This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [patch 6/7] [gold] Add initial source code for dwp utility.


Here's a revised patch, addressing all of your comments except the
base class to use for Sized_relobj_dwo. After thinking about it some
more, I'm now thinking multiple inheritance may not be necessary --
it may just be a simple matter of introducing a new abstract base
(say, Simple_elf_object) above Object, and promoting the basic interfaces
needed for dwp.cc and dwarf_reader.cc up to that class. I can then rename
Sized_relobj_dwo to something like Sized_simple_elf_object, and derive
that from Simple_elf_object. I'll play a bit with that idea, and
follow-up if that pans out.

I checked that bin_PROGRAMS does in fact pass the program name through
$(transform).

-cary


gold/
	* dwp.h: New header file.
	* dwp.cc: New source file.
	* gold.h: Move shared declarations to system.h.
	* system.h: New header file.
	* Makefile.am: Add dwp.
	* Makefile.in: Regenerate.
---
 gold/Makefile.am |    8 +
 gold/Makefile.in |  103 ++-
 gold/dwp.cc      | 2077 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 gold/dwp.h       |  120 ++++
 gold/gold.h      |  132 +---
 gold/system.h    |  158 +++++
 6 files changed, 2445 insertions(+), 153 deletions(-)
 create mode 100644 gold/dwp.cc
 create mode 100644 gold/dwp.h
 create mode 100644 gold/system.h

diff --git a/gold/Makefile.am b/gold/Makefile.am
index aa8caca..f82d4e9 100644
--- a/gold/Makefile.am
+++ b/gold/Makefile.am
@@ -34,6 +34,8 @@ AM_YFLAGS = -d
 am__skiplex =
 am__skipyacc =
 
+bin_PROGRAMS = dwp
+
 noinst_PROGRAMS = ld-new incremental-dump
 noinst_LIBRARIES = libgold.a
 
@@ -172,6 +174,12 @@ incremental_dump_DEPENDENCIES = $(TARGETOBJS) libgold.a $(LIBIBERTY) \
 incremental_dump_LDADD = $(TARGETOBJS) libgold.a $(LIBIBERTY) $(LIBINTL) \
 	 $(THREADSLIB) $(LIBDL)
 
+dwp_SOURCES = dwp.cc
+dwp_DEPENDENCIES = libgold.a $(LIBIBERTY) $(LIBINTL_DEP)
+dwp_LDADD = libgold.a $(LIBIBERTY) $(GOLD_LDADD) $(LIBINTL) $(THREADSLIB) \
+	$(LIBDL)
+dwp_LDFLAGS = $(GOLD_LDFLAGS)
+
 # Use an explicit dependency for the bison generated header file.
 expression.$(OBJEXT): yyscript.h
 script-sections.$(OBJEXT): yyscript.h
diff --git a/gold/dwp.cc b/gold/dwp.cc
new file mode 100644
index 0000000..3f52325
--- /dev/null
+++ b/gold/dwp.cc
@@ -0,0 +1,2077 @@
+// dwp.cc -- DWARF packaging utility
+
+// Copyright 2012 Free Software Foundation, Inc.
+// Written by Cary Coutant <ccoutant@google.com>.
+
+// This file is part of dwp, the DWARF packaging utility.
+
+// This program is free software; you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation; either version 3 of the License, or
+// (at your option) any later version.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with this program; if not, write to the Free Software
+// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+// MA 02110-1301, USA.
+
+#include "dwp.h"
+
+#include <cstdarg>
+#include <cstddef>
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <cerrno>
+
+#include <vector>
+#include <algorithm>
+
+#include "getopt.h"
+#include "libiberty.h"
+
+#include "elfcpp.h"
+#include "elfcpp_file.h"
+#include "dirsearch.h"
+#include "fileread.h"
+#include "object.h"
+#include "compressed_output.h"
+#include "stringpool.h"
+#include "dwarf_reader.h"
+
+static void
+usage() ATTRIBUTE_NORETURN;
+
+namespace gold {
+
+class Dwp_output_file;
+
+// An input file.
+// This class may represent either a .dwo file or a .dwp file
+// produced by an earlier run.
+
+template <int size, bool big_endian>
+class Sized_relobj_dwo;
+
+class Dwo_file
+{
+ public:
+  Dwo_file(const char* name)
+    : name_(name), obj_(NULL), input_file_(NULL), is_compressed_(),
+      str_offset_map_()
+  { }
+
+  ~Dwo_file();
+
+  // Read the input file and send its contents to OUTPUT_FILE.
+  void
+  read(Dwp_output_file* output_file);
+
+ private:
+  // Types for mapping input string offsets to output string offsets.
+  typedef std::pair<section_offset_type, section_offset_type>
+      Str_offset_map_entry;
+  typedef std::vector<Str_offset_map_entry> Str_offset_map;
+
+  // A less-than comparison routine for Str_offset_map.
+  struct Offset_compare
+  {
+    bool
+    operator()(const Str_offset_map_entry& i1,
+	       const Str_offset_map_entry& i2) const
+    { return i1.first < i2.first; }
+  };
+
+  // Create a Sized_relobj_dwo of the given size and endianness,
+  // and record the target info.  P is a pointer to the ELF header
+  // in memory.
+  Relobj*
+  make_object(int size, bool big_endian, const unsigned char* p,
+	      Input_file* input_file, Dwp_output_file* output_file);
+
+  template <int size, bool big_endian>
+  Relobj*
+  sized_make_object(const unsigned char* p, Input_file* input_file,
+		    Dwp_output_file* output_file);
+
+  // Return the number of sections in the input object file.
+  unsigned int
+  shnum() const
+  { return this->obj_->shnum(); }
+
+  // Return section type.
+  unsigned int
+  section_type(unsigned int shndx)
+  { return this->obj_->section_type(shndx); }
+
+  // Get the name of a section.
+  std::string
+  section_name(unsigned int shndx)
+  { return this->obj_->section_name(shndx); }
+
+  // Return a view of the contents of a section, decompressed if necessary.
+  // Set *PLEN to the size.  Set *IS_NEW to true if the contents need to be
+  // deleted by the caller.
+  const unsigned char*
+  section_contents(unsigned int shndx, section_size_type* plen, bool* is_new)
+  { return this->obj_->decompressed_section_contents(shndx, plen, is_new); }
+
+  // Read the .debug_cu_index section of a .dwp file,
+  // and process the CU sets.
+  void
+  read_compunit_index(unsigned int, Dwp_output_file*);
+
+  template <bool big_endian>
+  void
+  sized_read_compunit_index(unsigned int, Dwp_output_file*);
+
+  // Read the .debug_tu_index section of a .dwp file,
+  // and process the TU sets.
+  void
+  read_typeunit_index(unsigned int, Dwp_output_file*);
+
+  template <bool big_endian>
+  void
+  sized_read_typeunit_index(unsigned int, Dwp_output_file*);
+
+  // Merge the input string table section into the output file.
+  void
+  add_strings(Dwp_output_file*, unsigned int);
+
+  // Copy a section from the input file to the output file.
+  unsigned int
+  copy_section(Dwp_output_file* output_file, unsigned int shndx,
+	       const char* section_name, bool is_str_offsets);
+
+  // Remap the string offsets in the .debug_str_offsets.dwo section.
+  const unsigned char*
+  remap_str_offsets(const unsigned char* contents, section_size_type len);
+
+  template <bool big_endian>
+  const unsigned char*
+  sized_remap_str_offsets(const unsigned char* contents, section_size_type len);
+
+  // Remap a single string offsets from an offset in the input string table
+  // to an offset in the output string table.
+  unsigned int
+  remap_str_offset(unsigned int val);
+
+  // Add a set of .debug_info and related sections to OUTPUT_FILE.
+  void
+  add_cu_set(Dwp_output_file* output_file,
+	     uint64_t dwo_id,
+	     unsigned int debug_info,
+	     unsigned int debug_abbrev,
+	     unsigned int debug_line,
+	     unsigned int debug_loc,
+	     unsigned int debug_str_offsets,
+	     unsigned int debug_macinfo,
+	     unsigned int debug_macro);
+
+  // Add a set of .debug_types and related sections to OUTPUT_FILE.
+  void
+  add_tu_set(Dwp_output_file* output_file,
+	     uint64_t type_sig,
+	     unsigned int debug_types,
+	     unsigned int debug_abbrev,
+	     unsigned int debug_line,
+	     unsigned int debug_str_offsets);
+
+  // The filename.
+  const char* name_;
+  // The ELF file, represented as a gold Relobj instance.
+  Relobj* obj_;
+  // The Input_file object.
+  Input_file* input_file_;
+  // Flags indicating which sections are compressed.
+  std::vector<bool> is_compressed_;
+  // Map input section index onto output section index.
+  std::vector<unsigned int> shndx_map_;
+  // Map input string offsets to output string offsets.
+  Str_offset_map str_offset_map_;
+};
+
+// An ELF input file.
+// We derive from Sized_relobj so that we can use interfaces
+// in libgold to access the file.
+
+template <int size, bool big_endian>
+class Sized_relobj_dwo : public Sized_relobj<size, big_endian>
+{
+ public:
+  typedef typename elfcpp::Elf_types<size>::Elf_Addr Address;
+  typedef typename Sized_relobj<size, big_endian>::Symbols Symbols;
+
+  Sized_relobj_dwo(const char* name, Input_file* input_file,
+		   const elfcpp::Ehdr<size, big_endian>& ehdr)
+    : Sized_relobj<size, big_endian>(name, input_file),
+      elf_file_(this, ehdr)
+  { }
+
+  ~Sized_relobj_dwo()
+  { }
+
+  // Setup the section information.
+  void
+  setup();
+
+ protected:
+  // Return section type.
+  unsigned int
+  do_section_type(unsigned int shndx)
+  { return this->elf_file_.section_type(shndx); }
+
+  // Get the name of a section.
+  std::string
+  do_section_name(unsigned int shndx)
+  { return this->elf_file_.section_name(shndx); }
+
+  // Get the size of a section.
+  uint64_t
+  do_section_size(unsigned int shndx)
+  { return this->elf_file_.section_size(shndx); }
+
+  // Return a view of the contents of a section.
+  const unsigned char*
+  do_section_contents(unsigned int, section_size_type*, bool);
+
+  // Return a view of the uncompressed contents of a section.  Set *PLEN
+  // to the size.  Set *IS_NEW to true if the contents need to be deleted
+  // by the caller.
+  const unsigned char*
+  do_decompressed_section_contents(unsigned int shndx,
+				   section_size_type* plen,
+				   bool* is_new);
+
+  // The following virtual functions are abstract in the base classes,
+  // but are not used here.
+
+  // Read the symbols.
+  void
+  do_read_symbols(Read_symbols_data*)
+  { gold_unreachable(); }
+
+  // Lay out the input sections.
+  void
+  do_layout(Symbol_table*, Layout*, Read_symbols_data*)
+  { gold_unreachable(); }
+
+  // Layout sections whose layout was deferred while waiting for
+  // input files from a plugin.
