| /* BFD back-end for PDP-11 a.out binaries. | 
 |    Copyright (C) 2001-2023 Free Software Foundation, Inc. | 
 |  | 
 |    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. */ | 
 |  | 
 |  | 
 | /* BFD backend for PDP-11, running 2.11BSD in particular. | 
 |  | 
 |    This file was hacked up by looking hard at the existing vaxnetbsd | 
 |    back end and the header files in 2.11BSD.  The symbol table format | 
 |    of 2.11BSD has been extended to accommodate .stab symbols.  See | 
 |    struct pdp11_external_nlist below for details. | 
 |  | 
 |    TODO | 
 |    * support for V7 file formats | 
 |    * support for overlay object files (see 2.11 a.out(5)) | 
 |    * support for old and very old archives | 
 |    (see 2.11 ar(5), historical section) | 
 |  | 
 |    Search for TODO to find other areas needing more work.  */ | 
 |  | 
 | #define	BYTES_IN_WORD	2 | 
 | #define	BYTES_IN_LONG	4 | 
 | #define ARCH_SIZE	16 | 
 | #undef TARGET_IS_BIG_ENDIAN_P | 
 |  | 
 | #define	TARGET_PAGE_SIZE	8192 | 
 | #define	SEGMENT__SIZE	TARGET_PAGE_SIZE | 
 |  | 
 | #define	DEFAULT_ARCH	bfd_arch_pdp11 | 
 | #define	DEFAULT_MID	M_PDP11 | 
 |  | 
 | /* 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 (pdp11_aout_,OP) | 
 |  | 
 | /* This needs to start with a.out so GDB knows it is an a.out variant.  */ | 
 | #define TARGETNAME "a.out-pdp11" | 
 |  | 
 | /* This is the normal load address for executables.  */ | 
 | #define TEXT_START_ADDR		0 | 
 |  | 
 | /* The header is not included in the text segment.  */ | 
 | #define N_HEADER_IN_TEXT(x)	0 | 
 |  | 
 | /* There is no flags field.  */ | 
 | #define N_FLAGS(execp)		0 | 
 |  | 
 | #define N_SET_FLAGS(execp, flags) do { } while (0) | 
 | #define N_BADMAG(x) (N_MAGIC(x) != OMAGIC	\ | 
 | 		     && N_MAGIC(x) != NMAGIC	\ | 
 | 		     && N_MAGIC(x) != IMAGIC	\ | 
 | 		     && N_MAGIC(x) != ZMAGIC) | 
 |  | 
 | #include "sysdep.h" | 
 | #include <limits.h> | 
 | #include "bfd.h" | 
 |  | 
 | #define external_exec pdp11_external_exec | 
 | struct pdp11_external_exec | 
 | { | 
 |   bfd_byte e_info[2];		/* Magic number.  */ | 
 |   bfd_byte e_text[2];		/* Length of text section in bytes.  */ | 
 |   bfd_byte e_data[2];		/* Length of data section in bytes.  */ | 
 |   bfd_byte e_bss[2];		/* Length of bss area in bytes.  */ | 
 |   bfd_byte e_syms[2];		/* Length of symbol table in bytes.  */ | 
 |   bfd_byte e_entry[2];		/* Start address.  */ | 
 |   bfd_byte e_unused[2];		/* Not used.  */ | 
 |   bfd_byte e_flag[2];		/* Relocation info stripped.  */ | 
 |   bfd_byte e_relocatable;	/* Ugly hack.  */ | 
 | }; | 
 |  | 
 | #define	EXEC_BYTES_SIZE	(8 * 2) | 
 |  | 
 | #define	A_MAGIC1	OMAGIC | 
 | #define OMAGIC		0407	/* ...object file or impure executable.  */ | 
 | #define	A_MAGIC2	NMAGIC | 
 | #define NMAGIC		0410	/* Pure executable.  */ | 
 | #define ZMAGIC		0413	/* Demand-paged executable.  */ | 
 | #define	IMAGIC		0411	/* Separated I&D.  */ | 
 | #define	A_MAGIC3	IMAGIC | 
 | #define	A_MAGIC4	0405	/* Overlay.  */ | 
 | #define	A_MAGIC5	0430	/* Auto-overlay (nonseparate).  */ | 
 | #define	A_MAGIC6	0431	/* Auto-overlay (separate).  */ | 
 | #define QMAGIC		0 | 
 | #define BMAGIC		0 | 
 |  | 
 | #define A_FLAG_RELOC_STRIPPED	0x0001 | 
 |  | 
 | /* The following struct defines the format of an entry in the object file | 
 |    symbol table.  In the original 2.11BSD struct the index into the string | 
 |    table is stored as a long, but the PDP11 C convention for storing a long in | 
 |    memory placed the most significant word first even though the bytes within a | 
 |    word are stored least significant first.  So here the string table index is | 
 |    considered to be just 16 bits and the first two bytes of the struct were | 
 |    previously named e_unused.  To extend the symbol table format to accommodate | 
 |    .stab symbols, the e_unused bytes are renamed e_desc to store the desc field | 
 |    of the .stab symbol.  The GDP Project's STABS document says that the "other" | 
 |    field is almost always unused and can be set to zero; the only nonzero cases | 
 |    identified were for stabs in their own sections, which does not apply for | 
 |    pdp11 a.out format, and for a special case of GNU Modula2 which is not | 
 |    supported for the PDP11.  */ | 
 | #define external_nlist pdp11_external_nlist | 
 | struct pdp11_external_nlist | 
 | { | 
 |   bfd_byte e_desc[2];		/* The desc field for .stab symbols, else 0.  */ | 
 |   bfd_byte e_strx[2];		/* Index into string table of name.  */ | 
 |   bfd_byte e_type[1];		/* Type of symbol.  */ | 
 |   bfd_byte e_ovly[1];		/* Overlay number.  */ | 
 |   bfd_byte e_value[2];		/* Value of symbol.  */ | 
 | }; | 
 |  | 
 | #define	EXTERNAL_NLIST_SIZE	8 | 
 |  | 
 | #define N_TXTOFF(x)	(EXEC_BYTES_SIZE) | 
 | #define N_DATOFF(x)	(N_TXTOFF(x) + (x)->a_text) | 
 | #define N_TRELOFF(x)	(N_DATOFF(x) + (x)->a_data) | 
 | #define N_DRELOFF(x)	(N_TRELOFF(x) + (x)->a_trsize) | 
 | #define N_SYMOFF(x)	(N_DRELOFF(x) + (x)->a_drsize) | 
 | #define N_STROFF(x)	(N_SYMOFF(x) + (x)->a_syms) | 
 |  | 
 | #define WRITE_HEADERS(abfd, execp) pdp11_aout_write_headers (abfd, execp) | 
 |  | 
 | #include "libbfd.h" | 
 | #include "libaout.h" | 
 |  | 
 | #define SWAP_MAGIC(ext) bfd_getl16 (ext) | 
 |  | 
 | #define MY_entry_is_text_address 1 | 
 |  | 
 | #define MY_write_object_contents MY(write_object_contents) | 
 | static bool MY(write_object_contents) (bfd *); | 
 | #define MY_text_includes_header 1 | 
 |  | 
 | #define MY_BFD_TARGET | 
 |  | 
 | #include "aout-target.h" | 
 |  | 
 | /* Start of modified aoutx.h.  */ | 
 | #define KEEPIT udata.i | 
 |  | 
 | #include <string.h>		/* For strchr and friends.  */ | 
 | #include "bfd.h" | 
 | #include "sysdep.h" | 
 | #include "safe-ctype.h" | 
 | #include "bfdlink.h" | 
 |  | 
 | #include "libaout.h" | 
 | #include "aout/aout64.h" | 
 | #include "aout/stab_gnu.h" | 
 | #include "aout/ar.h" | 
 |  | 
 | /* The symbol type numbers for the 16-bit a.out format from 2.11BSD differ from | 
 |    those defined in aout64.h so we must redefine them here.  N_EXT changes from | 
 |    0x01 to 0x20 which creates a conflict with some .stab values, in particular | 
 |    between undefined externals (N_UNDF+N_EXT) vs. global variables (N_GYSM) and | 
 |    between external bss symbols (N_BSS+N_EXT) vs. function names (N_FUN).  We | 
 |    disambiguate those conflicts with a hack in is_stab() to look for the ':' in | 
 |    the global variable or function name string.  */ | 
 | #undef N_TYPE | 
 | #undef N_UNDF | 
 | #undef N_ABS | 
 | #undef N_TEXT | 
 | #undef N_DATA | 
 | #undef N_BSS | 
 | #undef N_REG | 
 | #undef N_FN | 
 | #undef N_EXT | 
 | #undef N_STAB | 
 | #define N_TYPE		0x1f	/* Type mask.  */ | 
 | #define N_UNDF		0x00	/* Undefined.  */ | 
 | #define N_ABS		0x01	/* Absolute.  */ | 
 | #define N_TEXT		0x02	/* Text segment.  */ | 
 | #define N_DATA		0x03	/* Data segment.  */ | 
 | #define N_BSS		0x04	/* Bss segment.  */ | 
 | #define N_REG		0x14	/* Register symbol.  */ | 
 | #define N_FN		0x1f	/* File name.  */ | 
 | #define N_EXT		0x20	/* External flag.  */ | 
 | /* Type numbers from .stab entries that could conflict: | 
 | 	N_GSYM		0x20	   Global variable [conflict with external undef] | 
 | 	N_FNAME		0x22	   Function name (for BSD Fortran) [ignored] | 
 | 	N_FUN		0x24	   Function name [conflict with external BSS] | 
 | 	N_NOMAP		0x34	   No DST map for sym. [ext. reg. doesn't exist] | 
 | */ | 
 |  | 
 | #define RELOC_SIZE 2 | 
 |  | 
 | #define RELFLG		0x0001	/* PC-relative flag.  */ | 
 | #define RTYPE		0x000e	/* Type mask.  */ | 
 | #define RIDXMASK	0xfff0	/* Index mask.  */ | 
 |  | 
 | #define RABS		0x00	/* Absolute.  */ | 
 | #define RTEXT		0x02	/* Text.  */ | 
 | #define RDATA		0x04	/* Data.  */ | 
 | #define RBSS		0x06	/* Bss.  */ | 
 | #define REXT		0x08	/* External.  */ | 
 |  | 
 | #define RINDEX(x)	(((x) & 0xfff0) >> 4) | 
 |  | 
 | #ifndef MY_final_link_relocate | 
 | #define MY_final_link_relocate _bfd_final_link_relocate | 
 | #endif | 
 |  | 
 | #ifndef MY_relocate_contents | 
 | #define MY_relocate_contents _bfd_relocate_contents | 
 | #endif | 
 |  | 
 | /* A hash table used for header files with N_BINCL entries.  */ | 
 |  | 
 | struct aout_link_includes_table | 
 | { | 
 |   struct bfd_hash_table root; | 
 | }; | 
 |  | 
 | /* A linked list of totals that we have found for a particular header | 
 |    file.  */ | 
 |  | 
 | struct aout_link_includes_totals | 
 | { | 
 |   struct aout_link_includes_totals *next; | 
 |   bfd_vma total; | 
 | }; | 
 |  | 
 | /* An entry in the header file hash table.  */ | 
 |  | 
 | struct aout_link_includes_entry | 
 | { | 
 |   struct bfd_hash_entry root; | 
 |   /* List of totals we have found for this file.  */ | 
 |   struct aout_link_includes_totals *totals; | 
 | }; | 
 |  | 
 | /* During the final link step we need to pass around a bunch of | 
 |    information, so we do it in an instance of this structure.  */ | 
 |  | 
 | struct aout_final_link_info | 
 | { | 
 |   /* General link information.  */ | 
 |   struct bfd_link_info *info; | 
 |   /* Output bfd.  */ | 
 |   bfd *output_bfd; | 
 |   /* Reloc file positions.  */ | 
 |   file_ptr treloff, dreloff; | 
 |   /* File position of symbols.  */ | 
 |   file_ptr symoff; | 
 |   /* String table.  */ | 
 |   struct bfd_strtab_hash *strtab; | 
 |   /* Header file hash table.  */ | 
 |   struct aout_link_includes_table includes; | 
 |   /* A buffer large enough to hold the contents of any section.  */ | 
 |   bfd_byte *contents; | 
 |   /* A buffer large enough to hold the relocs of any section.  */ | 
 |   void * relocs; | 
 |   /* A buffer large enough to hold the symbol map of any input BFD.  */ | 
 |   int *symbol_map; | 
 |   /* A buffer large enough to hold output symbols of any input BFD.  */ | 
 |   struct external_nlist *output_syms; | 
 | }; | 
 |  | 
 | /* Copy of the link_info.separate_code boolean to select the output format with | 
 |    separate instruction and data spaces selected by --imagic */ | 
 | static bool separate_i_d = false; | 
 |  | 
 | reloc_howto_type howto_table_pdp11[] = | 
 | { | 
 |   /* type	       rs size bsz  pcrel bitpos ovrf			  sf name     part_inpl readmask  setmask    pcdone */ | 
 | HOWTO( 0,	       0,  2,  16,  false, 0, complain_overflow_dont,0,"16",	true, 0x0000ffff,0x0000ffff, false), | 
 | HOWTO( 1,	       0,  2,  16,  true,  0, complain_overflow_dont,0,"DISP16",	true, 0x0000ffff,0x0000ffff, false), | 
 | HOWTO( 2,	       0,  4,  32,  false, 0, complain_overflow_dont,0,"32",	true, 0x0000ffff,0x0000ffff, false), | 
 | }; | 
 |  | 
 | #define TABLE_SIZE(TABLE)	(sizeof(TABLE)/sizeof(TABLE[0])) | 
 |  | 
 |  | 
 | static bool aout_link_check_archive_element (bfd *, struct bfd_link_info *, | 
 | 					     struct bfd_link_hash_entry *, | 
 | 					     const char *, bool *); | 
 | static bool aout_link_add_object_symbols (bfd *, struct bfd_link_info *); | 
 | static bool aout_link_add_symbols (bfd *, struct bfd_link_info *); | 
 | static bool aout_link_write_symbols (struct aout_final_link_info *, bfd *); | 
 |  | 
 |  | 
 | reloc_howto_type * | 
 | NAME (aout, reloc_type_lookup) (bfd * abfd ATTRIBUTE_UNUSED, | 
 | 				bfd_reloc_code_real_type code) | 
 | { | 
 |   switch (code) | 
 |     { | 
 |     case BFD_RELOC_16: | 
 |       return &howto_table_pdp11[0]; | 
 |     case BFD_RELOC_16_PCREL: | 
 |       return &howto_table_pdp11[1]; | 
 |     case BFD_RELOC_32: | 
 |       return &howto_table_pdp11[2]; | 
 |     default: | 
 |       return NULL; | 
 |     } | 
 | } | 
 |  | 
 | reloc_howto_type * | 
 | NAME (aout, reloc_name_lookup) (bfd *abfd ATTRIBUTE_UNUSED, | 
 | 				      const char *r_name) | 
 | { | 
 |   unsigned int i; | 
 |  | 
 |   for (i = 0; | 
 |        i < sizeof (howto_table_pdp11) / sizeof (howto_table_pdp11[0]); | 
 |        i++) | 
 |     if (howto_table_pdp11[i].name != NULL | 
 | 	&& strcasecmp (howto_table_pdp11[i].name, r_name) == 0) | 
 |       return &howto_table_pdp11[i]; | 
 |  | 
 |   return NULL; | 
 | } | 
 |  | 
 | /* Disambiguate conflicts between normal symbol types and .stab symbol types | 
 |    (undefined externals N_UNDF+N_EXT vs. global variables N_GYSM and external | 
 |    bss symbols N_BSS+N_EXT vs. function names N_FUN) with a hack to look for | 
 |    the ':' in the global variable or function name string.  */ | 
 |  | 
 | static int | 
 | is_stab (int type, const char *name) | 
 | { | 
 |   if (type == N_GSYM || type == N_FUN) | 
 |     return strchr (name, ':') != NULL; | 
 |   return type > N_FUN; | 
 | } | 
 |  | 
 | static int | 
 | pdp11_aout_write_headers (bfd *abfd, struct internal_exec *execp) | 
 | { | 
 |   struct external_exec exec_bytes; | 
 |  | 
 |   if (adata(abfd).magic == undecided_magic) | 
 |     NAME (aout, adjust_sizes_and_vmas) (abfd); | 
 |  | 
 |   execp->a_syms = bfd_get_symcount (abfd) * EXTERNAL_NLIST_SIZE; | 
 |   execp->a_entry = bfd_get_start_address (abfd); | 
 |  | 
 |   if (obj_textsec (abfd)->reloc_count > 0 | 
 |       || obj_datasec (abfd)->reloc_count > 0) | 
 |     { | 
 |       execp->a_trsize = execp->a_text; | 
 |       execp->a_drsize = execp->a_data; | 
 |     } | 
 |   else | 
 |     { | 
 |       execp->a_trsize = 0; | 
 |       execp->a_drsize = 0; | 
 |     } | 
 |  | 
 |   NAME (aout, swap_exec_header_out) (abfd, execp, & exec_bytes); | 
 |  | 
 |   if (bfd_seek (abfd, (file_ptr) 0, SEEK_SET) != 0) | 
 |     return false; | 
 |  | 
 |   if (bfd_bwrite ((void *) &exec_bytes, (bfd_size_type) EXEC_BYTES_SIZE, abfd) | 
 |       != EXEC_BYTES_SIZE) | 
 |     return false; | 
 |  | 
 |   /* Now write out reloc info, followed by syms and strings.  */ | 
 |   if (bfd_get_outsymbols (abfd) != NULL | 
 |       && bfd_get_symcount (abfd) != 0) | 
 |     { | 
 |       if (bfd_seek (abfd, (file_ptr) (N_SYMOFF (execp)), SEEK_SET) != 0) | 
 | 	return false; | 
 |  | 
 |       if (! NAME (aout, write_syms) (abfd)) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   if (obj_textsec (abfd)->reloc_count > 0 | 
 |       || obj_datasec (abfd)->reloc_count > 0) | 
 |     { | 
 |       if (bfd_seek (abfd, (file_ptr) (N_TRELOFF (execp)), SEEK_SET) != 0 | 
 | 	  || !NAME (aout, squirt_out_relocs) (abfd, obj_textsec (abfd)) | 
 | 	  || bfd_seek (abfd, (file_ptr) (N_DRELOFF (execp)), SEEK_SET) != 0 | 
 | 	  || !NAME (aout, squirt_out_relocs) (abfd, obj_datasec (abfd))) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Write an object file. | 
 |    Section contents have already been written.  We write the | 
 |    file header, symbols, and relocation.  */ | 
 |  | 
 | static bool | 
 | MY(write_object_contents) (bfd *abfd) | 
 | { | 
 |   struct internal_exec *execp = exec_hdr (abfd); | 
 |  | 
 |   /* We must make certain that the magic number has been set.  This | 
 |      will normally have been done by set_section_contents, but only if | 
 |      there actually are some section contents.  */ | 
 |   if (! abfd->output_has_begun) | 
 |     NAME (aout, adjust_sizes_and_vmas) (abfd); | 
 |  | 
 |   obj_reloc_entry_size (abfd) = RELOC_SIZE; | 
 |  | 
 |   return WRITE_HEADERS (abfd, execp); | 
 | } | 
 |  | 
 | /* Swap the information in an executable header @var{raw_bytes} taken | 
 |    from a raw byte stream memory image into the internal exec header | 
 |    structure "execp".  */ | 
 |  | 
 | #ifndef NAME_swap_exec_header_in | 
 | void | 
 | NAME (aout, swap_exec_header_in) (bfd *abfd, | 
 | 				  struct external_exec *bytes, | 
 | 				  struct internal_exec *execp) | 
 | { | 
 |   /* 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 ((void *) execp, 0, sizeof (struct internal_exec)); | 
 |   /* Now fill in fields in the execp, from the bytes in the raw data.  */ | 
 |   execp->a_info   = GET_MAGIC (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); | 
 |  | 
 |   if (GET_WORD (abfd, bytes->e_flag) & A_FLAG_RELOC_STRIPPED) | 
 |     { | 
 |       execp->a_trsize = 0; | 
 |       execp->a_drsize = 0; | 
 |     } | 
 |   else | 
 |     { | 
 |       execp->a_trsize = execp->a_text; | 
 |       execp->a_drsize = execp->a_data; | 
 |     } | 
 | } | 
 | #define NAME_swap_exec_header_in NAME (aout, swap_exec_header_in) | 
 | #endif | 
 |  | 
 | /*  Swap the information in an internal exec header structure | 
 |     "execp" into the buffer "bytes" ready for writing to disk.  */ | 
 | void | 
 | NAME (aout, swap_exec_header_out) (bfd *abfd, | 
 | 				   struct internal_exec *execp, | 
 | 				   struct external_exec *bytes) | 
 | { | 
 |   /* Now fill in fields in the raw data, from the fields in the exec struct.  */ | 
 |   PUT_MAGIC (abfd, execp->a_info,		bytes->e_info); | 
 |   PUT_WORD (abfd, execp->a_text,		bytes->e_text); | 
 |   PUT_WORD (abfd, execp->a_data,		bytes->e_data); | 
 |   PUT_WORD (abfd, execp->a_bss,			bytes->e_bss); | 
 |   PUT_WORD (abfd, execp->a_syms,		bytes->e_syms); | 
