std::string get_line_from_offset(StringRef buffer, std::size_t offset) {
assert(buffer.size() > offset);
auto line_start = buffer.find_last_of("\r\n", offset) + 1;
auto line_end = buffer.find_first_of("\r\n", offset);
return std::string(buffer.begin() + line_start, buffer.begin() + line_end);
}
DeclNameViewer::DeclNameViewer(StringRef Text): IsValid(true), HasParen(false) {
auto ArgStart = Text.find_first_of('(');
if (ArgStart == StringRef::npos) {
BaseName = Text;
return;
}
HasParen = true;
BaseName = Text.substr(0, ArgStart);
auto ArgEnd = Text.find_last_of(')');
if (ArgEnd == StringRef::npos) {
IsValid = false;
return;
}
StringRef AllArgs = Text.substr(ArgStart + 1, ArgEnd - ArgStart - 1);
AllArgs.split(Labels, ":");
if (Labels.empty())
return;
if ((IsValid = Labels.back().empty())) {
Labels.pop_back();
std::transform(Labels.begin(), Labels.end(), Labels.begin(),
[](StringRef Label) {
return Label == "_" ? StringRef() : Label;
});
}
}
/// \brief Parse \p Input as line sample.
///
/// \param Input input line.
/// \param IsCallsite true if the line represents an inlined callsite.
/// \param Depth the depth of the inline stack.
/// \param NumSamples total samples of the line/inlined callsite.
/// \param LineOffset line offset to the start of the function.
/// \param Discriminator discriminator of the line.
/// \param TargetCountMap map from indirect call target to count.
///
/// returns true if parsing is successful.
static bool ParseLine(const StringRef &Input, bool &IsCallsite, unsigned &Depth,
unsigned &NumSamples, unsigned &LineOffset,
unsigned &Discriminator, StringRef &CalleeName,
DenseMap<StringRef, unsigned> &TargetCountMap) {
for (Depth = 0; Input[Depth] == ' '; Depth++)
;
if (Depth == 0)
return false;
size_t n1 = Input.find(':');
StringRef Loc = Input.substr(Depth, n1 - Depth);
size_t n2 = Loc.find('.');
if (n2 == StringRef::npos) {
if (Loc.getAsInteger(10, LineOffset))
return false;
Discriminator = 0;
} else {
if (Loc.substr(0, n2).getAsInteger(10, LineOffset))
return false;
if (Loc.substr(n2 + 1).getAsInteger(10, Discriminator))
return false;
}
StringRef Rest = Input.substr(n1 + 2);
if (Rest[0] >= '0' && Rest[0] <= '9') {
IsCallsite = false;
size_t n3 = Rest.find(' ');
if (n3 == StringRef::npos) {
if (Rest.getAsInteger(10, NumSamples))
return false;
} else {
if (Rest.substr(0, n3).getAsInteger(10, NumSamples))
return false;
}
while (n3 != StringRef::npos) {
n3 += Rest.substr(n3).find_first_not_of(' ');
Rest = Rest.substr(n3);
n3 = Rest.find(' ');
StringRef pair = Rest;
if (n3 != StringRef::npos) {
pair = Rest.substr(0, n3);
}
int n4 = pair.find(':');
unsigned count;
if (pair.substr(n4 + 1).getAsInteger(10, count))
return false;
TargetCountMap[pair.substr(0, n4)] = count;
}
} else {
IsCallsite = true;
int n3 = Rest.find_last_of(':');
CalleeName = Rest.substr(0, n3);
if (Rest.substr(n3 + 1).getAsInteger(10, NumSamples))
return false;
}
return true;
}
std::string ModuleFile::Dependency::getPrettyPrintedPath() const {
StringRef pathWithoutScope = RawPath;
if (isScoped()) {
size_t splitPoint = pathWithoutScope.find_last_of('\0');
pathWithoutScope = pathWithoutScope.slice(0, splitPoint);
}
std::string output = pathWithoutScope.str();
std::replace(output.begin(), output.end(), '\0', '.');
return output;
}
StringRef extension(StringRef path, Style style) {
StringRef fname = filename(path, style);
size_t pos = fname.find_last_of('.');
if (pos == StringRef::npos)
return StringRef();
else
if ((fname.size() == 1 && fname == ".") ||
(fname.size() == 2 && fname == ".."))
