Subversion Repositories open8_urisc

/* BFD backend for hp-ux 9000/300

   Copyright 1990, 1991, 1993, 1994, 1995, 1997, 1999, 2000, 2001, 2002,

   2003, 2004, 2005, 2007, 2010 Free Software Foundation, Inc.

   Written by Glenn Engel.

   This file is part of BFD, the Binary File Descriptor library.

   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.  */

/*  hpux native  ------------> |               |

                               | hp300hpux bfd | ----------> hpux w/gnu ext

    hpux w/gnu extension ----> |               |

    Support for the 9000/[34]00 has several limitations.

      1. Shared libraries are not supported.

      2. The output format from this bfd is not usable by native tools.

    The primary motivation for writing this bfd was to allow use of

    gdb and gcc for host based debugging and not to mimic the hp-ux tools

    in every detail.  This leads to a significant simplification of the

    code and a leap in performance.  The decision to not output hp native

    compatible objects was further strengthened by the fact that the richness

    of the gcc compiled objects could not be represented without loss of

    information.  For example, while the hp format supports the concept of

    secondary symbols, it does not support indirect symbols.  Another

    reason is to maintain backwards compatibility with older implementations

    of gcc on hpux which used 'hpxt' to translate .a and .o files into a

    format which could be readily understood by the gnu linker and gdb.

    This allows reading hp secondary symbols and converting them into

    indirect symbols but the reverse it not always possible.

    Another example of differences is that the hp format stores symbol offsets

    in the object code while the gnu utilities use a field in the

    relocation record for this.  To support the hp native format, the object

    code would need to be patched with the offsets when producing .o files.

    The basic technique taken in this implementation is to #include the code

    from aoutx.h and aout-target.h with appropriate #defines to override

    code where a unique implementation is needed:

    {

        #define a bunch of stuff

        #include <aoutx.h>

        implement a bunch of functions

        #include "aout-target.h"

    }

    The hp symbol table is a bit different than other a.out targets.  Instead

    of having an array of nlist items and an array of strings, hp's format

    has them mixed together in one structure.  In addition, the strings are

    not null terminated.  It looks something like this:

    nlist element 1

    string1

    nlist element 2

    string2

    ...

    The whole symbol table is read as one chunk and then we march thru it

    and convert it to canonical form.  As we march thru the table, we copy

    the nlist data into the internal form and we compact the strings and null

    terminate them, using storage from the already allocated symbol table:

    string1

    null

    string2

    null

 */

/* @@ Is this really so different from normal a.out that it needs to include

   aoutx.h?  We should go through this file sometime and see what can be made

   more dependent on aout32.o and what might need to be broken off and accessed

   through the backend_data field.  Or, maybe we really do need such a

   completely separate implementation.  I don't have time to investigate this

   much further right now.  [raeburn:19930428.2124EST] */

/* @@ Also, note that there wind up being two versions of some routines, with

   different names, only one of which actually gets used.  For example:

        slurp_symbol_table

        swap_std_reloc_in

        slurp_reloc_table

        canonicalize_symtab

        get_symtab_upper_bound

        canonicalize_reloc

        mkobject

   This should also be fixed.  */

#define TARGETNAME "a.out-hp300hpux"

/* Do not "beautify" the CONCAT* macro args.  Traditional C will not

   remove whitespace added here, and thus will fail to concatenate

   the tokens.  */

#define MY(OP) CONCAT2 (hp300hpux_,OP)

#define external_exec hp300hpux_exec_bytes

#define external_nlist hp300hpux_nlist_bytes

#include "aout/hp300hpux.h"

/* define these so we can compile unused routines in aoutx.h */

#define e_strx  e_shlib

#define e_other e_length

#define e_desc  e_almod

#define AR_PAD_CHAR '/'