+  void
+  do_layout_deferred_sections(Layout*)
+  { gold_unreachable(); }
+
+  // Add the symbols to the symbol table.
+  void
+  do_add_symbols(Symbol_table*, Read_symbols_data*, Layout*)
+  { gold_unreachable(); }
+
+  Archive::Should_include
+  do_should_include_member(Symbol_table*, Layout*, Read_symbols_data*,
+                           std::string*)
+  { gold_unreachable(); }
+
+  // Iterate over global symbols, calling a visitor class V for each.
+  void
+  do_for_all_global_symbols(Read_symbols_data*,
+			    Library_base::Symbol_visitor_base*)
+  { gold_unreachable(); }
+
+  // Return section flags.
+  uint64_t
+  do_section_flags(unsigned int)
+  { gold_unreachable(); }
+
+  // Return section entsize.
+  uint64_t
+  do_section_entsize(unsigned int)
+  { gold_unreachable(); }
+
+  // Return section address.
+  uint64_t
+  do_section_address(unsigned int)
+  { gold_unreachable(); }
+
+  // Return the section link field.
+  unsigned int
+  do_section_link(unsigned int)
+  { gold_unreachable(); }
+
+  // Return the section link field.
+  unsigned int
+  do_section_info(unsigned int)
+  { gold_unreachable(); }
+
+  // Return the section alignment.
+  uint64_t
+  do_section_addralign(unsigned int)
+  { gold_unreachable(); }
+
+  // Return the Xindex structure to use.
+  Xindex*
+  do_initialize_xindex()
+  { gold_unreachable(); }
+
+  // Get symbol counts.
+  void
+  do_get_global_symbol_counts(const Symbol_table*, size_t*, size_t*) const
+  { gold_unreachable(); }
+
+  // Get global symbols.
+  const Symbols*
+  do_get_global_symbols() const
+  { return NULL; }
+
+  // Return the value of a local symbol.
+  uint64_t
+  do_local_symbol_value(unsigned int, uint64_t) const
+  { gold_unreachable(); }
+
+  unsigned int
+  do_local_plt_offset(unsigned int) const
+  { gold_unreachable(); }
+
+  // Return whether local symbol SYMNDX is a TLS symbol.
+  bool
+  do_local_is_tls(unsigned int) const
+  { gold_unreachable(); }
+
+  // Return the number of local symbols.
+  unsigned int
+  do_local_symbol_count() const
+  { gold_unreachable(); }
+
+  // Return the number of local symbols in the output symbol table.
+  unsigned int
+  do_output_local_symbol_count() const
+  { gold_unreachable(); }
+
+  // Return the file offset for local symbols in the output symbol table.
+  off_t
+  do_local_symbol_offset() const
+  { gold_unreachable(); }
+
+  // Read the relocs.
+  void
+  do_read_relocs(Read_relocs_data*)
+  { gold_unreachable(); }
+
+  // Process the relocs to find list of referenced sections. Used only
+  // during garbage collection.
+  void
+  do_gc_process_relocs(Symbol_table*, Layout*, Read_relocs_data*)
+  { gold_unreachable(); }
+
+  // Scan the relocs and adjust the symbol table.
+  void
+  do_scan_relocs(Symbol_table*, Layout*, Read_relocs_data*)
+  { gold_unreachable(); }
+
+  // Count the local symbols.
+  void
+  do_count_local_symbols(Stringpool_template<char>*,
+			 Stringpool_template<char>*)
+  { gold_unreachable(); }
+
+  // Finalize the local symbols.
+  unsigned int
+  do_finalize_local_symbols(unsigned int, off_t, Symbol_table*)
+  { gold_unreachable(); }
+
+  // Set the offset where local dynamic symbol information will be stored.
+  unsigned int
+  do_set_local_dynsym_indexes(unsigned int)
+  { gold_unreachable(); }
+
+  // Set the offset where local dynamic symbol information will be stored.
+  unsigned int
+  do_set_local_dynsym_offset(off_t)
+  { gold_unreachable(); }
+
+  // Relocate the input sections and write out the local symbols.
+  void
+  do_relocate(const Symbol_table*, const Layout*, Output_file*)
+  { gold_unreachable(); }
+
+ private:
+  // General access to the ELF file.
+  elfcpp::Elf_file<size, big_endian, Object> elf_file_;
+};
+
+// The output file.
+// This class is responsible for collecting the debug index information
+// and writing the .dwp file in ELF format.
+
+class Dwp_output_file
+{
+ public:
+  Dwp_output_file(const char* name)
+    : name_(name), machine_(0), size_(0), big_endian_(false), osabi_(0),
+      abiversion_(0), fd_(NULL), next_file_offset_(0), shnum_(1), sections_(),
+      shoff_(0), shstrndx_(0), have_strings_(false), stringpool_(),
+      shstrtab_(), cu_index_(), tu_index_(), last_type_sig_(0),
+      last_tu_slot_(0)
+  {
+    this->stringpool_.set_no_zero_null();
+  }
+
+  // Record the target info from an input file.
+  void
+  record_target_info(const char* name, int machine, int size, bool big_endian,
+		     int osabi, int abiversion);
+
+  // Add a string to the debug strings section.
+  section_offset_type
+  add_string(const char* str, size_t len);
+
+  // Add a section to the output file, and return the new section index.
+  unsigned int
+  add_section(const char* section_name, const unsigned char* contents,
+	      section_size_type len, int align);
+
+  // Add a set of .debug_info and related sections to the output file.
+  void
+  add_cu_set(uint64_t dwo_id, unsigned int debug_info,
+	     unsigned int debug_abbrev, unsigned int debug_line,
+	     unsigned int debug_loc, unsigned int debug_str_offsets,
+	     unsigned int debug_macinfo, unsigned int debug_macro);
+
+  // Lookup a type signature and return TRUE if we have already seen it.
+  bool
+  lookup_tu(uint64_t type_sig);
+
+  // Add a set of .debug_types and related sections to the output file.
+  void
+  add_tu_set(uint64_t type_sig, unsigned int debug_types,
+	     unsigned int debug_abbrev, unsigned int debug_line,
+	     unsigned int debug_str_offsets);
+
+  // Finalize the file, write the string tables and index sections,
+  // and close the file.
+  void
+  finalize();
+
+ private:
+  // Sections in the output file.
+  struct Section
+  {
+    const char* name;
+    off_t offset;
+    section_size_type size;
+    int align;
+  };
+
+  // A set of sections for a compilation unit or type unit.
+  struct Cu_or_tu_set
+  {
+    uint64_t signature;
+    unsigned int debug_info_or_types;
+    unsigned int debug_abbrev;
+    unsigned int debug_line;
+    unsigned int debug_loc;
+    unsigned int debug_str_offsets;
+    unsigned int debug_macinfo;
+    unsigned int debug_macro;
+  };
+
+  // The index sections defined by the DWARF Package File Format spec.
+  class Dwp_index
+  {
+   public:
+    // Vector for the section index pool.
+    typedef std::vector<unsigned int> Shndx_pool;
+
+    Dwp_index()
+      : capacity_(0), used_(0), hash_table_(NULL), shndx_pool_()
+    { }
+
+    ~Dwp_index()
+    { }
+
+    // Find a slot in the hash table for SIGNATURE.  Return TRUE
+    // if the entry already exists.
+    bool
+    find_or_add(uint64_t signature, unsigned int* slotp);
+
+    // Enter a CU or TU set at the given SLOT in the hash table.
+    void
+    enter_set(unsigned int slot, const Cu_or_tu_set& set);
+
+    // Return the contents of the given SLOT in the hash table of signatures.
+    uint64_t
+    hash_table(unsigned int slot) const
+    { return this->hash_table_[slot]; }
+
+    // Return the contents of the given SLOT in the parallel table of
+    // shndx pool indexes.
+    uint32_t
+    index_table(unsigned int slot) const
+    { return this->index_table_[slot]; }
+
+    // Return the total number of slots in the hash table.
+    unsigned int
+    hash_table_total_slots() const
+    { return this->capacity_; }
+
+    // Return the number of used slots in the hash table.
+    unsigned int
+    hash_table_used_slots() const
+    { return this->used_; }
+
+    // Return an iterator into the shndx pool.
+    Shndx_pool::const_iterator
+    shndx_pool() const
+    { return this->shndx_pool_.begin(); }
+
+    Shndx_pool::const_iterator
+    shndx_pool_end() const
+    { return this->shndx_pool_.end(); }
+
+    // Return the number of entries in the shndx pool.
+    unsigned int
+    shndx_pool_size() const
+    { return this->shndx_pool_.size(); }
+
+   private:
+    // Initialize the hash table.
+    void
+    initialize();
+
+    // Grow the hash table when we reach 2/3 capacity.
+    void
+    grow();
+
+    // The number of slots in the table, a power of 2 such that
+    // capacity > 3 * size / 2.
+    unsigned int capacity_;
+    // The current number of used slots in the hash table.
+    unsigned int used_;
+    // The storage for the hash table of signatures.
+    uint64_t* hash_table_;
+    // The storage for the parallel table of shndx pool indexes.
+    uint32_t* index_table_;
+    // The pool of section indexes.
+    Shndx_pool shndx_pool_;
+  };  // End class Dwp_output_file::Dwp_index.
+
+  // Initialize the output file.
+  void
+  initialize();
+
+  // Write the ELF header.
+  void
+  write_ehdr();
+
+  template<unsigned int size, bool big_endian>
+  void
+  sized_write_ehdr();
+
+  // Write a section header.
+  void
+  write_shdr(const char* name, unsigned int type, unsigned int flags,
+	     uint64_t addr, off_t offset, section_size_type sect_size,
+	     unsigned int link, unsigned int info,
+	     unsigned int align, unsigned int ent_size);
+
+  template<unsigned int size, bool big_endian>
+  void
+  sized_write_shdr(const char* name, unsigned int type, unsigned int flags,
+		   uint64_t addr, off_t offset, section_size_type sect_size,
+		   unsigned int link, unsigned int info,
+		   unsigned int align, unsigned int ent_size);
+
+  // Write a CU or TU index section.
+  template<bool big_endian>
+  void
+  write_index(const char* sect_name, const Dwp_index& index);
+
+  // The output filename.
+  const char* name_;
+  // ELF header parameters.
+  int machine_;
+  int size_;
+  int big_endian_;
+  int osabi_;
+  int abiversion_;
+  // The output file descriptor.
+  FILE* fd_;
+  // Next available file offset.
+  off_t next_file_offset_;
+  // The number of sections.
+  unsigned int shnum_;
+  // Section table. The first entry is shndx 1.
+  std::vector<Section> sections_;
+  // File offset of the section header table.
+  off_t shoff_;
+  // Section index of the section string table.
+  unsigned int shstrndx_;
+  // TRUE if we have added any strings to the string pool.
+  bool have_strings_;
+  // String pool for the output .debug_str.dwo section.
+  Stringpool stringpool_;
+  // String pool for the .shstrtab section.
+  Stringpool shstrtab_;
+  // The compilation unit index.
+  Dwp_index cu_index_;
+  // The type unit index.
