Add llvm-pdbdump to tools.

llvm-pdbdump is a tool which can be used to dump the contents
of Microsoft-generated PDB files.  It makes use of the Microsoft
DIA SDK, which is a COM based library designed specifically for
this purpose.

The initial commit of this tool dumps the raw bytes from PDB data
streams.  Future commits will dump more semantic information such
as types, symbols, source files, etc similar to the types of
information accessible via llvm-dwarfdump.

Reviewed by: Aaron Ballman, Reid Kleckner, Chandler Carruth
Differential Revision: http://reviews.llvm.org/D7153

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@227241 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Zachary Turner 2015-01-27 20:46:21 +00:00
parent a18b3e06fc
commit c1592bca1e
6 changed files with 495 additions and 1 deletions

View File

@ -61,6 +61,10 @@ add_llvm_tool_subdirectory(yaml2obj)
add_llvm_tool_subdirectory(llvm-go)
if(MSVC)
add_llvm_tool_subdirectory(llvm-pdbdump)
endif()
if(NOT CYGWIN AND LLVM_ENABLE_PIC)
add_llvm_tool_subdirectory(lto)
add_llvm_tool_subdirectory(llvm-lto)

View File

@ -16,7 +16,7 @@
;===------------------------------------------------------------------------===;
[common]
subdirectories = bugpoint llc lli llvm-ar llvm-as llvm-bcanalyzer llvm-cov llvm-diff llvm-dis llvm-dwarfdump llvm-extract llvm-jitlistener llvm-link llvm-lto llvm-mc llvm-nm llvm-objdump llvm-profdata llvm-rtdyld llvm-size macho-dump opt llvm-mcmarkup verify-uselistorder dsymutil
subdirectories = bugpoint llc lli llvm-ar llvm-as llvm-bcanalyzer llvm-cov llvm-diff llvm-dis llvm-dwarfdump llvm-extract llvm-jitlistener llvm-link llvm-lto llvm-mc llvm-nm llvm-objdump llvm-pdbdump llvm-profdata llvm-rtdyld llvm-size macho-dump opt llvm-mcmarkup verify-uselistorder dsymutil
[component_0]
type = Group

View File

@ -0,0 +1,26 @@
set(MSVC_DIA_SDK_DIR "$ENV{VSINSTALLDIR}DIA SDK")
# Due to a bug in MSVC 2013's installation software, it is possible
# for MSVC 2013 to write the DIA SDK into the Visual Studio 2012
# install directory. If this happens, the installation is corrupt
# and there's nothing we can do. It happens with enough frequency
# though that we should handle it. We do so by simply checking that
# the DIA SDK folder exists. Should this happen you will need to
# uninstall VS 2012 and then re-install VS 2013.
if (IS_DIRECTORY ${MSVC_DIA_SDK_DIR})
set(LLVM_LINK_COMPONENTS
Support
)
include_directories(${MSVC_DIA_SDK_DIR}/include)
if (CMAKE_SIZEOF_VOID_P EQUAL 8)
link_directories(${MSVC_DIA_SDK_DIR}/lib/amd64)
else()
link_directories(${MSVC_DIA_SDK_DIR}/lib)
endif()
add_llvm_tool(llvm-pdbdump
llvm-pdbdump.cpp
)
target_link_libraries(llvm-pdbdump diaguids)
endif()

