folly::symbolizer: add support for DWARF5

Summary:
folly::symbolizer now supports both DWARF4 and DWARF5 debug info
- http://www.dwarfstd.org/doc/DWARF4.pdf
- http://www.dwarfstd.org/doc/DWARF5.pdf

Split DWARF (Debug Fission)
- with `-fsplit-dwarf-inlining`: folly::symbolizer will use the debug info from the skeleton CU.
- without `-fsplit-dwarf-inlining`: folly::symbolizer won't read .dwo sections and can't symbolize it (yet).

Major things that changed between the standards that affect folly::symbolizer:
- new forms DW_FORM_addrx*, DW_FORM_loclistx, DW_FORM_rnglistx, DW_FORM_strx*, whose values are interpreted against new CU attributes DW_AT_addr_base, DW_AT_loclists_base, DW_AT_rnglists_base, DW_AT_str_offsets_base
- The .debug_line line header format is completely re-written and the constraint that the file register must be >= 1 was lifted.
- DWARF5 `.debug_loclists` replaces DWARF4 `.debug_ranges`

Differential Revision: D25521092

fbshipit-source-id: 5b48e8abed495f379d8b119177913f55ae11bd2b

folly/experimental/symbolizer/Dwarf.h

@@ -165,6 +165,9 @@ class Dwarf {
   /** Get an ELF section by name. */
   folly::StringPiece getSection(const char* name) const;
 
+  detail::CompilationUnit getCompilationUnit(uint64_t offset) const;
+  detail::CompilationUnit findCompilationUnit(uint64_t targetOffset) const;
+
   /** cu must exist during the life cycle of created detail::Die. */
   detail::Die getDieAtOffset(
       const detail::CompilationUnit& cu, uint64_t offset) const;
@@ -191,6 +194,12 @@ class Dwarf {
    */
   detail::DIEAbbreviation getAbbreviation(uint64_t code, uint64_t offset) const;
 
+  detail::Attribute readAttribute(
+      const detail::CompilationUnit& cu,
+      const detail::Die& die,
+      detail::AttributeSpec spec,
+      folly::StringPiece& info) const;
+
   /**
    * Iterates over all attributes of a debugging info entry,  calling the given
    * callable for each. If all attributes are visited, then return the offset of
@@ -212,18 +221,24 @@ class Dwarf {
    * .debug_ranges.
    */
   bool isAddrInRangeList(
+      const detail::CompilationUnit& cu,
       uint64_t address,
       folly::Optional<uint64_t> baseAddr,
       size_t offset,
       uint8_t addrSize) const;
 
   const ElfFile* elf_;
-  const folly::StringPiece debugInfo_; // .debug_info
   const folly::StringPiece debugAbbrev_; // .debug_abbrev
-  const folly::StringPiece debugLine_; // .debug_line
-  const folly::StringPiece debugStr_; // .debug_str
+  const folly::StringPiece debugAddr_; // .debug_addr (DWARF 5)
   const folly::StringPiece debugAranges_; // .debug_aranges
+  const folly::StringPiece debugInfo_; // .debug_info
+  const folly::StringPiece debugLine_; // .debug_line
+  const folly::StringPiece debugLineStr_; // .debug_line_str (DWARF 5)
+  const folly::StringPiece debugLoclists_; // .debug_loclists (DWARF 5)
   const folly::StringPiece debugRanges_; // .debug_ranges
+  const folly::StringPiece debugRnglists_; // .debug_rnglists (DWARF 5)
+  const folly::StringPiece debugStr_; // .debug_str
+  const folly::StringPiece debugStrOffsets_; // .debug_str_offsets (DWARF 5)
 };
 
 class Dwarf::Section {
@@ -250,15 +265,15 @@ class Dwarf::Section {
 class Dwarf::LineNumberVM {
  public:
   LineNumberVM(
-      folly::StringPiece data, folly::StringPiece compilationDirectory);
+      folly::StringPiece data,
+      folly::StringPiece compilationDirectory,
+      folly::StringPiece debugStr,
+      folly::StringPiece debugLineStr);
 
   bool findAddress(uintptr_t target, Path& file, uint64_t& line);
 
   /** Gets full file name at given index including directory. */
-  Path getFullFileName(uint64_t index) const {
-    auto fn = getFileName(index);
-    return Path({}, getIncludeDirectory(fn.directoryIndex), fn.relativeName);
-  }
+  Path getFullFileName(uint64_t index) const;
 
  private:
   void init();
@@ -304,6 +319,8 @@ class Dwarf::LineNumberVM {
   bool is64Bit_;
   folly::StringPiece data_;
   folly::StringPiece compilationDirectory_;
+  folly::StringPiece debugStr_; // needed for DWARF 5
+  folly::StringPiece debugLineStr_; // DWARF 5
 
   // Header
   uint16_t version_;
@@ -314,11 +331,25 @@ class Dwarf::LineNumberVM {
   uint8_t opcodeBase_;
   const uint8_t* standardOpcodeLengths_;
 
-  folly::StringPiece includeDirectories_;
-  size_t includeDirectoryCount_;
-
-  folly::StringPiece fileNames_;
-  size_t fileNameCount_;
+  // 6.2.4 The Line Number Program Header.
+  struct {
+    size_t includeDirectoryCount;
+    folly::StringPiece includeDirectories;
+    size_t fileNameCount;
+    folly::StringPiece fileNames;
+  } v4_;
+
+  struct {
+    uint8_t directoryEntryFormatCount;
+    folly::StringPiece directoryEntryFormat;
+    uint64_t directoriesCount;
+    folly::StringPiece directories;
+
+    uint8_t fileNameEntryFormatCount;
+    folly::StringPiece fileNameEntryFormat;
+    uint64_t fileNamesCount;
+    folly::StringPiece fileNames;
+  } v5_;
 
   // State machine registers
   uint64_t address_;

folly/experimental/symbolizer/Elf.h

@@ -261,6 +261,8 @@ class ElfFile {
   /** Find the section containing the given address */
   const ElfShdr* getSectionContainingAddress(ElfAddr addr) const noexcept;
 
+  const char* filepath() const { return filepath_; }
+
  private:
   OpenResult init() noexcept;
   void reset() noexcept;

folly/experimental/symbolizer/test/SymbolizerTest.cpp

@@ -151,9 +151,12 @@ void expectFrameEq(
   auto demangled = folly::demangle(frame.name);
   EXPECT_TRUE(demangled == shortName || demangled == fullName)
       << "Found: demangled=" << demangled
-      << " expecting shortName=" << shortName << " or fullName=" << fullName;
+      << " expecting shortName=" << shortName << " or fullName=" << fullName
+      << " address: " << frame.addr << " hex(address): " << std::hex
+      << frame.addr;
   EXPECT_EQ(normalizePath(frame.location.file.toString()), normalizePath(file))
-      << ' ' << fullName;
+      << ' ' << fullName << " address: " << frame.addr
+      << " hex(address): " << std::hex << frame.addr;
   EXPECT_EQ(frame.location.line, lineno) << ' ' << fullName;
 }
 
@@ -474,6 +477,12 @@ TEST(Dwarf, FindParameterNames) {
       extraInlineFrames,
       [&](const folly::StringPiece name) { names.push_back(name); });
 
+  if (names.empty()) {
+    // When using -fsplit-dwarf-inlining info about parameters will not be
+    // emitted for the inlined debug info.
+    return;
+  }
+
   ASSERT_EQ(2, names.size());
   ASSERT_EQ("a", names[0]);
   ASSERT_EQ("b", names[1]);