+  Dwp_index tu_index_;
+  // Cache of the last type signature looked up.
+  uint64_t last_type_sig_;
+  // Cache of the slot index for the last type signature.
+  unsigned int last_tu_slot_;
+};
+
+// A specialization of Dwarf_info_reader, for reading dwo_ids and
+// type signatures from DWARF CUs and TUs.
+
+class Dwo_id_info_reader : public Dwarf_info_reader
+{
+ public:
+  Dwo_id_info_reader(bool is_type_unit,
+		     Relobj* object,
+		     const unsigned char* symbols,
+		     off_t symbols_size,
+		     unsigned int shndx,
+		     unsigned int reloc_shndx,
+		     unsigned int reloc_type)
+    : Dwarf_info_reader(is_type_unit, object, symbols, symbols_size, shndx,
+			reloc_shndx, reloc_type),
+      dwo_id_found_(false), dwo_id_(0), type_sig_found_(false), type_sig_(0)
+  { }
+
+  ~Dwo_id_info_reader()
+  { }
+
+  // Return the dwo_id from a DWARF compilation unit DIE in *DWO_ID.
+  bool
+  get_dwo_id(uint64_t* dwo_id)
+  {
+    this->parse();
+    if (!this->dwo_id_found_)
+      return false;
+    *dwo_id = this->dwo_id_;
+    return true;
+  }
+
+  // Return the type signature from a DWARF type unit DIE in *TYPE_SIG.
+  bool
+  get_type_sig(uint64_t* type_sig)
+  {
+    this->parse();
+    if (!this->type_sig_found_)
+      return false;
+    *type_sig = this->type_sig_;
+    return true;
+  }
+
+ protected:
+  // Visit a compilation unit.
+  virtual void
+  visit_compilation_unit(off_t cu_offset, off_t cu_length, Dwarf_die*);
+
+  // Visit a type unit.
+  virtual void
+  visit_type_unit(off_t tu_offset, off_t type_offset, uint64_t signature,
+		  Dwarf_die*);
+
+ private:
+  // Visit a top-level DIE.
+  void
+  visit_top_die(Dwarf_die* die);
+
+  // TRUE if we found a dwo_id.
+  bool dwo_id_found_;
+  // The dwo_id.
+  uint64_t dwo_id_;
+  // TRUE if we found a type signature.
+  bool type_sig_found_;
+  // The type signature.
+  uint64_t type_sig_;
+};
+
+// Class Sized_relobj_dwo.
+
+// Setup the section information.
+
+template <int size, bool big_endian>
+void
+Sized_relobj_dwo<size, big_endian>::setup()
+{
+  const unsigned int shnum = this->elf_file_.shnum();
+  this->set_shnum(shnum);
+  this->section_offsets().resize(shnum);
+}
+
+// Return a view of the contents of a section.
+
+template <int size, bool big_endian>
+const unsigned char*
+Sized_relobj_dwo<size, big_endian>::do_section_contents(
+    unsigned int shndx,
+    section_size_type* plen,
+    bool cache)
+{
+  Object::Location loc(this->elf_file_.section_contents(shndx));
+  *plen = convert_to_section_size_type(loc.data_size);
+  if (*plen == 0)
+    {
+      static const unsigned char empty[1] = { '\0' };
+      return empty;
+    }
+  return this->get_view(loc.file_offset, *plen, true, cache);
+}
+
+// Return a view of the uncompressed contents of a section.  Set *PLEN
+// to the size.  Set *IS_NEW to true if the contents need to be deleted
+// by the caller.
+
+template <int size, bool big_endian>
+const unsigned char*
+Sized_relobj_dwo<size, big_endian>::do_decompressed_section_contents(
+    unsigned int shndx,
+    section_size_type* plen,
+    bool* is_new)
+{
+  section_size_type buffer_size;
+  const unsigned char* buffer = this->do_section_contents(shndx, &buffer_size,
+							  false);
+
+  std::string sect_name = this->do_section_name(shndx);
+  if (!is_prefix_of(".zdebug_", sect_name.c_str()))
+    {
+      *plen = buffer_size;
+      *is_new = false;
+      return buffer;
+    }
+
+  section_size_type uncompressed_size = get_uncompressed_size(buffer,
+							      buffer_size);
+  unsigned char* uncompressed_data = new unsigned char[uncompressed_size];
+  if (!decompress_input_section(buffer,
+				buffer_size,
+				uncompressed_data,
+				uncompressed_size))
+    this->error(_("could not decompress section %s"),
+		this->section_name(shndx).c_str());
+  *plen = uncompressed_size;
+  *is_new = true;
+  return uncompressed_data;
+}
+
+// Class Dwo_file.
+
+Dwo_file::~Dwo_file()
+{
+  if (this->input_file_ != NULL)
+    delete this->input_file_;
+  if (this->obj_ != NULL)
+    delete this->obj_;
+}
+
+// Read the input file and send its contents to OUTPUT_FILE.
+
+void
+Dwo_file::read(Dwp_output_file* output_file)
+{
+  // Open the input file.
+  this->input_file_ = new Input_file(this->name_);
+  Dirsearch dirpath;
+  int index;
+  if (!this->input_file_->open(dirpath, NULL, &index))
+    gold_fatal(_("%s: can't open"), this->name_);
+  
+  // Check that it's an ELF file.
+  off_t filesize = this->input_file_->file().filesize();
+  int hdrsize = elfcpp::Elf_recognizer::max_header_size;
+  if (filesize < hdrsize)
+    hdrsize = filesize;
+  const unsigned char* p =
+      this->input_file_->file().get_view(0, 0, hdrsize, true, false);
+  if (!elfcpp::Elf_recognizer::is_elf_file(p, hdrsize))
+    gold_fatal(_("%s: not an ELF object file"), this->name_);
+  
+  // Get the size, endianness, machine, etc. info from the header,
+  // make an appropriately-sized Relobj, and pass the target info
+  // to the output object.
+  int size;
+  bool big_endian;
+  std::string error;
+  if (!elfcpp::Elf_recognizer::is_valid_header(p, hdrsize, &size,
+					       &big_endian, &error))
+    gold_fatal(_("%s: %s"), this->name_, error.c_str());
+
+  this->obj_ = this->make_object(size, big_endian, p, this->input_file_,
+				 output_file);
+
+  unsigned int shnum = this->shnum();
+  this->is_compressed_.resize(shnum);
+  this->shndx_map_.resize(shnum);
+
+  typedef std::vector<unsigned int> Types_list;
+  Types_list debug_types;
+  unsigned int debug_info = 0;
+  unsigned int debug_abbrev = 0;
+  unsigned int debug_line = 0;
+  unsigned int debug_loc = 0;
+  unsigned int debug_str = 0;
+  unsigned int debug_str_offsets = 0;
+  unsigned int debug_macinfo = 0;
+  unsigned int debug_macro = 0;
+  unsigned int debug_cu_index = 0;
+  unsigned int debug_tu_index = 0;
+
+  // Scan the section table and look for .dwp index sections.
+  // (Section index 0 is a dummy section; skip it.)
+  for (unsigned int i = 1; i < shnum; i++)
+    {
+      if (this->section_type(i) != elfcpp::SHT_PROGBITS)
+	continue;
+      std::string sect_name = this->section_name(i);
+      const char* suffix = sect_name.c_str();
+      if (is_prefix_of(".debug_", suffix))
+	suffix += 7;
+      else if (is_prefix_of(".zdebug_", suffix))
+	{
+	  this->is_compressed_[i] = true;
+	  suffix += 8;
+	}
+      else
+	continue;
+      if (strcmp(suffix, "cu_index") == 0)
+	debug_cu_index = i;
+      else if (strcmp(suffix, "tu_index") == 0)
+	debug_tu_index = i;
+      else if (strcmp(suffix, "str.dwo") == 0)
+	debug_str = i;
+    }
+
+  // Merge the input string table into the output string table.
+  this->add_strings(output_file, debug_str);
+
+  // If we found any .dwp index sections, read those and add the section
+  // sets to the output file.
+  if (debug_cu_index > 0 || debug_tu_index > 0)
+    {
+      if (debug_cu_index > 0)
+	this->read_compunit_index(debug_cu_index, output_file);
+      if (debug_tu_index > 0)
+	this->read_typeunit_index(debug_tu_index, output_file);
+      return;
+    }
+
+  // If we found no index sections, this is a .dwo file.
+  // Scan the section table and collect the debug sections.
+  for (unsigned int i = 1; i < shnum; i++)
+    {
+      if (this->section_type(i) != elfcpp::SHT_PROGBITS)
+	continue;
+      std::string sect_name = this->section_name(i);
+      const char* suffix = sect_name.c_str();
+      if (is_prefix_of(".debug_", suffix))
+	suffix += 7;
+      else if (is_prefix_of(".zdebug_", suffix))
+	suffix += 8;
+      else
+	continue;
+      // TODO: Check for one of each section (except .debug_types).
+      if (strcmp(suffix, "info.dwo") == 0)
+	debug_info = i;
+      else if (strcmp(suffix, "types.dwo") == 0)
+	debug_types.push_back(i);
+      else if (strcmp(suffix, "abbrev.dwo") == 0)
+	debug_abbrev = i;
+      else if (strcmp(suffix, "line.dwo") == 0)
+	debug_line = i;
+      else if (strcmp(suffix, "loc.dwo") == 0)
+	debug_loc = i;
+      else if (strcmp(suffix, "str_offsets.dwo") == 0)
+	debug_str_offsets = i;
+      else if (strcmp(suffix, "macinfo.dwo") == 0)
+	debug_macinfo = i;
+      else if (strcmp(suffix, "macro.dwo") == 0)
+	debug_macro = i;
+    }
+
+  if (debug_info > 0)
+    {
+      // Extract the dwo_id from .debug_info.dwo section.
+      uint64_t dwo_id;
+      Dwo_id_info_reader dwarf_reader(false, this->obj_, NULL, 0, debug_info,
+				      0, 0);
+      dwarf_reader.set_abbrev_shndx(debug_abbrev);
+      if (!dwarf_reader.get_dwo_id(&dwo_id))
+	gold_fatal(_("%s: .debug_info.dwo section does not have DW_AT_GNU_dwo_id "
+		     "attribute"), this->name_);
+      this->add_cu_set(output_file, dwo_id, debug_info, debug_abbrev,
+		       debug_line, debug_loc, debug_str_offsets,
+		       debug_macinfo, debug_macro);
+    }
+
+  for (Types_list::const_iterator tp = debug_types.begin();
+       tp != debug_types.end();
+       ++tp)
+    {
+      // Extract the type signature from .debug_types.dwo section.
+      uint64_t type_sig;
+      gold_assert(*tp > 0);
+      Dwo_id_info_reader dwarf_reader(true, this->obj_, NULL, 0, *tp, 0, 0);
+      dwarf_reader.set_abbrev_shndx(debug_abbrev);
+      if (!dwarf_reader.get_type_sig(&type_sig))
+	gold_fatal(_("%s: .debug_types.dwo section does not have type signature"),
+		   this->name_);
+      this->add_tu_set(output_file, type_sig, *tp, debug_abbrev, debug_line,
+		       debug_str_offsets);
+    }
+}
+
+// Create a Sized_relobj_dwo of the given size and endianness,
+// and record the target info.  P is a pointer to the ELF header
+// in memory.
