gold/stringpool.h

   1 // stringpool.h -- a string pool for gold    -*- C++ -*-
   2
   3 // Copyright 2006, 2007 Free Software Foundation, Inc.
   4 // Written by Ian Lance Taylor <iant@google.com>.
   5
   6 // This file is part of gold.
   7
   8 // This program is free software; you can redistribute it and/or modify
   9 // it under the terms of the GNU General Public License as published by
  10 // the Free Software Foundation; either version 3 of the License, or
  11 // (at your option) any later version.
  12
  13 // This program is distributed in the hope that it will be useful,
  14 // but WITHOUT ANY WARRANTY; without even the implied warranty of
  15 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16 // GNU General Public License for more details.
  17
  18 // You should have received a copy of the GNU General Public License
  19 // along with this program; if not, write to the Free Software
  20 // Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
  21 // MA 02110-1301, USA.
  22
  23 #include <string>
  24 #include <list>
  25
  26 #ifndef GOLD_STRINGPOOL_H
  27 #define GOLD_STRINGPOOL_H
  28
  29 namespace gold
  30 {
  31
  32 class Output_file;
  33
  34 // A Stringpool is a pool of unique strings.  It provides the
  35 // following features:
  36
  37 // Every string in the pool is unique.  Thus, if you have two strings
  38 // in the Stringpool, you can compare them for equality by using
  39 // pointer comparison rather than string comparison.
  40
  41 // There is a key associated with every string in the pool.  If you
  42 // add strings to the Stringpool in the same order, then the key for
  43 // each string will always be the same for any run of the linker.
  44 // This is not true of the string pointers themselves, as they may
  45 // change due to address space randomization.  Some parts of the
  46 // linker (e.g., the symbol table) use the key value instead of the
  47 // string pointer so that repeated runs of the linker will generate
  48 // precisely the same output.
  49
  50 // When you add a string to a Stringpool, Stringpool will optionally
  51 // make a copy of it.  Thus there is no requirement to keep a copy
  52 // elsewhere.
  53
  54 // A Stringpool can be turned into a string table, a sequential series
  55 // of null terminated strings.  The first string may optionally be a
  56 // single zero byte, as required for SHT_STRTAB sections.  This
  57 // conversion is only permitted after all strings have been added to
  58 // the Stringpool.  After doing this conversion, you can ask for the
  59 // offset of any string in the stringpool in the string table, and you
  60 // can write the resulting string table to an output file.
  61
  62 // When a Stringpool is turned into a string table, then as an
  63 // optimization it will reuse string suffixes to avoid duplicating
  64 // strings.  That is, given the strings "abc" and "bc", only the
  65 // string "abc" will be stored, and "bc" will be represented by an
  66 // offset into the middle of the string "abc".
  67
  68 // Stringpools are implemented in terms of Stringpool_template, which
  69 // is generalized on the type of character used for the strings.  Most
  70 // uses will want the Stringpool type which uses char.  Other cases
  71 // are used for merging wide string constants.
  72
  73 template<typename Stringpool_char>
  74 class Stringpool_template
  75 {
  76  public:
  77   // The type of a key into the stringpool.  As described above, a key
  78   // value will always be the same during any run of the linker.  Zero
  79   // is never a valid key value.
  80   typedef size_t Key;
  81
  82   // Create a Stringpool.
  83   Stringpool_template();
  84
  85   ~Stringpool_template();
  86
  87   // Clear all the data from the stringpool.
  88   void
  89   clear();
  90
  91   // Indicate that we should not reserve offset 0 to hold the empty
  92   // string when converting the stringpool to a string table.  This
  93   // should not be called for a proper ELF SHT_STRTAB section.
  94   void
  95   set_no_zero_null()
  96   { this->zero_null_ = false; }
  97
  98   // Add the string S to the pool.  This returns a canonical permanent
  99   // pointer to the string in the pool.  If COPY is true, the string
 100   // is copied into permanent storage.  If PKEY is not NULL, this sets
 101   // *PKEY to the key for the string.
 102   const Stringpool_char*
 103   add(const Stringpool_char* s, bool copy, Key* pkey);
 104
 105   // Add the prefix of length LEN of string S to the pool.
 106   const Stringpool_char*
 107   add_prefix(const Stringpool_char* s, size_t len, Key* pkey);
 108
 109   // If the string S is present in the pool, return the canonical
 110   // string pointer.  Otherwise, return NULL.  If PKEY is not NULL,
 111   // set *PKEY to the key.
 112   const Stringpool_char*
 113   find(const Stringpool_char* s, Key* pkey) const;
 114
 115   // Turn the stringpool into a string table: determine the offsets of
 116   // all the strings.  After this is called, no more strings may be
 117   // added to the stringpool.
