),
+ std::string(Equal+1,
+ Str+PrimaryDefines.back().size())));
+ } else {
+ // Remember "macroname=1".
+ Map.insert(std::make_pair(PrimaryDefines.back(), std::string("1")));
+ }
+ PrimaryDefines.pop_back();
+ }
+}
+
+/// getTargetDefines - Appends the target-specific #define values for this
+/// target set to the specified buffer.
+void TargetInfo::getTargetDefines(std::vector<char> &Buffer) {
+ // This is tricky in the face of secondary targets. Specifically,
+ // target-specific #defines that are present and identical across all
+ // secondary targets are turned into #defines, #defines that are present in
+ // the primary target but are missing or different in the secondary targets
+ // are turned into #define_target, and #defines that are not defined in the
+ // primary, but are defined in a secondary are turned into
+ // #define_other_target. This allows the preprocessor to correctly track uses
+ // of target-specific macros.
+
+ // Get the set of primary #defines.
+ std::map<std::string, std::string> PrimaryDefines;
+ GetTargetDefineMap(PrimaryTarget, PrimaryDefines);
+
+ // If we have no secondary targets, be a bit more efficient.
+ if (SecondaryTargets.empty()) {
+ for (std::map<std::string, std::string>::iterator I =
+ PrimaryDefines.begin(), E = PrimaryDefines.end(); I != E; ++I) {
+ // If this define is non-portable, turn it into #define_target, otherwise
+ // just use #define.
+ const char *Command = "#define ";
+ Buffer.insert(Buffer.end(), Command, Command+strlen(Command));
+
+ // Insert "defname defvalue\n".
+ Buffer.insert(Buffer.end(), I->first.begin(), I->first.end());
+ Buffer.push_back(' ');
+ Buffer.insert(Buffer.end(), I->second.begin(), I->second.end());
+ Buffer.push_back('\n');
+ }
+ return;
+ }
+
+ // Get the sets of secondary #defines.
+ std::vector<std::map<std::string, std::string> > SecondaryDefines;
+ SecondaryDefines.resize(SecondaryTargets.size());
+ for (unsigned i = 0, e = SecondaryTargets.size(); i != e; ++i)
+ GetTargetDefineMap(SecondaryTargets[i], SecondaryDefines[i]);
+
+ // Loop over all defines in the primary target, processing them until we run
+ // out.
+ while (!PrimaryDefines.empty()) {
+ std::string DefineName = PrimaryDefines.begin()->first;
+ std::string DefineValue = PrimaryDefines.begin()->second;
+ PrimaryDefines.erase(PrimaryDefines.begin());
+
+ // Check to see whether all secondary targets have this #define and whether
+ // it is to the same value. Remember if not, but remove the #define from
+ // their collection in any case if they have it.
+ bool isPortable = true;
+
+ for (unsigned i = 0, e = SecondaryDefines.size(); i != e; ++i) {
+ std::map<std::string, std::string>::iterator I =
+ SecondaryDefines[i].find(DefineName);
+ if (I == SecondaryDefines[i].end()) {
+ // Secondary target doesn't have this #define.
+ isPortable = false;
+ } else {
+ // Secondary target has this define, remember if it disagrees.
+ if (isPortable)
+ isPortable = I->second == DefineValue;
+ // Remove it from the secondary target unconditionally.
+ SecondaryDefines[i].erase(I);
+ }
+ }
+
+ // If this define is non-portable, turn it into #define_target, otherwise
+ // just use #define.
+ const char *Command = isPortable ? "#define " : "#define_target ";
+ Buffer.insert(Buffer.end(), Command, Command+strlen(Command));
+
+ // Insert "defname defvalue\n".
+ Buffer.insert(Buffer.end(), DefineName.begin(), DefineName.end());
+ Buffer.push_back(' ');
+ Buffer.insert(Buffer.end(), DefineValue.begin(), DefineValue.end());
+ Buffer.push_back('\n');
+ }
+
+ // Now that all of the primary target's defines have been handled and removed
+ // from the secondary target's define sets, go through the remaining secondary
+ // target's #defines and taint them.
+ for (unsigned i = 0, e = SecondaryDefines.size(); i != e; ++i) {
+ std::map<std::string, std::string> &Defs = SecondaryDefines[i];
+ while (!Defs.empty()) {
+ const std::string &DefName = Defs.begin()->first;
+
+ // Insert "#define_other_target defname".