+
+Relobj*
+Dwo_file::make_object(int size, bool big_endian, const unsigned char* p,
+		      Input_file* input_file, Dwp_output_file* output_file)
+{
+  if (size == 32)
+    {
+      if (big_endian)
+#ifdef HAVE_TARGET_32_BIG
+	return this->sized_make_object<32, true>(p, input_file, output_file);
+#else
+	gold_unreachable();
+#endif
+      else
+#ifdef HAVE_TARGET_32_LITTLE
+	return this->sized_make_object<32, false>(p, input_file, output_file);
+#else
+	gold_unreachable();
+#endif
+    }
+  else if (size == 64)
+    {
+      if (big_endian)
+#ifdef HAVE_TARGET_64_BIG
+	return this->sized_make_object<64, true>(p, input_file, output_file);
+#else
+	gold_unreachable();
+#endif
+      else
+#ifdef HAVE_TARGET_64_LITTLE
+	return this->sized_make_object<64, false>(p, input_file, output_file);
+#else
+	gold_unreachable();
+#endif
+    }
+  else
+    gold_unreachable();
+}
+
+// Function template to create a Sized_relobj_dwo and record the target info.
+// P is a pointer to the ELF header in memory.
+
+template <int size, bool big_endian>
+Relobj*
+Dwo_file::sized_make_object(const unsigned char* p, Input_file* input_file,
+			    Dwp_output_file* output_file)
+{
+  elfcpp::Ehdr<size, big_endian> ehdr(p);
+  Sized_relobj_dwo<size, big_endian>* obj =
+      new Sized_relobj_dwo<size, big_endian>(this->name_, input_file, ehdr);
+  obj->setup();
+  output_file->record_target_info(
+      this->name_, ehdr.get_e_machine(), size, big_endian,
+      ehdr.get_e_ident()[elfcpp::EI_OSABI],
+      ehdr.get_e_ident()[elfcpp::EI_ABIVERSION]);
+  return obj;
+}
+
+// Read the .debug_cu_index section of a .dwp file,
+// and process the CU sets.
+
+void
+Dwo_file::read_compunit_index(unsigned int shndx, Dwp_output_file* output_file)
+{
+  if (this->obj_->is_big_endian())
+    this->sized_read_compunit_index<true>(shndx, output_file);
+  else
+    this->sized_read_compunit_index<false>(shndx, output_file);
+}
+
+template <bool big_endian>
+void
+Dwo_file::sized_read_compunit_index(unsigned int shndx,
+				    Dwp_output_file* output_file)
+{
+  section_size_type len;
+  bool is_new;
+  const unsigned char* contents = this->section_contents(shndx, &len, &is_new);
+
+  unsigned int version =
+      elfcpp::Swap_unaligned<32, big_endian>::readval(contents);
+  if (version != 1)
+    gold_fatal(_("%s: .debug_cu_index has unsupported version number %d"),
+	       this->name_, version);
+
+  unsigned int nused =
+      elfcpp::Swap_unaligned<32, big_endian>::readval(contents
+						      + 2 * sizeof(uint32_t));
+  if (nused == 0)
+    return;
+
+  unsigned int nslots =
+      elfcpp::Swap_unaligned<32, big_endian>::readval(contents
+						      + 3 * sizeof(uint32_t));
+
+  const unsigned char* phash = contents + 4 * sizeof(uint32_t);
+  const unsigned char* pindex = phash + nslots * sizeof(uint64_t);
+  const unsigned char* shndx_pool = pindex + nslots * sizeof(uint32_t);
+  const unsigned char* limit = contents + len;
+
+  if (shndx_pool >= limit)
+    gold_fatal(_("%s: .debug_cu_index is corrupt"), this->name_);
+
+  // Loop over the slots of the hash table.
+  for (unsigned int i = 0; i < nslots; ++i)
+    {
+      uint64_t dwo_id =
+          elfcpp::Swap_unaligned<64, big_endian>::readval(phash);
+      if (dwo_id != 0)
+	{
+	  unsigned int index =
+	      elfcpp::Swap_unaligned<32, big_endian>::readval(pindex);
+	  const unsigned char* shndx_list =
+	      shndx_pool + index * sizeof(uint32_t);
+
+	  // Collect the debug sections for this compilation unit set.
+	  unsigned int debug_info = 0;
+	  unsigned int debug_abbrev = 0;
+	  unsigned int debug_line = 0;
+	  unsigned int debug_loc = 0;
+	  unsigned int debug_str_offsets = 0;
+	  unsigned int debug_macinfo = 0;
+	  unsigned int debug_macro = 0;
+	  for (;;)
+	    {
+	      if (shndx_list >= limit)
+		gold_fatal(_("%s: .debug_cu_index is corrupt"),
+			   this->name_);
+	      unsigned int shndx =
+		  elfcpp::Swap_unaligned<32, big_endian>::readval(shndx_list);
+	      if (shndx == 0)
+	        break;
+	      if (shndx > this->shnum())
+		gold_fatal(_("%s: .debug_cu_index has bad shndx"),
+			   this->name_);
+	      std::string sect_name = this->section_name(shndx);
+	      const char* suffix = sect_name.c_str();
+	      if (is_prefix_of(".debug_", suffix))
+		suffix += 7;
+	      else if (is_prefix_of(".zdebug_", suffix))
+		suffix += 8;
+	      else
+		gold_fatal(_("%s: .debug_cu_index refers to "
+			     "non-debug section"), this->name_);
+	      if (strcmp(suffix, "info.dwo") == 0)
+		debug_info = shndx;
+	      else if (strcmp(suffix, "abbrev.dwo") == 0)
+		debug_abbrev = shndx;
+	      else if (strcmp(suffix, "line.dwo") == 0)
+		debug_line = shndx;
+	      else if (strcmp(suffix, "loc.dwo") == 0)
+		debug_loc = shndx;
+	      else if (strcmp(suffix, "str_offsets.dwo") == 0)
+		debug_str_offsets = shndx;
+	      else if (strcmp(suffix, "macinfo.dwo") == 0)
+		debug_macinfo = shndx;
+	      else if (strcmp(suffix, "macro.dwo") == 0)
+		debug_macro = shndx;
+	      shndx_list += sizeof(uint32_t);
+	    }
+	  this->add_cu_set(output_file, dwo_id, debug_info, debug_abbrev,
+			   debug_line, debug_loc, debug_str_offsets,
+			   debug_macinfo, debug_macro);
+	}
+      phash += sizeof(uint64_t);
+      pindex += sizeof(uint32_t);
+    }
+
+  if (is_new)
+    delete[] contents;
+}
+
+// Read the .debug_tu_index section of a .dwp file,
+// and process the TU sets.
+
+void
+Dwo_file::read_typeunit_index(unsigned int shndx, Dwp_output_file* output_file)
+{
+  if (this->obj_->is_big_endian())
+    this->sized_read_typeunit_index<true>(shndx, output_file);
+  else
+    this->sized_read_typeunit_index<false>(shndx, output_file);
+}
+
+template <bool big_endian>
+void
+Dwo_file::sized_read_typeunit_index(unsigned int shndx,
+				    Dwp_output_file* output_file)
+{
+  section_size_type len;
+  bool is_new;
+  const unsigned char* contents = this->section_contents(shndx, &len, &is_new);
+
+  unsigned int version =
+      elfcpp::Swap_unaligned<32, big_endian>::readval(contents);
+  if (version != 1)
+    gold_fatal(_("%s: .debug_tu_index has unsupported version number %d"),
+	       this->name_, version);
+
+  unsigned int nused =
+      elfcpp::Swap_unaligned<32, big_endian>::readval(contents
+						      + 2 * sizeof(uint32_t));
+  if (nused == 0)
+    return;
+
+  unsigned int nslots =
+      elfcpp::Swap_unaligned<32, big_endian>::readval(contents
+						      + 3 * sizeof(uint32_t));
+
+  const unsigned char* phash = contents + 4 * sizeof(uint32_t);
+  const unsigned char* pindex = phash + nslots * sizeof(uint64_t);
+  const unsigned char* shndx_pool = pindex + nslots * sizeof(uint32_t);
+  const unsigned char* limit = contents + len;
+
+  if (shndx_pool >= limit)
+    gold_fatal(_("%s: .debug_tu_index is corrupt"), this->name_);
+
+  // Loop over the slots of the hash table.
+  for (unsigned int i = 0; i < nslots; ++i)
+    {
+      uint64_t type_sig =
+          elfcpp::Swap_unaligned<64, big_endian>::readval(phash);
+      if (type_sig != 0)
+	{
+	  unsigned int index =
+	      elfcpp::Swap_unaligned<32, big_endian>::readval(pindex);
+	  const unsigned char* shndx_list =
+	      shndx_pool + index * sizeof(uint32_t);
+
+	  // Collect the debug sections for this type unit set.
+	  unsigned int debug_types = 0;
+	  unsigned int debug_abbrev = 0;
+	  unsigned int debug_line = 0;
+	  unsigned int debug_str_offsets = 0;
+	  for (;;)
+	    {
+	      if (shndx_list >= limit)
+		gold_fatal(_("%s: .debug_tu_index is corrupt"),
+			   this->name_);
+	      unsigned int shndx =
+		  elfcpp::Swap_unaligned<32, big_endian>::readval(shndx_list);
+	      if (shndx == 0)
+	        break;
+	      if (shndx > this->shnum())
+		gold_fatal(_("%s: .debug_tu_index has bad shndx"),
+			   this->name_);
+	      std::string sect_name = this->section_name(shndx);
+	      const char* suffix = sect_name.c_str();
+	      if (is_prefix_of(".debug_", suffix))
+		suffix += 7;
+	      else if (is_prefix_of(".zdebug_", suffix))
+		suffix += 8;
+	      else
+		gold_fatal(_("%s: .debug_tu_index refers to "
+			     "non-debug section"), this->name_);
+	      if (strcmp(suffix, "types.dwo") == 0)
+		debug_types = shndx;
+	      else if (strcmp(suffix, "abbrev.dwo") == 0)
+		debug_abbrev = shndx;
+	      else if (strcmp(suffix, "line.dwo") == 0)
+		debug_line = shndx;
+	      else if (strcmp(suffix, "str_offsets.dwo") == 0)
+		debug_str_offsets = shndx;
+	      shndx_list += sizeof(uint32_t);
+	    }
+	  this->add_tu_set(output_file, type_sig, debug_types, debug_abbrev,
+			   debug_line, debug_str_offsets);
+	}
+      phash += sizeof(uint64_t);
+      pindex += sizeof(uint32_t);
+    }
+
+  if (is_new)
+    delete[] contents;
+}
+
+// Merge the input string table section into the output file.