#define TARGET_IS_BIG_ENDIAN_P

#define DEFAULT_ARCH bfd_arch_m68k

#define MY_get_section_contents aout_32_get_section_contents

#define MY_slurp_armap bfd_slurp_bsd_armap_f2

/***********************************************/

/* provide overrides for routines in this file */

/***********************************************/

/* these don't use MY because that causes problems within JUMP_TABLE

   (CONCAT2 winds up being expanded recursively, which ANSI C compilers

   will not do).  */

#define MY_canonicalize_symtab hp300hpux_canonicalize_symtab

#define MY_get_symtab_upper_bound hp300hpux_get_symtab_upper_bound

#define MY_canonicalize_reloc hp300hpux_canonicalize_reloc

#define MY_write_object_contents hp300hpux_write_object_contents

#define MY_read_minisymbols _bfd_generic_read_minisymbols

#define MY_minisymbol_to_symbol _bfd_generic_minisymbol_to_symbol

#define MY_bfd_link_hash_table_create _bfd_generic_link_hash_table_create

#define MY_bfd_link_add_symbols _bfd_generic_link_add_symbols

#define MY_final_link_callback unused

#define MY_bfd_final_link _bfd_generic_final_link

/* Until and unless we convert the slurp_reloc and slurp_symtab

   routines in this file, we can not use the default aout

   free_cached_info routine which assumes that the relocs and symtabs

   were allocated using malloc.  */

#define MY_bfd_free_cached_info bfd_true

#define hp300hpux_write_syms aout_32_write_syms

#define MY_callback MY(callback)

#define MY_exec_hdr_flags 0x2

#define NAME_swap_exec_header_in NAME(hp300hpux_32_,swap_exec_header_in)

#define HP_SYMTYPE_UNDEFINED    0x00

#define HP_SYMTYPE_ABSOLUTE     0x01

#define HP_SYMTYPE_TEXT         0x02

#define HP_SYMTYPE_DATA         0x03

#define HP_SYMTYPE_BSS          0x04

#define HP_SYMTYPE_COMMON       0x05

#define HP_SYMTYPE_TYPE         0x0F

#define HP_SYMTYPE_FILENAME     0x1F

#define HP_SYMTYPE_ALIGN        0x10

#define HP_SYMTYPE_EXTERNAL     0x20

#define HP_SECONDARY_SYMBOL     0x40

/* RELOCATION DEFINITIONS */

#define HP_RSEGMENT_TEXT        0x00

#define HP_RSEGMENT_DATA        0x01

#define HP_RSEGMENT_BSS         0x02

#define HP_RSEGMENT_EXTERNAL    0x03

#define HP_RSEGMENT_PCREL       0x04

#define HP_RSEGMENT_RDLT        0x05

#define HP_RSEGMENT_RPLT        0x06

#define HP_RSEGMENT_NOOP        0x3F

#define HP_RLENGTH_BYTE         0x00

#define HP_RLENGTH_WORD         0x01

#define HP_RLENGTH_LONG         0x02

#define HP_RLENGTH_ALIGN        0x03

#define NAME(x,y) CONCAT3 (hp300hpux,_32_,y)

#define ARCH_SIZE 32

/* aoutx.h requires definitions for BMAGIC and QMAGIC.  */

#define BMAGIC HPUX_DOT_O_MAGIC

#define QMAGIC 0314

#include "aoutx.h"

static const bfd_target * MY (callback)

  PARAMS ((bfd *));

static bfd_boolean MY (write_object_contents)

  PARAMS ((bfd *));

static void convert_sym_type

  PARAMS ((struct external_nlist *, aout_symbol_type *, bfd *));

bfd_boolean MY (slurp_symbol_table)

  PARAMS ((bfd *));

void MY (swap_std_reloc_in)

  PARAMS ((bfd *, struct hp300hpux_reloc *, arelent *, asymbol **,

           bfd_size_type));

bfd_boolean MY (slurp_reloc_table)

  PARAMS ((bfd *, sec_ptr, asymbol **));

long MY (canonicalize_symtab)

  PARAMS ((bfd *, asymbol **));

long MY (get_symtab_upper_bound)

  PARAMS ((bfd *));

long MY (canonicalize_reloc)

  PARAMS ((bfd *, sec_ptr, arelent **, asymbol **));

/* Since the hpux symbol table has nlist elements interspersed with

   strings and we need to insert som strings for secondary symbols, we

   give ourselves a little extra padding up front to account for

   this.  Note that for each non-secondary symbol we process, we gain

   9 bytes of space for the discarded nlist element (one byte used for

   null).  SYM_EXTRA_BYTES is the extra space.  */

#define SYM_EXTRA_BYTES   1024

/* Set parameters about this a.out file that are machine-dependent.

