aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorCary Coutant <ccoutant@google.com>2012-10-24 02:26:39 +0000
committerCary Coutant <ccoutant@google.com>2012-10-24 02:26:39 +0000
commit34ce70fa99d7bc72b04be302800db348e99c33c0 (patch)
treeaf7a192d5ed00dbbf58e933ab8ccdf1e457fbb52
parent387c7a3fac1150418a8fa2f3fca672904b9a105b (diff)
downloadbinutils-34ce70fa99d7bc72b04be302800db348e99c33c0.tar.gz
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.
-rw-r--r--gold/ChangeLog9
-rw-r--r--gold/Makefile.am8
-rw-r--r--gold/Makefile.in103
-rw-r--r--gold/dwp.cc2074
-rw-r--r--gold/dwp.h120
-rw-r--r--gold/gold.h132
-rw-r--r--gold/system.h158
7 files changed, 2451 insertions, 153 deletions
diff --git a/gold/ChangeLog b/gold/ChangeLog
index 1e6032292..653f65f47 100644
--- a/gold/ChangeLog
+++ b/gold/ChangeLog
@@ -1,5 +1,14 @@
2012-10-23 Cary Coutant <ccoutant@google.com>
+ * 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.
+
+2012-10-23 Cary Coutant <ccoutant@google.com>
+
* dwarf_reader.cc (Dwarf_ranges_table::read_range_list): Call
Dwarf_info_reader::read_from_pointer.
(Dwarf_pubnames_table::read_header): Likewise.
diff --git a/gold/Makefile.am b/gold/Makefile.am
index aa8caca65..f82d4e9e1 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.$(OBJEXT): yyscript.h
diff --git a/gold/Makefile.in b/gold/Makefile.in
index c8d8453e4..16d2cf4b3 100644
--- a/gold/Makefile.in
+++ b/gold/Makefile.in
@@ -38,6 +38,7 @@ POST_UNINSTALL = :
build_triplet = @build@
host_triplet = @host@
target_triplet = @target@
+bin_PROGRAMS = dwp$(EXEEXT)
noinst_PROGRAMS = ld-new$(EXEEXT) incremental-dump$(EXEEXT)
@GCC_TRUE@@NATIVE_LINKER_TRUE@check_PROGRAMS = ld1$(EXEEXT) \
@GCC_TRUE@@NATIVE_LINKER_TRUE@ ld2$(EXEEXT) ld1-r$(EXEEXT) \
@@ -94,10 +95,15 @@ am__objects_3 = yyscript.$(OBJEXT)
am_libgold_a_OBJECTS = $(am__objects_1) $(am__objects_2) \
$(am__objects_3) $(am__objects_2)
libgold_a_OBJECTS = $(am_libgold_a_OBJECTS)
-PROGRAMS = $(noinst_PROGRAMS)
+am__installdirs = "$(DESTDIR)$(bindir)"
+PROGRAMS = $(bin_PROGRAMS) $(noinst_PROGRAMS)
+am_dwp_OBJECTS = dwp.$(OBJEXT)
+dwp_OBJECTS = $(am_dwp_OBJECTS)
+am__DEPENDENCIES_1 =
+dwp_LINK = $(CXXLD) $(AM_CXXFLAGS) $(CXXFLAGS) $(dwp_LDFLAGS) \
+ $(LDFLAGS) -o $@
am_incremental_dump_OBJECTS = incremental-dump.$(OBJEXT)
incremental_dump_OBJECTS = $(am_incremental_dump_OBJECTS)
-am__DEPENDENCIES_1 =
am__objects_4 = main.$(OBJEXT)
am_ld_new_OBJECTS = $(am__objects_4)
ld_new_OBJECTS = $(am_ld_new_OBJECTS)
@@ -138,9 +144,10 @@ CXXLINK = $(CXXLD) $(AM_CXXFLAGS) $(CXXFLAGS) $(AM_LDFLAGS) $(LDFLAGS) \
@MAINTAINER_MODE_FALSE@am__skipyacc = test -f $@ ||
YACCCOMPILE = $(YACC) $(YFLAGS) $(AM_YFLAGS)
YLWRAP = $(top_srcdir)/../ylwrap
-SOURCES = $(libgold_a_SOURCES) $(incremental_dump_SOURCES) \
- $(ld_new_SOURCES) $(EXTRA_ld_new_SOURCES) $(ld1_SOURCES) \
- $(ld1_r_SOURCES) $(ld2_SOURCES) $(ld2_r_SOURCES)
+SOURCES = $(libgold_a_SOURCES) $(dwp_SOURCES) \
+ $(incremental_dump_SOURCES) $(ld_new_SOURCES) \
+ $(EXTRA_ld_new_SOURCES) $(ld1_SOURCES) $(ld1_r_SOURCES) \
+ $(ld2_SOURCES) $(ld2_r_SOURCES)
RECURSIVE_TARGETS = all-recursive check-recursive dvi-recursive \
html-recursive info-recursive install-data-recursive \
install-dvi-recursive install-exec-recursive \
@@ -514,6 +521,12 @@ 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)
POTFILES = $(CCFILES) $(HFILES) $(TARGETSOURCES)
@GCC_TRUE@@NATIVE_LINKER_TRUE@ld1_SOURCES = $(sources_var)
