Skip to content

Dump CUs which are the same linkage name but differ in debug info contents #5594

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 10 commits into
base: experimental/cas/main
Choose a base branch
from
Draft

Dump CUs which are the same linkage name but differ in debug info contents #5594

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
7494837
WIP find CUs that do not dedupe and dwarfdump them
rastogishubham Nov 3, 2022
8908709
WIP dump similar CUs written
rastogishubham Nov 3, 2022
1074bd3
Small Fix
rastogishubham Nov 3, 2022
af1fc51
Small changes to make lives easier
rastogishubham Oct 19, 2022
87968e8
Fix bugs with llvm-cas-dump's --dwarf-dump option
rastogishubham Oct 28, 2022
32bd6e0
Few more bug fixes
rastogishubham Oct 30, 2022
4fee749
WIP dumpsimilarcus, need to change to use llvm set and map, need to p…
rastogishubham Oct 31, 2022
e5170ad
Change llvm-cas-dump to only print strp
rastogishubham Nov 1, 2022
e8104bb
Small bug fixes
rastogishubham Nov 10, 2022
161d243
Bug Fix
rastogishubham Nov 10, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions llvm/include/llvm/MC/CAS/MCCASObjectV1.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,6 +145,13 @@ class MCObjectProxy : public cas::ObjectProxy {

MCObjectProxy() = delete;

static Error encodeReferences(ArrayRef<cas::ObjectRef> Refs,
SmallVectorImpl<char> &Data,
SmallVectorImpl<cas::ObjectRef> &IDs);

static Expected<SmallVector<cas::ObjectRef>>
decodeReferences(const MCObjectProxy &Node, StringRef &Remaining);

protected:
MCObjectProxy(const MCSchema &Schema, const cas::ObjectProxy &Node)
: cas::ObjectProxy(Node), Schema(&Schema) {}
Expand Down
15 changes: 8 additions & 7 deletions llvm/lib/MC/MCCASObjectV1.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -398,9 +398,9 @@ static Error decodeRelocationsAndAddends(MCCASReader &Reader, StringRef Data) {
return Error::success();
}

static Error encodeReferences(ArrayRef<cas::ObjectRef> Refs,
SmallVectorImpl<char> &Data,
SmallVectorImpl<cas::ObjectRef> &IDs) {
Error MCObjectProxy::encodeReferences(ArrayRef<cas::ObjectRef> Refs,
SmallVectorImpl<char> &Data,
SmallVectorImpl<cas::ObjectRef> &IDs) {
DenseMap<cas::ObjectRef, unsigned> RefMap;
SmallVector<cas::ObjectRef> CompactRefs;
for (const auto &ID : Refs) {
Expand Down Expand Up @@ -430,8 +430,9 @@ static Error encodeReferences(ArrayRef<cas::ObjectRef> Refs,
return Error::success();
}

static Expected<SmallVector<cas::ObjectRef>>
decodeReferences(const MCObjectProxy &Node, StringRef &Remaining) {
Expected<SmallVector<cas::ObjectRef>>
MCObjectProxy::decodeReferences(const MCObjectProxy &Node,
StringRef &Remaining) {
Expected<SmallVector<cas::ObjectRef>> MaybeRefs = loadReferences(Node);
if (!MaybeRefs)
return MaybeRefs.takeError();
Expand Down Expand Up @@ -507,7 +508,7 @@ Expected<LoadedDebugAbbrevSection>
LoadedDebugAbbrevSection::load(DebugAbbrevSectionRef Section) {

StringRef Remaining = Section.getData();
auto Refs = decodeReferences(Section, Remaining);
auto Refs = MCObjectProxy::decodeReferences(Section, Remaining);
if (!Refs)
return Refs.takeError();

Expand Down Expand Up @@ -552,7 +553,7 @@ Expected<LoadedDebugLineSection>
LoadedDebugLineSection::load(DebugLineSectionRef Section) {

StringRef Remaining = Section.getData();
auto Refs = decodeReferences(Section, Remaining);
auto Refs = MCObjectProxy::decodeReferences(Section, Remaining);
if (!Refs)
return Refs.takeError();

Expand Down
161 changes: 153 additions & 8 deletions llvm/tools/llvm-cas-dump/CASDWARFObject.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,40 @@ using namespace llvm;
using namespace llvm::cas;
using namespace llvm::mccasformats::v1;

std::unordered_map<std::string, std::unordered_set<std::string>>
CASDWARFObject::MapOfLinkageNames =
std::unordered_map<std::string, std::unordered_set<std::string>>();

// template <> struct DenseMapInfo<std::string> {
// static inline std::string getEmptyKey() {
// return std::string(
// reinterpret_cast<const char *>(~static_cast<uintptr_t>(0)), 0);
// }

// static inline std::string getTombstoneKey() {
// return std::string(
// reinterpret_cast<const char *>(~static_cast<uintptr_t>(1)), 0);
// }

// static unsigned getHashValue(std::string Val);