+
+void
+Dwo_file::add_strings(Dwp_output_file* output_file, unsigned int debug_str)
+{
+  section_size_type len;
+  bool is_new;
+  const unsigned char* pdata = this->section_contents(debug_str, &len, &is_new);
+  const char* p = reinterpret_cast<const char*>(pdata);
+  const char* pend = p + len;
+
+  // Check that the last string is null terminated.
+  if (pend[-1] != '\0')
+    gold_fatal(_("%s: last entry in string section '%s' "
+		 "is not null terminated"),
+	       this->name_,
+	       this->section_name(debug_str).c_str());
+
+  // Count the number of strings in the section, and size the map.
+  size_t count = 0;
+  for (const char* pt = p; pt < pend; pt += strlen(pt) + 1)
+    ++count;
+  this->str_offset_map_.reserve(count + 1);
+
+  // Add the strings to the output string table, and record the new offsets
+  // in the map.
+  section_offset_type i = 0;
+  section_offset_type new_offset;
+  while (p < pend)
+    {
+      size_t len = strlen(p);
+      new_offset = output_file->add_string(p, len);
+      this->str_offset_map_.push_back(std::make_pair(i, new_offset));
+      p += len + 1;
+      i += len + 1;
+    }
+  new_offset = 0;
+  this->str_offset_map_.push_back(std::make_pair(i, new_offset));
+  if (is_new)
+    delete[] pdata;
+}
+
+// Copy a section from the input file to the output file.
+// If IS_STR_OFFSETS is true, remap the string offsets for the
+// output string table.
+
+unsigned int
+Dwo_file::copy_section(Dwp_output_file* output_file, unsigned int shndx,
+		       const char* section_name, bool is_str_offsets)
+{
+  // Some sections may be referenced from more than one set.
+  // Don't copy a section more than once.
+  if (this->shndx_map_[shndx] > 0)
+    return this->shndx_map_[shndx];
+
+  section_size_type len;
+  bool is_new;
+  const unsigned char* contents = this->section_contents(shndx, &len, &is_new);
+
+  if (is_str_offsets)
+    {
+      const unsigned char* remapped = this->remap_str_offsets(contents, len);
+      if (is_new)
+	delete[] contents;
+      contents = remapped;
+      is_new = true;
+    }
+
+  this->shndx_map_[shndx] = output_file->add_section(section_name, contents,
+						     len, 1);
+  if (is_new)
+    delete[] contents;
+
+  return this->shndx_map_[shndx];
+}
+
+// Remap the 
+const unsigned char*
+Dwo_file::remap_str_offsets(const unsigned char* contents,
+			    section_size_type len)
+{
+  if ((len & 3) != 0)
+    gold_fatal(_("%s: .debug_str_offsets.dwo section size not a multiple of 4"),
+	       this->name_);
+
+  if (this->obj_->is_big_endian())
+    return this->sized_remap_str_offsets<true>(contents, len);
+  else
+    return this->sized_remap_str_offsets<false>(contents, len);
+}
+
+template <bool big_endian>
+const unsigned char*
+Dwo_file::sized_remap_str_offsets(const unsigned char* contents,
+				  section_size_type len)
+{
+  unsigned char* remapped = new unsigned char[len];
+  const unsigned char* p = contents;
+  unsigned char* q = remapped;
+  while (len > 0)
+    {
+      unsigned int val = elfcpp::Swap_unaligned<32, big_endian>::readval(p);
+      val = this->remap_str_offset(val);
+      elfcpp::Swap_unaligned<32, big_endian>::writeval(q, val);
+      len -= 4;
+      p += 4;
+      q += 4;
+    }
+  return remapped;
+}
+
+unsigned int
+Dwo_file::remap_str_offset(unsigned int val)
+{
+  Str_offset_map_entry entry;
+  entry.first = val;
+
+  Str_offset_map::const_iterator p =
+      std::lower_bound(this->str_offset_map_.begin(),
+		       this->str_offset_map_.end(),
+		       entry, Offset_compare());
+
+  if (p == this->str_offset_map_.end() || p->first > val)
+    {
+      if (p == this->str_offset_map_.begin())
+	return 0;
+      --p;
+      gold_assert(p->first <= val);
+    }
+
+  return p->second + (val - p->first);
+}
+
+// Add a set of .debug_info and related sections to OUTPUT_FILE.
+
+void
+Dwo_file::add_cu_set(Dwp_output_file* output_file,
+		     uint64_t dwo_id,
+		     unsigned int debug_info,
+		     unsigned int debug_abbrev,
+		     unsigned int debug_line,
+		     unsigned int debug_loc,
+		     unsigned int debug_str_offsets,
+		     unsigned int debug_macinfo,
+		     unsigned int debug_macro)
+{
+  if (debug_info == 0)
+    gold_fatal(_("%s: no .debug_info.dwo section found"), this->name_);
+  if (debug_abbrev == 0)
+    gold_fatal(_("%s: no .debug_abbrev.dwo section found"), this->name_);
+
+  debug_abbrev = this->copy_section(output_file, debug_abbrev,
+				    ".debug_abbrev.dwo", false);
+  if (debug_line > 0)
+    debug_line = this->copy_section(output_file, debug_line,
+				    ".debug_line.dwo", false);
+  if (debug_loc > 0)
+    debug_loc = this->copy_section(output_file, debug_loc, ".debug_loc.dwo",
+				   false);
+  if (debug_macinfo > 0)
+    debug_macinfo = this->copy_section(output_file, debug_macinfo,
+				       ".debug_macinfo.dwo", false);
+  if (debug_macro > 0)
+    debug_macro = this->copy_section(output_file, debug_macro,
+				     ".debug_macro.dwo", false);
+
+  if (debug_str_offsets > 0)
+    debug_str_offsets = this->copy_section(output_file, debug_str_offsets,
+					   ".debug_str_offsets.dwo", true);
+
+  debug_info = this->copy_section(output_file, debug_info, ".debug_info.dwo",
+				  false);
+
+  output_file->add_cu_set(dwo_id, debug_info, debug_abbrev, debug_line,
+			  debug_loc, debug_str_offsets, debug_macinfo,
+			  debug_macro);
+}
+
+// Add a set of .debug_types and related sections to OUTPUT_FILE.
+
+void
+Dwo_file::add_tu_set(Dwp_output_file* output_file,
+		     uint64_t type_sig,
+		     unsigned int debug_types,
+		     unsigned int debug_abbrev,
+		     unsigned int debug_line,
+		     unsigned int debug_str_offsets)
+{
+  if (debug_types == 0)
+    gold_fatal(_("%s: no .debug_types.dwo section found"), this->name_);
+  if (debug_abbrev == 0)
+    gold_fatal(_("%s: no .debug_abbrev.dwo section found"), this->name_);
+
+  // Ignore duplicate type signatures.
+  if (output_file->lookup_tu(type_sig))
+    return;
+
+  debug_abbrev = this->copy_section(output_file, debug_abbrev,
+				    ".debug_abbrev.dwo", false);
+  if (debug_line > 0)
+    debug_line = this->copy_section(output_file, debug_line,
+				    ".debug_line.dwo", false);
+
+  if (debug_str_offsets > 0)
+    debug_str_offsets = this->copy_section(output_file, debug_str_offsets,
+					   ".debug_str_offsets.dwo", true);
+
+  debug_types = this->copy_section(output_file, debug_types,
+				   ".debug_types.dwo", false);
+
+  output_file->add_tu_set(type_sig, debug_types, debug_abbrev, debug_line,
+			  debug_str_offsets);
+}
+
+// Class Dwp_output_file.
+
+// Record the target info from an input file.  On first call, we
+// set the ELF header values for the output file.  On subsequent
+// calls, we just verify that the values match.
+
+void
+Dwp_output_file::record_target_info(const char*, int machine,
+				    int size, bool big_endian,
+				    int osabi, int abiversion)
+{
+  // TODO: Check the values on subsequent calls.
+  if (this->size_ > 0)
+    return;
+
+  this->machine_ = machine;
+  this->size_ = size;
+  this->big_endian_ = big_endian;
+  this->osabi_ = osabi;
+  this->abiversion_ = abiversion;
+
+  if (size == 32)
+    this->next_file_offset_ = elfcpp::Elf_sizes<32>::ehdr_size;
+  else if (size == 64)
+    this->next_file_offset_ = elfcpp::Elf_sizes<64>::ehdr_size;
+  else
+    gold_unreachable();
+
+  this->fd_ = ::fopen(this->name_, "wb");
+  if (this->fd_ == NULL)
+    gold_fatal(_("%s: %s"), this->name_, strerror(errno));
+
+  // Write zeroes for the ELF header initially.  We'll write
+  // the actual header during finalize().
+  static const char buf[elfcpp::Elf_sizes<64>::ehdr_size] = { 0 };
+  ::fwrite(buf, 1, this->next_file_offset_, this->fd_);
+}
+
+// Add a string to the debug strings section.
+
+section_offset_type
+Dwp_output_file::add_string(const char* str, size_t len)
+{
+  Stringpool::Key key;
+  this->stringpool_.add_with_length(str, len, true, &key);
+  this->have_strings_ = true;
+  // We aren't supposed to call get_offset() until after
+  // calling set_string_offsets(), but the offsets will
+  // not change unless optimizing the string pool.
+  return this->stringpool_.get_offset_from_key(key);
+}
+
+// Align the file offset to the given boundary.
+
+static inline off_t
+align_offset(off_t off, int align)
+{
+  return (off + align - 1) & ~(align - 1);
+}
+
+// Add a section to the output file, and return the new section index.
+
+unsigned int
+Dwp_output_file::add_section(const char* section_name,
+			     const unsigned char* contents,
+			     section_size_type len,
+			     int align)
+{
+  off_t file_offset = this->next_file_offset_;
+  gold_assert(this->size_ > 0 && file_offset > 0);
+
+  file_offset = align_offset(file_offset, align);
+
+  ::fseek(this->fd_, file_offset, SEEK_SET);
+  if (::fwrite(contents, 1, len, this->fd_) < len)
+    gold_fatal(_("%s: error writing section '%s'"), this->name_, section_name);
+
+  section_name = this->shstrtab_.add_with_length(section_name,
+						 strlen(section_name),
+						 false, NULL);
+  Section sect = { section_name, file_offset, len, align };
+  this->sections_.push_back(sect);
+
+  this->next_file_offset_ = file_offset + len;
+  return this->shnum_++;
+}
+
+// Add a set of .debug_info and related sections to the output file.
+
+void
+Dwp_output_file::add_cu_set(uint64_t dwo_id,
+			    unsigned int debug_info,
+			    unsigned int debug_abbrev,
+			    unsigned int debug_line,
+			    unsigned int debug_loc,
+			    unsigned int debug_str_offsets,
+			    unsigned int debug_macinfo,
+			    unsigned int debug_macro)
+{
+  Cu_or_tu_set cu_set = { dwo_id, debug_info, debug_abbrev, debug_line,
+			  debug_loc, debug_str_offsets, debug_macinfo,
+			  debug_macro };
+  unsigned int slot;
+  this->cu_index_.find_or_add(dwo_id, &slot);
+  this->cu_index_.enter_set(slot, cu_set);
+}
+
+// Lookup a type signature and return TRUE if we have already seen it.