   This routine is called from some_aout_object_p just before it returns.  */

static const bfd_target *

MY (callback) (abfd)

     bfd *abfd;

{

  struct internal_exec *execp = exec_hdr (abfd);

  /* Calculate the file positions of the parts of a newly read aout header */

  obj_textsec (abfd)->size = N_TXTSIZE (*execp);

  /* The virtual memory addresses of the sections */

  obj_textsec (abfd)->vma = N_TXTADDR (*execp);

  obj_datasec (abfd)->vma = N_DATADDR (*execp);

  obj_bsssec (abfd)->vma = N_BSSADDR (*execp);

  obj_textsec (abfd)->lma = obj_textsec (abfd)->vma;

  obj_datasec (abfd)->lma = obj_datasec (abfd)->vma;

  obj_bsssec (abfd)->lma = obj_bsssec (abfd)->vma;

  /* The file offsets of the sections */

  obj_textsec (abfd)->filepos = N_TXTOFF (*execp);

  obj_datasec (abfd)->filepos = N_DATOFF (*execp);

  /* The file offsets of the relocation info */

  obj_textsec (abfd)->rel_filepos = N_TRELOFF (*execp);

  obj_datasec (abfd)->rel_filepos = N_DRELOFF (*execp);

  /* The file offsets of the string table and symbol table.  */

  obj_sym_filepos (abfd) = N_SYMOFF (*execp);

  obj_str_filepos (abfd) = N_STROFF (*execp);

  /* Determine the architecture and machine type of the object file.  */

#ifdef SET_ARCH_MACH

  SET_ARCH_MACH (abfd, *execp);

#else

  bfd_default_set_arch_mach (abfd, DEFAULT_ARCH, 0);

#endif

  if (obj_aout_subformat (abfd) == gnu_encap_format)

    {

      /* The file offsets of the relocation info */

      obj_textsec (abfd)->rel_filepos = N_GNU_TRELOFF (*execp);

      obj_datasec (abfd)->rel_filepos = N_GNU_DRELOFF (*execp);

      /* The file offsets of the string table and symbol table.  */

      obj_sym_filepos (abfd) = N_GNU_SYMOFF (*execp);

      obj_str_filepos (abfd) = (obj_sym_filepos (abfd) + execp->a_syms);

      abfd->flags |= HAS_LINENO | HAS_DEBUG | HAS_SYMS | HAS_LOCALS;

      bfd_get_symcount (abfd) = execp->a_syms / 12;

      obj_symbol_entry_size (abfd) = 12;

      obj_reloc_entry_size (abfd) = RELOC_STD_SIZE;

    }

  return abfd->xvec;

}

extern bfd_boolean aout_32_write_syms

  PARAMS ((bfd * abfd));

static bfd_boolean

MY (write_object_contents) (abfd)

     bfd *abfd;

{

  struct external_exec exec_bytes;

  struct internal_exec *execp = exec_hdr (abfd);

  bfd_size_type text_size;      /* dummy vars */

  file_ptr text_end;

  memset (&exec_bytes, 0, sizeof (exec_bytes));

  obj_reloc_entry_size (abfd) = RELOC_STD_SIZE;

  if (adata (abfd).magic == undecided_magic)

    NAME (aout,adjust_sizes_and_vmas) (abfd, &text_size, &text_end);

  execp->a_syms = 0;

  execp->a_entry = bfd_get_start_address (abfd);

  execp->a_trsize = ((obj_textsec (abfd)->reloc_count) *

                     obj_reloc_entry_size (abfd));

  execp->a_drsize = ((obj_datasec (abfd)->reloc_count) *

                     obj_reloc_entry_size (abfd));

  N_SET_MACHTYPE (*execp, 0xc);

  N_SET_FLAGS (*execp, aout_backend_info (abfd)->exec_hdr_flags);

  NAME (aout,swap_exec_header_out) (abfd, execp, &exec_bytes);

  /* update fields not covered by default swap_exec_header_out */

  /* this is really the sym table size but we store it in drelocs */

  H_PUT_32 (abfd, (bfd_get_symcount (abfd) * 12), exec_bytes.e_drelocs);

  if (bfd_seek (abfd, (file_ptr) 0, FALSE) != 0

      || (bfd_bwrite ((PTR) &exec_bytes, (bfd_size_type) EXEC_BYTES_SIZE, abfd)