 |   PUT_WORD (abfd, execp->a_entry,		bytes->e_entry); | 
 |   PUT_WORD (abfd, 0,				bytes->e_unused); | 
 |  | 
 |   if ((execp->a_trsize == 0 || execp->a_text == 0) | 
 |       && (execp->a_drsize == 0 || execp->a_data == 0)) | 
 |     PUT_WORD (abfd, A_FLAG_RELOC_STRIPPED, bytes->e_flag); | 
 |   else if (execp->a_trsize == execp->a_text | 
 | 	   && execp->a_drsize == execp->a_data) | 
 |     PUT_WORD (abfd, 0, bytes->e_flag); | 
 |   else | 
 |     { | 
 |       /* TODO: print a proper warning message.  */ | 
 |       fprintf (stderr, "BFD:%s:%d: internal error\n", __FILE__, __LINE__); | 
 |       PUT_WORD (abfd, 0,			bytes->e_flag); | 
 |     } | 
 | } | 
 |  | 
 | /* Make all the section for an a.out file.  */ | 
 |  | 
 | bool | 
 | NAME (aout, make_sections) (bfd *abfd) | 
 | { | 
 |   if (obj_textsec (abfd) == NULL && bfd_make_section (abfd, ".text") == NULL) | 
 |     return false; | 
 |   if (obj_datasec (abfd) == NULL && bfd_make_section (abfd, ".data") == NULL) | 
 |     return false; | 
 |   if (obj_bsssec (abfd) == NULL  && bfd_make_section (abfd, ".bss") == NULL) | 
 |     return false; | 
 |   return true; | 
 | } | 
 |  | 
 | /* Some a.out variant thinks that the file open in ABFD | 
 |    checking is an a.out file.  Do some more checking, and set up | 
 |    for access if it really is.  Call back to the calling | 
 |    environment's "finish up" function just before returning, to | 
 |    handle any last-minute setup.  */ | 
 |  | 
 | bfd_cleanup | 
 | NAME (aout, some_aout_object_p) (bfd *abfd, | 
 | 				 struct internal_exec *execp, | 
 | 				 bfd_cleanup (*callback_to_real_object_p) (bfd *)) | 
 | { | 
 |   struct aout_data_struct *rawptr, *oldrawptr; | 
 |   bfd_cleanup cleanup; | 
 |   size_t amt = sizeof (struct aout_data_struct); | 
 |  | 
 |   rawptr = bfd_zalloc (abfd, amt); | 
 |   if (rawptr == NULL) | 
 |     return 0; | 
 |  | 
 |   oldrawptr = abfd->tdata.aout_data; | 
 |   abfd->tdata.aout_data = rawptr; | 
 |  | 
 |   /* Copy the contents of the old tdata struct.  */ | 
 |   if (oldrawptr != NULL) | 
 |     *abfd->tdata.aout_data = *oldrawptr; | 
 |  | 
 |   abfd->tdata.aout_data->a.hdr = &rawptr->e; | 
 |   *(abfd->tdata.aout_data->a.hdr) = *execp;	/* Copy in the internal_exec struct.  */ | 
 |   execp = abfd->tdata.aout_data->a.hdr; | 
 |  | 
 |   /* Set the file flags.  */ | 
 |   abfd->flags = BFD_NO_FLAGS; | 
 |   if (execp->a_drsize || execp->a_trsize) | 
 |     abfd->flags |= HAS_RELOC; | 
 |   /* Setting of EXEC_P has been deferred to the bottom of this function.  */ | 
 |   if (execp->a_syms) | 
 |     abfd->flags |= HAS_LINENO | HAS_DEBUG | HAS_SYMS | HAS_LOCALS; | 
 |   if (N_DYNAMIC (execp)) | 
 |     abfd->flags |= DYNAMIC; | 
 |  | 
 |   if (N_MAGIC (execp) == ZMAGIC) | 
 |     { | 
 |       abfd->flags |= D_PAGED | WP_TEXT; | 
 |       adata (abfd).magic = z_magic; | 
 |     } | 
 |   else if (N_MAGIC (execp) == NMAGIC) | 
 |     { | 
 |       abfd->flags |= WP_TEXT; | 
 |       adata (abfd).magic = n_magic; | 
 |     } | 
 |   else if (N_MAGIC (execp) == OMAGIC) | 
 |     adata (abfd).magic = o_magic; | 
 |   else if (N_MAGIC (execp) == IMAGIC) | 
 |     adata (abfd).magic = i_magic; | 
 |   else | 
 |     { | 
 |       /* Should have been checked with N_BADMAG before this routine | 
 | 	 was called.  */ | 
 |       abort (); | 
 |     } | 
 |  | 
 |   abfd->start_address = execp->a_entry; | 
 |  | 
 |   obj_aout_symbols (abfd) = NULL; | 
 |   abfd->symcount = execp->a_syms / sizeof (struct external_nlist); | 
 |  | 
 |   /* The default relocation entry size is that of traditional V7 Unix.  */ | 
 |   obj_reloc_entry_size (abfd) = RELOC_SIZE; | 
 |  | 
 |   /* The default symbol entry size is that of traditional Unix.  */ | 
 |   obj_symbol_entry_size (abfd) = EXTERNAL_NLIST_SIZE; | 
 |  | 
 | #ifdef USE_MMAP | 
 |   bfd_init_window (&obj_aout_sym_window (abfd)); | 
 |   bfd_init_window (&obj_aout_string_window (abfd)); | 
 | #endif | 
 |  | 
 |   obj_aout_external_syms (abfd) = NULL; | 
 |   obj_aout_external_strings (abfd) = NULL; | 
 |   obj_aout_sym_hashes (abfd) = NULL; | 
 |  | 
 |   if (! NAME (aout, make_sections) (abfd)) | 
 |     return NULL; | 
 |  | 
 |   obj_datasec (abfd)->size = execp->a_data; | 
 |   obj_bsssec (abfd)->size = execp->a_bss; | 
 |  | 
 |   obj_textsec (abfd)->flags = | 
 |     (execp->a_trsize != 0 | 
 |      ? (SEC_ALLOC | SEC_LOAD | SEC_CODE | SEC_HAS_CONTENTS | SEC_RELOC) | 
 |      : (SEC_ALLOC | SEC_LOAD | SEC_CODE | SEC_HAS_CONTENTS)); | 
 |   obj_datasec (abfd)->flags = | 
 |     (execp->a_drsize != 0 | 
 |      ? (SEC_ALLOC | SEC_LOAD | SEC_DATA | SEC_HAS_CONTENTS | SEC_RELOC) | 
 |      : (SEC_ALLOC | SEC_LOAD | SEC_DATA | SEC_HAS_CONTENTS)); | 
 |   obj_bsssec (abfd)->flags = SEC_ALLOC; | 
 |  | 
 | #ifdef THIS_IS_ONLY_DOCUMENTATION | 
 |   /* The common code can't fill in these things because they depend | 
 |      on either the start address of the text segment, the rounding | 
 |      up of virtual addresses between segments, or the starting file | 
 |      position of the text segment -- all of which varies among different | 
 |      versions of a.out.  */ | 
 |  | 
 |   /* Call back to the format-dependent code to fill in the rest of the | 
 |      fields and do any further cleanup.  Things that should be filled | 
 |      in by the callback:  */ | 
 |   struct exec *execp = exec_hdr (abfd); | 
 |  | 
 |   obj_textsec (abfd)->size = N_TXTSIZE (execp); | 
 |   /* Data and bss are already filled in since they're so standard.  */ | 
 |  | 
 |   /* 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); | 
 |  | 
 |   /* 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_str_filepos (abfd) = N_STROFF (execp); | 
 |   obj_sym_filepos (abfd) = N_SYMOFF (execp); | 
 |  | 
 |   /* Determine the architecture and machine type of the object file.  */ | 
 |   abfd->obj_arch = bfd_arch_obscure; | 
 |  | 
 |   adata(abfd)->page_size = TARGET_PAGE_SIZE; | 
 |   adata(abfd)->segment_size = SEGMENT_SIZE; | 
 |   adata(abfd)->exec_bytes_size = EXEC_BYTES_SIZE; | 
 |  | 
 |   return _bfd_no_cleanup; | 
 |  | 
 |   /* The architecture is encoded in various ways in various a.out variants, | 
 |      or is not encoded at all in some of them.  The relocation size depends | 
 |      on the architecture and the a.out variant.  Finally, the return value | 
 |      is the bfd_target vector in use.  If an error occurs, return zero and | 
 |      set bfd_error to the appropriate error code. | 
 |  | 
 |      Formats such as b.out, which have additional fields in the a.out | 
 |      header, should cope with them in this callback as well.  */ | 
 | #endif	/* DOCUMENTATION */ | 
 |  | 
 |   cleanup = (*callback_to_real_object_p)(abfd); | 
 |  | 
 |   /* Now that the segment addresses have been worked out, take a better | 
 |      guess at whether the file is executable.  If the entry point | 
 |      is within the text segment, assume it is.  (This makes files | 
 |      executable even if their entry point address is 0, as long as | 
 |      their text starts at zero.). | 
 |  | 
 |      This test had to be changed to deal with systems where the text segment | 
 |      runs at a different location than the default.  The problem is that the | 
 |      entry address can appear to be outside the text segment, thus causing an | 
 |      erroneous conclusion that the file isn't executable. | 
 |  | 
 |      To fix this, we now accept any non-zero entry point as an indication of | 
 |      executability.  This will work most of the time, since only the linker | 
 |      sets the entry point, and that is likely to be non-zero for most systems. */ | 
 |  | 
 |   if (execp->a_entry != 0 | 
 |       || (execp->a_entry >= obj_textsec (abfd)->vma | 
 | 	  && execp->a_entry < (obj_textsec (abfd)->vma | 
 | 			       + obj_textsec (abfd)->size) | 
 | 	  && execp->a_trsize == 0 | 
 | 	  && execp->a_drsize == 0)) | 
 |     abfd->flags |= EXEC_P; | 
 | #ifdef STAT_FOR_EXEC | 
 |   else | 
 |     { | 
 |       struct stat stat_buf; | 
 |  | 
 |       /* The original heuristic doesn't work in some important cases. | 
 | 	The a.out file has no information about the text start | 
 | 	address.  For files (like kernels) linked to non-standard | 
 | 	addresses (ld -Ttext nnn) the entry point may not be between | 
 | 	the default text start (obj_textsec(abfd)->vma) and | 
 | 	(obj_textsec(abfd)->vma) + text size.  This is not just a mach | 
 | 	issue.  Many kernels are loaded at non standard addresses.  */ | 
 |       if (abfd->iostream != NULL | 
 | 	  && (abfd->flags & BFD_IN_MEMORY) == 0 | 
 | 	  && (fstat(fileno((FILE *) (abfd->iostream)), &stat_buf) == 0) | 
 | 	  && ((stat_buf.st_mode & 0111) != 0)) | 
 | 	abfd->flags |= EXEC_P; | 
 |     } | 
 | #endif /* STAT_FOR_EXEC */ | 
 |  | 
 |   if (!cleanup) | 
 |     { | 
 |       free (rawptr); | 
 |       abfd->tdata.aout_data = oldrawptr; | 
 |     } | 
 |   return cleanup; | 
 | } | 
 |  | 
 | /* Initialize ABFD for use with a.out files.  */ | 
 |  | 
 | bool | 
 | NAME (aout, mkobject) (bfd *abfd) | 
 | { | 
 |   struct aout_data_struct  *rawptr; | 
 |   size_t amt = sizeof (struct aout_data_struct); | 
 |  | 
 |   bfd_set_error (bfd_error_system_call); | 
 |  | 
 |   /* Use an intermediate variable for clarity.  */ | 
 |   rawptr = bfd_zalloc (abfd, amt); | 
 |  | 
 |   if (rawptr == NULL) | 
 |     return false; | 
 |  | 
 |   abfd->tdata.aout_data = rawptr; | 
 |   exec_hdr (abfd) = &(rawptr->e); | 
 |  | 
 |   obj_textsec (abfd) = NULL; | 
 |   obj_datasec (abfd) = NULL; | 
 |   obj_bsssec (abfd)  = NULL; | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Keep track of machine architecture and machine type for | 
 |    a.out's. Return the <<machine_type>> for a particular | 
 |    architecture and machine, or <<M_UNKNOWN>> if that exact architecture | 
 |    and machine can't be represented in a.out format. | 
 |  | 
 |    If the architecture is understood, machine type 0 (default) | 
 |    is always understood.  */ | 
 |  | 
 | enum machine_type | 
 | NAME (aout, machine_type) (enum bfd_architecture arch, | 
 | 			   unsigned long machine, | 
 | 			   bool *unknown) | 
 | { | 
 |   enum machine_type arch_flags; | 
 |  | 
 |   arch_flags = M_UNKNOWN; | 
 |   *unknown = true; | 
 |  | 
 |   switch (arch) | 
 |     { | 
 |     case bfd_arch_sparc: | 
 |       if (machine == 0 | 
 | 	  || machine == bfd_mach_sparc | 
 | 	  || machine == bfd_mach_sparc_sparclite | 
 | 	  || machine == bfd_mach_sparc_v9) | 
 | 	arch_flags = M_SPARC; | 
 |       else if (machine == bfd_mach_sparc_sparclet) | 
 | 	arch_flags = M_SPARCLET; | 
 |       break; | 
 |  | 
 |     case bfd_arch_i386: | 
 |       if (machine == 0 | 
 | 	  || machine == bfd_mach_i386_i386 | 
 | 	  || machine == bfd_mach_i386_i386_intel_syntax) | 
 | 	arch_flags = M_386; | 
 |       break; | 
 |  | 
 |     case bfd_arch_arm: | 
 |       if (machine == 0)	arch_flags = M_ARM; | 
 |       break; | 
 |  | 
 |     case bfd_arch_mips: | 
 |       switch (machine) | 
 | 	{ | 
 | 	case 0: | 
 | 	case 2000: | 
 | 	case bfd_mach_mips3000: | 
 | 	  arch_flags = M_MIPS1; | 
 | 	  break; | 
 | 	case bfd_mach_mips4000: /* MIPS3 */ | 
 | 	case bfd_mach_mips4400: | 
 | 	case bfd_mach_mips8000: /* MIPS4 */ | 
 | 	case bfd_mach_mips6000: /* Real MIPS2: */ | 
 | 	  arch_flags = M_MIPS2; | 
 | 	  break; | 
 | 	default: | 
 | 	  arch_flags = M_UNKNOWN; | 
 | 	  break; | 
 | 	} | 
 |       break; | 
 |  | 
 |     case bfd_arch_ns32k: | 
 |       switch (machine) | 
 | 	{ | 
 | 	case 0:			arch_flags = M_NS32532; break; | 
 | 	case 32032:		arch_flags = M_NS32032; break; | 
 | 	case 32532:		arch_flags = M_NS32532; break; | 
 | 	default:		arch_flags = M_UNKNOWN; break; | 
 | 	} | 
 |       break; | 
 |  | 
 |     case bfd_arch_pdp11: | 
 |       /* TODO: arch_flags = M_PDP11; */ | 
 |       *unknown = false; | 
 |       break; | 
 |  | 
 |     case bfd_arch_vax: | 
 |       *unknown = false; | 
 |       break; | 
 |  | 
 |     default: | 
 |       arch_flags = M_UNKNOWN; | 
 |     } | 
 |  | 
 |   if (arch_flags != M_UNKNOWN) | 
 |     *unknown = false; | 
 |  | 
 |   return arch_flags; | 
 | } | 
 |  | 
 | /* Set the architecture and the machine of the ABFD to the | 
 |    values ARCH and MACHINE.  Verify that @ABFD's format | 
 |    can support the architecture required.  */ | 
 |  | 
 | bool | 
 | NAME (aout, set_arch_mach) (bfd *abfd, | 
 | 			    enum bfd_architecture arch, | 
 | 			    unsigned long machine) | 
 | { | 
 |   if (! bfd_default_set_arch_mach (abfd, arch, machine)) | 
 |     return false; | 
 |  | 
 |   if (arch != bfd_arch_unknown) | 
 |     { | 
 |       bool unknown; | 
 |  | 
 |       NAME (aout, machine_type) (arch, machine, &unknown); | 
 |       if (unknown) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   obj_reloc_entry_size (abfd) = RELOC_SIZE; | 
 |  | 
 |   return (*aout_backend_info(abfd)->set_sizes) (abfd); | 
 | } | 
 |  | 
 | static void | 
 | adjust_o_magic (bfd *abfd, struct internal_exec *execp) | 
 | { | 
 |   file_ptr pos = adata (abfd).exec_bytes_size; | 
 |   bfd_vma vma = 0; | 
 |   int pad = 0; | 
 |   asection *text = obj_textsec (abfd); | 
 |   asection *data = obj_datasec (abfd); | 
 |   asection *bss = obj_bsssec (abfd); | 
 |  | 
 |   /* Text.  */ | 
 |   text->filepos = pos; | 
 |   if (!text->user_set_vma) | 
 |     text->vma = vma; | 
 |   else | 
 |     vma = text->vma; | 
 |  | 
 |   pos += execp->a_text; | 
 |   vma += execp->a_text; | 
 |  | 
 |   /* Data.  */ | 
 |   if (!data->user_set_vma) | 
 |     { | 
 |       pos += pad; | 
 |       vma += pad; | 
 |       data->vma = vma; | 
 |     } | 
 |   else | 
 |     vma = data->vma; | 
 |   execp->a_text += pad; | 
 |  | 
 |   data->filepos = pos; | 
 |   pos += data->size; | 
 |   vma += data->size; | 
 |  | 
 |   /* BSS.  */ | 
 |   if (!bss->user_set_vma) | 
 |     { | 
 |       pos += pad; | 
 |       vma += pad; | 
 |       bss->vma = vma; | 
 |     } | 
 |   else if (data->size > 0 || bss->size > 0) /* PR25677: for objcopy --extract-symbol */ | 
 |     { | 
 |       /* The VMA of the .bss section is set by the VMA of the | 
 | 	 .data section plus the size of the .data section.  We may | 
 | 	 need to add padding bytes to make this true.  */ | 
 |       pad = bss->vma - vma; | 
 |       if (pad < 0) | 
 | 	pad = 0; | 
 |       pos += pad; | 
 |     } | 
 |   execp->a_data = data->size + pad; | 
 |   bss->filepos = pos; | 
 |   execp->a_bss = bss->size; | 
 |  | 
 |   N_SET_MAGIC (execp, OMAGIC); | 
 | } | 
 |  | 
 | static void | 
 | adjust_z_magic (bfd *abfd, struct internal_exec *execp) | 
 | { | 
 |   bfd_size_type data_pad, text_pad; | 
 |   file_ptr text_end; | 
 |   const struct aout_backend_data *abdp; | 
 |   /* TRUE if text includes exec header.  */ | 
 |   bool ztih; | 
 |   asection *text = obj_textsec (abfd); | 
 |   asection *data = obj_datasec (abfd); | 
 |   asection *bss = obj_bsssec (abfd); | 
 |  | 
 |   abdp = aout_backend_info (abfd); | 
 |  | 
 |   /* Text.  */ | 
 |   ztih = (abdp != NULL | 
 | 	  && (abdp->text_includes_header | 
 | 	      || obj_aout_subformat (abfd) == q_magic_format)); | 
 |   text->filepos = (ztih | 
 | 		   ? adata (abfd).exec_bytes_size | 
 | 		   : adata (abfd).zmagic_disk_block_size); | 
 |   if (!text->user_set_vma) | 
 |     { | 
 |       /* ?? Do we really need to check for relocs here?  */ | 
 |       text->vma = ((abfd->flags & HAS_RELOC) | 
 | 		   ? 0 | 
 | 		   : (ztih | 
 | 		      ? abdp->default_text_vma + adata (abfd).exec_bytes_size | 
 | 		      : abdp->default_text_vma)); | 
 |       text_pad = 0; | 
 |     } | 
 |   else | 
 |     { | 
 |       /* The .text section is being loaded at an unusual address.  We | 
 | 	 may need to pad it such that the .data section starts at a page | 
 | 	 boundary.  */ | 
 |       if (ztih) | 
 | 	text_pad = ((text->filepos - text->vma) | 
 | 		    & (adata (abfd).page_size - 1)); | 
 |       else | 
 | 	text_pad = (-text->vma | 
 | 		    & (adata (abfd).page_size - 1)); | 
 |     } | 
 |  | 
 |   /* Find start of data.  */ | 
 |   if (ztih) | 
 |     { | 
 |       text_end = text->filepos + execp->a_text; | 
 |       text_pad += BFD_ALIGN (text_end, adata (abfd).page_size) - text_end; | 
 |     } | 
 |   else | 
 |     { | 
 |       /* Note that if page_size == zmagic_disk_block_size, then | 
 | 	 filepos == page_size, and this case is the same as the ztih | 
 | 	 case.  */ | 
 |       text_end = execp->a_text; | 
 |       text_pad += BFD_ALIGN (text_end, adata (abfd).page_size) - text_end; | 
 |       text_end += text->filepos; | 
 |     } | 
 |   execp->a_text += text_pad; | 
 |  | 
 |   /* Data.  */ | 
 |   if (!data->user_set_vma) | 
 |     { | 
 |       bfd_vma vma; | 
 |       vma = text->vma + execp->a_text; | 
 |       data->vma = BFD_ALIGN (vma, adata (abfd).segment_size); | 
 |     } | 
 |   if (abdp && abdp->zmagic_mapped_contiguous) | 