 118   void
 119   set_string_offsets();
 120
 121   // Get the offset of the string S in the string table.  This returns
 122   // the offset in bytes, not in units of Stringpool_char.  This may
 123   // only be called after set_string_offsets has been called.
 124   off_t
 125   get_offset(const Stringpool_char* s) const;
 126
 127   // Get the offset of the string S in the string table.
 128   off_t
 129   get_offset(const std::basic_string<Stringpool_char>& s) const
 130   { return this->get_offset(s.c_str()); }
 131
 132   // Get the size of the string table.  This returns the number of
 133   // bytes, not in units of Stringpool_char.
 134   off_t
 135   get_strtab_size() const
 136   {
 137     gold_assert(this->strtab_size_ != 0);
 138     return this->strtab_size_;
 139   }
 140
 141   // Write the string table into the output file at the specified
 142   // offset.
 143   void
 144   write(Output_file*, off_t offset);
 145
 146   // Write the string table into the specified buffer, of the
 147   // specified size.  buffer_size should be at least
 148   // get_strtab_size().
 149   void
 150   write_to_buffer(unsigned char* buffer, size_t buffer_size);
 151
 152   // Dump statistical information to stderr.
 153   void
 154   print_stats(const char*) const;
 155
 156  private:
 157   Stringpool_template(const Stringpool_template&);
 158   Stringpool_template& operator=(const Stringpool_template&);
 159
 160   // Return the length of a string in units of Stringpool_char.
 161   static size_t
 162   string_length(const Stringpool_char*);
 163
 164   // We store the actual data in a list of these buffers.
 165   struct Stringdata
 166   {
 167     // Length of data in buffer.
 168     size_t len;
 169     // Allocated size of buffer.
 170     size_t alc;
 171     // Buffer index.
 172     unsigned int index;
 173     // Buffer.
 174     char data[1];
 175   };
 176
 177   // Copy a string into the buffers, returning a canonical string.
 178   const Stringpool_char*
 179   add_string(const Stringpool_char*, Key*);
 180
 181   // Hash function.
 182   struct Stringpool_hash
 183   {
 184     size_t
 185     operator()(const Stringpool_char*) const;
 186   };
 187
 188   // Equality comparison function for hash table.
 189   struct Stringpool_eq
 190   {
 191     bool
 192     operator()(const Stringpool_char* p1, const Stringpool_char* p2) const;
 193   };
 194
 195   // Return whether s1 is a suffix of s2.
 196   static bool
 197   is_suffix(const Stringpool_char* s1, size_t len1,
 198             const Stringpool_char* s2, size_t len2);
 199
 200   // The hash table is a map from string names to a pair of Key and
 201   // string table offsets.  We only use the offsets if we turn this
 202   // into an string table section.
 203
 204   typedef std::pair<Key, off_t> Val;
 205
 206 #ifdef HAVE_TR1_UNORDERED_SET
 207   typedef Unordered_map<const Stringpool_char*, Val, Stringpool_hash,
 208                         Stringpool_eq,
 209                         std::allocator<std::pair<const Stringpool_char* const,
 210                                                  Val> >,
 211                         true> String_set_type;
 212 #else
 213   typedef Unordered_map<const Stringpool_char*, Val, Stringpool_hash,
 214                         Stringpool_eq> String_set_type;
 215 #endif
 216
 217   // Comparison routine used when sorting into a string table.  We
 218   // store string-sizes in the sort-vector so we don't have to
 219   // recompute them log(n) times as we sort.
 220   struct Stringpool_sort_info
 221   {
 222     typename String_set_type::iterator it;
 223     size_t string_length;
 224     Stringpool_sort_info(typename String_set_type::iterator i, size_t s)
 225       : it(i), string_length(s)
 226     { }
 227   };
 228
 229   struct Stringpool_sort_comparison
 230   {
 231     bool
 232     operator()(const Stringpool_sort_info&, const Stringpool_sort_info&) const;
 233   };
 234
 235   // List of Stringdata structures.
 236   typedef std::list<Stringdata*> Stringdata_list;
 237
 238   // Mapping from const char* to namepool entry.
 239   String_set_type string_set_;
 240   // List of buffers.
 241   Stringdata_list strings_;
 242   // Size of string table.
 243   off_t strtab_size_;
 244   // Next Stringdata index.
 245   unsigned int next_index_;
 246   // Next key value for a string we don't copy.
 247   int next_uncopied_key_;
 248   // Whether to reserve offset 0 to hold the null string.
 249   bool zero_null_;
 250 };
 251
 252 // The most common type of Stringpool.
 253 typedef Stringpool_template<char> Stringpool;
 254
 255 } // End namespace gold.
 256
 257 #endif // !defined(GOLD_STRINGPOOL_H)