          != EXEC_BYTES_SIZE))

    return FALSE;

  /* Write out the symbols, and then the relocs.  We must write out

       the symbols first so that we know the symbol indices.  */

  if (bfd_get_symcount (abfd) != 0)

    {

      /* Skip the relocs to where we want to put the symbols.  */

      if (bfd_seek (abfd, (file_ptr) (N_DRELOFF (*execp) + execp->a_drsize),

                    SEEK_SET) != 0)

        return FALSE;

    }

  if (!MY (write_syms) (abfd))

    return FALSE;

  if (bfd_get_symcount (abfd) != 0)

    {

      if (bfd_seek (abfd, (file_ptr) N_TRELOFF (*execp), SEEK_CUR) != 0)

        return FALSE;

      if (!NAME (aout,squirt_out_relocs) (abfd, obj_textsec (abfd)))

        return FALSE;

      if (bfd_seek (abfd, (file_ptr) N_DRELOFF (*execp), SEEK_CUR) != 0)

        return FALSE;

      if (!NAME (aout,squirt_out_relocs) (abfd, obj_datasec (abfd)))

        return FALSE;

    }

  return TRUE;

}

/* convert the hp symbol type to be the same as aout64.h usage so we */

/* can piggyback routines in aoutx.h.                                */

static void

convert_sym_type (sym_pointer, cache_ptr, abfd)

     struct external_nlist *sym_pointer ATTRIBUTE_UNUSED;

     aout_symbol_type *cache_ptr;

     bfd *abfd ATTRIBUTE_UNUSED;

{

  int name_type;

  int new_type;

  name_type = (cache_ptr->type);

  new_type = 0;

  if ((name_type & HP_SYMTYPE_ALIGN) != 0)

    {

      /* iou_error ("aligned symbol encountered: %s", name);*/

      name_type = 0;

    }

  if (name_type == HP_SYMTYPE_FILENAME)

    new_type = N_FN;

  else

    {

      switch (name_type & HP_SYMTYPE_TYPE)

        {

        case HP_SYMTYPE_UNDEFINED:

          new_type = N_UNDF;

          break;

        case HP_SYMTYPE_ABSOLUTE:

          new_type = N_ABS;

          break;

        case HP_SYMTYPE_TEXT:

          new_type = N_TEXT;

          break;

        case HP_SYMTYPE_DATA:

          new_type = N_DATA;

          break;

        case HP_SYMTYPE_BSS:

          new_type = N_BSS;

          break;

        case HP_SYMTYPE_COMMON:

          new_type = N_COMM;

          break;

        default:

          abort ();

          break;

        }

      if (name_type & HP_SYMTYPE_EXTERNAL)

        new_type |= N_EXT;

      if (name_type & HP_SECONDARY_SYMBOL)

        {

          switch (new_type)

            {

            default:

              abort ();

            case N_UNDF | N_EXT:

              /* If the value is nonzero, then just treat this as a

                 common symbol.  I don't know if this is correct in

                 all cases, but it is more correct than treating it as

                 a weak undefined symbol.  */

              if (cache_ptr->symbol.value == 0)

                new_type = N_WEAKU;

              break;

            case N_ABS | N_EXT:

              new_type = N_WEAKA;

              break;

            case N_TEXT | N_EXT:

              new_type = N_WEAKT;

              break;

            case N_DATA | N_EXT:

              new_type = N_WEAKD;

              break;

            case N_BSS | N_EXT:

              new_type = N_WEAKB;

              break;

            }

        }

    }

  cache_ptr->type = new_type;

}

/*

DESCRIPTION

        Swaps the information in an executable header taken from a raw

        byte stream memory image, into the internal exec_header

        structure.