 |     { | 
 |       text_pad = data->vma - (text->vma + execp->a_text); | 
 |       /* Only pad the text section if the data | 
 | 	 section is going to be placed after it.  */ | 
 |       if (text_pad > 0) | 
 | 	execp->a_text += text_pad; | 
 |     } | 
 |   data->filepos = text->filepos + execp->a_text; | 
 |  | 
 |   /* Fix up exec header while we're at it.  */ | 
 |   if (ztih && (!abdp || (abdp && !abdp->exec_header_not_counted))) | 
 |     execp->a_text += adata (abfd).exec_bytes_size; | 
 |   N_SET_MAGIC (execp, ZMAGIC); | 
 |  | 
 |   /* Spec says data section should be rounded up to page boundary.  */ | 
 |   execp->a_data = align_power (data->size, bss->alignment_power); | 
 |   execp->a_data = BFD_ALIGN (execp->a_data, adata (abfd).page_size); | 
 |   data_pad = execp->a_data - data->size; | 
 |  | 
 |   /* BSS.  */ | 
 |   if (!bss->user_set_vma) | 
 |     bss->vma = data->vma + execp->a_data; | 
 |   /* If the BSS immediately follows the data section and extra space | 
 |      in the page is left after the data section, fudge data | 
 |      in the header so that the bss section looks smaller by that | 
 |      amount.  We'll start the bss section there, and lie to the OS. | 
 |      (Note that a linker script, as well as the above assignment, | 
 |      could have explicitly set the BSS vma to immediately follow | 
 |      the data section.)  */ | 
 |   if (align_power (bss->vma, bss->alignment_power) == data->vma + execp->a_data) | 
 |     execp->a_bss = data_pad > bss->size ? 0 : bss->size - data_pad; | 
 |   else | 
 |     execp->a_bss = bss->size; | 
 | } | 
 |  | 
 | static void | 
 | adjust_n_magic (bfd *abfd, struct internal_exec *execp) | 
 | { | 
 |   file_ptr pos = adata (abfd).exec_bytes_size; | 
 |   bfd_vma vma = 0; | 
 |   int pad; | 
 |   asection *text = obj_textsec (abfd); | 
 |   asection *data = obj_datasec (abfd); | 
 |   asection *bss = obj_bsssec (abfd); | 
 |  | 
 |   /* Text.  */ | 
 |   text->filepos = pos; | 
 |   if (!text->user_set_vma) | 
 |     text->vma = vma; | 
 |   else | 
 |     vma = text->vma; | 
 |   pos += execp->a_text; | 
 |   vma += execp->a_text; | 
 |  | 
 |   /* Data.  */ | 
 |   data->filepos = pos; | 
 |   if (!data->user_set_vma) | 
 |     data->vma = BFD_ALIGN (vma, adata (abfd).segment_size); | 
 |   vma = data->vma; | 
 |  | 
 |   /* Since BSS follows data immediately, see if it needs alignment.  */ | 
 |   vma += data->size; | 
 |   pad = align_power (vma, bss->alignment_power) - vma; | 
 |   execp->a_data = data->size + pad; | 
 |   pos += execp->a_data; | 
 |  | 
 |   /* BSS.  */ | 
 |   if (!bss->user_set_vma) | 
 |     bss->vma = vma; | 
 |   else | 
 |     vma = bss->vma; | 
 |  | 
 |   /* Fix up exec header.  */ | 
 |   execp->a_bss = bss->size; | 
 |   N_SET_MAGIC (execp, NMAGIC); | 
 | } | 
 |  | 
 | static void | 
 | adjust_i_magic (bfd *abfd, struct internal_exec *execp) | 
 | { | 
 |   file_ptr pos = adata (abfd).exec_bytes_size; | 
 |   bfd_vma vma = 0; | 
 |   int pad; | 
 |   asection *text = obj_textsec (abfd); | 
 |   asection *data = obj_datasec (abfd); | 
 |   asection *bss = obj_bsssec (abfd); | 
 |  | 
 |   /* Text.  */ | 
 |   text->filepos = pos; | 
 |   if (!text->user_set_vma) | 
 |     text->vma = vma; | 
 |   else | 
 |     vma = text->vma; | 
 |   pos += execp->a_text; | 
 |  | 
 |   /* Data.  */ | 
 |   data->filepos = pos; | 
 |   if (!data->user_set_vma) | 
 |     data->vma = 0; | 
 |   vma = data->vma; | 
 |  | 
 |   /* Since BSS follows data immediately, see if it needs alignment.  */ | 
 |   vma += data->size; | 
 |   pad = align_power (vma, bss->alignment_power) - vma; | 
 |   execp->a_data = data->size + pad; | 
 |   pos += execp->a_data; | 
 |  | 
 |   /* BSS.  */ | 
 |   if (!bss->user_set_vma) | 
 |     bss->vma = vma; | 
 |   else | 
 |     vma = bss->vma; | 
 |  | 
 |   /* Fix up exec header.  */ | 
 |   execp->a_bss = bss->size; | 
 |   N_SET_MAGIC (execp, IMAGIC); | 
 | } | 
 |  | 
 | bool | 
 | NAME (aout, adjust_sizes_and_vmas) (bfd *abfd) | 
 | { | 
 |   struct internal_exec *execp = exec_hdr (abfd); | 
 |  | 
 |   if (! NAME (aout, make_sections) (abfd)) | 
 |     return false; | 
 |  | 
 |   if (adata (abfd).magic != undecided_magic) | 
 |     return true; | 
 |  | 
 |   execp->a_text = align_power (obj_textsec (abfd)->size, | 
 | 			       obj_textsec (abfd)->alignment_power); | 
 |  | 
 |   /* Rule (heuristic) for when to pad to a new page.  Note that there | 
 |      are (at least) two ways demand-paged (ZMAGIC) files have been | 
 |      handled.  Most Berkeley-based systems start the text segment at | 
 |      (TARGET_PAGE_SIZE).  However, newer versions of SUNOS start the text | 
 |      segment right after the exec header; the latter is counted in the | 
 |      text segment size, and is paged in by the kernel with the rest of | 
 |      the text.  */ | 
 |  | 
 |   /* This perhaps isn't the right way to do this, but made it simpler for me | 
 |      to understand enough to implement it.  Better would probably be to go | 
 |      right from BFD flags to alignment/positioning characteristics.  But the | 
 |      old code was sloppy enough about handling the flags, and had enough | 
 |      other magic, that it was a little hard for me to understand.  I think | 
 |      I understand it better now, but I haven't time to do the cleanup this | 
 |      minute.  */ | 
 |  | 
 |   if (separate_i_d) | 
 |     adata (abfd).magic = i_magic; | 
 |   else if (abfd->flags & WP_TEXT) | 
 |     adata (abfd).magic = n_magic; | 
 |   else | 
 |     adata (abfd).magic = o_magic; | 
 |  | 
 | #ifdef BFD_AOUT_DEBUG /* requires gcc2 */ | 
 | #if __GNUC__ >= 2 | 
 |   fprintf (stderr, "%s text=<%x,%x,%x> data=<%x,%x,%x> bss=<%x,%x,%x>\n", | 
 | 	   ({ char *str; | 
 | 	      switch (adata (abfd).magic) | 
 | 		{ | 
 | 		case n_magic: str = "NMAGIC"; break; | 
 | 		case o_magic: str = "OMAGIC"; break; | 
 | 		case i_magic: str = "IMAGIC"; break; | 
 | 		case z_magic: str = "ZMAGIC"; break; | 
 | 		default: abort (); | 
 | 		} | 
 | 	      str; | 
 | 	    }), | 
 | 	   obj_textsec (abfd)->vma, obj_textsec (abfd)->size, | 
 | 		obj_textsec (abfd)->alignment_power, | 
 | 	   obj_datasec (abfd)->vma, obj_datasec (abfd)->size, | 
 | 		obj_datasec (abfd)->alignment_power, | 
 | 	   obj_bsssec (abfd)->vma, obj_bsssec (abfd)->size, | 
 | 		obj_bsssec (abfd)->alignment_power); | 
 | #endif | 
 | #endif | 
 |  | 
 |   switch (adata (abfd).magic) | 
 |     { | 
 |     case o_magic: | 
 |       adjust_o_magic (abfd, execp); | 
 |       break; | 
 |     case z_magic: | 
 |       adjust_z_magic (abfd, execp); | 
 |       break; | 
 |     case n_magic: | 
 |       adjust_n_magic (abfd, execp); | 
 |       break; | 
 |     case i_magic: | 
 |       adjust_i_magic (abfd, execp); | 
 |       break; | 
 |     default: | 
 |       abort (); | 
 |     } | 
 |  | 
 | #ifdef BFD_AOUT_DEBUG | 
 |   fprintf (stderr, "       text=<%x,%x,%x> data=<%x,%x,%x> bss=<%x,%x>\n", | 
 | 	   obj_textsec (abfd)->vma, execp->a_text, | 
 | 		obj_textsec (abfd)->filepos, | 
 | 	   obj_datasec (abfd)->vma, execp->a_data, | 
 | 		obj_datasec (abfd)->filepos, | 
 | 	   obj_bsssec (abfd)->vma, execp->a_bss); | 
 | #endif | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Called by the BFD in response to a bfd_make_section request.  */ | 
 |  | 
 | bool | 
 | NAME (aout, new_section_hook) (bfd *abfd, asection *newsect) | 
 | { | 
 |   /* Align to double at least.  */ | 
 |   newsect->alignment_power = bfd_get_arch_info(abfd)->section_align_power; | 
 |  | 
 |   if (bfd_get_format (abfd) == bfd_object) | 
 |     { | 
 |       if (obj_textsec (abfd) == NULL | 
 | 	  && !strcmp (newsect->name, ".text")) | 
 | 	{ | 
 | 	  obj_textsec(abfd)= newsect; | 
 | 	  newsect->target_index = N_TEXT; | 
 | 	} | 
 |       else if (obj_datasec (abfd) == NULL | 
 | 	       && !strcmp (newsect->name, ".data")) | 
 | 	{ | 
 | 	  obj_datasec (abfd) = newsect; | 
 | 	  newsect->target_index = N_DATA; | 
 | 	} | 
 |       else if (obj_bsssec (abfd) == NULL | 
 | 	       && !strcmp (newsect->name, ".bss")) | 
 | 	{ | 
 | 	  obj_bsssec (abfd) = newsect; | 
 | 	  newsect->target_index = N_BSS; | 
 | 	} | 
 |     } | 
 |  | 
 |   /* We allow more than three sections internally.  */ | 
 |   return _bfd_generic_new_section_hook (abfd, newsect); | 
 | } | 
 |  | 
 | bool | 
 | NAME (aout, set_section_contents) (bfd *abfd, | 
 | 				   sec_ptr section, | 
 | 				   const void * location, | 
 | 				   file_ptr offset, | 
 | 				   bfd_size_type count) | 
 | { | 
 |   if (! abfd->output_has_begun) | 
 |     { | 
 |       if (! NAME (aout, adjust_sizes_and_vmas) (abfd)) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   if (section == obj_bsssec (abfd)) | 
 |     { | 
 |       bfd_set_error (bfd_error_no_contents); | 
 |       return false; | 
 |     } | 
 |  | 
 |   if (section != obj_textsec (abfd) | 
 |       && section != obj_datasec (abfd)) | 
 |     { | 
 |       _bfd_error_handler | 
 | 	/* xgettext:c-format */ | 
 | 	(_("%pB: can not represent section `%pA' in a.out object file format"), | 
 | 	 abfd, section); | 
 |       bfd_set_error (bfd_error_nonrepresentable_section); | 
 |       return false; | 
 |     } | 
 |  | 
 |   if (count != 0) | 
 |     { | 
 |       if (bfd_seek (abfd, section->filepos + offset, SEEK_SET) != 0 | 
 | 	  || bfd_bwrite (location, count, abfd) != count) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Read the external symbols from an a.out file.  */ | 
 |  | 
 | static bool | 
 | aout_get_external_symbols (bfd *abfd) | 
 | { | 
 |   if (obj_aout_external_syms (abfd) == NULL) | 
 |     { | 
 |       bfd_size_type count; | 
 |       struct external_nlist *syms; | 
 |  | 
 |       count = exec_hdr (abfd)->a_syms / EXTERNAL_NLIST_SIZE; | 
 |  | 
 |       /* PR 17512: file: 011f5a08.  */ | 
 |       if (count == 0) | 
 | 	{ | 
 | 	  obj_aout_external_syms (abfd) = NULL; | 
 | 	  obj_aout_external_sym_count (abfd) = count; | 
 | 	  return true; | 
 | 	} | 
 |  | 
 | #ifdef USE_MMAP | 
 |       if (! bfd_get_file_window (abfd, obj_sym_filepos (abfd), | 
 | 				 exec_hdr (abfd)->a_syms, | 
 | 				 &obj_aout_sym_window (abfd), true)) | 
 | 	return false; | 
 |       syms = (struct external_nlist *) obj_aout_sym_window (abfd).data; | 
 | #else | 
 |       /* We allocate using malloc to make the values easy to free | 
 | 	 later on.  If we put them on the objalloc it might not be | 
 | 	 possible to free them.  */ | 
 |       if (bfd_seek (abfd, obj_sym_filepos (abfd), SEEK_SET) != 0) | 
 | 	return false; | 
 |       syms = (struct external_nlist *) | 
 | 	_bfd_malloc_and_read (abfd, count * EXTERNAL_NLIST_SIZE, | 
 | 			      count * EXTERNAL_NLIST_SIZE); | 
 |       if (syms == NULL) | 
 | 	return false; | 
 | #endif | 
 |  | 
 |       obj_aout_external_syms (abfd) = syms; | 
 |       obj_aout_external_sym_count (abfd) = count; | 
 |     } | 
 |  | 
 |   if (obj_aout_external_strings (abfd) == NULL | 
 |       && exec_hdr (abfd)->a_syms != 0) | 
 |     { | 
 |       unsigned char string_chars[BYTES_IN_LONG]; | 
 |       bfd_size_type stringsize; | 
 |       char *strings; | 
 |       bfd_size_type amt = BYTES_IN_LONG; | 
 |  | 
 |       /* Get the size of the strings.  */ | 
 |       if (bfd_seek (abfd, obj_str_filepos (abfd), SEEK_SET) != 0 | 
 | 	  || bfd_bread ((void *) string_chars, amt, abfd) != amt) | 
 | 	return false; | 
 |       stringsize = H_GET_32 (abfd, string_chars); | 
 |       if (stringsize == 0) | 
 | 	stringsize = 1; | 
 |       else if (stringsize < BYTES_IN_LONG | 
 | 	       || (size_t) stringsize != stringsize) | 
 | 	{ | 
 | 	  bfd_set_error (bfd_error_bad_value); | 
 | 	  return false; | 
 | 	} | 
 |  | 
 | #ifdef USE_MMAP | 
 |       if (stringsize >= BYTES_IN_LONG) | 
 | 	{ | 
 | 	  if (! bfd_get_file_window (abfd, obj_str_filepos (abfd), stringsize + 1, | 
 | 				     &obj_aout_string_window (abfd), true)) | 
 | 	    return false; | 
 | 	  strings = (char *) obj_aout_string_window (abfd).data; | 
 | 	} | 
 |       else | 
 | #endif | 
 | 	{ | 
 | 	  strings = (char *) bfd_malloc (stringsize + 1); | 
 | 	  if (strings == NULL) | 
 | 	    return false; | 
 |  | 
 | 	  if (stringsize >= BYTES_IN_LONG) | 
 | 	    { | 
 | 	      amt = stringsize - BYTES_IN_LONG; | 
 | 	      if (bfd_bread (strings + BYTES_IN_LONG, amt, abfd) != amt) | 
 | 		{ | 
 | 		  free (strings); | 
 | 		  return false; | 
 | 		} | 
 | 	    } | 
 | 	} | 
 |       /* Ensure that a zero index yields an empty string.  */ | 
 |       if (stringsize >= BYTES_IN_WORD) | 
 | 	memset (strings, 0, BYTES_IN_LONG); | 
 |  | 
 |       /* Ensure that the string buffer is NUL terminated.  */ | 
 |       strings[stringsize] = 0; | 
 |  | 
 |       obj_aout_external_strings (abfd) = strings; | 
 |       obj_aout_external_string_size (abfd) = stringsize; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Translate an a.out symbol into a BFD symbol.  The desc, other, type | 
 |    and symbol->value fields of CACHE_PTR will be set from the a.out | 
 |    nlist structure.  This function is responsible for setting | 
 |    symbol->flags and symbol->section, and adjusting symbol->value.  */ | 
 |  | 
 | static bool | 
 | translate_from_native_sym_flags (bfd *abfd, | 
 | 				 aout_symbol_type *cache_ptr) | 
 | { | 
 |   flagword visible; | 
 |  | 
 |   if (is_stab (cache_ptr->type, cache_ptr->symbol.name)) | 
 |     { | 
 |       asection *sec; | 
 |  | 
 |       /* This is a debugging symbol.  */ | 
 |       cache_ptr->symbol.flags = BSF_DEBUGGING; | 
 |  | 
 |       /* Work out the symbol section.  */ | 
 |       switch (cache_ptr->type) | 
 | 	{ | 
 | 	case N_SO: | 
 | 	case N_SOL: | 
 | 	case N_FUN: | 
 | 	case N_ENTRY: | 
 | 	case N_SLINE: | 
 | 	case N_FN: | 
 | 	  sec = obj_textsec (abfd); | 
 | 	  break; | 
 | 	case N_STSYM: | 
 | 	case N_DSLINE: | 
 | 	  sec = obj_datasec (abfd); | 
 | 	  break; | 
 | 	case N_LCSYM: | 
 | 	case N_BSLINE: | 
 | 	  sec = obj_bsssec (abfd); | 
 | 	  break; | 
 | 	default: | 
 | 	  sec = bfd_abs_section_ptr; | 
 | 	  break; | 
 | 	} | 
 |  | 
 |       cache_ptr->symbol.section = sec; | 
 |       cache_ptr->symbol.value -= sec->vma; | 
 |  | 
 |       return true; | 
 |     } | 
 |  | 
 |   /* Get the default visibility.  This does not apply to all types, so | 
 |      we just hold it in a local variable to use if wanted.  */ | 
 |   if ((cache_ptr->type & N_EXT) == 0) | 
 |     visible = BSF_LOCAL; | 
 |   else | 
 |     visible = BSF_GLOBAL; | 
 |  | 
 |   switch (cache_ptr->type) | 
 |     { | 
 |     default: | 
 |     case N_ABS: case N_ABS | N_EXT: | 
 |       cache_ptr->symbol.section = bfd_abs_section_ptr; | 
 |       cache_ptr->symbol.flags = visible; | 
 |       break; | 
 |  | 
 |     case N_UNDF | N_EXT: | 
 |       if (cache_ptr->symbol.value != 0) | 
 | 	{ | 
 | 	  /* This is a common symbol.  */ | 
 | 	  cache_ptr->symbol.flags = BSF_GLOBAL; | 
 | 	  cache_ptr->symbol.section = bfd_com_section_ptr; | 
 | 	} | 
 |       else | 
 | 	{ | 
 | 	  cache_ptr->symbol.flags = 0; | 
 | 	  cache_ptr->symbol.section = bfd_und_section_ptr; | 
 | 	} | 
 |       break; | 
 |  | 
 |     case N_TEXT: case N_TEXT | N_EXT: | 
 |       cache_ptr->symbol.section = obj_textsec (abfd); | 
 |       cache_ptr->symbol.value -= cache_ptr->symbol.section->vma; | 
 |       cache_ptr->symbol.flags = visible; | 
 |       break; | 
 |  | 
 |     case N_DATA: case N_DATA | N_EXT: | 
 |       cache_ptr->symbol.section = obj_datasec (abfd); | 
 |       cache_ptr->symbol.value -= cache_ptr->symbol.section->vma; | 
 |       cache_ptr->symbol.flags = visible; | 
 |       break; | 
 |  | 
 |     case N_BSS: case N_BSS | N_EXT: | 
 |       cache_ptr->symbol.section = obj_bsssec (abfd); | 
 |       cache_ptr->symbol.value -= cache_ptr->symbol.section->vma; | 
 |       cache_ptr->symbol.flags = visible; | 
 |       break; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Set the fields of SYM_POINTER according to CACHE_PTR.  */ | 
 |  | 
 | static bool | 
 | translate_to_native_sym_flags (bfd *abfd, | 
 | 			       asymbol *cache_ptr, | 
 | 			       struct external_nlist *sym_pointer) | 
 | { | 
 |   bfd_vma value = cache_ptr->value; | 
 |   asection *sec; | 
 |   bfd_vma off; | 
 |   const char *name = cache_ptr->name != NULL ? cache_ptr->name : "*unknown*"; | 
 |  | 
 |   /* Mask out any existing type bits in case copying from one section | 
 |      to another.  */ | 
 |   if (!is_stab (sym_pointer->e_type[0], name)) | 
 |     sym_pointer->e_type[0] &= ~N_TYPE; | 
 |  | 
 |   sec = bfd_asymbol_section (cache_ptr); | 
 |   off = 0; | 
 |  | 
 |   if (sec == NULL) | 
 |     { | 
 |       /* This case occurs, e.g., for the *DEBUG* section of a COFF | 
 | 	 file.  */ | 
 |       _bfd_error_handler | 
 | 	/* xgettext:c-format */ | 
 | 	(_("%pB: can not represent section for symbol `%s' in a.out object file format"), | 