// static bool isEqual(std::string LHS, std::string RHS) {
// if (RHS.data() == getEmptyKey().data())
// return LHS.data() == getEmptyKey().data();
// if (RHS.data() == getTombstoneKey().data())
// return LHS.data() == getTombstoneKey().data();
// return LHS == RHS;
// }
// };

// unsigned DenseMapInfo<std::string>::getHashValue(std::string Val) {
// assert(Val.data() != getEmptyKey().data() &&
// "Cannot hash the empty key!");
// assert(Val.data() != getTombstoneKey().data() &&
// "Cannot hash the tombstone key!");
// return (unsigned)(hash_value(Val));
// }

namespace {
/// Parse the MachO header to extract details such as endianness.
/// Unfortunately object::MachOObjectfile() doesn't support parsing
Expand Down Expand Up @@ -150,24 +184,99 @@ Error CASDWARFObject::discoverDwarfSections(MCObjectProxy MCObj) {
else if (DebugStrRef::Cast(MCObj)) {
DebugStringSection.append(Data.begin(), Data.end());
DebugStringSection.push_back(0);
} else if (DebugInfoCURef::Cast(MCObj))
return Error::success();
if (DebugAbbrevSectionRef::Cast(MCObj) || GroupRef::Cast(MCObj) ||
SymbolTableRef::Cast(MCObj) || SectionRef::Cast(MCObj) ||
DebugLineSectionRef::Cast(MCObj) || AtomRef::Cast(MCObj)) {
auto Refs = MCObjectProxy::decodeReferences(MCObj, Data);
if (!Refs)
return Refs.takeError();
for (auto Ref : *Refs) {
if (Error E = discoverDwarfSections(Ref))
return E;
}
return Error::success();
}
return MCObj.forEachReference(
[this](ObjectRef CASObj) { return discoverDwarfSections(CASObj); });
}

static Optional<StringRef> getLinkageName(DWARFDie &CUDie) {
if (CUDie.getTag() == dwarf::DW_TAG_subprogram) {
auto Decl = CUDie.findRecursively({dwarf::DW_AT_declaration});
if (!Decl) {
StringRef LinkageName = CUDie.getLinkageName();
if (LinkageName.size())
return LinkageName;
}
}
DWARFDie Child = CUDie.getFirstChild();
while(Child) {
auto Name = getLinkageName(Child);
Child = Child.getSibling();
if (Name)
return Name;
}
return None;
}

static void findStrpsInCompileUnit(DWARFDie &CUDie, raw_ostream &OS) {
for (auto Attr : CUDie.attributes()) {
if (Attr.Value.getForm() == llvm::dwarf::DW_FORM_strp) {
Attr.Value.dump(OS);
OS << "\n";
}
}
DWARFDie Child = CUDie.getFirstChild();
while (Child && Child.getAbbreviationDeclarationPtr()) {
findStrpsInCompileUnit(Child, OS);
Child = Child.getSibling();
}
}

// void CASDWARFObject::addLinkageNameAndObjectRefToMap(DWARFDie &CUDie,
// MCObjectProxy MCObj,
// bool &LinkageFound, DWARFUnit &U, DIDumpOptions &DumpOpts) {
// if (CUDie.getTag() == dwarf::DW_TAG_subprogram) {
// auto Decl = CUDie.findRecursively({dwarf::DW_AT_declaration});
// if (!Decl) {
// StringRef LinkageName = CUDie.getLinkageName();
// if (LinkageName.size()) {
// if (MapOfLinkageNames.find(LinkageName.data()) ==
// MapOfLinkageNames.end())
// MapOfLinkageNames.try_emplace(LinkageName.data(),
// std::unordered_set<std::string>());
// SmallVector<char, 0> DumpContents;
// raw_svector_ostream DumpStream(DumpContents);
// U.dump(DumpStream, DumpOpts);
// MapOfLinkageNames[LinkageName.data()].insert(DumpStream.str().data());
// LinkageFound = true;
// }
// }
// }
// DWARFDie Child = CUDie.getFirstChild();
// while (Child && !LinkageFound) {
// addLinkageNameAndObjectRefToMap(Child, MCObj, LinkageFound, U, DumpOpts);
// Child = Child.getSibling();
// }
// }