*/

void

NAME (aout,swap_exec_header_in) (abfd, raw_bytes, execp)

     bfd *abfd;

     struct external_exec *raw_bytes;

     struct internal_exec *execp;

{

  struct external_exec *bytes = (struct external_exec *) raw_bytes;

  /* The internal_exec structure has some fields that are unused in this

     configuration (IE for i960), so ensure that all such uninitialized

     fields are zero'd out.  There are places where two of these structs

     are memcmp'd, and thus the contents do matter. */

  memset (execp, 0, sizeof (struct internal_exec));

  /* Now fill in fields in the execp, from the bytes in the raw data.  */

  execp->a_info = H_GET_32 (abfd, bytes->e_info);

  execp->a_text = GET_WORD (abfd, bytes->e_text);

  execp->a_data = GET_WORD (abfd, bytes->e_data);

  execp->a_bss = GET_WORD (abfd, bytes->e_bss);

  execp->a_syms = GET_WORD (abfd, bytes->e_syms);

  execp->a_entry = GET_WORD (abfd, bytes->e_entry);

  execp->a_trsize = GET_WORD (abfd, bytes->e_trsize);

  execp->a_drsize = GET_WORD (abfd, bytes->e_drsize);

  /***************************************************************/

  /* check the header to see if it was generated by a bfd output */

  /* this is detected rather bizarrely by requiring a bunch of   */

  /* header fields to be zero and an old unused field (now used) */

  /* to be set.                                                  */

  /***************************************************************/

  do

    {

      long syms;

      struct aout_data_struct *rawptr;

      bfd_size_type amt;

      if (H_GET_32 (abfd, bytes->e_passize) != 0)

        break;

      if (H_GET_32 (abfd, bytes->e_syms) != 0)

        break;

      if (H_GET_32 (abfd, bytes->e_supsize) != 0)

        break;

      syms = H_GET_32 (abfd, bytes->e_drelocs);

      if (syms == 0)

        break;

      /* OK, we've passed the test as best as we can determine */

      execp->a_syms = syms;

      /* allocate storage for where we will store this result */

      amt = sizeof (*rawptr);

      rawptr = (struct aout_data_struct *) bfd_zalloc (abfd, amt);

      if (rawptr == NULL)

        return;

      abfd->tdata.aout_data = rawptr;

      obj_aout_subformat (abfd) = gnu_encap_format;

    }

  while (0);

}

/* The hp symbol table is a bit different than other a.out targets.  Instead

   of having an array of nlist items and an array of strings, hp's format

   has them mixed together in one structure.  In addition, the strings are

   not null terminated.  It looks something like this:

   nlist element 1

   string1

   nlist element 2

   string2

   ...

   The whole symbol table is read as one chunk and then we march thru it

   and convert it to canonical form.  As we march thru the table, we copy

   the nlist data into the internal form and we compact the strings and null

   terminate them, using storage from the already allocated symbol table:

   string1

   null

   string2

   null

   ...

*/

bfd_boolean

MY (slurp_symbol_table) (abfd)

     bfd *abfd;

{

  bfd_size_type symbol_bytes;

  struct external_nlist *syms;

  struct external_nlist *sym_pointer;

  struct external_nlist *sym_end;

  char *strings;

  aout_symbol_type *cached;

  unsigned num_syms = 0;

  bfd_size_type amt;

  /* If there's no work to be done, don't do any */

  if (obj_aout_symbols (abfd) != (aout_symbol_type *) NULL)

    return TRUE;

  symbol_bytes = exec_hdr (abfd)->a_syms;

  amt = symbol_bytes + SYM_EXTRA_BYTES;

  strings = (char *) bfd_alloc (abfd, amt);

  if (!strings)

    return FALSE;

  syms = (struct external_nlist *) (strings + SYM_EXTRA_BYTES);

  if (bfd_seek (abfd, obj_sym_filepos (abfd), SEEK_SET) != 0

      || bfd_bread ((PTR) syms, symbol_bytes, abfd) != symbol_bytes)

    {

      bfd_release (abfd, syms);

      return FALSE;

    }

  sym_end = (struct external_nlist *) (((char *) syms) + symbol_bytes);

  /* first, march thru the table and figure out how many symbols there are */

  for (sym_pointer = syms; sym_pointer < sym_end; sym_pointer++, num_syms++)

    {

      /* skip over the embedded symbol. */

      sym_pointer = (struct external_nlist *) (((char *) sym_pointer) +

                                               sym_pointer->e_length[0]);

    }

  /* now that we know the symbol count, update the bfd header */

  bfd_get_symcount (abfd) = num_syms;

  amt = num_syms;

  amt *= sizeof (aout_symbol_type);

  cached = (aout_symbol_type *) bfd_zalloc (abfd, amt);

  if (cached == NULL && num_syms != 0)

    return FALSE;