return StringRef();
else
return fname.substr(pos);
}
static unsigned getColumnNumber(StringRef buffer, llvm::SMLoc loc) {
assert(loc.getPointer() >= buffer.data());
assert((size_t)(loc.getPointer() - buffer.data()) <= buffer.size());
StringRef UpToLoc = buffer.slice(0, loc.getPointer() - buffer.data());
size_t ColumnNo = UpToLoc.size();
size_t NewlinePos = UpToLoc.find_last_of("\r\n");
if (NewlinePos != StringRef::npos)
ColumnNo -= NewlinePos;
return static_cast<unsigned>(ColumnNo);
}
bool Punycode::decodePunycode(StringRef InputPunycode,
std::vector<uint32_t> &OutCodePoints) {
OutCodePoints.clear();
OutCodePoints.reserve(InputPunycode.size());
// -- Build the decoded string as UTF32 first because we need random access.
uint32_t n = initial_n;
int i = 0;
int bias = initial_bias;
/// let output = an empty string indexed from 0
// consume all code points before the last delimiter (if there is one)
// and copy them to output,
size_t lastDelimiter = InputPunycode.find_last_of(delimiter);
if (lastDelimiter != StringRef::npos) {
for (char c : InputPunycode.slice(0, lastDelimiter)) {
// fail on any non-basic code point
if (static_cast<unsigned char>(c) > 0x7f)
return true;
OutCodePoints.push_back(c);
}
// if more than zero code points were consumed then consume one more
// (which will be the last delimiter)
InputPunycode =
InputPunycode.slice(lastDelimiter + 1, InputPunycode.size());
}
while (!InputPunycode.empty()) {
int oldi = i;
int w = 1;
for (int k = base; ; k += base) {
// consume a code point, or fail if there was none to consume
if (InputPunycode.empty())
return true;
char codePoint = InputPunycode.front();
InputPunycode = InputPunycode.slice(1, InputPunycode.size());
// let digit = the code point's digit-value, fail if it has none
int digit = digit_index(codePoint);
if (digit < 0)
return true;
i = i + digit * w;
int t = k <= bias ? tmin
: k >= bias + tmax ? tmax
: k - bias;
if (digit < t)
break;
w = w * (base - t);
}
bias = adapt(i - oldi, OutCodePoints.size() + 1, oldi == 0);
n = n + i / (OutCodePoints.size() + 1);
i = i % (OutCodePoints.size() + 1);
// if n is a basic code point then fail
if (n < 0x80)
return true;
// insert n into output at position i
OutCodePoints.insert(OutCodePoints.begin() + i, n);
i++;
}
return true;
}
Status ModuleFile::associateWithFileContext(FileUnit *file,
SourceLoc diagLoc) {
PrettyModuleFileDeserialization stackEntry(*this);
assert(getStatus() == Status::Valid && "invalid module file");
assert(!FileContext && "already associated with an AST module");
FileContext = file;
if (file->getParentModule()->getName().str() != Name)
return error(Status::NameMismatch);
ASTContext &ctx = getContext();
llvm::Triple moduleTarget(llvm::Triple::normalize(TargetTriple));
if (!areCompatibleArchitectures(moduleTarget, ctx.LangOpts.Target) ||
!areCompatibleOSs(moduleTarget, ctx.LangOpts.Target)) {
return error(Status::TargetIncompatible);
}
if (ctx.LangOpts.EnableTargetOSChecking &&
isTargetTooNew(moduleTarget, ctx.LangOpts.Target)) {
return error(Status::TargetTooNew);
}
for (const auto &searchPathPair : SearchPaths)
ctx.addSearchPath(searchPathPair.first, searchPathPair.second);
auto clangImporter = static_cast<ClangImporter *>(ctx.getClangModuleLoader());
bool missingDependency = false;
for (auto &dependency : Dependencies) {
assert(!dependency.isLoaded() && "already loaded?");
if (dependency.isHeader()) {
// The path may be empty if the file being loaded is a partial AST,
// and the current compiler invocation is a merge-modules step.