+bool
+Dwp_output_file::lookup_tu(uint64_t type_sig)
+{
+  this->last_type_sig_ = type_sig;
+  return this->tu_index_.find_or_add(type_sig, &this->last_tu_slot_);
+}
+
+// Add a set of .debug_types and related sections to the output file.
+
+void
+Dwp_output_file::add_tu_set(uint64_t type_sig,
+			    unsigned int debug_types,
+			    unsigned int debug_abbrev,
+			    unsigned int debug_line,
+			    unsigned int debug_str_offsets)
+{
+  Cu_or_tu_set tu_set = { type_sig, debug_types, debug_abbrev, debug_line,
+			  0, debug_str_offsets, 0, 0 };
+  unsigned int slot;
+  if (type_sig == this->last_type_sig_)
+    slot = this->last_tu_slot_;
+  else
+    this->tu_index_.find_or_add(type_sig, &slot);
+  this->tu_index_.enter_set(slot, tu_set);
+}
+
+// Find a slot in the hash table for SIGNATURE.  Return TRUE
+// if the entry already exists.
+
+bool
+Dwp_output_file::Dwp_index::find_or_add(uint64_t signature,
+					unsigned int* slotp)
+{
+  if (this->capacity_ == 0)
+    this->initialize();
+  unsigned int slot =
+      static_cast<unsigned int>(signature) & (this->capacity_ - 1);
+  unsigned int secondary_hash;
+  uint64_t probe = this->hash_table_[slot];
+  if (probe != 0 && probe != signature)
+    {
+      secondary_hash = (static_cast<unsigned int>(signature >> 32)
+			& (this->capacity_ - 1)) | 1;
+      do
+	{
+	  slot = (slot + secondary_hash) & (this->capacity_ - 1);
+	  probe = this->hash_table_[slot];
+	} while (probe != 0 && probe != signature);
+    }
+  *slotp = slot;
+  return (probe != 0);
+}
+
+// Enter a CU or TU set at the given SLOT in the hash table.
+
+void
+Dwp_output_file::Dwp_index::enter_set(unsigned int slot,
+				      const Cu_or_tu_set& set)
+{
+  gold_assert(slot < this->capacity_);
+  gold_assert(set.debug_info_or_types > 0);
+  gold_assert(set.debug_abbrev > 0);
+
+  // Add the section indexes to the pool.
+  uint32_t pool_index = this->shndx_pool_.size();
+  this->shndx_pool_.push_back(set.debug_info_or_types);
+  this->shndx_pool_.push_back(set.debug_abbrev);
+  if (set.debug_line > 0)
+    this->shndx_pool_.push_back(set.debug_line);
+  if (set.debug_loc > 0)
+    this->shndx_pool_.push_back(set.debug_loc);
+  if (set.debug_str_offsets > 0)
+    this->shndx_pool_.push_back(set.debug_str_offsets);
+  if (set.debug_macinfo > 0)
+    this->shndx_pool_.push_back(set.debug_macinfo);
+  if (set.debug_macro > 0)
+    this->shndx_pool_.push_back(set.debug_macro);
+  this->shndx_pool_.push_back(0);
+
+  // Enter the signature and pool index into the hash table.
+  this->hash_table_[slot] = set.signature;
+  this->index_table_[slot] = pool_index;
+  ++this->used_;
+
+  // Grow the hash table when we exceed 2/3 capacity.
+  if (this->used_ * 3 > this->capacity_ * 2)
+    this->grow();
+}
+
+// Initialize the hash table.
+
+void
+Dwp_output_file::Dwp_index::initialize()
+{
+  this->capacity_ = 16;
+  this->hash_table_ = new uint64_t[this->capacity_];
+  memset(this->hash_table_, 0, this->capacity_ * sizeof(uint64_t));
+  this->index_table_ = new uint32_t[this->capacity_];
+  memset(this->index_table_, 0, this->capacity_ * sizeof(uint32_t));
+}
+
+// Grow the hash table when we reach 2/3 capacity.
+
+void
+Dwp_output_file::Dwp_index::grow()
+{
+  unsigned int old_capacity = this->capacity_;
+  uint64_t* old_hash_table = this->hash_table_;
+  uint32_t* old_index_table = this->index_table_;
+  unsigned int old_used = this->used_;
+
+  this->capacity_ = old_capacity * 2;
+  this->hash_table_ = new uint64_t[this->capacity_];
+  memset(this->hash_table_, 0, this->capacity_ * sizeof(uint64_t));
+  this->index_table_ = new uint32_t[this->capacity_];
+  memset(this->index_table_, 0, this->capacity_ * sizeof(uint32_t));
+  this->used_ = 0;
+
+  for (unsigned int i = 0; i < old_capacity; ++i)
+    {
+      uint64_t signature = old_hash_table[i];
+      if (signature != 0)
+        {
+	  unsigned int slot;
+	  bool found = this->find_or_add(signature, &slot);
+	  gold_assert(!found);
+	  this->hash_table_[slot] = signature;
+	  this->index_table_[slot] = old_index_table[i];
+	  ++this->used_;
+        }
+    }
+  gold_assert(this->used_ == old_used);
+
+  delete[] old_hash_table;
+  delete[] old_index_table;
+}
+
+// Initialize the output file.
+
+void
+Dwp_output_file::initialize()
+{
+  // We can't initialize the output file until we've recorded the
+  // target info from the first input file.
+  gold_assert(this->size_ > 0);
+}
+
+// Finalize the file, write the string tables and index sections,
+// and close the file.
+
+void
+Dwp_output_file::finalize()
+{
+  unsigned char* buf;
+
+  // Write the debug string table.
+  if (this->have_strings_)
+    {
+      this->stringpool_.set_string_offsets();
+      section_size_type len = this->stringpool_.get_strtab_size();
+      buf = new unsigned char[len];
+      this->stringpool_.write_to_buffer(buf, len);
+      this->add_section(".debug_str.dwo", buf, len, 1);
+      delete[] buf;
+    }
+
+  // Write the CU and TU indexes.
+  if (this->big_endian_)
+    {
+      this->write_index<true>(".debug_cu_index", this->cu_index_);
+      this->write_index<true>(".debug_tu_index", this->tu_index_);
+    }
+  else
+    {
+      this->write_index<false>(".debug_cu_index", this->cu_index_);
+      this->write_index<false>(".debug_tu_index", this->tu_index_);
+    }
+
+  off_t file_offset = this->next_file_offset_;
+
+  // Write the section string table.
+  this->shstrndx_ = this->shnum_++;
+  const char* shstrtab_name =
+      this->shstrtab_.add_with_length(".shstrtab",
+					   sizeof(".shstrtab") - 1,
+					   false, NULL);
+  this->shstrtab_.set_string_offsets();
+  section_size_type shstrtab_len = this->shstrtab_.get_strtab_size();
+  buf = new unsigned char[shstrtab_len];
+  this->shstrtab_.write_to_buffer(buf, shstrtab_len);
+  off_t shstrtab_off = file_offset;
+  ::fseek(this->fd_, file_offset, 0);
+  if (::fwrite(buf, 1, shstrtab_len, this->fd_) < shstrtab_len)
+    gold_fatal(_("%s: error writing section '.shstrtab'"), this->name_);
+  delete[] buf;
+  file_offset += shstrtab_len;
+
+  // Write the section header table.  The first entry is a NULL entry.
+  // This is followed by the debug sections, and finally we write the
+  // .shstrtab section header.
+  file_offset = align_offset(file_offset, this->size_ == 32 ? 4 : 8);
+  this->shoff_ = file_offset;
+  ::fseek(this->fd_, file_offset, 0);
+  section_size_type sh0_size = 0;
+  unsigned int sh0_link = 0;
+  if (this->shnum_ >= elfcpp::SHN_LORESERVE)
+    sh0_size = this->shnum_;
+  if (this->shstrndx_ >= elfcpp::SHN_LORESERVE)
+    sh0_link = this->shstrndx_;
+  this->write_shdr(NULL, 0, 0, 0, 0, sh0_size, sh0_link, 0, 0, 0);
+  for (unsigned int i = 0; i < this->sections_.size(); ++i)
+    {
+      Section& sect = this->sections_[i];
+      this->write_shdr(sect.name, elfcpp::SHT_PROGBITS, 0, 0, sect.offset,
+		       sect.size, 0, 0, sect.align, 0);
+    }
+  this->write_shdr(shstrtab_name, elfcpp::SHT_STRTAB, 0, 0,
+		   shstrtab_off, shstrtab_len, 0, 0, 1, 0);
+
+  // Write the ELF header.
+  this->write_ehdr();
+
+  // Close the file.
+  if (this->fd_ != NULL)
+    {
+      if (::fclose(this->fd_) != 0)
+	gold_fatal(_("%s: %s"), this->name_, strerror(errno));
+    }
+  this->fd_ = NULL;
+}
+
+// Write a CU or TU index section.
+template<bool big_endian>
+void
+Dwp_output_file::write_index(const char* sect_name, const Dwp_index& index)
+{
+  const unsigned int nslots = index.hash_table_total_slots();
+  const unsigned int nused = index.hash_table_used_slots();
+  const unsigned int npool = index.shndx_pool_size();
+  const section_size_type index_size = (4 * sizeof(uint32_t)
+					+ nslots * sizeof(uint64_t)
+					+ nslots * sizeof(uint32_t)
+					+ npool * sizeof(uint32_t));
+
+  // Allocate a buffer for the section contents.
+  unsigned char* buf = new unsigned char[index_size];
+  unsigned char* p = buf;
+
+  // Write the section header: version number, padding,
+  // number of used slots and total number of slots.
+  elfcpp::Swap_unaligned<32, big_endian>::writeval(p, 1);
+  p += sizeof(uint32_t);
+  elfcpp::Swap_unaligned<32, big_endian>::writeval(p, 0);
+  p += sizeof(uint32_t);
+  elfcpp::Swap_unaligned<32, big_endian>::writeval(p, nused);
+  p += sizeof(uint32_t);
+  elfcpp::Swap_unaligned<32, big_endian>::writeval(p, nslots);
+  p += sizeof(uint32_t);
+
+  // Write the hash table.
+  for (unsigned int i = 0; i < nslots; ++i)
+    {
+      elfcpp::Swap_unaligned<64, big_endian>::writeval(p, index.hash_table(i));
+      p += sizeof(uint64_t);
+    }
+
+  // Write the parallel index table.
+  for (unsigned int i = 0; i < nslots; ++i)
+    {
+      elfcpp::Swap_unaligned<32, big_endian>::writeval(p, index.index_table(i));
+      p += sizeof(uint32_t);
+    }
+
+  // Write the section index pool.
+  Dwp_index::Shndx_pool::const_iterator pool = index.shndx_pool();
+  for (unsigned int i = 0; i < npool; ++i)
+    {
+      gold_assert(pool != index.shndx_pool_end());
+      elfcpp::Swap_unaligned<32, big_endian>::writeval(p, *pool);
+      p += sizeof(uint32_t);
+      ++pool;
+    }
+
+  gold_assert(p == buf + index_size);
+
+  this->add_section(sect_name, buf, index_size, sizeof(uint64_t));
+
+  delete[] buf;
+}
+
+// Write the ELF header.