  /* as we march thru the hp symbol table, convert it into a list of

     null terminated strings to hold the symbol names.  Make sure any

     assignment to the strings pointer is done after we're thru using

     the nlist so we don't overwrite anything important. */

  /* OK, now walk the new symtable, caching symbol properties */

  {

    aout_symbol_type *cache_ptr = cached;

    /* Run through table and copy values */

    for (sym_pointer = syms, cache_ptr = cached;

         sym_pointer < sym_end; sym_pointer++, cache_ptr++)

      {

        unsigned int length;

        cache_ptr->symbol.the_bfd = abfd;

        cache_ptr->symbol.value = GET_SWORD (abfd, sym_pointer->e_value);

        cache_ptr->desc = bfd_get_16 (abfd, sym_pointer->e_almod);

        cache_ptr->type = bfd_get_8 (abfd, sym_pointer->e_type);

        cache_ptr->symbol.udata.p = NULL;

        length = bfd_get_8 (abfd, sym_pointer->e_length);

        cache_ptr->other = length;      /* other not used, save length here */

        convert_sym_type (sym_pointer, cache_ptr, abfd);

        if (!translate_from_native_sym_flags (abfd, cache_ptr))

          return FALSE;

        /********************************************************/

        /* for hpux, the 'length' value indicates the length of */

        /* the symbol name which follows the nlist entry.       */

        /********************************************************/

        if (length)

          {

            /**************************************************************/

            /* the hp string is not null terminated so we create a new one*/

            /* by copying the string to overlap the just vacated nlist    */

            /* structure before it in memory.                             */

            /**************************************************************/

            cache_ptr->symbol.name = strings;

            memcpy (strings, sym_pointer + 1, length);

            strings[length] = '\0';

            strings += length + 1;

          }

        else

          cache_ptr->symbol.name = (char *) NULL;

        /* skip over the embedded symbol. */

        sym_pointer = (struct external_nlist *) (((char *) sym_pointer) +

                                                 length);

      }

  }

  obj_aout_symbols (abfd) = cached;

  return TRUE;

}

void

MY (swap_std_reloc_in) (abfd, bytes, cache_ptr, symbols, symcount)

     bfd *abfd;

     struct hp300hpux_reloc *bytes;

     arelent *cache_ptr;

     asymbol **symbols;

     bfd_size_type symcount ATTRIBUTE_UNUSED;

{

  int r_index;

  int r_extern = 0;

  unsigned int r_length;

  int r_pcrel = 0;

  struct aoutdata *su = &(abfd->tdata.aout_data->a);

  cache_ptr->address = H_GET_32 (abfd, bytes->r_address);

  r_index = H_GET_16 (abfd, bytes->r_index);

  switch (bytes->r_type[0])

    {

    case HP_RSEGMENT_TEXT:

      r_index = N_TEXT;

      break;

    case HP_RSEGMENT_DATA:

      r_index = N_DATA;

      break;

    case HP_RSEGMENT_BSS:

      r_index = N_BSS;

      break;

    case HP_RSEGMENT_EXTERNAL:

      r_extern = 1;

      break;

    case HP_RSEGMENT_PCREL:

      r_extern = 1;

      r_pcrel = 1;

      break;

    case HP_RSEGMENT_RDLT:

      break;

    case HP_RSEGMENT_RPLT:

      break;

    case HP_RSEGMENT_NOOP:

      break;

    default:

      abort ();

      break;

    }

  switch (bytes->r_length[0])

    {

    case HP_RLENGTH_BYTE:

      r_length = 0;

      break;

    case HP_RLENGTH_WORD:

      r_length = 1;

      break;

    case HP_RLENGTH_LONG:

      r_length = 2;

      break;

    default:

      abort ();

      break;

    }

  cache_ptr->howto = howto_table_std + r_length + 4 * r_pcrel;

  /* FIXME-soon:  Roll baserel, jmptable, relative bits into howto setting */

  /* This macro uses the r_index value computed above */

  if (r_pcrel && r_extern)

    {

      /* The GNU linker assumes any offset from beginning of section */

      /* is already incorporated into the image while the HP linker  */

      /* adds this in later.  Add it in now...                       */

      MOVE_ADDRESS (-cache_ptr->address);

    }

  else

    {

      MOVE_ADDRESS (0);

    }

}