+ const char *Command = "#define_other_target ";
+ Buffer.insert(Buffer.end(), Command, Command+strlen(Command));
+ Buffer.insert(Buffer.end(), DefName.begin(), DefName.end());
+ Buffer.push_back('\n');
+
+ // If any other secondary targets have this same define, remove it from
+ // them to avoid duplicate #define_other_target directives.
+ for (unsigned j = i+1; j != e; ++j)
+ SecondaryDefines[j].erase(DefName);
+
+ Defs.erase(Defs.begin());
+ }
+ }
+}
+
+/// ComputeWCharWidth - Determine the width of the wchar_t type for the primary
+/// target, diagnosing whether this is non-portable across the secondary
+/// targets.
+void TargetInfo::ComputeWCharWidth(SourceLocation Loc) {
+ WCharWidth = PrimaryTarget->getWCharWidth();
+
+ // Check whether this is portable across the secondary targets if the T-U is
+ // portable so far.
+ for (unsigned i = 0, e = SecondaryTargets.size(); i != e; ++i)
+ if (SecondaryTargets[i]->getWCharWidth() != WCharWidth)
+ return DiagnoseNonPortability(Loc, diag::port_wchar_t);
+}
+
+
+/// getTargetBuiltins - Return information about target-specific builtins for
+/// the current primary target, and info about which builtins are non-portable
+/// across the current set of primary and secondary targets.
+void TargetInfo::getTargetBuiltins(const Builtin::Info *&Records,
+ unsigned &NumRecords,
+ std::vector<const char *> &NPortable) const {
+ // Get info about what actual builtins we will expose.
+ PrimaryTarget->getTargetBuiltins(Records, NumRecords);
+ if (SecondaryTargets.empty()) return;
+
+ // Compute the set of non-portable builtins.
+
+ // Start by computing a mapping from the primary target's builtins to their
+ // info records for efficient lookup.
+ std::map<std::string, const Builtin::Info*> PrimaryRecs;
+ for (unsigned i = 0, e = NumRecords; i != e; ++i)
+ PrimaryRecs[Records[i].Name] = Records+i;
+
+ for (unsigned i = 0, e = SecondaryTargets.size(); i != e; ++i) {
+ // Get the builtins for this secondary target.
+ const Builtin::Info *Records2nd;
+ unsigned NumRecords2nd;
+ SecondaryTargets[i]->getTargetBuiltins(Records2nd, NumRecords2nd);
+
+ // Remember all of the secondary builtin names.
+ std::set<std::string> BuiltinNames2nd;
+
+ for (unsigned j = 0, e = NumRecords2nd; j != e; ++j) {
+ BuiltinNames2nd.insert(Records2nd[j].Name);
+
+ // Check to see if the primary target has this builtin.
+ if (const Builtin::Info *PrimBI = PrimaryRecs[Records2nd[j].Name]) {
+ // If does. If they are not identical, mark the builtin as being
+ // non-portable.
+ if (Records2nd[j] != *PrimBI)
+ NPortable.push_back(PrimBI->Name);
+ } else {
+ // The primary target doesn't have this, it is non-portable.
+ NPortable.push_back(Records2nd[j].Name);
+ }
+ }
+
+ // Now that we checked all the secondary builtins, check to see if the
+ // primary target has any builtins that the secondary one doesn't. If so,
+ // then those are non-portable.
+ for (unsigned j = 0, e = NumRecords; j != e; ++j) {
+ if (!BuiltinNames2nd.count(Records[j].Name))
+ NPortable.push_back(Records[j].Name);
+ }
+ }
+}
+
+
diff --git a/Basic/TokenKinds.cpp b/Basic/TokenKinds.cpp
new file mode 100644
index 0000000..772925b
--- /dev/null
+++ b/Basic/TokenKinds.cpp
@@ -0,0 +1,28 @@
+//===--- TokenKinds.cpp - Token Kinds Support -----------------------------===//
+//
+// The LLVM Compiler Infrastructure
+//
+// This file was developed by Chris Lattner and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements the TokenKind enum and support functions.
+//
+//===----------------------------------------------------------------------===//
+
+#include "clang/Basic/TokenKinds.h"
+#include <cassert>
+using namespace clang;
+
+static const char * const TokNames[] = {
+#define TOK(X) #X,
+#define KEYWORD(X,Y) #X,
+#include "clang/Basic/TokenKinds.def"
+ 0
+};
+
+const char *tok::getTokenName(enum TokenKind Kind) {
+ assert(Kind < tok::NUM_TOKENS);
+ return TokNames[Kind];
+}