 | 	 abfd, name); | 
 |       bfd_set_error (bfd_error_nonrepresentable_section); | 
 |       return false; | 
 |     } | 
 |  | 
 |   if (sec->output_section != NULL) | 
 |     { | 
 |       off = sec->output_offset; | 
 |       sec = sec->output_section; | 
 |     } | 
 |  | 
 |   if (bfd_is_abs_section (sec)) | 
 |     sym_pointer->e_type[0] |= N_ABS; | 
 |   else if (sec == obj_textsec (abfd)) | 
 |     sym_pointer->e_type[0] |= N_TEXT; | 
 |   else if (sec == obj_datasec (abfd)) | 
 |     sym_pointer->e_type[0] |= N_DATA; | 
 |   else if (sec == obj_bsssec (abfd)) | 
 |     sym_pointer->e_type[0] |= N_BSS; | 
 |   else if (bfd_is_und_section (sec)) | 
 |     sym_pointer->e_type[0] = N_UNDF | N_EXT; | 
 |   else if (bfd_is_com_section (sec)) | 
 |     sym_pointer->e_type[0] = N_UNDF | N_EXT; | 
 |   else | 
 |     { | 
 |       _bfd_error_handler | 
 | 	/* xgettext:c-format */ | 
 | 	(_("%pB: can not represent section `%pA' in a.out object file format"), | 
 | 	 abfd, sec); | 
 |       bfd_set_error (bfd_error_nonrepresentable_section); | 
 |       return false; | 
 |     } | 
 |  | 
 |   /* Turn the symbol from section relative to absolute again */ | 
 |   value += sec->vma + off; | 
 |  | 
 |   if ((cache_ptr->flags & BSF_DEBUGGING) != 0) | 
 |     sym_pointer->e_type[0] = ((aout_symbol_type *) cache_ptr)->type; | 
 |   else if ((cache_ptr->flags & BSF_GLOBAL) != 0) | 
 |     sym_pointer->e_type[0] |= N_EXT; | 
 |  | 
 |   PUT_WORD(abfd, value, sym_pointer->e_value); | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Native-level interface to symbols. */ | 
 |  | 
 | asymbol * | 
 | NAME (aout, make_empty_symbol) (bfd *abfd) | 
 | { | 
 |   size_t amt = sizeof (aout_symbol_type); | 
 |   aout_symbol_type *new_symbol_type = bfd_zalloc (abfd, amt); | 
 |  | 
 |   if (!new_symbol_type) | 
 |     return NULL; | 
 |   new_symbol_type->symbol.the_bfd = abfd; | 
 |  | 
 |   return &new_symbol_type->symbol; | 
 | } | 
 |  | 
 | /* Translate a set of external symbols into internal symbols.  */ | 
 |  | 
 | bool | 
 | NAME (aout, translate_symbol_table) (bfd *abfd, | 
 | 				     aout_symbol_type *in, | 
 | 				     struct external_nlist *ext, | 
 | 				     bfd_size_type count, | 
 | 				     char *str, | 
 | 				     bfd_size_type strsize, | 
 | 				     bool dynamic) | 
 | { | 
 |   struct external_nlist *ext_end; | 
 |  | 
 |   ext_end = ext + count; | 
 |   for (; ext < ext_end; ext++, in++) | 
 |     { | 
 |       bfd_vma x; | 
 |       int ovly; | 
 |  | 
 |       x = GET_WORD (abfd, ext->e_strx); | 
 |       in->symbol.the_bfd = abfd; | 
 |  | 
 |       /* For the normal symbols, the zero index points at the number | 
 | 	 of bytes in the string table but is to be interpreted as the | 
 | 	 null string.  For the dynamic symbols, the number of bytes in | 
 | 	 the string table is stored in the __DYNAMIC structure and the | 
 | 	 zero index points at an actual string.  */ | 
 |       if (x == 0 && ! dynamic) | 
 | 	in->symbol.name = ""; | 
 |       else if (x < strsize) | 
 | 	in->symbol.name = str + x; | 
 |       else | 
 | 	{ | 
 | 	  _bfd_error_handler | 
 | 	    (_("%pB: invalid string offset %" PRIu64 " >= %" PRIu64), | 
 | 	     abfd, (uint64_t) x, (uint64_t) strsize); | 
 | 	  bfd_set_error (bfd_error_bad_value); | 
 | 	  return false; | 
 | 	} | 
 |  | 
 |       ovly = H_GET_8 (abfd, ext->e_ovly); | 
 |       if (ovly != 0) | 
 | 	{ | 
 | 	  _bfd_error_handler | 
 | 	    (_("%pB: symbol indicates overlay (not supported)"), abfd); | 
 | 	  bfd_set_error (bfd_error_bad_value); | 
 | 	  return false; | 
 | 	} | 
 |  | 
 |       in->symbol.value = GET_WORD (abfd,  ext->e_value); | 
 |       /* e_desc is zero for normal symbols but for .stab symbols it | 
 | 	 carries the desc field in our extended 2.11BSD format. */ | 
 |       in->desc = H_GET_16 (abfd, ext->e_desc); | 
 |       in->other = 0; | 
 |       in->type = H_GET_8 (abfd,  ext->e_type); | 
 |       in->symbol.udata.p = NULL; | 
 |  | 
 |       if (! translate_from_native_sym_flags (abfd, in)) | 
 | 	return false; | 
 |  | 
 |       if (dynamic) | 
 | 	in->symbol.flags |= BSF_DYNAMIC; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* We read the symbols into a buffer, which is discarded when this | 
 |    function exits.  We read the strings into a buffer large enough to | 
 |    hold them all plus all the cached symbol entries.  */ | 
 |  | 
 | bool | 
 | NAME (aout, slurp_symbol_table) (bfd *abfd) | 
 | { | 
 |   struct external_nlist *old_external_syms; | 
 |   aout_symbol_type *cached; | 
 |   bfd_size_type cached_size; | 
 |  | 
 |   /* If there's no work to be done, don't do any.  */ | 
 |   if (obj_aout_symbols (abfd) != NULL) | 
 |     return true; | 
 |  | 
 |   old_external_syms = obj_aout_external_syms (abfd); | 
 |  | 
 |   if (! aout_get_external_symbols (abfd)) | 
 |     return false; | 
 |  | 
 |   cached_size = obj_aout_external_sym_count (abfd); | 
 |   cached_size *= sizeof (aout_symbol_type); | 
 |   cached = bfd_zmalloc (cached_size); | 
 |   if (cached == NULL && cached_size != 0) | 
 |     return false; | 
 |  | 
 |   /* Convert from external symbol information to internal.  */ | 
 |   if (! (NAME (aout, translate_symbol_table) | 
 | 	 (abfd, cached, | 
 | 	  obj_aout_external_syms (abfd), | 
 | 	  obj_aout_external_sym_count (abfd), | 
 | 	  obj_aout_external_strings (abfd), | 
 | 	  obj_aout_external_string_size (abfd), | 
 | 	  false))) | 
 |     { | 
 |       free (cached); | 
 |       return false; | 
 |     } | 
 |  | 
 |   abfd->symcount = obj_aout_external_sym_count (abfd); | 
 |  | 
 |   obj_aout_symbols (abfd) = cached; | 
 |  | 
 |   /* It is very likely that anybody who calls this function will not | 
 |      want the external symbol information, so if it was allocated | 
 |      because of our call to aout_get_external_symbols, we free it up | 
 |      right away to save space.  */ | 
 |   if (old_external_syms == NULL | 
 |       && obj_aout_external_syms (abfd) != NULL) | 
 |     { | 
 | #ifdef USE_MMAP | 
 |       bfd_free_window (&obj_aout_sym_window (abfd)); | 
 | #else | 
 |       free (obj_aout_external_syms (abfd)); | 
 | #endif | 
 |       obj_aout_external_syms (abfd) = NULL; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* We use a hash table when writing out symbols so that we only write | 
 |    out a particular string once.  This helps particularly when the | 
 |    linker writes out stabs debugging entries, because each different | 
 |    contributing object file tends to have many duplicate stabs | 
 |    strings. | 
 |  | 
 |    This hash table code breaks dbx on SunOS 4.1.3, so we don't do it | 
 |    if BFD_TRADITIONAL_FORMAT is set.  */ | 
 |  | 
 | /* Get the index of a string in a strtab, adding it if it is not | 
 |    already present.  */ | 
 |  | 
 | static inline bfd_size_type | 
 | add_to_stringtab (bfd *abfd, | 
 | 		  struct bfd_strtab_hash *tab, | 
 | 		  const char *str, | 
 | 		  bool copy) | 
 | { | 
 |   bool hash; | 
 |   bfd_size_type str_index; | 
 |  | 
 |   /* An index of 0 always means the empty string.  */ | 
 |   if (str == 0 || *str == '\0') | 
 |     return 0; | 
 |  | 
 |   /* Don't hash if BFD_TRADITIONAL_FORMAT is set, because SunOS dbx | 
 |      doesn't understand a hashed string table.  */ | 
 |   hash = true; | 
 |   if ((abfd->flags & BFD_TRADITIONAL_FORMAT) != 0) | 
 |     hash = false; | 
 |  | 
 |   str_index = _bfd_stringtab_add (tab, str, hash, copy); | 
 |  | 
 |   if (str_index != (bfd_size_type) -1) | 
 |     /* Add BYTES_IN_LONG to the return value to account for the | 
 |        space taken up by the string table size.  */ | 
 |     str_index += BYTES_IN_LONG; | 
 |  | 
 |   return str_index; | 
 | } | 
 |  | 
 | /* Write out a strtab.  ABFD is already at the right location in the | 
 |    file.  */ | 
 |  | 
 | static bool | 
 | emit_stringtab (bfd *abfd, struct bfd_strtab_hash *tab) | 
 | { | 
 |   bfd_byte buffer[BYTES_IN_LONG]; | 
 |  | 
 |   /* The string table starts with the size.  */ | 
 |   H_PUT_32 (abfd, _bfd_stringtab_size (tab) + BYTES_IN_LONG, buffer); | 
 |   if (bfd_bwrite ((void *) buffer, (bfd_size_type) BYTES_IN_LONG, abfd) | 
 |       != BYTES_IN_LONG) | 
 |     return false; | 
 |  | 
 |   return _bfd_stringtab_emit (abfd, tab); | 
 | } | 
 |  | 
 | bool | 
 | NAME (aout, write_syms) (bfd *abfd) | 
 | { | 
 |   unsigned int count ; | 
 |   asymbol **generic = bfd_get_outsymbols (abfd); | 
 |   struct bfd_strtab_hash *strtab; | 
 |  | 
 |   strtab = _bfd_stringtab_init (); | 
 |   if (strtab == NULL) | 
 |     return false; | 
 |  | 
 |   for (count = 0; count < bfd_get_symcount (abfd); count++) | 
 |     { | 
 |       asymbol *g = generic[count]; | 
 |       bfd_size_type indx; | 
 |       struct external_nlist nsp; | 
 |  | 
 |       indx = add_to_stringtab (abfd, strtab, g->name, false); | 
 |       if (indx == (bfd_size_type) -1) | 
 | 	goto error_return; | 
 |       PUT_WORD (abfd, indx, nsp.e_strx); | 
 |  | 
 |       if (bfd_asymbol_flavour(g) == abfd->xvec->flavour) | 
 | 	{ | 
 | 	  H_PUT_16 (abfd, aout_symbol (g)->desc,  nsp.e_desc); | 
 | 	  H_PUT_8 (abfd, 0, nsp.e_ovly); | 
 | 	  H_PUT_8 (abfd, aout_symbol (g)->type,  nsp.e_type); | 
 | 	} | 
 |       else | 
 | 	{ | 
 | 	  H_PUT_16 (abfd, 0, nsp.e_desc); | 
 | 	  H_PUT_8 (abfd, 0, nsp.e_ovly); | 
 | 	  H_PUT_8 (abfd, 0, nsp.e_type); | 
 | 	} | 
 |  | 
 |       if (! translate_to_native_sym_flags (abfd, g, &nsp)) | 
 | 	goto error_return; | 
 |  | 
 |       if (bfd_bwrite ((void *)&nsp, (bfd_size_type) EXTERNAL_NLIST_SIZE, abfd) | 
 | 	  != EXTERNAL_NLIST_SIZE) | 
 | 	goto error_return; | 
 |  | 
 |       /* NB: `KEEPIT' currently overlays `udata.p', so set this only | 
 | 	 here, at the end.  */ | 
 |       g->KEEPIT = count; | 
 |     } | 
 |  | 
 |   if (! emit_stringtab (abfd, strtab)) | 
 |     goto error_return; | 
 |  | 
 |   _bfd_stringtab_free (strtab); | 
 |  | 
 |   return true; | 
 |  | 
 |  error_return: | 
 |   _bfd_stringtab_free (strtab); | 
 |   return false; | 
 | } | 
 |  | 
 |  | 
 | long | 
 | NAME (aout, canonicalize_symtab) (bfd *abfd, asymbol **location) | 
 | { | 
 |   unsigned int counter = 0; | 
 |   aout_symbol_type *symbase; | 
 |  | 
 |   if (!NAME (aout, slurp_symbol_table) (abfd)) | 
 |     return -1; | 
 |  | 
 |   for (symbase = obj_aout_symbols (abfd); counter++ < bfd_get_symcount (abfd);) | 
 |     *(location++) = (asymbol *)(symbase++); | 
 |   *location++ =0; | 
 |   return bfd_get_symcount (abfd); | 
 | } | 
 |  | 
 |  | 
 | /* Output extended relocation information to a file in target byte order.  */ | 
 |  | 
 | static void | 
 | pdp11_aout_swap_reloc_out (bfd *abfd, arelent *g, bfd_byte *natptr) | 
 | { | 
 |   int r_index; | 
 |   int r_pcrel; | 
 |   int reloc_entry; | 
 |   int r_type; | 
 |   asymbol *sym = *(g->sym_ptr_ptr); | 
 |   asection *output_section = sym->section->output_section; | 
 |  | 
 |   if (g->addend != 0) | 
 |     fprintf (stderr, "BFD: can't do this reloc addend stuff\n"); | 
 |  | 
 |   r_pcrel = g->howto->pc_relative; | 
 |  | 
 |   if (bfd_is_abs_section (output_section)) | 
 |     r_type = RABS; | 
 |   else if (output_section == obj_textsec (abfd)) | 
 |     r_type = RTEXT; | 
 |   else if (output_section == obj_datasec (abfd)) | 
 |     r_type = RDATA; | 
 |   else if (output_section == obj_bsssec (abfd)) | 
 |     r_type = RBSS; | 
 |   else if (bfd_is_und_section (output_section)) | 
 |     r_type = REXT; | 
 |   else if (bfd_is_com_section (output_section)) | 
 |     r_type = REXT; | 
 |   else | 
 |     r_type = -1; | 
 |  | 
 |   BFD_ASSERT (r_type != -1); | 
 |  | 
 |   if (r_type == RABS) | 
 |     r_index = 0; | 
 |   else | 
 |     r_index = (*(g->sym_ptr_ptr))->KEEPIT; | 
 |  | 
 |   reloc_entry = r_index << 4 | r_type | r_pcrel; | 
 |  | 
 |   PUT_WORD (abfd, reloc_entry, natptr); | 
 | } | 
 |  | 
 | /* BFD deals internally with all things based from the section they're | 
 |    in. so, something in 10 bytes into a text section  with a base of | 
 |    50 would have a symbol (.text+10) and know .text vma was 50. | 
 |  | 
 |    Aout keeps all it's symbols based from zero, so the symbol would | 
 |    contain 60. This macro subs the base of each section from the value | 
 |    to give the true offset from the section */ | 
 |  | 
 |  | 
 | #define MOVE_ADDRESS(ad)						\ | 
 |   if (r_extern)								\ | 
 |     {									\ | 
 |       /* Undefined symbol.  */						\ | 
 |       if (symbols != NULL && r_index < bfd_get_symcount (abfd))		\ | 
 | 	cache_ptr->sym_ptr_ptr = symbols + r_index;			\ | 
 |       else								\ | 
 | 	cache_ptr->sym_ptr_ptr = bfd_abs_section_ptr->symbol_ptr_ptr;	\ | 
 |       cache_ptr->addend = ad;						\ | 
 |     }									\ | 
 |   else									\ | 
 |     {									\ | 
 |       /* Defined, section relative. replace symbol with pointer to	\ | 
 | 	 symbol which points to section.  */				\ | 
 |       switch (r_index)							\ | 
 | 	{								\ | 
 | 	case N_TEXT:							\ | 
 | 	case N_TEXT | N_EXT:						\ | 
 | 	  cache_ptr->sym_ptr_ptr  = obj_textsec (abfd)->symbol_ptr_ptr;	\ | 
 | 	  cache_ptr->addend = ad  - su->textsec->vma;			\ | 
 | 	  break;							\ | 
 | 	case N_DATA:							\ | 
 | 	case N_DATA | N_EXT:						\ | 
 | 	  cache_ptr->sym_ptr_ptr  = obj_datasec (abfd)->symbol_ptr_ptr;	\ | 
 | 	  cache_ptr->addend = ad - su->datasec->vma;			\ | 
 | 	  break;							\ | 
 | 	case N_BSS:							\ | 
 | 	case N_BSS | N_EXT:						\ | 
 | 	  cache_ptr->sym_ptr_ptr  = obj_bsssec (abfd)->symbol_ptr_ptr;	\ | 
 | 	  cache_ptr->addend = ad - su->bsssec->vma;			\ | 
 | 	  break;							\ | 
 | 	default:							\ | 
 | 	case N_ABS:							\ | 
 | 	case N_ABS | N_EXT:						\ | 
 | 	  cache_ptr->sym_ptr_ptr = bfd_abs_section_ptr->symbol_ptr_ptr;	\ | 
 | 	  cache_ptr->addend = ad;					\ | 
 | 	  break;							\ | 
 | 	}								\ | 
 |     } | 
 |  | 
 | static void | 
 | pdp11_aout_swap_reloc_in (bfd *		 abfd, | 
 | 			  bfd_byte *	 bytes, | 
 | 			  arelent *	 cache_ptr, | 
 | 			  bfd_size_type	 offset, | 
 | 			  asymbol **	 symbols, | 
 | 			  bfd_size_type	 symcount) | 
 | { | 
 |   struct aoutdata *su = &(abfd->tdata.aout_data->a); | 
 |   unsigned int r_index; | 
 |   int reloc_entry; | 
 |   int r_extern; | 
 |   int r_pcrel; | 
 |  | 
 |   reloc_entry = GET_WORD (abfd, (void *) bytes); | 
 |  | 
 |   r_pcrel = reloc_entry & RELFLG; | 
 |  | 
 |   cache_ptr->address = offset; | 
 |   cache_ptr->howto = howto_table_pdp11 + (r_pcrel ? 1 : 0); | 
 |  | 
 |   if ((reloc_entry & RTYPE) == RABS) | 
 |     r_index = N_ABS; | 
 |   else | 
 |     r_index = RINDEX (reloc_entry); | 
 |  | 
 |   /* r_extern reflects whether the symbol the reloc is against is | 
 |      local or global.  */ | 
 |   r_extern = (reloc_entry & RTYPE) == REXT; | 
 |  | 
 |   if (r_extern && r_index >= symcount) | 
 |     { | 
 |       /* We could arrange to return an error, but it might be useful | 
 | 	 to see the file even if it is bad.  FIXME: Of course this | 
 | 	 means that objdump -r *doesn't* see the actual reloc, and | 
 | 	 objcopy silently writes a different reloc.  */ | 
 |       r_extern = 0; | 
 |       r_index = N_ABS; | 
 |     } | 
 |  | 
 |   MOVE_ADDRESS(0); | 
 | } | 
 |  | 
 | /* Read and swap the relocs for a section.  */ | 
 |  | 
 | bool | 
 | NAME (aout, slurp_reloc_table) (bfd *abfd, sec_ptr asect, asymbol **symbols) | 
 | { | 
 |   bfd_byte *rptr; | 
 |   bfd_size_type count; | 
 |   bfd_size_type reloc_size; | 
 |   void * relocs; | 
 |   arelent *reloc_cache; | 
 |   size_t each_size; | 
 |   unsigned int counter = 0; | 
 |   arelent *cache_ptr; | 
 |  | 
 |   if (asect->relocation) | 
 |     return true; | 
 |  | 
 |   if (asect->flags & SEC_CONSTRUCTOR) | 
 |     return true; | 
 |  | 
 |   if (asect == obj_datasec (abfd)) | 
 |     reloc_size = exec_hdr(abfd)->a_drsize; | 
 |   else if (asect == obj_textsec (abfd)) | 
 |     reloc_size = exec_hdr(abfd)->a_trsize; | 
 |   else if (asect == obj_bsssec (abfd)) | 
 |     reloc_size = 0; | 
 |   else | 
 |     { | 
 |       bfd_set_error (bfd_error_invalid_operation); | 
 |       return false; | 
 |     } | 
 |  | 
 |   if (bfd_seek (abfd, asect->rel_filepos, SEEK_SET) != 0) | 
 |     return false; | 
 |   relocs = _bfd_malloc_and_read (abfd, reloc_size, reloc_size); | 
 |   if (relocs == NULL && reloc_size != 0) | 
 |     return false; | 
 |  | 
 |   each_size = obj_reloc_entry_size (abfd); | 
 |   count = reloc_size / each_size; | 
 |  | 
 |   /* Count the number of NON-ZERO relocs, this is the count we want.  */ | 
 |   { | 
 |     unsigned int real_count = 0; | 
 |  | 
 |     for (counter = 0; counter < count; counter++) | 
 |       { | 
 | 	int x; | 
 |  | 
 | 	x = GET_WORD (abfd, (char *) relocs + each_size * counter); | 
 | 	if (x != 0) | 
 | 	  real_count++; | 
 |       } | 
 |  | 
 |     count = real_count; | 