if (!dependency.RawPath.empty()) {
bool hadError =
clangImporter->importHeader(dependency.RawPath,
file->getParentModule(),
importedHeaderInfo.fileSize,
importedHeaderInfo.fileModTime,
importedHeaderInfo.contents,
diagLoc);
if (hadError)
return error(Status::FailedToLoadBridgingHeader);
}
Module *importedHeaderModule = clangImporter->getImportedHeaderModule();
dependency.Import = { {}, importedHeaderModule };
continue;
}
StringRef modulePathStr = dependency.RawPath;
StringRef scopePath;
if (dependency.isScoped()) {
auto splitPoint = modulePathStr.find_last_of('\0');
assert(splitPoint != StringRef::npos);
scopePath = modulePathStr.substr(splitPoint+1);
modulePathStr = modulePathStr.slice(0, splitPoint);
}
SmallVector<Identifier, 4> modulePath;
while (!modulePathStr.empty()) {
StringRef nextComponent;
std::tie(nextComponent, modulePathStr) = modulePathStr.split('\0');
modulePath.push_back(ctx.getIdentifier(nextComponent));
assert(!modulePath.back().empty() &&
"invalid module name (submodules not yet supported)");
}
auto module = getModule(modulePath);
if (!module) {
// If we're missing the module we're shadowing, treat that specially.
if (modulePath.size() == 1 &&
modulePath.front() == file->getParentModule()->getName()) {
return error(Status::MissingShadowedModule);
}
// Otherwise, continue trying to load dependencies, so that we can list
// everything that's missing.
missingDependency = true;
continue;
}
// This is for backwards-compatibility with modules that still rely on the
// "HasUnderlyingModule" flag.
if (Bits.HasUnderlyingModule && module == ShadowedModule)
dependency.forceExported();
if (scopePath.empty()) {
dependency.Import = { {}, module };
} else {
auto scopeID = ctx.getIdentifier(scopePath);
assert(!scopeID.empty() &&
"invalid decl name (non-top-level decls not supported)");
auto path = Module::AccessPathTy({scopeID, SourceLoc()});
dependency.Import = { ctx.AllocateCopy(path), module };
}
}
if (missingDependency) {
return error(Status::MissingDependency);
}
if (Bits.HasEntryPoint) {
FileContext->getParentModule()->registerEntryPointFile(FileContext,
SourceLoc(),
None);
}
return getStatus();
}
// Returns the last element of a path, which is supposed to be a filename.
static StringRef getBasename(StringRef Path) {
size_t Pos = Path.find_last_of("\\/");
if (Pos == StringRef::npos)
return Path;
return Path.substr(Pos + 1);
}
void MCObjectDisassembler::buildCFG(MCModule *Module) {
typedef std::map<uint64_t, BBInfo> BBInfoByAddrTy;
BBInfoByAddrTy BBInfos;
AddressSetTy Splits;
AddressSetTy Calls;
error_code ec;
for (symbol_iterator SI = Obj.begin_symbols(), SE = Obj.end_symbols();
SI != SE; SI.increment(ec)) {
if (ec)
break;
SymbolRef::Type SymType;
SI->getType(SymType);
if (SymType == SymbolRef::ST_Function) {
uint64_t SymAddr;
SI->getAddress(SymAddr);
SymAddr = getEffectiveLoadAddr(SymAddr);
Calls.push_back(SymAddr);
Splits.push_back(SymAddr);
}
}
assert(Module->func_begin() == Module->func_end()
&& "Module already has a CFG!");
// First, determine the basic block boundaries and call targets.