+
+void
+Dwp_output_file::write_ehdr()
+{
+  if (this->size_ == 32)
+    {
+      if (this->big_endian_)
+	return this->sized_write_ehdr<32, true>();
+      else
+	return this->sized_write_ehdr<32, false>();
+    }
+  else if (this->size_ == 64)
+    {
+      if (this->big_endian_)
+	return this->sized_write_ehdr<64, true>();
+      else
+	return this->sized_write_ehdr<64, false>();
+    }
+  else
+    gold_unreachable();
+}
+
+template<unsigned int size, bool big_endian>
+void
+Dwp_output_file::sized_write_ehdr()
+{
+  const unsigned int ehdr_size = elfcpp::Elf_sizes<size>::ehdr_size;
+  unsigned char buf[ehdr_size];
+  elfcpp::Ehdr_write<size, big_endian> ehdr(buf);
+
+  unsigned char e_ident[elfcpp::EI_NIDENT];
+  memset(e_ident, 0, elfcpp::EI_NIDENT);
+  e_ident[elfcpp::EI_MAG0] = elfcpp::ELFMAG0;
+  e_ident[elfcpp::EI_MAG1] = elfcpp::ELFMAG1;
+  e_ident[elfcpp::EI_MAG2] = elfcpp::ELFMAG2;
+  e_ident[elfcpp::EI_MAG3] = elfcpp::ELFMAG3;
+  if (size == 32)
+    e_ident[elfcpp::EI_CLASS] = elfcpp::ELFCLASS32;
+  else if (size == 64)
+    e_ident[elfcpp::EI_CLASS] = elfcpp::ELFCLASS64;
+  else
+    gold_unreachable();
+  e_ident[elfcpp::EI_DATA] = (big_endian
+			      ? elfcpp::ELFDATA2MSB
+			      : elfcpp::ELFDATA2LSB);
+  e_ident[elfcpp::EI_VERSION] = elfcpp::EV_CURRENT;
+  ehdr.put_e_ident(e_ident);
+
+  ehdr.put_e_type(elfcpp::ET_REL);
+  ehdr.put_e_machine(this->machine_);
+  ehdr.put_e_version(elfcpp::EV_CURRENT);
+  ehdr.put_e_entry(0);
+  ehdr.put_e_phoff(0);
+  ehdr.put_e_shoff(this->shoff_);
+  ehdr.put_e_flags(0);
+  ehdr.put_e_ehsize(elfcpp::Elf_sizes<size>::ehdr_size);
+  ehdr.put_e_phentsize(0);
+  ehdr.put_e_phnum(0);
+  ehdr.put_e_shentsize(elfcpp::Elf_sizes<size>::shdr_size);
+  ehdr.put_e_shnum(this->shnum_ < elfcpp::SHN_LORESERVE ? this->shnum_ : 0);
+  ehdr.put_e_shstrndx(this->shstrndx_ < elfcpp::SHN_LORESERVE
+		      ? this->shstrndx_
+		      : static_cast<unsigned int>(elfcpp::SHN_XINDEX));
+
+  ::fseek(this->fd_, 0, 0);
+  if (::fwrite(buf, 1, ehdr_size, this->fd_) < ehdr_size)
+    gold_fatal(_("%s: error writing ELF header"), this->name_);
+}
+
+// Write a section header.
+
+void
+Dwp_output_file::write_shdr(const char* name, unsigned int type,
+			    unsigned int flags, uint64_t addr, off_t offset,
+			    section_size_type sect_size, unsigned int link,
+			    unsigned int info, unsigned int align,
+			    unsigned int ent_size)
+{
+  if (this->size_ == 32)
+    {
+      if (this->big_endian_)
+	return this->sized_write_shdr<32, true>(name, type, flags, addr,
+						offset, sect_size, link, info,
+						align, ent_size);
+      else
+	return this->sized_write_shdr<32, false>(name, type, flags, addr,
+						 offset, sect_size, link, info,
+						 align, ent_size);
+    }
+  else if (this->size_ == 64)
+    {
+      if (this->big_endian_)
+	return this->sized_write_shdr<64, true>(name, type, flags, addr,
+						offset, sect_size, link, info,
+						align, ent_size);
+      else
+	return this->sized_write_shdr<64, false>(name, type, flags, addr,
+						 offset, sect_size, link, info,
+						 align, ent_size);
+    }
+  else
+    gold_unreachable();
+}
+
+template<unsigned int size, bool big_endian>
+void
+Dwp_output_file::sized_write_shdr(const char* name, unsigned int type,
+				  unsigned int flags, uint64_t addr,
+				  off_t offset, section_size_type sect_size,
+				  unsigned int link, unsigned int info,
+				  unsigned int align, unsigned int ent_size)
+{
+  const unsigned int shdr_size = elfcpp::Elf_sizes<size>::shdr_size;
+  unsigned char buf[shdr_size];
+  elfcpp::Shdr_write<size, big_endian> shdr(buf);
+
+  shdr.put_sh_name(name == NULL ? 0 : this->shstrtab_.get_offset(name));
+  shdr.put_sh_type(type);
+  shdr.put_sh_flags(flags);
+  shdr.put_sh_addr(addr);
+  shdr.put_sh_offset(offset);
+  shdr.put_sh_size(sect_size);
+  shdr.put_sh_link(link);
+  shdr.put_sh_info(info);
+  shdr.put_sh_addralign(align);
+  shdr.put_sh_entsize(ent_size);
+  if (::fwrite(buf, 1, shdr_size, this->fd_) < shdr_size)
+    gold_fatal(_("%s: error writing section header table"), this->name_);
+}
+
+// Class Dwo_id_info_reader.
+
+// Visit a compilation unit.
+
+void
+Dwo_id_info_reader::visit_compilation_unit(off_t, off_t, Dwarf_die* die)
+{
+  this->dwo_id_ = die->uint_attribute(elfcpp::DW_AT_GNU_dwo_id);
+  if (this->dwo_id_ != 0)
+    this->dwo_id_found_ = true;
+}
+
+// Visit a type unit.
+
+void
+Dwo_id_info_reader::visit_type_unit(off_t, off_t, uint64_t signature, Dwarf_die*)
+{
+  this->type_sig_ = signature;
+  this->type_sig_found_ = true;
+}
+
+}; // End namespace gold
+
+using namespace gold;
+
+// Options.
+
+extern int optind;
+extern char* optarg;
+
+struct option dwp_options[] =
+  {
+    { "verbose", no_argument, NULL, 'v' },
+    { "output", required_argument, NULL, 'o' },
+    { NULL, 0, NULL, 0 }
+  };
+
+// Print usage message and exit.
+
+static void
+usage()
+{
+  fprintf(stderr, _("Usage: %s [options] file...\n"), program_name);
+  fprintf(stderr, _("  -v, --verbose            Verbose output\n"));
+  fprintf(stderr, _("  -o FILE, --output FILE   Set output dwp file name"
+		    " (required)\n"));
+  exit(1);
+}
+
+// Main program.
+
+int
+main(int argc, char** argv)
+{
+#if defined (HAVE_SETLOCALE) && defined (HAVE_LC_MESSAGES)
+  setlocale(LC_MESSAGES, "");
+#endif
+#if defined (HAVE_SETLOCALE)
+  setlocale(LC_CTYPE, "");
+#endif
+  bindtextdomain(PACKAGE, LOCALEDIR);
+  textdomain(PACKAGE);
+
+  program_name = argv[0];
+
+  // Initialize the global parameters, to let random code get to the
+  // errors object.
+  Errors errors(program_name);
+  set_parameters_errors(&errors);
+
+  // Initialize gold's global options.  We don't use these in
+  // this program, but they need to be initialized so that
+  // functions we call from libgold work properly.
+  General_options options;
+  set_parameters_options(&options);
+
+  // In libiberty; expands @filename to the args in "filename".
+  expandargv(&argc, &argv);
+
+  // Collect file names and options.
+  typedef std::vector<char*> File_list;
+  File_list files;
+  const char* output_filename = NULL;
+  bool verbose = false;
+  int c;
+  while ((c = getopt_long(argc, argv, "vo:", dwp_options, NULL)) != -1)
+    {
+      switch (c)
+        {
+	  case 'v':
+	    verbose = true;
+	    break;
+	  case 'o':
+	    output_filename = optarg;
+	    break;
+	  case '?':
+	  default:
+	    usage();
+	}
+    }
+  for (int i = optind; i < argc; ++i)
+    files.push_back(argv[i]);
+
+  if (files.empty())
+    gold_fatal(_("no input files"));
+  if (output_filename == NULL)
+    gold_fatal(_("no output file specified"));
+
+  Dwp_output_file output_file(output_filename);
+  
+  // Process each file, adding its contents to the output file.
+  for (File_list::const_iterator f = files.begin(); f != files.end(); ++f)
+    {
+      if (verbose)
+        fprintf(stderr, "%s\n", *f);
+      Dwo_file dwo_file(*f);
+      dwo_file.read(&output_file);
+    }
+
+  output_file.finalize();
+
+  return EXIT_SUCCESS;
+}
diff --git a/gold/dwp.h b/gold/dwp.h
new file mode 100644
index 0000000..25d4f68
--- /dev/null
+++ b/gold/dwp.h
@@ -0,0 +1,120 @@
+// dwp.h -- general definitions for dwp.
+
+// Copyright 2012 Free Software Foundation, Inc.
+// Written by Cary Coutant <ccoutant@google.com>.
+
+// This file is part of dwp, the DWARF packaging utility.
+
+// This program is free software; you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation; either version 3 of the License, or
+// (at your option) any later version.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with this program; if not, write to the Free Software
+// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+// MA 02110-1301, USA.
+
+#ifndef DWP_DWP_H
+#define DWP_DWP_H 1
+
+#include "config.h"
+#include "ansidecl.h"
+
+#include <cstddef>
+#include <cstdlib>
+#include <cstring>
+#include <stdint.h>
+#include <sys/types.h>
+
+#include "system.h"
+
+namespace gold
+{
+
+extern const char* program_name;
+
+class General_options;
+class Command_line;
+class Dirsearch;
+class Input_objects;
+class Mapfile;
+class Symbol;
+class Symbol_table;
+class Layout;
+class Task;
+class Workqueue;
+class Output_file;
+template<int size, bool big_endian>
+struct Relocate_info;
+
+// The size of a section if we are going to look at the contents.
+typedef size_t section_size_type;
+
+// An offset within a section when we are looking at the contents.
+typedef ptrdiff_t section_offset_type;
+
+inline bool
+is_prefix_of(const char* prefix, const char* str)
+{
+  return strncmp(prefix, str, strlen(prefix)) == 0;
+}
+
+// Exit status codes.