Error CASDWARFObject::dump(raw_ostream &OS, int Indent, DWARFContext &DWARFCtx,
MCObjectProxy MCObj, bool ShowForm, bool Verbose) {
OS.indent(Indent);
MCObjectProxy MCObj, bool ShowForm, bool Verbose,
bool DumpSameLinkageDifferentCU) {
if (!DumpSameLinkageDifferentCU)
OS.indent(Indent);
DIDumpOptions DumpOpts;
DumpOpts.ShowChildren = true;
DumpOpts.ShowForm = ShowForm;
DumpOpts.Verbose = Verbose;
Error Err = Error::success();
StringRef Data = MCObj.getData();
if (Data.empty())
return Err;
if (DebugStrRef::Cast(MCObj)) {
return Error::success();
if (DebugStrRef::Cast(MCObj) && !DumpSameLinkageDifferentCU) {
// Dump __debug_str data.
Error Err = Error::success();
assert(Data.data()[Data.size()] == 0);
DataExtractor StrData(StringRef(Data.data(), Data.size() + 1),
isLittleEndian(), 0);
Expand All @@ -184,7 +293,7 @@ Error CASDWARFObject::dump(raw_ostream &OS, int Indent, DWARFContext &DWARFCtx,
OS << "\"\n";
StrOffset = Offset;
}
} else if (DebugLineRef::Cast(MCObj)) {
} else if (DebugLineRef::Cast(MCObj) && !DumpSameLinkageDifferentCU) {
// Dump __debug_line data.
uint64_t Address = 0;
DWARFDataExtractor LineData(*this, {Data, Address}, isLittleEndian(), 0);
Expand Down Expand Up @@ -215,7 +324,43 @@ Error CASDWARFObject::dump(raw_ostream &OS, int Indent, DWARFContext &DWARFCtx,
&getLocSection(), getStrSection(),
getStrOffsetsSection(), &getAddrSection(),
getLocSection(), isLittleEndian(), false, UV);
U.dump(OS, DumpOpts);
if (DumpSameLinkageDifferentCU) {
if (DWARFDie CUDie = U.getUnitDIE(false)) {
auto Name = getLinkageName(CUDie);
if (!Name)
return Error::success();
SmallVector<char, 0> StrpData;
raw_svector_ostream OS(StrpData);
findStrpsInCompileUnit(CUDie, OS);
if (MapOfLinkageNames.find(Name->data()) == MapOfLinkageNames.end())
MapOfLinkageNames.try_emplace(Name->data(), std::unordered_set<std::string>());
MapOfLinkageNames[Name->data()].insert(StrpData.data());
}

} else {
U.dump(OS, DumpOpts);
}
}
return Err;
return Error::success();
}

Error CASDWARFObject::dumpSimilarCUs() {
for (auto KeyValue : MapOfLinkageNames) {
if (KeyValue.second.size() == 2) {
for (auto ID : KeyValue.second) {
llvm::outs() << KeyValue.first << " " << ID << "\n";
break;
}
}
}
for (auto KeyValue : MapOfLinkageNames) {
if (KeyValue.second.size() == 2) {
int count = 0;
for (auto ID : KeyValue.second) {
if (count == 0) { count++; continue; }
llvm::outs() << KeyValue.first << " " << ID << "\n";
}
}
}
return Error::success();
}
19 changes: 17 additions & 2 deletions llvm/tools/llvm-cas-dump/CASDWARFObject.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,8 @@
#include "llvm/DebugInfo/DWARF/DWARFDebugLine.h"
#include "llvm/DebugInfo/DWARF/DWARFObject.h"
#include "llvm/MC/CAS/MCCASObjectV1.h"
#include <unordered_map>
#include <unordered_set>

namespace llvm {

Expand All @@ -32,6 +34,9 @@ class CASDWARFObject : public DWARFObject {
SmallVector<StringRef> CUDataVec;
SmallVector<uint32_t> SecOffsetVals;
unsigned LineTableOffset = 0;
DenseMap<cas::ObjectRef, unsigned> MapOfStringOffsets;
static std::unordered_map<std::string, std::unordered_set<std::string>>
MapOfLinkageNames;
unsigned CompileUnitIndex = 0;

const mccasformats::v1::MCSchema &Schema;
Expand All @@ -45,6 +50,10 @@ class CASDWARFObject : public DWARFObject {
/// raw_ostream &OS);
Error discoverDebugInfoSection(cas::ObjectRef CASObj, raw_ostream &OS);

void addLinkageNameAndObjectRefToMap(DWARFDie &CUDie,
mccasformats::v1::MCObjectProxy MCObj,
bool &LinkageFound, DWARFUnit &U, DIDumpOptions &DumpOpts);

public:
CASDWARFObject(const mccasformats::v1::MCSchema &Schema) : Schema(Schema) {}

Expand All @@ -57,10 +66,14 @@ class CASDWARFObject : public DWARFObject {
Error discoverDebugInfoSection(mccasformats::v1::MCObjectProxy MCObj,
raw_ostream &OS);

void addLinkageNameAndObjectRefToMap(DWARFDie CUDie,
mccasformats::v1::MCObjectProxy MCObj,
bool &LinkageFound);

/// Dump MCObj as textual DWARF output.
Error dump(raw_ostream &OS, int Indent, DWARFContext &DWARFCtx,
mccasformats::v1::MCObjectProxy MCObj, bool ShowForm,
bool Verbose);
mccasformats::v1::MCObjectProxy MCObj, bool ShowForm, bool Verbose,
bool DumpSameLinkageDifferentCU);

StringRef getFileName() const override { return "CAS"; }
ArrayRef<SectionName> getSectionNames() const override { return {}; }
Expand All @@ -77,6 +90,8 @@ class CASDWARFObject : public DWARFObject {
return {};
};
ArrayRef<uint32_t> getSecOffsetVals() { return SecOffsetVals; }

static Error dumpSimilarCUs();
};
} // namespace llvm

Expand Down
Loading