for (MCModule::atom_iterator AI = Module->atom_begin(),
AE = Module->atom_end();
AI != AE; ++AI) {
MCTextAtom *TA = dyn_cast<MCTextAtom>(*AI);
if (!TA) continue;
Calls.push_back(TA->getBeginAddr());
BBInfos[TA->getBeginAddr()].Atom = TA;
for (MCTextAtom::const_iterator II = TA->begin(), IE = TA->end();
II != IE; ++II) {
if (MIA.isTerminator(II->Inst))
Splits.push_back(II->Address + II->Size);
uint64_t Target;
if (MIA.evaluateBranch(II->Inst, II->Address, II->Size, Target)) {
if (MIA.isCall(II->Inst))
Calls.push_back(Target);
Splits.push_back(Target);
}
}
}
RemoveDupsFromAddressVector(Splits);
RemoveDupsFromAddressVector(Calls);
// Split text atoms into basic block atoms.
for (AddressSetTy::const_iterator SI = Splits.begin(), SE = Splits.end();
SI != SE; ++SI) {
MCAtom *A = Module->findAtomContaining(*SI);
if (!A) continue;
MCTextAtom *TA = cast<MCTextAtom>(A);
if (TA->getBeginAddr() == *SI)
continue;
MCTextAtom *NewAtom = TA->split(*SI);
BBInfos[NewAtom->getBeginAddr()].Atom = NewAtom;
StringRef BBName = TA->getName();
BBName = BBName.substr(0, BBName.find_last_of(':'));
NewAtom->setName((BBName + ":" + utohexstr(*SI)).str());
}
// Compute succs/preds.
for (MCModule::atom_iterator AI = Module->atom_begin(),
AE = Module->atom_end();
AI != AE; ++AI) {
MCTextAtom *TA = dyn_cast<MCTextAtom>(*AI);
if (!TA) continue;
BBInfo &CurBB = BBInfos[TA->getBeginAddr()];
const MCDecodedInst &LI = TA->back();
if (MIA.isBranch(LI.Inst)) {
uint64_t Target;
if (MIA.evaluateBranch(LI.Inst, LI.Address, LI.Size, Target))
CurBB.addSucc(BBInfos[Target]);
if (MIA.isConditionalBranch(LI.Inst))
CurBB.addSucc(BBInfos[LI.Address + LI.Size]);
} else if (!MIA.isTerminator(LI.Inst))
CurBB.addSucc(BBInfos[LI.Address + LI.Size]);
}
// Create functions and basic blocks.
for (AddressSetTy::const_iterator CI = Calls.begin(), CE = Calls.end();
CI != CE; ++CI) {
BBInfo &BBI = BBInfos[*CI];
if (!BBI.Atom) continue;
MCFunction &MCFN = *Module->createFunction(BBI.Atom->getName());
// Create MCBBs.
SmallSetVector<BBInfo*, 16> Worklist;
Worklist.insert(&BBI);
for (size_t wi = 0; wi < Worklist.size(); ++wi) {
BBInfo *BBI = Worklist[wi];
if (!BBI->Atom)
continue;
BBI->BB = &MCFN.createBlock(*BBI->Atom);
// Add all predecessors and successors to the worklist.
for (BBInfoSetTy::iterator SI = BBI->Succs.begin(), SE = BBI->Succs.end();
SI != SE; ++SI)
Worklist.insert(*SI);
for (BBInfoSetTy::iterator PI = BBI->Preds.begin(), PE = BBI->Preds.end();
PI != PE; ++PI)
Worklist.insert(*PI);
}
// Set preds/succs.
for (size_t wi = 0; wi < Worklist.size(); ++wi) {
BBInfo *BBI = Worklist[wi];
MCBasicBlock *MCBB = BBI->BB;
if (!MCBB)
continue;
for (BBInfoSetTy::iterator SI = BBI->Succs.begin(), SE = BBI->Succs.end();
SI != SE; ++SI)
if ((*SI)->BB)
MCBB->addSuccessor((*SI)->BB);
for (BBInfoSetTy::iterator PI = BBI->Preds.begin(), PE = BBI->Preds.end();
PI != PE; ++PI)
if ((*PI)->BB)
MCBB->addPredecessor((*PI)->BB);
}
}
}
// Drop directory components and replace extension with ".exe".
static std::string getOutputPath(StringRef Path) {
auto P = Path.find_last_of("\\/");
StringRef S = (P == StringRef::npos) ? Path : Path.substr(P + 1);
return (S.substr(0, S.rfind('.')) + ".exe").str();
}
/// Parse \p Input as line sample.