+
+enum Exit_status
+{
+  GOLD_OK = EXIT_SUCCESS,
+  GOLD_ERR = EXIT_FAILURE,
+  GOLD_FALLBACK = EXIT_FAILURE + 1
+};
+
+// This function is called to exit the program.  Status is true to
+// exit success (0) and false to exit failure (1).
+extern void
+gold_exit(Exit_status status) ATTRIBUTE_NORETURN;
+
+// This function is called to emit an error message and then
+// immediately exit with failure.
+extern void
+gold_fatal(const char* format, ...) ATTRIBUTE_NORETURN ATTRIBUTE_PRINTF_1;
+
+// This function is called to issue a warning.
+extern void
+gold_warning(const char* msg, ...) ATTRIBUTE_PRINTF_1;
+
+#define gold_unreachable() \
+  (gold::do_gold_unreachable(__FILE__, __LINE__, \
+			     static_cast<const char*>(__FUNCTION__)))
+
+extern void do_gold_unreachable(const char*, int, const char*)
+  ATTRIBUTE_NORETURN;
+
+// Assertion check.
+
+#define gold_assert(expr) ((void)(!(expr) ? gold_unreachable(), 0 : 0))
+
+// Convert numeric types without unnoticed loss of precision.
+template<typename To, typename From>
+inline To
+convert_types(const From from)
+{
+  To to = from;
+  gold_assert(static_cast<From>(to) == from);
+  return to;
+}
+
+// A common case of convert_types<>: convert to section_size_type.
+template<typename From>
+inline section_size_type
+convert_to_section_size_type(const From from)
+{ return convert_types<section_size_type, From>(from); }
+
+}; // End namespace gold.
+
+#endif // !defined(DWP_DWP_H)
diff --git a/gold/gold.h b/gold/gold.h
index 9aa7cf3..ef95f53 100644
--- a/gold/gold.h
+++ b/gold/gold.h
@@ -32,137 +32,7 @@
 #include <stdint.h>
 #include <sys/types.h>
 
-#ifndef ENABLE_NLS
-  // The Solaris version of locale.h always includes libintl.h.  If we
-  // have been configured with --disable-nls then ENABLE_NLS will not
-  // be defined and the dummy definitions of bindtextdomain (et al)
-  // below will conflict with the definitions in libintl.h.  So we
-  // define these values to prevent the bogus inclusion of libintl.h.
-# define _LIBINTL_H
-# define _LIBGETTEXT_H
-#endif
-
-// Always include <clocale> first to avoid conflicts with the macros
-// used when ENABLE_NLS is not defined.
-#include <clocale>
-
-#ifdef ENABLE_NLS
-# include <libintl.h>
-# define _(String) gettext (String)
-# ifdef gettext_noop
-#  define N_(String) gettext_noop (String)
-# else
-#  define N_(String) (String)
-# endif
-#else
-# define gettext(Msgid) (Msgid)
-# define dgettext(Domainname, Msgid) (Msgid)
-# define dcgettext(Domainname, Msgid, Category) (Msgid)
-# define textdomain(Domainname) do {} while (0) /* nothing */
-# define bindtextdomain(Domainname, Dirname) do {} while (0) /* nothing */
-# define _(String) (String)
-# define N_(String) (String)
-#endif
-
-// Figure out how to get a hash set and a hash map.
-
-#if defined(HAVE_TR1_UNORDERED_SET) && defined(HAVE_TR1_UNORDERED_MAP) \
-    && defined(HAVE_TR1_UNORDERED_MAP_REHASH)
-
-#include <tr1/unordered_set>
-#include <tr1/unordered_map>
-
-// We need a template typedef here.
-
-#define Unordered_set std::tr1::unordered_set
-#define Unordered_map std::tr1::unordered_map
-#define Unordered_multimap std::tr1::unordered_multimap
-
-#define reserve_unordered_map(map, n) ((map)->rehash(n))
-
-#ifndef HAVE_TR1_HASH_OFF_T
-// The library does not support hashes of off_t values.  Add support
-// here.  This is likely to be specific to libstdc++.  This issue
-// arises with GCC 4.1.x when compiling in 32-bit mode with a 64-bit
-// off_t type.
-namespace std { namespace tr1 {
-template<>
-struct hash<off_t> : public std::unary_function<off_t, std::size_t>
-{
-  std::size_t
-  operator()(off_t val) const
-  { return static_cast<std::size_t>(val); }
-};
-} } // Close namespaces.
-#endif // !defined(HAVE_TR1_HASH_OFF_T)
-
-#elif defined(HAVE_EXT_HASH_MAP) && defined(HAVE_EXT_HASH_SET)
-
-#include <ext/hash_map>
-#include <ext/hash_set>
-#include <string>
-
-#define Unordered_set __gnu_cxx::hash_set
-#define Unordered_map __gnu_cxx::hash_map
-#define Unordered_multimap __gnu_cxx::hash_multimap
-
-namespace __gnu_cxx
-{
-
-template<>
-struct hash<std::string>
-{
-  size_t
-  operator()(std::string s) const
-  { return __stl_hash_string(s.c_str()); }
-};
-
-template<typename T>
-struct hash<T*>
-{
-  size_t
-  operator()(T* p) const
-  { return reinterpret_cast<size_t>(p); }
-};
-
-}
-
-#define reserve_unordered_map(map, n) ((map)->resize(n))
-
-#else
-
-// The fallback is to just use set and map.
-
-#include <set>
-#include <map>
-
-#define Unordered_set std::set
-#define Unordered_map std::map
-#define Unordered_multimap std::multimap
-
-#define reserve_unordered_map(map, n)
-
-#endif
-
-#ifndef HAVE_PREAD
-extern "C" ssize_t pread(int, void*, size_t, off_t);
-#endif
-
-#ifndef HAVE_FTRUNCATE
-extern "C" int ftruncate(int, off_t);
-#endif
-
-#ifndef HAVE_FFSLL
-extern "C" int ffsll(long long);
-#endif
-
-#if !HAVE_DECL_MEMMEM
-extern "C" void *memmem(const void *, size_t, const void *, size_t);
-#endif
-
-#if !HAVE_DECL_STRNDUP
-extern "C" char *strndup(const char *, size_t);
-#endif
+#include "system.h"
 
 namespace gold
 {
diff --git a/gold/system.h b/gold/system.h
new file mode 100644
index 0000000..f84a94a
--- /dev/null
+++ b/gold/system.h
@@ -0,0 +1,158 @@
+// gold.h -- general definitions for gold   -*- C++ -*-
+
+// Copyright 2006, 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc.
+// Written by Ian Lance Taylor <iant@google.com>.
+
+// This file is part of gold.
+
+// This program is free software; you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation; either version 3 of the License, or
+// (at your option) any later version.
+
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License
+// along with this program; if not, write to the Free Software
+// Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
+// MA 02110-1301, USA.
+
+#ifndef SYSTEM_H
+#define SYSTEM_H 1
+
+#ifndef ENABLE_NLS
+  // The Solaris version of locale.h always includes libintl.h.  If we
+  // have been configured with --disable-nls then ENABLE_NLS will not
+  // be defined and the dummy definitions of bindtextdomain (et al)
+  // below will conflict with the definitions in libintl.h.  So we
+  // define these values to prevent the bogus inclusion of libintl.h.
+# define _LIBINTL_H
+# define _LIBGETTEXT_H
+#endif
+
+// Always include <clocale> first to avoid conflicts with the macros
+// used when ENABLE_NLS is not defined.
+#include <clocale>
+
+#ifdef ENABLE_NLS
+# include <libintl.h>
+# define _(String) gettext (String)
+# ifdef gettext_noop
+#  define N_(String) gettext_noop (String)
+# else
+#  define N_(String) (String)
+# endif
+#else
+# define gettext(Msgid) (Msgid)
+# define dgettext(Domainname, Msgid) (Msgid)
+# define dcgettext(Domainname, Msgid, Category) (Msgid)
+# define textdomain(Domainname) do {} while (0) /* nothing */
+# define bindtextdomain(Domainname, Dirname) do {} while (0) /* nothing */
+# define _(String) (String)
+# define N_(String) (String)
+#endif
+
+// Figure out how to get a hash set and a hash map.
+
+#if defined(HAVE_TR1_UNORDERED_SET) && defined(HAVE_TR1_UNORDERED_MAP) \
+    && defined(HAVE_TR1_UNORDERED_MAP_REHASH)
+
+#include <tr1/unordered_set>
+#include <tr1/unordered_map>
+
+// We need a template typedef here.
+
+#define Unordered_set std::tr1::unordered_set
+#define Unordered_map std::tr1::unordered_map
+#define Unordered_multimap std::tr1::unordered_multimap
+
+#define reserve_unordered_map(map, n) ((map)->rehash(n))
+
+#ifndef HAVE_TR1_HASH_OFF_T
+// The library does not support hashes of off_t values.  Add support
+// here.  This is likely to be specific to libstdc++.  This issue
+// arises with GCC 4.1.x when compiling in 32-bit mode with a 64-bit
+// off_t type.
+namespace std { namespace tr1 {
+template<>
+struct hash<off_t> : public std::unary_function<off_t, std::size_t>
+{
+  std::size_t
+  operator()(off_t val) const
+  { return static_cast<std::size_t>(val); }
+};
+} } // Close namespaces.
+#endif // !defined(HAVE_TR1_HASH_OFF_T)
+
+#elif defined(HAVE_EXT_HASH_MAP) && defined(HAVE_EXT_HASH_SET)
+
+#include <ext/hash_map>
+#include <ext/hash_set>
+#include <string>
+
+#define Unordered_set __gnu_cxx::hash_set
+#define Unordered_map __gnu_cxx::hash_map
+#define Unordered_multimap __gnu_cxx::hash_multimap
+
+namespace __gnu_cxx
+{
+
+template<>
+struct hash<std::string>
+{
+  size_t
+  operator()(std::string s) const
+  { return __stl_hash_string(s.c_str()); }
+};
+
+template<typename T>
+struct hash<T*>
+{
+  size_t
+  operator()(T* p) const
+  { return reinterpret_cast<size_t>(p); }
+};
+
+}
+
+#define reserve_unordered_map(map, n) ((map)->resize(n))
+
+#else
+
+// The fallback is to just use set and map.
+
+#include <set>
+#include <map>
+
+#define Unordered_set std::set
+#define Unordered_map std::map
+#define Unordered_multimap std::multimap
+
+#define reserve_unordered_map(map, n)
+
+#endif
+
+#ifndef HAVE_PREAD
+extern "C" ssize_t pread(int, void*, size_t, off_t);
+#endif
+
+#ifndef HAVE_FTRUNCATE
+extern "C" int ftruncate(int, off_t);
+#endif
+
+#ifndef HAVE_FFSLL
+extern "C" int ffsll(long long);
+#endif
+
+#if !HAVE_DECL_MEMMEM
+extern "C" void *memmem(const void *, size_t, const void *, size_t);
+#endif
+
+#if !HAVE_DECL_STRNDUP
+extern "C" char *strndup(const char *, size_t);
+#endif
+
+#endif // !defined(SYSTEM_H)
-- 
1.7.9.6 (Apple Git-31.1)


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]