View File

@ -0,0 +1,286 @@
//===- COMExtras.h - Helper files for COM operations -------------*- C++-*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
#ifndef LLVM_TOOLS_LLVM_PDBDUMP_COMEXTRAS_H
#define LLVM_TOOLS_LLVM_PDBDUMP_COMEXTRAS_H
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/SmallVector.h"
#include <tuple>
namespace llvm {
template <typename F> struct function_traits;
#if LLVM_HAS_VARIADIC_TEMPLATES
template <typename R, typename... Args>
struct function_traits<R (*)(Args...)> : public function_traits<R(Args...)> {};
template <typename C, typename R, typename... Args>
struct function_traits<R (__stdcall C::*)(Args...)> {
using args_tuple = std::tuple<Args...>;
};
#else
// For the sake of COM, we only need a 3 argument version and a 5 argument
// version. We could provide 1, 2, 4, and other length of argument lists if
// this were intended to be more generic. Alternatively, this will "just work"
// if VS2012 support is dropped and we can use the variadic template case
// exclusively.
template <typename C, typename R, typename A1, typename A2, typename A3>
struct function_traits<R (__stdcall C::*)(A1, A2, A3)> {
using args_tuple = std::tuple<A1, A2, A3>;
};
template <typename C, typename R, typename A1, typename A2, typename A3,
typename A4, typename A5>
struct function_traits<R (__stdcall C::*)(A1, A2, A3, A4, A5)> {
using args_tuple = std::tuple<A1, A2, A3, A4, A5>;
};
#endif
template <class FuncTraits, std::size_t arg> struct function_arg {
// Writing function_arg as a separate class that accesses the tuple from
// function_traits is necessary due to what appears to be a bug in MSVC.
// If you write a nested class inside function_traits like this:
// template<std::size_t ArgIndex>
// struct Argument
// {
// typedef typename
// std::tuple_element<ArgIndex, std::tuple<Args...>>::type type;
// };
// MSVC encounters a parsing error.
typedef
typename std::tuple_element<arg, typename FuncTraits::args_tuple>::type
type;
};
template <class T> struct remove_double_pointer {};
template <class T> struct remove_double_pointer<T **> { typedef T type; };
namespace sys {
namespace windows {
/// A helper class for allowing the use of COM enumerators in range-based
/// for loops.
///
/// A common idiom in the COM world is to have an enumerator interface, say
/// IMyEnumerator. It's responsible for enumerating over some child data type,
/// say IChildType. You do the enumeration by calling IMyEnumerator::Next()
/// one of whose arguments will be an IChildType**. Eventually Next() fails,
/// indicating that there are no more items.
///
/// com_iterator represents a single point-in-time of this iteration. It is
/// used by ComEnumerator to support iterating in this fashion via range-based
/// for loops and other common C++ paradigms.
template <class EnumeratorType, std::size_t ArgIndex> class com_iterator {
using FunctionTraits = function_traits<decltype(&EnumeratorType::Next)>;
typedef typename function_arg<FunctionTraits, ArgIndex>::type FuncArgType;
// FuncArgType is now something like ISomeCOMInterface **. Remove both
// pointers, so we can make a CComPtr<T> out of it.
typedef typename remove_double_pointer<FuncArgType>::type EnumDataType;
CComPtr<EnumeratorType> EnumeratorObject;
CComPtr<EnumDataType> CurrentItem;
public:
typedef CComPtr<EnumDataType> value_type;
typedef std::ptrdiff_t difference_type;
typedef value_type *pointer_type;
typedef value_type &reference_type;
typedef std::forward_iterator_tag iterator_category;
explicit com_iterator(CComPtr<EnumeratorType> Enumerator,
CComPtr<EnumDataType> Current)
: EnumeratorObject(Enumerator), CurrentItem(Current) {}
com_iterator() {}
com_iterator &operator++() {
// EnumeratorObject->Next() expects CurrentItem to be NULL.
CurrentItem.Release();
ULONG Count = 0;
HRESULT hr = EnumeratorObject->Next(1, &CurrentItem, &Count);
if (FAILED(hr) || Count == 0)
*this = com_iterator();
return *this;
}
value_type operator*() { return CurrentItem; }
bool operator==(const com_iterator &other) const {
return (EnumeratorObject == other.EnumeratorObject) &&
(CurrentItem == other.CurrentItem);
}
bool operator!=(const com_iterator &other) const { return !(*this == other); }
com_iterator &operator=(const com_iterator &other) {
EnumeratorObject = other.EnumeratorObject;
CurrentItem = other.CurrentItem;
return *this;
}
};
/// ComEnumerator implements the interfaced required for C++ to allow its use
/// in range-based for loops. In particular, a begin() and end() method.
/// These methods simply construct and return an appropriate ComIterator
/// instance.
template <class EnumeratorType, std::size_t ArgIndex> class com_enumerator {