///
/// \param Input input line.
/// \param IsCallsite true if the line represents an inlined callsite.
/// \param Depth the depth of the inline stack.
/// \param NumSamples total samples of the line/inlined callsite.
/// \param LineOffset line offset to the start of the function.
/// \param Discriminator discriminator of the line.
/// \param TargetCountMap map from indirect call target to count.
///
/// returns true if parsing is successful.
static bool ParseLine(const StringRef &Input, bool &IsCallsite, uint32_t &Depth,
uint64_t &NumSamples, uint32_t &LineOffset,
uint32_t &Discriminator, StringRef &CalleeName,
DenseMap<StringRef, uint64_t> &TargetCountMap) {
for (Depth = 0; Input[Depth] == ' '; Depth++)
;
if (Depth == 0)
return false;
size_t n1 = Input.find(':');
StringRef Loc = Input.substr(Depth, n1 - Depth);
size_t n2 = Loc.find('.');
if (n2 == StringRef::npos) {
if (Loc.getAsInteger(10, LineOffset) || !isOffsetLegal(LineOffset))
return false;
Discriminator = 0;
} else {
if (Loc.substr(0, n2).getAsInteger(10, LineOffset))
return false;
if (Loc.substr(n2 + 1).getAsInteger(10, Discriminator))
return false;
}
StringRef Rest = Input.substr(n1 + 2);
if (Rest[0] >= '0' && Rest[0] <= '9') {
IsCallsite = false;
size_t n3 = Rest.find(' ');
if (n3 == StringRef::npos) {
if (Rest.getAsInteger(10, NumSamples))
return false;
} else {
if (Rest.substr(0, n3).getAsInteger(10, NumSamples))
return false;
}
// Find call targets and their sample counts.
// Note: In some cases, there are symbols in the profile which are not
// mangled. To accommodate such cases, use colon + integer pairs as the
// anchor points.
// An example:
// _M_construct<char *>:1000 string_view<std::allocator<char> >:437
// ":1000" and ":437" are used as anchor points so the string above will
// be interpreted as
// target: _M_construct<char *>
// count: 1000
// target: string_view<std::allocator<char> >
// count: 437
while (n3 != StringRef::npos) {
n3 += Rest.substr(n3).find_first_not_of(' ');
Rest = Rest.substr(n3);
n3 = Rest.find_first_of(':');
if (n3 == StringRef::npos || n3 == 0)
return false;
StringRef Target;
uint64_t count, n4;
while (true) {
// Get the segment after the current colon.
StringRef AfterColon = Rest.substr(n3 + 1);
// Get the target symbol before the current colon.
Target = Rest.substr(0, n3);
// Check if the word after the current colon is an integer.
n4 = AfterColon.find_first_of(' ');
n4 = (n4 != StringRef::npos) ? n3 + n4 + 1 : Rest.size();
StringRef WordAfterColon = Rest.substr(n3 + 1, n4 - n3 - 1);
if (!WordAfterColon.getAsInteger(10, count))
break;
// Try to find the next colon.
uint64_t n5 = AfterColon.find_first_of(':');
if (n5 == StringRef::npos)
return false;
n3 += n5 + 1;
}
// An anchor point is found. Save the {target, count} pair
TargetCountMap[Target] = count;
if (n4 == Rest.size())
break;
// Change n3 to the next blank space after colon + integer pair.
n3 = n4;
}
} else {
IsCallsite = true;
size_t n3 = Rest.find_last_of(':');
CalleeName = Rest.substr(0, n3);
if (Rest.substr(n3 + 1).getAsInteger(10, NumSamples))
return false;
}
return true;
}