 |   } | 
 |  | 
 |   reloc_cache = bfd_zmalloc (count * sizeof (arelent)); | 
 |   if (reloc_cache == NULL && count != 0) | 
 |     return false; | 
 |  | 
 |   cache_ptr = reloc_cache; | 
 |  | 
 |   rptr = relocs; | 
 |   for (counter = 0; | 
 |        counter < count; | 
 |        counter++, rptr += RELOC_SIZE, cache_ptr++) | 
 |     { | 
 |       while (GET_WORD (abfd, (void *) rptr) == 0) | 
 | 	{ | 
 | 	  rptr += RELOC_SIZE; | 
 | 	  if ((char *) rptr >= (char *) relocs + reloc_size) | 
 | 	    goto done; | 
 | 	} | 
 |  | 
 |       pdp11_aout_swap_reloc_in (abfd, rptr, cache_ptr, | 
 | 				(bfd_size_type) ((char *) rptr - (char *) relocs), | 
 | 				symbols, | 
 | 				(bfd_size_type) bfd_get_symcount (abfd)); | 
 |     } | 
 |  done: | 
 |   /* Just in case, if rptr >= relocs + reloc_size should happen | 
 |      too early.  */ | 
 |   BFD_ASSERT (counter == count); | 
 |  | 
 |   free (relocs); | 
 |  | 
 |   asect->relocation = reloc_cache; | 
 |   asect->reloc_count = cache_ptr - reloc_cache; | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Write out a relocation section into an object file.  */ | 
 |  | 
 | bool | 
 | NAME (aout, squirt_out_relocs) (bfd *abfd, asection *section) | 
 | { | 
 |   arelent **generic; | 
 |   unsigned char *native; | 
 |   unsigned int count = section->reloc_count; | 
 |   bfd_size_type natsize; | 
 |  | 
 |   natsize = section->size; | 
 |   native = bfd_zalloc (abfd, natsize); | 
 |   if (!native) | 
 |     return false; | 
 |  | 
 |   generic = section->orelocation; | 
 |   if (generic != NULL) | 
 |     { | 
 |       while (count > 0) | 
 | 	{ | 
 | 	  bfd_byte *r; | 
 |  | 
 | 	  if ((*generic)->howto == NULL | 
 | 	      || (*generic)->sym_ptr_ptr == NULL) | 
 | 	    { | 
 | 	      bfd_set_error (bfd_error_invalid_operation); | 
 | 	      _bfd_error_handler (_("%pB: attempt to write out " | 
 | 				    "unknown reloc type"), abfd); | 
 | 	      bfd_release (abfd, native); | 
 | 	      return false; | 
 | 	    } | 
 | 	  r = native + (*generic)->address; | 
 | 	  pdp11_aout_swap_reloc_out (abfd, *generic, r); | 
 | 	  count--; | 
 | 	  generic++; | 
 | 	} | 
 |     } | 
 |  | 
 |   if (bfd_bwrite ((void *) native, natsize, abfd) != natsize) | 
 |     { | 
 |       bfd_release (abfd, native); | 
 |       return false; | 
 |     } | 
 |  | 
 |   bfd_release (abfd, native); | 
 |   return true; | 
 | } | 
 |  | 
 | /* This is stupid.  This function should be a boolean predicate.  */ | 
 |  | 
 | long | 
 | NAME (aout, canonicalize_reloc) (bfd *abfd, | 
 | 				 sec_ptr section, | 
 | 				 arelent **relptr, | 
 | 				 asymbol **symbols) | 
 | { | 
 |   arelent *tblptr = section->relocation; | 
 |   unsigned int count; | 
 |  | 
 |   if (section == obj_bsssec (abfd)) | 
 |     { | 
 |       *relptr = NULL; | 
 |       return 0; | 
 |     } | 
 |  | 
 |   if (!(tblptr || NAME (aout, slurp_reloc_table)(abfd, section, symbols))) | 
 |     return -1; | 
 |  | 
 |   if (section->flags & SEC_CONSTRUCTOR) | 
 |     { | 
 |       arelent_chain *chain = section->constructor_chain; | 
 |  | 
 |       for (count = 0; count < section->reloc_count; count ++) | 
 | 	{ | 
 | 	  *relptr ++ = &chain->relent; | 
 | 	  chain = chain->next; | 
 | 	} | 
 |     } | 
 |   else | 
 |     { | 
 |       tblptr = section->relocation; | 
 |  | 
 |       for (count = 0; count++ < section->reloc_count;) | 
 | 	*relptr++ = tblptr++; | 
 |     } | 
 |  | 
 |   *relptr = 0; | 
 |  | 
 |   return section->reloc_count; | 
 | } | 
 |  | 
 | long | 
 | NAME (aout, get_reloc_upper_bound) (bfd *abfd, sec_ptr asect) | 
 | { | 
 |   size_t count, raw; | 
 |  | 
 |   if (asect->flags & SEC_CONSTRUCTOR) | 
 |     count = asect->reloc_count; | 
 |   else if (asect == obj_datasec (abfd)) | 
 |     count = exec_hdr (abfd)->a_drsize / obj_reloc_entry_size (abfd); | 
 |   else if (asect == obj_textsec (abfd)) | 
 |     count = exec_hdr (abfd)->a_trsize / obj_reloc_entry_size (abfd); | 
 |   else if (asect == obj_bsssec (abfd)) | 
 |     count = 0; | 
 |   else | 
 |     { | 
 |       bfd_set_error (bfd_error_invalid_operation); | 
 |       return -1; | 
 |     } | 
 |  | 
 |   if (count >= LONG_MAX / sizeof (arelent *) | 
 |       || _bfd_mul_overflow (count, obj_reloc_entry_size (abfd), &raw)) | 
 |     { | 
 |       bfd_set_error (bfd_error_file_too_big); | 
 |       return -1; | 
 |     } | 
 |   if (!bfd_write_p (abfd)) | 
 |     { | 
 |       ufile_ptr filesize = bfd_get_file_size (abfd); | 
 |       if (filesize != 0 && raw > filesize) | 
 | 	{ | 
 | 	  bfd_set_error (bfd_error_file_truncated); | 
 | 	  return -1; | 
 | 	} | 
 |     } | 
 |   return (count + 1) * sizeof (arelent *); | 
 | } | 
 |  | 
 |  | 
 | long | 
 | NAME (aout, get_symtab_upper_bound) (bfd *abfd) | 
 | { | 
 |   if (!NAME (aout, slurp_symbol_table) (abfd)) | 
 |     return -1; | 
 |  | 
 |   return (bfd_get_symcount (abfd) + 1) * (sizeof (aout_symbol_type *)); | 
 | } | 
 |  | 
 | alent * | 
 | NAME (aout, get_lineno) (bfd * abfd ATTRIBUTE_UNUSED, | 
 | 			 asymbol * symbol ATTRIBUTE_UNUSED) | 
 | { | 
 |   return NULL; | 
 | } | 
 |  | 
 | void | 
 | NAME (aout, get_symbol_info) (bfd * abfd ATTRIBUTE_UNUSED, | 
 | 			      asymbol *symbol, | 
 | 			      symbol_info *ret) | 
 | { | 
 |   bfd_symbol_info (symbol, ret); | 
 |  | 
 |   if (ret->type == '?') | 
 |     { | 
 |       int type_code = aout_symbol(symbol)->type & 0xff; | 
 |       const char *stab_name = bfd_get_stab_name (type_code); | 
 |       static char buf[10]; | 
 |  | 
 |       if (stab_name == NULL) | 
 | 	{ | 
 | 	  sprintf(buf, "(%d)", type_code); | 
 | 	  stab_name = buf; | 
 | 	} | 
 |       ret->type = '-'; | 
 |       ret->stab_type  = type_code; | 
 |       ret->stab_other = (unsigned) (aout_symbol(symbol)->other & 0xff); | 
 |       ret->stab_desc  = (unsigned) (aout_symbol(symbol)->desc & 0xffff); | 
 |       ret->stab_name  = stab_name; | 
 |     } | 
 | } | 
 |  | 
 | void | 
 | NAME (aout, print_symbol) (bfd * abfd, | 
 | 			   void * afile, | 
 | 			   asymbol *symbol, | 
 | 			   bfd_print_symbol_type how) | 
 | { | 
 |   FILE *file = (FILE *) afile; | 
 |  | 
 |   switch (how) | 
 |     { | 
 |     case bfd_print_symbol_name: | 
 |       if (symbol->name) | 
 | 	fprintf(file,"%s", symbol->name); | 
 |       break; | 
 |     case bfd_print_symbol_more: | 
 |       fprintf(file,"%4x %2x %2x", | 
 | 	      (unsigned) (aout_symbol (symbol)->desc & 0xffff), | 
 | 	      (unsigned) (aout_symbol (symbol)->other & 0xff), | 
 | 	      (unsigned) (aout_symbol (symbol)->type)); | 
 |       break; | 
 |     case bfd_print_symbol_all: | 
 |       { | 
 | 	const char *section_name = symbol->section->name; | 
 |  | 
 | 	bfd_print_symbol_vandf (abfd, (void *) file, symbol); | 
 |  | 
 | 	fprintf (file," %-5s %04x %02x %02x", | 
 | 		 section_name, | 
 | 		 (unsigned) (aout_symbol (symbol)->desc & 0xffff), | 
 | 		 (unsigned) (aout_symbol (symbol)->other & 0xff), | 
 | 		 (unsigned) (aout_symbol (symbol)->type  & 0xff)); | 
 | 	if (symbol->name) | 
 | 	  fprintf(file," %s", symbol->name); | 
 |       } | 
 |       break; | 
 |     } | 
 | } | 
 |  | 
 | /* If we don't have to allocate more than 1MB to hold the generic | 
 |    symbols, we use the generic minisymbol method: it's faster, since | 
 |    it only translates the symbols once, not multiple times.  */ | 
 | #define MINISYM_THRESHOLD (1000000 / sizeof (asymbol)) | 
 |  | 
 | /* Read minisymbols.  For minisymbols, we use the unmodified a.out | 
 |    symbols.  The minisymbol_to_symbol function translates these into | 
 |    BFD asymbol structures.  */ | 
 |  | 
 | long | 
 | NAME (aout, read_minisymbols) (bfd *abfd, | 
 | 			       bool dynamic, | 
 | 			       void * *minisymsp, | 
 | 			       unsigned int *sizep) | 
 | { | 
 |   if (dynamic) | 
 |     /* We could handle the dynamic symbols here as well, but it's | 
 |        easier to hand them off.  */ | 
 |     return _bfd_generic_read_minisymbols (abfd, dynamic, minisymsp, sizep); | 
 |  | 
 |   if (! aout_get_external_symbols (abfd)) | 
 |     return -1; | 
 |  | 
 |   if (obj_aout_external_sym_count (abfd) < MINISYM_THRESHOLD) | 
 |     return _bfd_generic_read_minisymbols (abfd, dynamic, minisymsp, sizep); | 
 |  | 
 |   *minisymsp = (void *) obj_aout_external_syms (abfd); | 
 |  | 
 |   /* By passing the external symbols back from this routine, we are | 
 |      giving up control over the memory block.  Clear | 
 |      obj_aout_external_syms, so that we do not try to free it | 
 |      ourselves.  */ | 
 |   obj_aout_external_syms (abfd) = NULL; | 
 |  | 
 |   *sizep = EXTERNAL_NLIST_SIZE; | 
 |   return obj_aout_external_sym_count (abfd); | 
 | } | 
 |  | 
 | /* Convert a minisymbol to a BFD asymbol.  A minisymbol is just an | 
 |    unmodified a.out symbol.  The SYM argument is a structure returned | 
 |    by bfd_make_empty_symbol, which we fill in here.  */ | 
 |  | 
 | asymbol * | 
 | NAME (aout, minisymbol_to_symbol) (bfd *abfd, | 
 | 				   bool dynamic, | 
 | 				   const void * minisym, | 
 | 				   asymbol *sym) | 
 | { | 
 |   if (dynamic | 
 |       || obj_aout_external_sym_count (abfd) < MINISYM_THRESHOLD) | 
 |     return _bfd_generic_minisymbol_to_symbol (abfd, dynamic, minisym, sym); | 
 |  | 
 |   memset (sym, 0, sizeof (aout_symbol_type)); | 
 |  | 
 |   /* We call translate_symbol_table to translate a single symbol.  */ | 
 |   if (! (NAME (aout, translate_symbol_table) | 
 | 	 (abfd, | 
 | 	  (aout_symbol_type *) sym, | 
 | 	  (struct external_nlist *) minisym, | 
 | 	  (bfd_size_type) 1, | 
 | 	  obj_aout_external_strings (abfd), | 
 | 	  obj_aout_external_string_size (abfd), | 
 | 	  false))) | 
 |     return NULL; | 
 |  | 
 |   return sym; | 
 | } | 
 |  | 
 | /* Provided a BFD, a section and an offset into the section, calculate | 
 |    and return the name of the source file and the line nearest to the | 
 |    wanted location.  */ | 
 |  | 
 | bool | 
 | NAME (aout, find_nearest_line) (bfd *abfd, | 
 | 				asymbol **symbols, | 
 | 				asection *section, | 
 | 				bfd_vma offset, | 
 | 				const char **filename_ptr, | 
 | 				const char **functionname_ptr, | 
 | 				unsigned int *line_ptr, | 
 | 				unsigned int *discriminator_ptr) | 
 | { | 
 |   /* Run down the file looking for the filename, function and linenumber.  */ | 
 |   asymbol **p; | 
 |   const char *directory_name = NULL; | 
 |   const char *main_file_name = NULL; | 
 |   const char *current_file_name = NULL; | 
 |   const char *line_file_name = NULL; /* Value of current_file_name at line number.  */ | 
 |   bfd_vma low_line_vma = 0; | 
 |   bfd_vma low_func_vma = 0; | 
 |   asymbol *func = 0; | 
 |   size_t filelen, funclen; | 
 |   char *buf; | 
 |  | 
 |   *filename_ptr = bfd_get_filename (abfd); | 
 |   *functionname_ptr = NULL; | 
 |   *line_ptr = 0; | 
 |   if (discriminator_ptr) | 
 |     *discriminator_ptr = 0; | 
 |  | 
 |   if (symbols != NULL) | 
 |     { | 
 |       for (p = symbols; *p; p++) | 
 | 	{ | 
 | 	  aout_symbol_type  *q = (aout_symbol_type *)(*p); | 
 | 	next: | 
 | 	  switch (q->type) | 
 | 	    { | 
 | 	    case N_TEXT: | 
 | 	      /* If this looks like a file name symbol, and it comes after | 
 | 		 the line number we have found so far, but before the | 
 | 		 offset, then we have probably not found the right line | 
 | 		 number.  */ | 
 | 	      if (q->symbol.value <= offset | 
 | 		  && ((q->symbol.value > low_line_vma | 
 | 		       && (line_file_name != NULL | 
 | 			   || *line_ptr != 0)) | 
 | 		      || (q->symbol.value > low_func_vma | 
 | 			  && func != NULL))) | 
 | 		{ | 
 | 		  const char * symname; | 
 |  | 
 | 		  symname = q->symbol.name; | 
 |  | 
 | 		  if (symname != NULL | 
 | 		      && strlen (symname) > 2 | 
 | 		      && strcmp (symname + strlen (symname) - 2, ".o") == 0) | 
 | 		    { | 
 | 		      if (q->symbol.value > low_line_vma) | 
 | 			{ | 
 | 			  *line_ptr = 0; | 
 | 			  line_file_name = NULL; | 
 | 			} | 
 | 		      if (q->symbol.value > low_func_vma) | 
 | 			func = NULL; | 
 | 		    } | 
 | 		} | 
 | 	      break; | 
 |  | 
 | 	    case N_SO: | 
 | 	      /* If this symbol is less than the offset, but greater than | 
 | 		 the line number we have found so far, then we have not | 
 | 		 found the right line number.  */ | 
 | 	      if (q->symbol.value <= offset) | 
 | 		{ | 
 | 		  if (q->symbol.value > low_line_vma) | 
 | 		    { | 
 | 		      *line_ptr = 0; | 
 | 		      line_file_name = NULL; | 
 | 		    } | 
 | 		  if (q->symbol.value > low_func_vma) | 
 | 		    func = NULL; | 
 | 		} | 
 |  | 
 | 	      main_file_name = current_file_name = q->symbol.name; | 
 | 	      /* Look ahead to next symbol to check if that too is an N_SO.  */ | 
 | 	      p++; | 
 | 	      if (*p == NULL) | 
 | 		goto done; | 
 | 	      q = (aout_symbol_type *)(*p); | 
 | 	      if (q->type != (int) N_SO) | 
 | 		goto next; | 
 |  | 
 | 	      /* Found a second N_SO  First is directory; second is filename.  */ | 
 | 	      directory_name = current_file_name; | 
 | 	      main_file_name = current_file_name = q->symbol.name; | 
 | 	      if (obj_textsec(abfd) != section) | 
 | 		goto done; | 
 | 	      break; | 
 | 	    case N_SOL: | 
 | 	      current_file_name = q->symbol.name; | 
 | 	      break; | 
 |  | 
 | 	    case N_SLINE: | 
 | 	    case N_DSLINE: | 
 | 	    case N_BSLINE: | 
 | 	      /* We'll keep this if it resolves nearer than the one we have | 
 | 		 already.  */ | 
 | 	      if (q->symbol.value >= low_line_vma | 
 | 		  && q->symbol.value <= offset) | 
 | 		{ | 
 | 		  *line_ptr = q->desc; | 
 | 		  low_line_vma = q->symbol.value; | 
 | 		  line_file_name = current_file_name; | 
 | 		} | 
 | 	      break; | 
 |  | 
 | 	    case N_FUN: | 
 | 	      { | 
 | 		/* We'll keep this if it is nearer than the one we have already.  */ | 
 | 		if (q->symbol.value >= low_func_vma && | 
 | 		    q->symbol.value <= offset) | 
 | 		  { | 
 | 		    low_func_vma = q->symbol.value; | 
 | 		    func = (asymbol *) q; | 
 | 		  } | 
 | 		else if (q->symbol.value > offset) | 
 | 		  goto done; | 
 | 	      } | 
 | 	      break; | 
 | 	    } | 
 | 	} | 
 |     } | 
 |  | 
 |  done: | 
 |   if (*line_ptr != 0) | 
 |     main_file_name = line_file_name; | 
 |  | 
 |   if (main_file_name == NULL | 
 |       || main_file_name[0] == '/' | 
 |       || directory_name == NULL) | 
 |     filelen = 0; | 
 |   else | 
 |     filelen = strlen (directory_name) + strlen (main_file_name); | 
 |   if (func == NULL) | 
 |     funclen = 0; | 
 |   else | 
 |     funclen = strlen (bfd_asymbol_name (func)); | 
 |  | 
 |   free (adata (abfd).line_buf); | 
 |   if (filelen + funclen == 0) | 
 |     adata (abfd).line_buf = buf = NULL; | 
 |   else | 
 |     { | 
 |       buf = bfd_malloc ((bfd_size_type) filelen + funclen + 3); | 
 |       adata (abfd).line_buf = buf; | 
 |       if (buf == NULL) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   if (main_file_name != NULL) | 
 |     { | 
 |       if (main_file_name[0] == '/' || directory_name == NULL) | 
 | 	*filename_ptr = main_file_name; | 
 |       else | 
 | 	{ | 
 | 	  if (buf == NULL) | 
 | 	    /* PR binutils/20891: In a corrupt input file both | 
 | 	       main_file_name and directory_name can be empty...  */ | 
 | 	    * filename_ptr = NULL; | 
 | 	  else | 
 | 	    { | 
 | 	      snprintf (buf, filelen + 1, "%s%s", directory_name, | 
 | 			main_file_name); | 
 | 	      *filename_ptr = buf; | 
 | 	      buf += filelen + 1; | 
 | 	    } | 
 | 	} | 
 |     } | 
 |  | 
 |   if (func) | 
 |     { | 
 |       const char *function = func->name; | 
 |       char *colon; | 
 |  | 
 |       if (buf == NULL) | 
 | 	{ | 
 | 	  /* PR binutils/20892: In a corrupt input file func can be empty.  */ | 
 | 	  * functionname_ptr = NULL; | 
 | 	  return true; | 
 | 	} | 
 |       /* The caller expects a symbol name.  We actually have a | 
 | 	 function name, without the leading underscore.  Put the | 
 | 	 underscore back in, so that the caller gets a symbol name.  */ | 
 |       if (bfd_get_symbol_leading_char (abfd) == '\0') | 
 | 	strcpy (buf, function); | 
 |       else | 
 | 	{ | 
 | 	  buf[0] = bfd_get_symbol_leading_char (abfd); | 
 | 	  strcpy (buf + 1, function); | 
 | 	} | 
 |  | 
 |       /* Have to remove : stuff.  */ | 
 |       colon = strchr (buf, ':'); | 
 |       if (colon != NULL) | 
 | 	*colon = '\0'; | 
 |       *functionname_ptr = buf; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | int | 
 | NAME (aout, sizeof_headers) (bfd *abfd, | 
 | 			     struct bfd_link_info *info ATTRIBUTE_UNUSED) | 
 | { | 
 |   return adata (abfd).exec_bytes_size; | 
 | } | 
 |  | 
 | /* Free all information we have cached for this BFD.  We can always | 
 |    read it again later if we need it.  */ | 
 |  | 
 | bool | 
 | NAME (aout, bfd_free_cached_info) (bfd *abfd) | 
 | { | 
 |   asection *o; | 
 |  | 
 |   if (bfd_get_format (abfd) != bfd_object | 