typedef function_traits<decltype(&EnumeratorType::Next)> FunctionTraits;
typedef typename function_arg<FunctionTraits, ArgIndex>::type FuncArgType;
typedef typename remove_double_pointer<FuncArgType>::type EnumDataType;
CComPtr<EnumeratorType> EnumeratorObject;
public:
com_enumerator(CComPtr<EnumeratorType> Enumerator)
: EnumeratorObject(Enumerator) {}
com_iterator<EnumeratorType, ArgIndex> begin() {
if (!EnumeratorObject)
return end();
EnumeratorObject->Reset();
ULONG Count = 0;
CComPtr<EnumDataType> FirstItem;
HRESULT hr = EnumeratorObject->Next(1, &FirstItem, &Count);
return (FAILED(hr) || Count == 0) ? end()
: com_iterator<EnumeratorType, ArgIndex>(
EnumeratorObject, FirstItem);
}
com_iterator<EnumeratorType, ArgIndex> end() {
return com_iterator<EnumeratorType, ArgIndex>();
}
};
/// A helper class for allowing the use of COM record enumerators in range-
/// based for loops.
///
/// A record enumerator is almost the same as a regular enumerator, except
/// that it returns raw byte-data instead of interfaces to other COM objects.
/// As a result, the enumerator's Next() method has a slightly different
/// signature, and an iterator dereferences to an ArrayRef instead of a
/// CComPtr.
template <class EnumeratorType> class com_data_record_iterator {
public:
typedef llvm::ArrayRef<uint8_t> value_type;
typedef std::ptrdiff_t difference_type;
typedef value_type *pointer_type;
typedef value_type &reference_type;
typedef std::forward_iterator_tag iterator_category;
explicit com_data_record_iterator(CComPtr<EnumeratorType> enumerator)
: Enumerator(enumerator), CurrentRecord(0) {
// Make sure we start at the beginning. If there are no records,
// immediately set ourselves equal to end().
if (enumerator)
enumerator->Reset();
if (!ReadNextRecord())
*this = com_data_record_iterator();
}
com_data_record_iterator() {}
com_data_record_iterator &operator++() {
++CurrentRecord;
// If we can't read any more records, either because there are no more
// or because we encountered an error, we should compare equal to end.
if (!ReadNextRecord())
*this = com_data_record_iterator();
return *this;
}
value_type operator*() {
return llvm::ArrayRef<uint8_t>(RecordData.begin(), RecordData.end());
}
bool operator==(const com_data_record_iterator &other) const {
return (Enumerator == other.Enumerator) &&
(CurrentRecord == other.CurrentRecord);
}
bool operator!=(const com_data_record_iterator &other) const {
return !(*this == other);
}
private:
bool ReadNextRecord() {
RecordData.clear();
ULONG Count = 0;
DWORD RequiredBufferSize;
HRESULT hr = Enumerator->Next(1, 0, &RequiredBufferSize, nullptr, &Count);
if (hr == S_OK) {
RecordData.resize(RequiredBufferSize);
DWORD BytesRead = 0;
hr = Enumerator->Next(1, RequiredBufferSize, &BytesRead,
RecordData.data(), &Count);
}
if (hr != S_OK)
RecordData.clear();
return (hr == S_OK);
}
CComPtr<EnumeratorType> Enumerator;
uint32_t CurrentRecord;
llvm::SmallVector<uint8_t, 32> RecordData;
};
/// Similar to ComEnumerator, com_data_record_enumerator implements the range
/// interface for ComDataRecordIterators.
template <class EnumeratorType> class com_data_record_enumerator {
public:
com_data_record_enumerator(CComPtr<EnumeratorType> enumerator)
: Enumerator(enumerator) {}
com_data_record_iterator<EnumeratorType> begin() {
return com_data_record_iterator<EnumeratorType>(Enumerator);
}
com_data_record_iterator<EnumeratorType> end() {
LONG NumElts = 0;
HRESULT hr = Enumerator->get_Count(&NumElts);
return (FAILED(hr)) ? com_data_record_iterator<EnumeratorType>(Enumerator)
: com_data_record_iterator<EnumeratorType>();
}
private:
CComPtr<EnumeratorType> Enumerator;
};
/// com_enumerator is a simple helper function to allow the enumerator
/// class's type to be inferred automatically.
/// This allows you to write simply:
/// for (auto item : com_enumerator(MyEnumerator)) {
/// }
template <class EnumeratorType>
com_enumerator<EnumeratorType, 1>
make_com_enumerator(CComPtr<EnumeratorType> Enumerator) {
return com_enumerator<EnumeratorType, 1>(Enumerator);
}
/// com_data_record_enumerator is a simple helper function to allow the
/// enumerator class's type to be inferred automatically.
/// This allows you to write simply:
/// for (auto item : com_data_record_enumerator(MyEnumerator)) {
/// }
//=============================================================================
template <class EnumeratorType>
com_data_record_enumerator<EnumeratorType>
make_com_data_record_enumerator(CComPtr<EnumeratorType> Enumerator) {
return com_data_record_enumerator<EnumeratorType>(Enumerator);
}
} // namespace windows
} // namespace sys
} // namespace llvm
#endif