@GCC_TRUE@@NATIVE_LINKER_TRUE@ld1_DEPENDENCIES = $(deps_var) gcctestdir1/ld
@@ -602,12 +615,52 @@ libgold.a: $(libgold_a_OBJECTS) $(libgold_a_DEPENDENCIES)
-rm -f libgold.a
$(libgold_a_AR) libgold.a $(libgold_a_OBJECTS) $(libgold_a_LIBADD)
$(RANLIB) libgold.a
+install-binPROGRAMS: $(bin_PROGRAMS)
+ @$(NORMAL_INSTALL)
+ test -z "$(bindir)" || $(MKDIR_P) "$(DESTDIR)$(bindir)"
+ @list='$(bin_PROGRAMS)'; test -n "$(bindir)" || list=; \
+ for p in $$list; do echo "$$p $$p"; done | \
+ sed 's/$(EXEEXT)$$//' | \
+ while read p p1; do if test -f $$p; \
+ then echo "$$p"; echo "$$p"; else :; fi; \
+ done | \
+ sed -e 'p;s,.*/,,;n;h' -e 's|.*|.|' \
+ -e 'p;x;s,.*/,,;s/$(EXEEXT)$$//;$(transform);s/$$/$(EXEEXT)/' | \
+ sed 'N;N;N;s,\n, ,g' | \
+ $(AWK) 'BEGIN { files["."] = ""; dirs["."] = 1 } \
+ { d=$$3; if (dirs[d] != 1) { print "d", d; dirs[d] = 1 } \
+ if ($$2 == $$4) files[d] = files[d] " " $$1; \
+ else { print "f", $$3 "/" $$4, $$1; } } \
+ END { for (d in files) print "f", d, files[d] }' | \
+ while read type dir files; do \
+ if test "$$dir" = .; then dir=; else dir=/$$dir; fi; \
+ test -z "$$files" || { \
+ echo " $(INSTALL_PROGRAM_ENV) $(INSTALL_PROGRAM) $$files '$(DESTDIR)$(bindir)$$dir'"; \
+ $(INSTALL_PROGRAM_ENV) $(INSTALL_PROGRAM) $$files "$(DESTDIR)$(bindir)$$dir" || exit $$?; \
+ } \
+ ; done
+
+uninstall-binPROGRAMS:
+ @$(NORMAL_UNINSTALL)
+ @list='$(bin_PROGRAMS)'; test -n "$(bindir)" || list=; \
+ files=`for p in $$list; do echo "$$p"; done | \
+ sed -e 'h;s,^.*/,,;s/$(EXEEXT)$$//;$(transform)' \
+ -e 's/$$/$(EXEEXT)/' `; \
+ test -n "$$list" || exit 0; \
+ echo " ( cd '$(DESTDIR)$(bindir)' && rm -f" $$files ")"; \
+ cd "$(DESTDIR)$(bindir)" && rm -f $$files
+
+clean-binPROGRAMS:
+ -test -z "$(bin_PROGRAMS)" || rm -f $(bin_PROGRAMS)
clean-checkPROGRAMS:
-test -z "$(check_PROGRAMS)" || rm -f $(check_PROGRAMS)
clean-noinstPROGRAMS:
-test -z "$(noinst_PROGRAMS)" || rm -f $(noinst_PROGRAMS)
+dwp$(EXEEXT): $(dwp_OBJECTS) $(dwp_DEPENDENCIES)
+ @rm -f dwp$(EXEEXT)
+ $(dwp_LINK) $(dwp_OBJECTS) $(dwp_LDADD) $(LIBS)
incremental-dump$(EXEEXT): $(incremental_dump_OBJECTS) $(incremental_dump_DEPENDENCIES)
@rm -f incremental-dump$(EXEEXT)
$(CXXLINK) $(incremental_dump_OBJECTS) $(incremental_dump_LDADD) $(LIBS)
@@ -650,6 +703,7 @@ distclean-compile:
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/descriptors.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/dirsearch.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/dwarf_reader.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/dwp.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/dynobj.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/ehframe.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/errors.Po@am__quote@
@@ -686,6 +740,7 @@ distclean-compile:
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/symtab.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/target-select.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/target.Po@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tilegx.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/timer.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/version.Po@am__quote@
@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/workqueue-threads.Po@am__quote@
@@ -1026,6 +1081,9 @@ check: check-recursive
all-am: Makefile $(LIBRARIES) $(PROGRAMS) config.h
installdirs: installdirs-recursive
installdirs-am:
+ for dir in "$(DESTDIR)$(bindir)"; do \
+ test -z "$$dir" || $(MKDIR_P) "$$dir"; \
+ done
install: install-recursive
install-exec: install-exec-recursive
install-data: install-data-recursive
@@ -1059,8 +1117,8 @@ maintainer-clean-generic:
-rm -f yyscript.h
clean: clean-recursive
-clean-am: clean-checkPROGRAMS clean-generic clean-noinstLIBRARIES \
- clean-noinstPROGRAMS mostlyclean-am
+clean-am: clean-binPROGRAMS clean-checkPROGRAMS clean-generic \
+ clean-noinstLIBRARIES clean-noinstPROGRAMS mostlyclean-am
distclean: distclean-recursive
-rm -f $(am__CONFIG_DISTCLEAN_FILES)
@@ -1087,7 +1145,7 @@ install-dvi: install-dvi-recursive
install-dvi-am:
-install-exec-am: install-exec-local
+install-exec-am: install-binPROGRAMS install-exec-local
install-html: install-html-recursive
@@ -1128,7 +1186,7 @@ ps: ps-recursive
ps-am:
-uninstall-am:
+uninstall-am: uninstall-binPROGRAMS
.MAKE: $(RECURSIVE_CLEAN_TARGETS) $(RECURSIVE_TARGETS) all check-am \
check-html ctags-recursive install-am install-strip recheck \
@@ -1136,19 +1194,20 @@ uninstall-am:
.PHONY: $(RECURSIVE_CLEAN_TARGETS) $(RECURSIVE_TARGETS) CTAGS GTAGS \
all all-am am--refresh check check-TESTS check-am check-html \
- clean clean-checkPROGRAMS clean-generic clean-noinstLIBRARIES \
- clean-noinstPROGRAMS ctags ctags-recursive distclean \
- distclean-compile distclean-generic distclean-hdr \
- distclean-tags dvi dvi-am html html-am info info-am install \
- install-am install-data install-data-am install-data-local \
- install-dvi install-dvi-am install-exec install-exec-am \
- install-exec-local install-html install-html-am install-info \
- install-info-am install-man install-pdf install-pdf-am \
- install-ps install-ps-am install-strip installcheck \
- installcheck-am installdirs installdirs-am maintainer-clean \
- maintainer-clean-generic mostlyclean mostlyclean-compile \
- mostlyclean-generic pdf pdf-am ps ps-am recheck recheck-html \
- tags tags-recursive uninstall uninstall-am
+ clean clean-binPROGRAMS clean-checkPROGRAMS clean-generic \
+ clean-noinstLIBRARIES clean-noinstPROGRAMS ctags \
+ ctags-recursive distclean distclean-compile distclean-generic \
+ distclean-hdr distclean-tags dvi dvi-am html html-am info \
+ info-am install install-am install-binPROGRAMS install-data \
+ install-data-am install-data-local install-dvi install-dvi-am \
+ install-exec install-exec-am install-exec-local install-html \
+ install-html-am install-info install-info-am install-man \
+ install-pdf install-pdf-am install-ps install-ps-am \
+ install-strip installcheck installcheck-am installdirs \
+ installdirs-am maintainer-clean maintainer-clean-generic \
+ mostlyclean mostlyclean-compile mostlyclean-generic pdf pdf-am \
+ ps ps-am recheck recheck-html tags tags-recursive uninstall \
+ uninstall-am uninstall-binPROGRAMS
# Use an explicit dependency for the bison generated header file.
diff --git a/gold/dwp.cc b/gold/dwp.cc
new file mode 100644
index 000000000..445d1afc5
--- /dev/null
+++ b/gold/dwp.cc
@@ -0,0 +1,2074 @@
+// 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.
+
+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 000000000..25d4f680b
--- /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 9aa7cf350..ef95f5302 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 000000000..ce2d93979
--- /dev/null
+++ b/gold/system.h
@@ -0,0 +1,158 @@
+// system.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)