 |       || abfd->tdata.aout_data == NULL) | 
 |     return true; | 
 |  | 
 | #define BFCI_FREE(x) do { free (x); x = NULL; } while (0) | 
 |   BFCI_FREE (adata (abfd).line_buf); | 
 |   BFCI_FREE (obj_aout_symbols (abfd)); | 
 | #ifdef USE_MMAP | 
 |   obj_aout_external_syms (abfd) = 0; | 
 |   bfd_free_window (&obj_aout_sym_window (abfd)); | 
 |   bfd_free_window (&obj_aout_string_window (abfd)); | 
 |   obj_aout_external_strings (abfd) = 0; | 
 | #else | 
 |   BFCI_FREE (obj_aout_external_syms (abfd)); | 
 |   BFCI_FREE (obj_aout_external_strings (abfd)); | 
 | #endif | 
 |   for (o = abfd->sections; o != NULL; o = o->next) | 
 |     BFCI_FREE (o->relocation); | 
 | #undef BFCI_FREE | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Routine to create an entry in an a.out link hash table.  */ | 
 |  | 
 | struct bfd_hash_entry * | 
 | NAME (aout, link_hash_newfunc) (struct bfd_hash_entry *entry, | 
 | 				struct bfd_hash_table *table, | 
 | 				const char *string) | 
 | { | 
 |   struct aout_link_hash_entry *ret = (struct aout_link_hash_entry *) entry; | 
 |  | 
 |   /* Allocate the structure if it has not already been allocated by a | 
 |      subclass.  */ | 
 |   if (ret == NULL) | 
 |     ret = bfd_hash_allocate (table, sizeof (* ret)); | 
 |   if (ret == NULL) | 
 |     return NULL; | 
 |  | 
 |   /* Call the allocation method of the superclass.  */ | 
 |   ret = (struct aout_link_hash_entry *) | 
 | 	 _bfd_link_hash_newfunc ((struct bfd_hash_entry *) ret, table, string); | 
 |   if (ret) | 
 |     { | 
 |       /* Set local fields.  */ | 
 |       ret->written = false; | 
 |       ret->indx = -1; | 
 |     } | 
 |  | 
 |   return (struct bfd_hash_entry *) ret; | 
 | } | 
 |  | 
 | /* Initialize an a.out link hash table.  */ | 
 |  | 
 | bool | 
 | NAME (aout, link_hash_table_init) (struct aout_link_hash_table *table, | 
 | 				   bfd *abfd, | 
 | 				   struct bfd_hash_entry *(*newfunc) (struct bfd_hash_entry *, | 
 | 								     struct bfd_hash_table *, | 
 | 								     const char *), | 
 | 				   unsigned int entsize) | 
 | { | 
 |   return _bfd_link_hash_table_init (&table->root, abfd, newfunc, entsize); | 
 | } | 
 |  | 
 | /* Create an a.out link hash table.  */ | 
 |  | 
 | struct bfd_link_hash_table * | 
 | NAME (aout, link_hash_table_create) (bfd *abfd) | 
 | { | 
 |   struct aout_link_hash_table *ret; | 
 |   size_t amt = sizeof (struct aout_link_hash_table); | 
 |  | 
 |   ret = bfd_malloc (amt); | 
 |   if (ret == NULL) | 
 |     return NULL; | 
 |   if (! NAME (aout, link_hash_table_init) (ret, abfd, | 
 | 					   NAME (aout, link_hash_newfunc), | 
 | 					   sizeof (struct aout_link_hash_entry))) | 
 |     { | 
 |       free (ret); | 
 |       return NULL; | 
 |     } | 
 |   return &ret->root; | 
 | } | 
 |  | 
 | /* Free up the internal symbols read from an a.out file.  */ | 
 |  | 
 | static bool | 
 | aout_link_free_symbols (bfd *abfd) | 
 | { | 
 |   if (obj_aout_external_syms (abfd) != NULL) | 
 |     { | 
 | #ifdef USE_MMAP | 
 |       bfd_free_window (&obj_aout_sym_window (abfd)); | 
 | #else | 
 |       free ((void *) obj_aout_external_syms (abfd)); | 
 | #endif | 
 |       obj_aout_external_syms (abfd) = NULL; | 
 |     } | 
 |  | 
 |   if (obj_aout_external_strings (abfd) != NULL) | 
 |     { | 
 | #ifdef USE_MMAP | 
 |       bfd_free_window (&obj_aout_string_window (abfd)); | 
 | #else | 
 |       free ((void *) obj_aout_external_strings (abfd)); | 
 | #endif | 
 |       obj_aout_external_strings (abfd) = NULL; | 
 |     } | 
 |   return true; | 
 | } | 
 |  | 
 | /* Given an a.out BFD, add symbols to the global hash table as | 
 |    appropriate.  */ | 
 |  | 
 | bool | 
 | NAME (aout, link_add_symbols) (bfd *abfd, struct bfd_link_info *info) | 
 | { | 
 |   switch (bfd_get_format (abfd)) | 
 |     { | 
 |     case bfd_object: | 
 |       return aout_link_add_object_symbols (abfd, info); | 
 |     case bfd_archive: | 
 |       return _bfd_generic_link_add_archive_symbols | 
 | 	(abfd, info, aout_link_check_archive_element); | 
 |     default: | 
 |       bfd_set_error (bfd_error_wrong_format); | 
 |       return false; | 
 |     } | 
 | } | 
 |  | 
 | /* Add symbols from an a.out object file.  */ | 
 |  | 
 | static bool | 
 | aout_link_add_object_symbols (bfd *abfd, struct bfd_link_info *info) | 
 | { | 
 |   if (! aout_get_external_symbols (abfd)) | 
 |     return false; | 
 |   if (! aout_link_add_symbols (abfd, info)) | 
 |     return false; | 
 |   if (! info->keep_memory) | 
 |     { | 
 |       if (! aout_link_free_symbols (abfd)) | 
 | 	return false; | 
 |     } | 
 |   return true; | 
 | } | 
 |  | 
 | /* Look through the internal symbols to see if this object file should | 
 |    be included in the link.  We should include this object file if it | 
 |    defines any symbols which are currently undefined.  If this object | 
 |    file defines a common symbol, then we may adjust the size of the | 
 |    known symbol but we do not include the object file in the link | 
 |    (unless there is some other reason to include it).  */ | 
 |  | 
 | static bool | 
 | aout_link_check_ar_symbols (bfd *abfd, | 
 | 			    struct bfd_link_info *info, | 
 | 			    bool *pneeded, | 
 | 			    bfd **subsbfd) | 
 | { | 
 |   struct external_nlist *p; | 
 |   struct external_nlist *pend; | 
 |   char *strings; | 
 |  | 
 |   *pneeded = false; | 
 |  | 
 |   /* Look through all the symbols.  */ | 
 |   p = obj_aout_external_syms (abfd); | 
 |   pend = p + obj_aout_external_sym_count (abfd); | 
 |   strings = obj_aout_external_strings (abfd); | 
 |   for (; p < pend; p++) | 
 |     { | 
 |       int type = H_GET_8 (abfd, p->e_type); | 
 |       const char *name = strings + GET_WORD (abfd, p->e_strx); | 
 |       struct bfd_link_hash_entry *h; | 
 |  | 
 |       /* Ignore symbols that are not externally visible.  This is an | 
 | 	 optimization only, as we check the type more thoroughly | 
 | 	 below.  */ | 
 |       if ((type & N_EXT) == 0 | 
 | 	  || is_stab(type, name) | 
 | 	  || type == N_FN) | 
 | 	continue; | 
 |  | 
 |       h = bfd_link_hash_lookup (info->hash, name, false, false, true); | 
 |  | 
 |       /* We are only interested in symbols that are currently | 
 | 	 undefined or common.  */ | 
 |       if (h == NULL | 
 | 	  || (h->type != bfd_link_hash_undefined | 
 | 	      && h->type != bfd_link_hash_common)) | 
 | 	continue; | 
 |  | 
 |       if (type == (N_TEXT | N_EXT) | 
 | 	  || type == (N_DATA | N_EXT) | 
 | 	  || type == (N_BSS | N_EXT) | 
 | 	  || type == (N_ABS | N_EXT)) | 
 | 	{ | 
 | 	  /* This object file defines this symbol.  We must link it | 
 | 	     in.  This is true regardless of whether the current | 
 | 	     definition of the symbol is undefined or common.  If the | 
 | 	     current definition is common, we have a case in which we | 
 | 	     have already seen an object file including | 
 | 		 int a; | 
 | 	     and this object file from the archive includes | 
 | 		 int a = 5; | 
 | 	     In such a case we must include this object file. | 
 |  | 
 | 	     FIXME: The SunOS 4.1.3 linker will pull in the archive | 
 | 	     element if the symbol is defined in the .data section, | 
 | 	     but not if it is defined in the .text section.  That | 
 | 	     seems a bit crazy to me, and I haven't implemented it. | 
 | 	     However, it might be correct.  */ | 
 | 	  if (!(*info->callbacks | 
 | 		->add_archive_element) (info, abfd, name, subsbfd)) | 
 | 	    continue; | 
 | 	  *pneeded = true; | 
 | 	  return true; | 
 | 	} | 
 |  | 
 |       if (type == (N_UNDF | N_EXT)) | 
 | 	{ | 
 | 	  bfd_vma value; | 
 |  | 
 | 	  value = GET_WORD (abfd, p->e_value); | 
 | 	  if (value != 0) | 
 | 	    { | 
 | 	      /* This symbol is common in the object from the archive | 
 | 		 file.  */ | 
 | 	      if (h->type == bfd_link_hash_undefined) | 
 | 		{ | 
 | 		  bfd *symbfd; | 
 | 		  unsigned int power; | 
 |  | 
 | 		  symbfd = h->u.undef.abfd; | 
 | 		  if (symbfd == NULL) | 
 | 		    { | 
 | 		      /* This symbol was created as undefined from | 
 | 			 outside BFD.  We assume that we should link | 
 | 			 in the object file.  This is done for the -u | 
 | 			 option in the linker.  */ | 
 | 		      if (!(*info->callbacks | 
 | 			    ->add_archive_element) (info, abfd, name, subsbfd)) | 
 | 			return false; | 
 | 		      *pneeded = true; | 
 | 		      return true; | 
 | 		    } | 
 | 		  /* Turn the current link symbol into a common | 
 | 		     symbol.  It is already on the undefs list.  */ | 
 | 		  h->type = bfd_link_hash_common; | 
 | 		  h->u.c.p = bfd_hash_allocate (&info->hash->table, | 
 | 						sizeof (struct bfd_link_hash_common_entry)); | 
 | 		  if (h->u.c.p == NULL) | 
 | 		    return false; | 
 |  | 
 | 		  h->u.c.size = value; | 
 |  | 
 | 		  /* FIXME: This isn't quite right.  The maximum | 
 | 		     alignment of a common symbol should be set by the | 
 | 		     architecture of the output file, not of the input | 
 | 		     file.  */ | 
 | 		  power = bfd_log2 (value); | 
 | 		  if (power > bfd_get_arch_info (abfd)->section_align_power) | 
 | 		    power = bfd_get_arch_info (abfd)->section_align_power; | 
 | 		  h->u.c.p->alignment_power = power; | 
 |  | 
 | 		  h->u.c.p->section = bfd_make_section_old_way (symbfd, | 
 | 								"COMMON"); | 
 | 		} | 
 | 	      else | 
 | 		{ | 
 | 		  /* Adjust the size of the common symbol if | 
 | 		     necessary.  */ | 
 | 		  if (value > h->u.c.size) | 
 | 		    h->u.c.size = value; | 
 | 		} | 
 | 	    } | 
 | 	} | 
 |     } | 
 |  | 
 |   /* We do not need this object file.  */ | 
 |   return true; | 
 | } | 
 |  | 
 | /* Check a single archive element to see if we need to include it in | 
 |    the link.  *PNEEDED is set according to whether this element is | 
 |    needed in the link or not.  This is called from | 
 |    _bfd_generic_link_add_archive_symbols.  */ | 
 |  | 
 | static bool | 
 | aout_link_check_archive_element (bfd *abfd, | 
 | 				 struct bfd_link_info *info, | 
 | 				 struct bfd_link_hash_entry *h ATTRIBUTE_UNUSED, | 
 | 				 const char *name ATTRIBUTE_UNUSED, | 
 | 				 bool *pneeded) | 
 | { | 
 |   bfd *oldbfd; | 
 |   bool needed; | 
 |  | 
 |   if (!aout_get_external_symbols (abfd)) | 
 |     return false; | 
 |  | 
 |   oldbfd = abfd; | 
 |   if (!aout_link_check_ar_symbols (abfd, info, pneeded, &abfd)) | 
 |     return false; | 
 |  | 
 |   needed = *pneeded; | 
 |   if (needed) | 
 |     { | 
 |       /* Potentially, the add_archive_element hook may have set a | 
 | 	 substitute BFD for us.  */ | 
 |       if (abfd != oldbfd) | 
 | 	{ | 
 | 	  if (!info->keep_memory | 
 | 	      && !aout_link_free_symbols (oldbfd)) | 
 | 	    return false; | 
 | 	  if (!aout_get_external_symbols (abfd)) | 
 | 	    return false; | 
 | 	} | 
 |       if (!aout_link_add_symbols (abfd, info)) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   if (!info->keep_memory || !needed) | 
 |     { | 
 |       if (!aout_link_free_symbols (abfd)) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Add all symbols from an object file to the hash table.  */ | 
 |  | 
 | static bool | 
 | aout_link_add_symbols (bfd *abfd, struct bfd_link_info *info) | 
 | { | 
 |   bool (*add_one_symbol) | 
 |     (struct bfd_link_info *, bfd *, const char *, flagword, asection *, | 
 |      bfd_vma, const char *, bool, bool, struct bfd_link_hash_entry **); | 
 |   struct external_nlist *syms; | 
 |   bfd_size_type sym_count; | 
 |   char *strings; | 
 |   bool copy; | 
 |   struct aout_link_hash_entry **sym_hash; | 
 |   struct external_nlist *p; | 
 |   struct external_nlist *pend; | 
 |  | 
 |   syms = obj_aout_external_syms (abfd); | 
 |   sym_count = obj_aout_external_sym_count (abfd); | 
 |   strings = obj_aout_external_strings (abfd); | 
 |   if (info->keep_memory) | 
 |     copy = false; | 
 |   else | 
 |     copy = true; | 
 |  | 
 |   if (aout_backend_info (abfd)->add_dynamic_symbols != NULL) | 
 |     { | 
 |       if (! ((*aout_backend_info (abfd)->add_dynamic_symbols) | 
 | 	     (abfd, info, &syms, &sym_count, &strings))) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   /* We keep a list of the linker hash table entries that correspond | 
 |      to particular symbols.  We could just look them up in the hash | 
 |      table, but keeping the list is more efficient.  Perhaps this | 
 |      should be conditional on info->keep_memory.  */ | 
 |   sym_hash = bfd_alloc (abfd, | 
 | 			sym_count * sizeof (struct aout_link_hash_entry *)); | 
 |   if (sym_hash == NULL && sym_count != 0) | 
 |     return false; | 
 |   obj_aout_sym_hashes (abfd) = sym_hash; | 
 |  | 
 |   add_one_symbol = aout_backend_info (abfd)->add_one_symbol; | 
 |   if (add_one_symbol == NULL) | 
 |     add_one_symbol = _bfd_generic_link_add_one_symbol; | 
 |  | 
 |   p = syms; | 
 |   pend = p + sym_count; | 
 |   for (; p < pend; p++, sym_hash++) | 
 |     { | 
 |       int type; | 
 |       const char *name; | 
 |       bfd_vma value; | 
 |       asection *section; | 
 |       flagword flags; | 
 |       const char *string; | 
 |  | 
 |       *sym_hash = NULL; | 
 |  | 
 |       type = H_GET_8 (abfd, p->e_type); | 
 |  | 
 |       /* PR 19629: Corrupt binaries can contain illegal string offsets.  */ | 
 |       if (GET_WORD (abfd, p->e_strx) >= obj_aout_external_string_size (abfd)) | 
 | 	return false; | 
 |       name = strings + GET_WORD (abfd, p->e_strx); | 
 |  | 
 |       /* Ignore debugging symbols.  */ | 
 |       if (is_stab (type, name)) | 
 | 	continue; | 
 |  | 
 |       value = GET_WORD (abfd, p->e_value); | 
 |       flags = BSF_GLOBAL; | 
 |       string = NULL; | 
 |       switch (type) | 
 | 	{ | 
 | 	default: | 
 | 	  /* Shouldn't be any types not covered.  */ | 
 | 	  BFD_ASSERT (0); | 
 | 	  continue; | 
 |  | 
 | 	case N_UNDF: | 
 | 	case N_ABS: | 
 | 	case N_TEXT: | 
 | 	case N_DATA: | 
 | 	case N_BSS: | 
 | 	case N_REG: | 
 | 	case N_FN: | 
 | 	  /* Ignore symbols that are not externally visible.  */ | 
 | 	  continue; | 
 |  | 
 | 	case N_UNDF | N_EXT: | 
 | 	  if (value == 0) | 
 | 	    { | 
 | 	      section = bfd_und_section_ptr; | 
 | 	      flags = 0; | 
 | 	    } | 
 | 	  else | 
 | 	    section = bfd_com_section_ptr; | 
 | 	  break; | 
 | 	case N_ABS | N_EXT: | 
 | 	  section = bfd_abs_section_ptr; | 
 | 	  break; | 
 | 	case N_TEXT | N_EXT: | 
 | 	  section = obj_textsec (abfd); | 
 | 	  value -= bfd_section_vma (section); | 
 | 	  break; | 
 | 	case N_DATA | N_EXT: | 
 | 	  /* Treat N_SETV symbols as N_DATA symbol; see comment in | 
 | 	     translate_from_native_sym_flags.  */ | 
 | 	  section = obj_datasec (abfd); | 
 | 	  value -= bfd_section_vma (section); | 
 | 	  break; | 
 | 	case N_BSS | N_EXT: | 
 | 	  section = obj_bsssec (abfd); | 
 | 	  value -= bfd_section_vma (section); | 
 | 	  break; | 
 | 	} | 
 |  | 
 |       if (! ((*add_one_symbol) | 
 | 	     (info, abfd, name, flags, section, value, string, copy, false, | 
 | 	      (struct bfd_link_hash_entry **) sym_hash))) | 
 | 	return false; | 
 |  | 
 |       /* Restrict the maximum alignment of a common symbol based on | 
 | 	 the architecture, since a.out has no way to represent | 
 | 	 alignment requirements of a section in a .o file.  FIXME: | 
 | 	 This isn't quite right: it should use the architecture of the | 
 | 	 output file, not the input files.  */ | 
 |       if ((*sym_hash)->root.type == bfd_link_hash_common | 
 | 	  && ((*sym_hash)->root.u.c.p->alignment_power > | 
 | 	      bfd_get_arch_info (abfd)->section_align_power)) | 
 | 	(*sym_hash)->root.u.c.p->alignment_power = | 
 | 	  bfd_get_arch_info (abfd)->section_align_power; | 
 |  | 
 |       /* If this is a set symbol, and we are not building sets, then | 
 | 	 it is possible for the hash entry to not have been set.  In | 
 | 	 such a case, treat the symbol as not globally defined.  */ | 
 |       if ((*sym_hash)->root.type == bfd_link_hash_new) | 
 | 	{ | 
 | 	  BFD_ASSERT ((flags & BSF_CONSTRUCTOR) != 0); | 
 | 	  *sym_hash = NULL; | 
 | 	} | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Look up an entry in an the header file hash table.  */ | 
 |  | 
 | #define aout_link_includes_lookup(table, string, create, copy) \ | 
 |   ((struct aout_link_includes_entry *) \ | 
 |    bfd_hash_lookup (&(table)->root, (string), (create), (copy))) | 
 |  | 
 | /* The function to create a new entry in the header file hash table.  */ | 
 |  | 
 | static struct bfd_hash_entry * | 
 | aout_link_includes_newfunc (struct bfd_hash_entry *entry, | 
 | 			    struct bfd_hash_table *table, | 
 | 			    const char *string) | 
 | { | 
 |   struct aout_link_includes_entry * ret = | 
 |     (struct aout_link_includes_entry *) entry; | 
 |  | 
 |   /* Allocate the structure if it has not already been allocated by a | 
 |      subclass.  */ | 
 |   if (ret == NULL) | 
 |     ret = bfd_hash_allocate (table, | 