View File

@ -0,0 +1,23 @@
;===- ./tools/llvm-pdbdump/LLVMBuild.txt -----------------------*- Conf -*--===;
;
; The LLVM Compiler Infrastructure
;
; This file is distributed under the University of Illinois Open Source
; License. See LICENSE.TXT for details.
;
;===------------------------------------------------------------------------===;
;
; This is an LLVMBuild description file for the components in this subdirectory.
;
; For more information on the LLVMBuild system, please see:
;
; http://llvm.org/docs/LLVMBuild.html
;
;===------------------------------------------------------------------------===;
[component_0]
type = Tool
name = llvm-pdbdump
parent = Tools
required_libraries =

View File

@ -0,0 +1,155 @@
//===- llvm-pdbdump.cpp - Dump debug info from a PDB file -------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// Dumps debug information present in PDB files. This utility makes use of
// the Microsoft Windows SDK, so will not compile or run on non-Windows
// platforms.
//
//===----------------------------------------------------------------------===//
#define NTDDI_VERSION NTDDI_VISTA
#define _WIN32_WINNT _WIN32_WINNT_VISTA
#define WINVER _WIN32_WINNT_VISTA
#ifndef NOMINMAX
#define NOMINMAX
#endif
#include <atlbase.h>
#include <windows.h>
#include <dia2.h>
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/ConvertUTF.h"
#include "llvm/Support/Format.h"
#include "llvm/Support/ManagedStatic.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Support/Process.h"
#include "llvm/Support/PrettyStackTrace.h"
#include "llvm/Support/Signals.h"
#include "COMExtras.h"
using namespace llvm;
using namespace llvm::sys::windows;
namespace opts {
cl::list<std::string> InputFilenames(cl::Positional,
cl::desc("<input PDB files>"),
cl::OneOrMore);
cl::opt<bool> Streams("streams", cl::desc("Display data stream information"));
cl::alias StreamsShort("s", cl::desc("Alias for --streams"),
cl::aliasopt(Streams));
cl::opt<bool> StreamData("stream-data",
cl::desc("Dumps stream record data as bytes"));
cl::alias StreamDataShort("S", cl::desc("Alias for --stream-data"),
cl::aliasopt(StreamData));
}
namespace {
bool BSTRToUTF8(BSTR String16, std::string &String8) {
UINT ByteLength = ::SysStringByteLen(String16);
char *Bytes = reinterpret_cast<char *>(String16);
String8.clear();
return llvm::convertUTF16ToUTF8String(ArrayRef<char>(Bytes, ByteLength),
String8);
}
}
static void dumpDataStreams(IDiaSession *session) {
CComPtr<IDiaEnumDebugStreams> DebugStreams = nullptr;
if (FAILED(session->getEnumDebugStreams(&DebugStreams)))
return;
LONG Count = 0;
if (FAILED(DebugStreams->get_Count(&Count)))
return;
outs() << "Data Streams [count=" << Count << "]\n";
std::string Name8;
for (auto Stream : make_com_enumerator(DebugStreams)) {
BSTR Name16;
if (FAILED(Stream->get_name(&Name16)))
continue;
if (BSTRToUTF8(Name16, Name8))
outs() << " " << Name8;
::SysFreeString(Name16);
if (FAILED(Stream->get_Count(&Count))) {
outs() << "\n";
continue;
}
outs() << " [" << Count << " records]\n";
if (opts::StreamData) {
int RecordIndex = 0;
for (auto StreamRecord : make_com_data_record_enumerator(Stream)) {
outs() << " Record " << RecordIndex << " [" << StreamRecord.size()
<< " bytes]";
for (uint8_t byte : StreamRecord) {
outs() << " " << llvm::format_hex_no_prefix(byte, 2, true);
}
outs() << "\n";
++RecordIndex;
}
}
}
outs().flush();
}
static void dumpInput(StringRef Path) {
SmallVector<UTF16, 128> Path16String;
llvm::convertUTF8ToUTF16String(Path, Path16String);
wchar_t *Path16 = reinterpret_cast<wchar_t *>(Path16String.data());
CComPtr<IDiaDataSource> source;
HRESULT hr =
::CoCreateInstance(CLSID_DiaSource, nullptr, CLSCTX_INPROC_SERVER,
__uuidof(IDiaDataSource), (void **)&source);
if (FAILED(hr))
return;
if (FAILED(source->loadDataFromPdb(Path16)))
return;
CComPtr<IDiaSession> session;
if (FAILED(source->openSession(&session)))
return;
if (opts::Streams || opts::StreamData) {
dumpDataStreams(session);
}
}
int main(int argc_, const char *argv_[]) {
// Print a stack trace if we signal out.
sys::PrintStackTraceOnErrorSignal();
PrettyStackTraceProgram X(argc_, argv_);
SmallVector<const char *, 256> argv;
llvm::SpecificBumpPtrAllocator<char> ArgAllocator;
std::error_code EC = llvm::sys::Process::GetArgumentVector(
argv, llvm::makeArrayRef(argv_, argc_), ArgAllocator);
if (EC) {
llvm::errs() << "error: couldn't get arguments: " << EC.message() << '\n';
return 1;
}
llvm_shutdown_obj Y; // Call llvm_shutdown() on exit.
cl::ParseCommandLineOptions(argv.size(), argv.data(), "LLVM PDB Dumper\n");
CoInitializeEx(nullptr, COINIT_MULTITHREADED);
std::for_each(opts::InputFilenames.begin(), opts::InputFilenames.end(),
dumpInput);
CoUninitialize();
return 0;
}