 | 			     sizeof (struct aout_link_includes_entry)); | 
 |   if (ret == NULL) | 
 |     return NULL; | 
 |  | 
 |   /* Call the allocation method of the superclass.  */ | 
 |   ret = ((struct aout_link_includes_entry *) | 
 | 	 bfd_hash_newfunc ((struct bfd_hash_entry *) ret, table, string)); | 
 |   if (ret) | 
 |     /* Set local fields.  */ | 
 |     ret->totals = NULL; | 
 |  | 
 |   return (struct bfd_hash_entry *) ret; | 
 | } | 
 |  | 
 | /* Write out a symbol that was not associated with an a.out input | 
 |    object.  */ | 
 |  | 
 | static bool | 
 | aout_link_write_other_symbol (struct bfd_hash_entry *bh, void *data) | 
 | { | 
 |   struct aout_link_hash_entry *h = (struct aout_link_hash_entry *) bh; | 
 |   struct aout_final_link_info *flaginfo = (struct aout_final_link_info *) data; | 
 |   bfd *output_bfd; | 
 |   int type; | 
 |   bfd_vma val; | 
 |   struct external_nlist outsym; | 
 |   bfd_size_type indx; | 
 |   size_t amt; | 
 |  | 
 |   if (h->root.type == bfd_link_hash_warning) | 
 |     { | 
 |       h = (struct aout_link_hash_entry *) h->root.u.i.link; | 
 |       if (h->root.type == bfd_link_hash_new) | 
 | 	return true; | 
 |     } | 
 |  | 
 |   output_bfd = flaginfo->output_bfd; | 
 |  | 
 |   if (aout_backend_info (output_bfd)->write_dynamic_symbol != NULL) | 
 |     { | 
 |       if (! ((*aout_backend_info (output_bfd)->write_dynamic_symbol) | 
 | 	     (output_bfd, flaginfo->info, h))) | 
 | 	{ | 
 | 	  /* FIXME: No way to handle errors.  */ | 
 | 	  abort (); | 
 | 	} | 
 |     } | 
 |  | 
 |   if (h->written) | 
 |     return true; | 
 |  | 
 |   h->written = true; | 
 |  | 
 |   /* An indx of -2 means the symbol must be written.  */ | 
 |   if (h->indx != -2 | 
 |       && (flaginfo->info->strip == strip_all | 
 | 	  || (flaginfo->info->strip == strip_some | 
 | 	      && bfd_hash_lookup (flaginfo->info->keep_hash, h->root.root.string, | 
 | 				  false, false) == NULL))) | 
 |     return true; | 
 |  | 
 |   switch (h->root.type) | 
 |     { | 
 |     default: | 
 |       abort (); | 
 |       /* Avoid variable not initialized warnings.  */ | 
 |       return true; | 
 |     case bfd_link_hash_new: | 
 |       /* This can happen for set symbols when sets are not being | 
 | 	 built.  */ | 
 |       return true; | 
 |     case bfd_link_hash_undefined: | 
 |       type = N_UNDF | N_EXT; | 
 |       val = 0; | 
 |       break; | 
 |     case bfd_link_hash_defined: | 
 |     case bfd_link_hash_defweak: | 
 |       { | 
 | 	asection *sec; | 
 |  | 
 | 	sec = h->root.u.def.section->output_section; | 
 | 	BFD_ASSERT (bfd_is_abs_section (sec) | 
 | 		    || sec->owner == output_bfd); | 
 | 	if (sec == obj_textsec (output_bfd)) | 
 | 	  type = h->root.type == bfd_link_hash_defined ? N_TEXT : N_WEAKT; | 
 | 	else if (sec == obj_datasec (output_bfd)) | 
 | 	  type = h->root.type == bfd_link_hash_defined ? N_DATA : N_WEAKD; | 
 | 	else if (sec == obj_bsssec (output_bfd)) | 
 | 	  type = h->root.type == bfd_link_hash_defined ? N_BSS : N_WEAKB; | 
 | 	else | 
 | 	  type = h->root.type == bfd_link_hash_defined ? N_ABS : N_WEAKA; | 
 | 	type |= N_EXT; | 
 | 	val = (h->root.u.def.value | 
 | 	       + sec->vma | 
 | 	       + h->root.u.def.section->output_offset); | 
 |       } | 
 |       break; | 
 |     case bfd_link_hash_common: | 
 |       type = N_UNDF | N_EXT; | 
 |       val = h->root.u.c.size; | 
 |       break; | 
 |     case bfd_link_hash_undefweak: | 
 |       type = N_WEAKU; | 
 |       val = 0; | 
 |       /* Fall through.  */ | 
 |     case bfd_link_hash_indirect: | 
 |     case bfd_link_hash_warning: | 
 |       /* FIXME: Ignore these for now.  The circumstances under which | 
 | 	 they should be written out are not clear to me.  */ | 
 |       return true; | 
 |     } | 
 |  | 
 |   H_PUT_8 (output_bfd, type, outsym.e_type); | 
 |   H_PUT_8 (output_bfd, 0, outsym.e_ovly); | 
 |   indx = add_to_stringtab (output_bfd, flaginfo->strtab, h->root.root.string, | 
 | 			   false); | 
 |   if (indx == (bfd_size_type) -1) | 
 |     /* FIXME: No way to handle errors.  */ | 
 |     abort (); | 
 |  | 
 |   PUT_WORD (output_bfd, 0, outsym.e_desc); | 
 |   PUT_WORD (output_bfd, indx, outsym.e_strx); | 
 |   PUT_WORD (output_bfd, val, outsym.e_value); | 
 |  | 
 |   amt = EXTERNAL_NLIST_SIZE; | 
 |   if (bfd_seek (output_bfd, flaginfo->symoff, SEEK_SET) != 0 | 
 |       || bfd_bwrite ((void *) &outsym, amt, output_bfd) != amt) | 
 |     /* FIXME: No way to handle errors.  */ | 
 |     abort (); | 
 |  | 
 |   flaginfo->symoff += amt; | 
 |   h->indx = obj_aout_external_sym_count (output_bfd); | 
 |   ++obj_aout_external_sym_count (output_bfd); | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Handle a link order which is supposed to generate a reloc.  */ | 
 |  | 
 | static bool | 
 | aout_link_reloc_link_order (struct aout_final_link_info *flaginfo, | 
 | 			    asection *o, | 
 | 			    struct bfd_link_order *p) | 
 | { | 
 |   struct bfd_link_order_reloc *pr; | 
 |   int r_index; | 
 |   int r_extern; | 
 |   reloc_howto_type *howto; | 
 |   file_ptr *reloff_ptr; | 
 |   struct reloc_std_external srel; | 
 |   void * rel_ptr; | 
 |   bfd_size_type rel_size; | 
 |  | 
 |   pr = p->u.reloc.p; | 
 |  | 
 |   if (p->type == bfd_section_reloc_link_order) | 
 |     { | 
 |       r_extern = 0; | 
 |       if (bfd_is_abs_section (pr->u.section)) | 
 | 	r_index = N_ABS | N_EXT; | 
 |       else | 
 | 	{ | 
 | 	  BFD_ASSERT (pr->u.section->owner == flaginfo->output_bfd); | 
 | 	  r_index = pr->u.section->target_index; | 
 | 	} | 
 |     } | 
 |   else | 
 |     { | 
 |       struct aout_link_hash_entry *h; | 
 |  | 
 |       BFD_ASSERT (p->type == bfd_symbol_reloc_link_order); | 
 |       r_extern = 1; | 
 |       h = ((struct aout_link_hash_entry *) | 
 | 	   bfd_wrapped_link_hash_lookup (flaginfo->output_bfd, flaginfo->info, | 
 | 					 pr->u.name, false, false, true)); | 
 |       if (h != NULL | 
 | 	  && h->indx >= 0) | 
 | 	r_index = h->indx; | 
 |       else if (h != NULL) | 
 | 	{ | 
 | 	  /* We decided to strip this symbol, but it turns out that we | 
 | 	     can't.  Note that we lose the other and desc information | 
 | 	     here.  I don't think that will ever matter for a global | 
 | 	     symbol.  */ | 
 | 	  h->indx = -2; | 
 | 	  h->written = false; | 
 | 	  if (!aout_link_write_other_symbol (&h->root.root, flaginfo)) | 
 | 	    return false; | 
 | 	  r_index = h->indx; | 
 | 	} | 
 |       else | 
 | 	{ | 
 | 	  (*flaginfo->info->callbacks->unattached_reloc) | 
 | 	    (flaginfo->info, pr->u.name, NULL, NULL, (bfd_vma) 0); | 
 | 	  r_index = 0; | 
 | 	} | 
 |     } | 
 |  | 
 |   howto = bfd_reloc_type_lookup (flaginfo->output_bfd, pr->reloc); | 
 |   if (howto == 0) | 
 |     { | 
 |       bfd_set_error (bfd_error_bad_value); | 
 |       return false; | 
 |     } | 
 |  | 
 |   if (o == obj_textsec (flaginfo->output_bfd)) | 
 |     reloff_ptr = &flaginfo->treloff; | 
 |   else if (o == obj_datasec (flaginfo->output_bfd)) | 
 |     reloff_ptr = &flaginfo->dreloff; | 
 |   else | 
 |     abort (); | 
 |  | 
 | #ifdef MY_put_reloc | 
 |   MY_put_reloc(flaginfo->output_bfd, r_extern, r_index, p->offset, howto, | 
 | 	       &srel); | 
 | #else | 
 |   { | 
 |     int r_pcrel; | 
 |     int r_baserel; | 
 |     int r_jmptable; | 
 |     int r_relative; | 
 |     int r_length; | 
 |  | 
 |     fprintf (stderr, "TODO: line %d in bfd/pdp11.c\n", __LINE__); | 
 |  | 
 |     r_pcrel = howto->pc_relative; | 
 |     r_baserel = (howto->type & 8) != 0; | 
 |     r_jmptable = (howto->type & 16) != 0; | 
 |     r_relative = (howto->type & 32) != 0; | 
 |     r_length = bfd_log2 (bfd_get_reloc_size (howto)); | 
 |  | 
 |     PUT_WORD (flaginfo->output_bfd, p->offset, srel.r_address); | 
 |     if (bfd_header_big_endian (flaginfo->output_bfd)) | 
 |       { | 
 | 	srel.r_index[0] = r_index >> 16; | 
 | 	srel.r_index[1] = r_index >> 8; | 
 | 	srel.r_index[2] = r_index; | 
 | 	srel.r_type[0] = | 
 | 	  ((r_extern ?     RELOC_STD_BITS_EXTERN_BIG : 0) | 
 | 	   | (r_pcrel ?    RELOC_STD_BITS_PCREL_BIG : 0) | 
 | 	   | (r_baserel ?  RELOC_STD_BITS_BASEREL_BIG : 0) | 
 | 	   | (r_jmptable ? RELOC_STD_BITS_JMPTABLE_BIG : 0) | 
 | 	   | (r_relative ? RELOC_STD_BITS_RELATIVE_BIG : 0) | 
 | 	   | (r_length <<  RELOC_STD_BITS_LENGTH_SH_BIG)); | 
 |       } | 
 |     else | 
 |       { | 
 | 	srel.r_index[2] = r_index >> 16; | 
 | 	srel.r_index[1] = r_index >> 8; | 
 | 	srel.r_index[0] = r_index; | 
 | 	srel.r_type[0] = | 
 | 	  ((r_extern ?     RELOC_STD_BITS_EXTERN_LITTLE : 0) | 
 | 	   | (r_pcrel ?    RELOC_STD_BITS_PCREL_LITTLE : 0) | 
 | 	   | (r_baserel ?  RELOC_STD_BITS_BASEREL_LITTLE : 0) | 
 | 	   | (r_jmptable ? RELOC_STD_BITS_JMPTABLE_LITTLE : 0) | 
 | 	   | (r_relative ? RELOC_STD_BITS_RELATIVE_LITTLE : 0) | 
 | 	   | (r_length <<  RELOC_STD_BITS_LENGTH_SH_LITTLE)); | 
 |       } | 
 |   } | 
 | #endif | 
 |   rel_ptr = (void *) &srel; | 
 |  | 
 |   /* We have to write the addend into the object file, since | 
 |      standard a.out relocs are in place.  It would be more | 
 |      reliable if we had the current contents of the file here, | 
 |      rather than assuming zeroes, but we can't read the file since | 
 |      it was opened using bfd_openw.  */ | 
 |   if (pr->addend != 0) | 
 |     { | 
 |       bfd_size_type size; | 
 |       bfd_reloc_status_type r; | 
 |       bfd_byte *buf; | 
 |       bool ok; | 
 |  | 
 |       size = bfd_get_reloc_size (howto); | 
 |       buf = bfd_zmalloc (size); | 
 |       if (buf == NULL && size != 0) | 
 | 	return false; | 
 |       r = MY_relocate_contents (howto, flaginfo->output_bfd, | 
 | 				pr->addend, buf); | 
 |       switch (r) | 
 | 	{ | 
 | 	case bfd_reloc_ok: | 
 | 	  break; | 
 | 	default: | 
 | 	case bfd_reloc_outofrange: | 
 | 	  abort (); | 
 | 	case bfd_reloc_overflow: | 
 | 	  (*flaginfo->info->callbacks->reloc_overflow) | 
 | 	    (flaginfo->info, NULL, | 
 | 	     (p->type == bfd_section_reloc_link_order | 
 | 	      ? bfd_section_name (pr->u.section) | 
 | 	      : pr->u.name), | 
 | 	     howto->name, pr->addend, NULL, | 
 | 	     (asection *) NULL, (bfd_vma) 0); | 
 | 	  break; | 
 | 	} | 
 |       ok = bfd_set_section_contents (flaginfo->output_bfd, o, | 
 | 				     (void *) buf, | 
 | 				     (file_ptr) p->offset, | 
 | 				     size); | 
 |       free (buf); | 
 |       if (! ok) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   rel_size = obj_reloc_entry_size (flaginfo->output_bfd); | 
 |   if (bfd_seek (flaginfo->output_bfd, *reloff_ptr, SEEK_SET) != 0 | 
 |       || bfd_bwrite (rel_ptr, rel_size, flaginfo->output_bfd) != rel_size) | 
 |     return false; | 
 |  | 
 |   *reloff_ptr += rel_size; | 
 |  | 
 |   /* Assert that the relocs have not run into the symbols, and that n | 
 |      the text relocs have not run into the data relocs.  */ | 
 |   BFD_ASSERT (*reloff_ptr <= obj_sym_filepos (flaginfo->output_bfd) | 
 | 	      && (reloff_ptr != &flaginfo->treloff | 
 | 		  || (*reloff_ptr | 
 | 		      <= obj_datasec (flaginfo->output_bfd)->rel_filepos))); | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Get the section corresponding to a reloc index.  */ | 
 |  | 
 | static inline asection * | 
 | aout_reloc_type_to_section (bfd *abfd, int type) | 
 | { | 
 |   switch (type) | 
 |     { | 
 |     case RTEXT:	return obj_textsec (abfd); | 
 |     case RDATA: return obj_datasec (abfd); | 
 |     case RBSS:  return obj_bsssec (abfd); | 
 |     case RABS:  return bfd_abs_section_ptr; | 
 |     case REXT:  return bfd_und_section_ptr; | 
 |     default:    abort (); | 
 |     } | 
 | } | 
 |  | 
 | static bool | 
 | pdp11_aout_link_input_section (struct aout_final_link_info *flaginfo, | 
 | 			       bfd *input_bfd, | 
 | 			       asection *input_section, | 
 | 			       bfd_byte *relocs, | 
 | 			       bfd_size_type rel_size, | 
 | 			       bfd_byte *contents) | 
 | { | 
 |   bool (*check_dynamic_reloc) | 
 |     (struct bfd_link_info *, bfd *, asection *, | 
 |      struct aout_link_hash_entry *, void *, bfd_byte *, bool *, bfd_vma *); | 
 |   bfd *output_bfd; | 
 |   bool relocatable; | 
 |   struct external_nlist *syms; | 
 |   char *strings; | 
 |   struct aout_link_hash_entry **sym_hashes; | 
 |   int *symbol_map; | 
 |   bfd_byte *rel; | 
 |   bfd_byte *rel_end; | 
 |  | 
 |   output_bfd = flaginfo->output_bfd; | 
 |   check_dynamic_reloc = aout_backend_info (output_bfd)->check_dynamic_reloc; | 
 |  | 
 |   BFD_ASSERT (obj_reloc_entry_size (input_bfd) == RELOC_SIZE); | 
 |   BFD_ASSERT (input_bfd->xvec->header_byteorder | 
 | 	      == output_bfd->xvec->header_byteorder); | 
 |  | 
 |   relocatable = bfd_link_relocatable (flaginfo->info); | 
 |   syms = obj_aout_external_syms (input_bfd); | 
 |   strings = obj_aout_external_strings (input_bfd); | 
 |   sym_hashes = obj_aout_sym_hashes (input_bfd); | 
 |   symbol_map = flaginfo->symbol_map; | 
 |  | 
 |   rel = relocs; | 
 |   rel_end = rel + rel_size; | 
 |   for (; rel < rel_end; rel += RELOC_SIZE) | 
 |     { | 
 |       bfd_vma r_addr; | 
 |       int r_index; | 
 |       int r_type; | 
 |       int r_pcrel; | 
 |       int r_extern; | 
 |       reloc_howto_type *howto; | 
 |       struct aout_link_hash_entry *h = NULL; | 
 |       bfd_vma relocation; | 
 |       bfd_reloc_status_type r; | 
 |       int reloc_entry; | 
 |  | 
 |       reloc_entry = GET_WORD (input_bfd, (void *) rel); | 
 |       if (reloc_entry == 0) | 
 | 	continue; | 
 |  | 
 |       { | 
 | 	unsigned int howto_idx; | 
 |  | 
 | 	r_index = (reloc_entry & RIDXMASK) >> 4; | 
 | 	r_type = reloc_entry & RTYPE; | 
 | 	r_pcrel = reloc_entry & RELFLG; | 
 | 	r_addr = (char *) rel - (char *) relocs; | 
 |  | 
 | 	r_extern = (r_type == REXT); | 
 |  | 
 | 	howto_idx = r_pcrel; | 
 | 	if (howto_idx < TABLE_SIZE (howto_table_pdp11)) | 
 | 	  howto = howto_table_pdp11 + howto_idx; | 
 | 	else | 
 | 	  { | 
 | 	    _bfd_error_handler (_("%pB: unsupported relocation type"), | 
 | 				input_bfd); | 
 | 	    bfd_set_error (bfd_error_bad_value); | 
 | 	    return false; | 
 | 	  } | 
 |       } | 
 |  | 
 |       if (relocatable) | 
 | 	{ | 
 | 	  /* We are generating a relocatable output file, and must | 
 | 	     modify the reloc accordingly.  */ | 
 | 	  if (r_extern) | 
 | 	    { | 
 | 	      /* If we know the symbol this relocation is against, | 
 | 		 convert it into a relocation against a section.  This | 
 | 		 is what the native linker does.  */ | 
 | 	      h = sym_hashes[r_index]; | 
 | 	      if (h != NULL | 
 | 		  && (h->root.type == bfd_link_hash_defined | 
 | 		      || h->root.type == bfd_link_hash_defweak)) | 
 | 		{ | 
 | 		  asection *output_section; | 
 |  | 
 | 		  /* Compute a new r_index.  */ | 
 | 		  output_section = h->root.u.def.section->output_section; | 
 | 		  if (output_section == obj_textsec (output_bfd)) | 
 | 		    r_type = N_TEXT; | 
 | 		  else if (output_section == obj_datasec (output_bfd)) | 
 | 		    r_type = N_DATA; | 
 | 		  else if (output_section == obj_bsssec (output_bfd)) | 
 | 		    r_type = N_BSS; | 
 | 		  else | 
 | 		    r_type = N_ABS; | 
 |  | 
 | 		  /* Add the symbol value and the section VMA to the | 
 | 		     addend stored in the contents.  */ | 
 | 		  relocation = (h->root.u.def.value | 
 | 				+ output_section->vma | 
 | 				+ h->root.u.def.section->output_offset); | 
 | 		} | 
 | 	      else | 
 | 		{ | 
 | 		  /* We must change r_index according to the symbol | 
 | 		     map.  */ | 
 | 		  r_index = symbol_map[r_index]; | 
 |  | 
 | 		  if (r_index == -1) | 
 | 		    { | 
 | 		      if (h != NULL) | 
 | 			{ | 
 | 			  /* We decided to strip this symbol, but it | 
 | 			     turns out that we can't.  Note that we | 
 | 			     lose the other and desc information here. | 
 | 			     I don't think that will ever matter for a | 
 | 			     global symbol.  */ | 
 | 			  if (h->indx < 0) | 
 | 			    { | 
 | 			      h->indx = -2; | 
 | 			      h->written = false; | 
 | 			      if (!aout_link_write_other_symbol (&h->root.root, | 
 | 								 flaginfo)) | 
 | 				return false; | 
 | 			    } | 
 | 			  r_index = h->indx; | 
 | 			} | 
 | 		      else | 
 | 			{ | 
 | 			  const char *name; | 
 |  | 
 | 			  name = strings + GET_WORD (input_bfd, | 
 | 						     syms[r_index].e_strx); | 
 | 			  (*flaginfo->info->callbacks->unattached_reloc) | 
 | 			    (flaginfo->info, name, input_bfd, input_section, | 
 | 			     r_addr); | 
 | 			  r_index = 0; | 
 | 			} | 
 | 		    } | 
 |  | 
 | 		  relocation = 0; | 
 | 		} | 
 |  | 
 | 	      /* Write out the new r_index value.  */ | 
 | 	      reloc_entry = GET_WORD (input_bfd, rel); | 
 | 	      reloc_entry &= RIDXMASK; | 
 | 	      reloc_entry |= r_index << 4; | 
 | 	      PUT_WORD (input_bfd, reloc_entry, rel); | 
 | 	    } | 
 | 	  else | 
 | 	    { | 
 | 	      asection *section; | 
 |  | 
 | 	      /* This is a relocation against a section.  We must | 
 | 		 adjust by the amount that the section moved.  */ | 
 | 	      section = aout_reloc_type_to_section (input_bfd, r_type); | 
 | 	      relocation = (section->output_section->vma | 
 | 			    + section->output_offset | 
 | 			    - section->vma); | 
 | 	    } | 
 |  | 
 | 	  /* Change the address of the relocation.  */ | 
 | 	  fprintf (stderr, "TODO: change the address of the relocation\n"); | 
 |  | 
 | 	  /* Adjust a PC relative relocation by removing the reference | 
 | 	     to the original address in the section and including the | 
 | 	     reference to the new address.  */ | 
 | 	  if (r_pcrel) | 
 | 	    relocation -= (input_section->output_section->vma | 
 | 			   + input_section->output_offset | 
 | 			   - input_section->vma); | 
 |  | 
 | #ifdef MY_relocatable_reloc | 
 | 	  MY_relocatable_reloc (howto, output_bfd, rel, relocation, r_addr); | 
 | #endif | 
 |  | 
 | 	  if (relocation == 0) | 
 | 	    r = bfd_reloc_ok; | 
 | 	  else | 
 | 	    r = MY_relocate_contents (howto, | 
 | 				      input_bfd, relocation, | 
 | 				      contents + r_addr); | 
 | 	} | 
 |       else | 
 | 	{ | 
 | 	  bool hundef; | 
 |  | 
 | 	  /* We are generating an executable, and must do a full | 
 | 	     relocation.  */ | 
 | 	  hundef = false; | 
 | 	  if (r_extern) | 
 | 	    { | 
 | 	      h = sym_hashes[r_index]; | 
 |  | 
 | 	      if (h != NULL | 
 | 		  && (h->root.type == bfd_link_hash_defined | 
 | 		      || h->root.type == bfd_link_hash_defweak)) | 
 | 		{ | 
 | 		  relocation = (h->root.u.def.value | 
 | 				+ h->root.u.def.section->output_section->vma | 
 | 				+ h->root.u.def.section->output_offset); | 
 | 		} | 
 | 	      else if (h != NULL | 
 | 		       && h->root.type == bfd_link_hash_undefweak) | 
 | 		relocation = 0; | 
 | 	      else | 
 | 		{ | 
 | 		  hundef = true; | 
 | 		  relocation = 0; | 
 | 		} | 
 | 	    } | 
 | 	  else | 
 | 	    { | 
 | 	      asection *section; | 
 |  | 
 | 	      section = aout_reloc_type_to_section (input_bfd, r_type); | 
 | 	      relocation = (section->output_section->vma | 
 | 			    + section->output_offset | 
 | 			    - section->vma); | 
 | 	      if (r_pcrel) | 
 | 		relocation += input_section->vma; | 
 | 	    } | 
 |  | 
 | 	  if (check_dynamic_reloc != NULL) | 
 | 	    { | 
 | 	      bool skip; | 
 |  | 
 | 	      if (! ((*check_dynamic_reloc) | 
 | 		     (flaginfo->info, input_bfd, input_section, h, | 
 | 		      (void *) rel, contents, &skip, &relocation))) | 
 | 		return false; | 
 | 	      if (skip) | 
 | 		continue; | 
 | 	    } | 
 |  | 
 | 	  /* Now warn if a global symbol is undefined.  We could not | 
 | 	     do this earlier, because check_dynamic_reloc might want | 
 | 	     to skip this reloc.  */ | 
 | 	  if (hundef && ! bfd_link_pic (flaginfo->info)) | 
 | 	    { | 
 | 	      const char *name; | 
 |  | 
 | 	      if (h != NULL) | 
 | 		name = h->root.root.string; | 
 | 	      else | 
 | 		name = strings + GET_WORD (input_bfd, syms[r_index].e_strx); | 
 | 	      (*flaginfo->info->callbacks->undefined_symbol) | 
 | 		(flaginfo->info, name, input_bfd, input_section, | 
 | 		 r_addr, true); | 
 | 	    } | 
 |  | 
 | 	  r = MY_final_link_relocate (howto, | 
 | 				      input_bfd, input_section, | 
 | 				      contents, r_addr, relocation, | 
 | 				      (bfd_vma) 0); | 
 | 	} | 
 |  | 
 |       if (r != bfd_reloc_ok) | 
 | 	{ | 
 | 	  switch (r) | 
 | 	    { | 
 | 	    default: | 
 | 	    case bfd_reloc_outofrange: | 
 | 	      abort (); | 
 | 	    case bfd_reloc_overflow: | 
 | 	      { | 
 | 		const char *name; | 
 |  | 
 | 		if (h != NULL) | 
 | 		  name = NULL; | 
 | 		else if (r_extern) | 
 | 		  name = strings + GET_WORD (input_bfd, | 
 | 					     syms[r_index].e_strx); | 
 | 		else | 
 | 		  { | 
 | 		    asection *s; | 
 |  | 
 | 		    s = aout_reloc_type_to_section (input_bfd, r_type); | 
 | 		    name = bfd_section_name (s); | 
 | 		  } | 
 | 		(*flaginfo->info->callbacks->reloc_overflow) | 
 | 		  (flaginfo->info, (h ? &h->root : NULL), name, howto->name, | 
 | 		   (bfd_vma) 0, input_bfd, input_section, r_addr); | 
 | 	      } | 
 | 	      break; | 
 | 	    } | 
 | 	} | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Link an a.out section into the output file.  */ | 
 |  | 
 | static bool | 
 | aout_link_input_section (struct aout_final_link_info *flaginfo, | 
 | 			 bfd *input_bfd, | 
 | 			 asection *input_section, | 
 | 			 file_ptr *reloff_ptr, | 
 | 			 bfd_size_type rel_size) | 
 | { | 
 |   bfd_size_type input_size; | 
 |   void * relocs; | 
 |  | 
 |   /* Get the section contents.  */ | 
 |   input_size = input_section->size; | 
 |   if (! bfd_get_section_contents (input_bfd, input_section, | 
 | 				  (void *) flaginfo->contents, | 
 | 				  (file_ptr) 0, input_size)) | 
 |     return false; | 
 |  | 
 |   /* Read in the relocs if we haven't already done it.  */ | 
 |   if (aout_section_data (input_section) != NULL | 
 |       && aout_section_data (input_section)->relocs != NULL) | 
 |     relocs = aout_section_data (input_section)->relocs; | 
 |   else | 
 |     { | 
 |       relocs = flaginfo->relocs; | 
 |       if (rel_size > 0) | 
 | 	{ | 
 | 	  if (bfd_seek (input_bfd, input_section->rel_filepos, SEEK_SET) != 0 | 
 | 	      || bfd_bread (relocs, rel_size, input_bfd) != rel_size) | 
 | 	    return false; | 
 | 	} | 
 |     } | 
 |  | 
 |   /* Relocate the section contents.  */ | 
 |   if (! pdp11_aout_link_input_section (flaginfo, input_bfd, input_section, | 
 | 				       (bfd_byte *) relocs, | 
 | 				       rel_size, flaginfo->contents)) | 
 |     return false; | 
 |  | 
 |   /* Write out the section contents.  */ | 
 |   if (! bfd_set_section_contents (flaginfo->output_bfd, | 
 | 				  input_section->output_section, | 
 | 				  (void *) flaginfo->contents, | 
 | 				  (file_ptr) input_section->output_offset, | 
 | 				  input_size)) | 
 |     return false; | 
 |  | 
 |   /* If we are producing relocatable output, the relocs were | 
 |      modified, and we now write them out.  */ | 
 |   if (bfd_link_relocatable (flaginfo->info) && rel_size > 0) | 
 |     { | 
 |       if (bfd_seek (flaginfo->output_bfd, *reloff_ptr, SEEK_SET) != 0) | 
 | 	return false; | 
 |       if (bfd_bwrite (relocs, rel_size, flaginfo->output_bfd) != rel_size) | 
 | 	return false; | 
 |       *reloff_ptr += rel_size; | 
 |  | 
 |       /* Assert that the relocs have not run into the symbols, and | 
 | 	 that if these are the text relocs they have not run into the | 
 | 	 data relocs.  */ | 
 |       BFD_ASSERT (*reloff_ptr <= obj_sym_filepos (flaginfo->output_bfd) | 
 | 		  && (reloff_ptr != &flaginfo->treloff | 
 | 		      || (*reloff_ptr | 
 | 			  <= obj_datasec (flaginfo->output_bfd)->rel_filepos))); | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Link an a.out input BFD into the output file.  */ | 
 |  | 
 | static bool | 
 | aout_link_input_bfd (struct aout_final_link_info *flaginfo, bfd *input_bfd) | 
 | { | 
 |   BFD_ASSERT (bfd_get_format (input_bfd) == bfd_object); | 
 |  | 
 |   /* If this is a dynamic object, it may need special handling.  */ | 
 |   if ((input_bfd->flags & DYNAMIC) != 0 | 
 |       && aout_backend_info (input_bfd)->link_dynamic_object != NULL) | 
 |     return ((*aout_backend_info (input_bfd)->link_dynamic_object) | 
 | 	    (flaginfo->info, input_bfd)); | 
 |  | 
 |   /* Get the symbols.  We probably have them already, unless | 
 |      flaginfo->info->keep_memory is FALSE.  */ | 
 |   if (! aout_get_external_symbols (input_bfd)) | 
 |     return false; | 
 |  | 
 |   /* Write out the symbols and get a map of the new indices.  The map | 
 |      is placed into flaginfo->symbol_map.  */ | 
 |   if (! aout_link_write_symbols (flaginfo, input_bfd)) | 
 |     return false; | 
 |  | 
 |   /* Relocate and write out the sections.  These functions use the | 
 |      symbol map created by aout_link_write_symbols.  The linker_mark | 
 |      field will be set if these sections are to be included in the | 
 |      link, which will normally be the case.  */ | 
 |   if (obj_textsec (input_bfd)->linker_mark) | 
 |     { | 
 |       if (! aout_link_input_section (flaginfo, input_bfd, | 
 | 				     obj_textsec (input_bfd), | 
 | 				     &flaginfo->treloff, | 
 | 				     exec_hdr (input_bfd)->a_trsize)) | 
 | 	return false; | 
 |     } | 
 |   if (obj_datasec (input_bfd)->linker_mark) | 
 |     { | 
 |       if (! aout_link_input_section (flaginfo, input_bfd, | 
 | 				     obj_datasec (input_bfd), | 
 | 				     &flaginfo->dreloff, | 
 | 				     exec_hdr (input_bfd)->a_drsize)) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   /* If we are not keeping memory, we don't need the symbols any | 
 |      longer.  We still need them if we are keeping memory, because the | 
 |      strings in the hash table point into them.  */ | 
 |   if (! flaginfo->info->keep_memory) | 
 |     { | 
 |       if (! aout_link_free_symbols (input_bfd)) | 
 | 	return false; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Do the final link step.  This is called on the output BFD.  The | 
 |    INFO structure should point to a list of BFDs linked through the | 
 |    link.next field which can be used to find each BFD which takes part | 
 |    in the output.  Also, each section in ABFD should point to a list | 
 |    of bfd_link_order structures which list all the input sections for | 
 |    the output section.  */ | 
 |  | 
 | bool | 
 | NAME (aout, final_link) (bfd *abfd, | 
 | 			 struct bfd_link_info *info, | 
 | 			 void (*callback) (bfd *, file_ptr *, file_ptr *, file_ptr *)) | 
 | { | 
 |   struct aout_final_link_info aout_info; | 
 |   bool includes_hash_initialized = false; | 
 |   bfd *sub; | 
 |   bfd_size_type trsize, drsize; | 
 |   bfd_size_type max_contents_size; | 
 |   bfd_size_type max_relocs_size; | 
 |   bfd_size_type max_sym_count; | 
 |   struct bfd_link_order *p; | 
 |   asection *o; | 
 |   bool have_link_order_relocs; | 
 |  | 
 |   if (bfd_link_pic (info)) | 
 |     abfd->flags |= DYNAMIC; | 
 |  | 
 |   separate_i_d = info->separate_code; | 
 |   aout_info.info = info; | 
 |   aout_info.output_bfd = abfd; | 
 |   aout_info.contents = NULL; | 
 |   aout_info.relocs = NULL; | 
 |   aout_info.symbol_map = NULL; | 
 |   aout_info.output_syms = NULL; | 
 |  | 
 |   if (!bfd_hash_table_init_n (&aout_info.includes.root, | 
 | 			      aout_link_includes_newfunc, | 
 | 			      sizeof (struct aout_link_includes_entry), | 
 | 			      251)) | 
 |     goto error_return; | 
 |   includes_hash_initialized = true; | 
 |  | 
 |   /* Figure out the largest section size.  Also, if generating | 
 |      relocatable output, count the relocs.  */ | 
 |   trsize = 0; | 
 |   drsize = 0; | 
 |   max_contents_size = 0; | 
 |   max_relocs_size = 0; | 
 |   max_sym_count = 0; | 
 |   for (sub = info->input_bfds; sub != NULL; sub = sub->link.next) | 
 |     { | 
 |       size_t sz; | 
 |  | 
 |       if (bfd_link_relocatable (info)) | 
 | 	{ | 
 | 	  if (bfd_get_flavour (sub) == bfd_target_aout_flavour) | 
 | 	    { | 
 | 	      trsize += exec_hdr (sub)->a_trsize; | 
 | 	      drsize += exec_hdr (sub)->a_drsize; | 
 | 	    } | 
 | 	  else | 
 | 	    { | 
 | 	      /* FIXME: We need to identify the .text and .data sections | 
 | 		 and call get_reloc_upper_bound and canonicalize_reloc to | 
 | 		 work out the number of relocs needed, and then multiply | 
 | 		 by the reloc size.  */ | 
 | 	      _bfd_error_handler | 
 | 		/* xgettext:c-format */ | 
 | 		(_("%pB: relocatable link from %s to %s not supported"), | 
 | 		 abfd, sub->xvec->name, abfd->xvec->name); | 
 | 	      bfd_set_error (bfd_error_invalid_operation); | 
 | 	      goto error_return; | 
 | 	    } | 
 | 	} | 
 |  | 
 |       if (bfd_get_flavour (sub) == bfd_target_aout_flavour) | 
 | 	{ | 
 | 	  sz = obj_textsec (sub)->size; | 
 | 	  if (sz > max_contents_size) | 
 | 	    max_contents_size = sz; | 
 | 	  sz = obj_datasec (sub)->size; | 
 | 	  if (sz > max_contents_size) | 
 | 	    max_contents_size = sz; | 
 |  | 
 | 	  sz = exec_hdr (sub)->a_trsize; | 
 | 	  if (sz > max_relocs_size) | 
 | 	    max_relocs_size = sz; | 
 | 	  sz = exec_hdr (sub)->a_drsize; | 
 | 	  if (sz > max_relocs_size) | 
 | 	    max_relocs_size = sz; | 
 |  | 
 | 	  sz = obj_aout_external_sym_count (sub); | 
 | 	  if (sz > max_sym_count) | 
 | 	    max_sym_count = sz; | 
 | 	} | 
 |     } | 
 |  | 
 |   if (bfd_link_relocatable (info)) | 
 |     { | 
 |       if (obj_textsec (abfd) != NULL) | 
 | 	trsize += (_bfd_count_link_order_relocs (obj_textsec (abfd) | 
 | 						 ->map_head.link_order) | 
 | 		   * obj_reloc_entry_size (abfd)); | 
 |       if (obj_datasec (abfd) != NULL) | 
 | 	drsize += (_bfd_count_link_order_relocs (obj_datasec (abfd) | 
 | 						 ->map_head.link_order) | 
 | 		   * obj_reloc_entry_size (abfd)); | 
 |     } | 
 |  | 
 |   exec_hdr (abfd)->a_trsize = trsize; | 
 |   exec_hdr (abfd)->a_drsize = drsize; | 
 |   exec_hdr (abfd)->a_entry = bfd_get_start_address (abfd); | 
 |  | 
 |   /* Adjust the section sizes and vmas according to the magic number. | 
 |      This sets a_text, a_data and a_bss in the exec_hdr and sets the | 
 |      filepos for each section.  */ | 
 |   if (! NAME (aout, adjust_sizes_and_vmas) (abfd)) | 
 |     goto error_return; | 
 |  | 
 |   /* The relocation and symbol file positions differ among a.out | 
 |      targets.  We are passed a callback routine from the backend | 
 |      specific code to handle this. | 
 |      FIXME: At this point we do not know how much space the symbol | 
 |      table will require.  This will not work for any (nonstandard) | 
 |      a.out target that needs to know the symbol table size before it | 
 |      can compute the relocation file positions.  */ | 
 |   (*callback) (abfd, &aout_info.treloff, &aout_info.dreloff, | 
 | 	       &aout_info.symoff); | 
 |   obj_textsec (abfd)->rel_filepos = aout_info.treloff; | 
 |   obj_datasec (abfd)->rel_filepos = aout_info.dreloff; | 
 |   obj_sym_filepos (abfd) = aout_info.symoff; | 
 |  | 
 |   /* We keep a count of the symbols as we output them.  */ | 
 |   obj_aout_external_sym_count (abfd) = 0; | 
 |  | 
 |   /* We accumulate the string table as we write out the symbols.  */ | 
 |   aout_info.strtab = _bfd_stringtab_init (); | 
 |   if (aout_info.strtab == NULL) | 
 |     goto error_return; | 
 |  | 
 |   /* Allocate buffers to hold section contents and relocs.  */ | 
 |   aout_info.contents = bfd_malloc (max_contents_size); | 
 |   aout_info.relocs = bfd_malloc (max_relocs_size); | 
 |   aout_info.symbol_map = bfd_malloc (max_sym_count * sizeof (int *)); | 
 |   aout_info.output_syms = bfd_malloc ((max_sym_count + 1) | 
 | 				      * sizeof (struct external_nlist)); | 
 |   if ((aout_info.contents == NULL && max_contents_size != 0) | 
 |       || (aout_info.relocs == NULL && max_relocs_size != 0) | 
 |       || (aout_info.symbol_map == NULL && max_sym_count != 0) | 
 |       || aout_info.output_syms == NULL) | 
 |     goto error_return; | 
 |  | 
 |   /* If we have a symbol named __DYNAMIC, force it out now.  This is | 
 |      required by SunOS.  Doing this here rather than in sunos.c is a | 
 |      hack, but it's easier than exporting everything which would be | 
 |      needed.  */ | 
 |   { | 
 |     struct aout_link_hash_entry *h; | 
 |  | 
 |     h = aout_link_hash_lookup (aout_hash_table (info), "__DYNAMIC", | 
 | 			       false, false, false); | 
 |     if (h != NULL) | 
 |       aout_link_write_other_symbol (&h->root.root, &aout_info); | 
 |   } | 
 |  | 
 |   /* The most time efficient way to do the link would be to read all | 
 |      the input object files into memory and then sort out the | 
 |      information into the output file.  Unfortunately, that will | 
 |      probably use too much memory.  Another method would be to step | 
 |      through everything that composes the text section and write it | 
 |      out, and then everything that composes the data section and write | 
 |      it out, and then write out the relocs, and then write out the | 
 |      symbols.  Unfortunately, that requires reading stuff from each | 
 |      input file several times, and we will not be able to keep all the | 
 |      input files open simultaneously, and reopening them will be slow. | 
 |  | 
 |      What we do is basically process one input file at a time.  We do | 
 |      everything we need to do with an input file once--copy over the | 
 |      section contents, handle the relocation information, and write | 
 |      out the symbols--and then we throw away the information we read | 
 |      from it.  This approach requires a lot of lseeks of the output | 
 |      file, which is unfortunate but still faster than reopening a lot | 
 |      of files. | 
 |  | 
 |      We use the output_has_begun field of the input BFDs to see | 
 |      whether we have already handled it.  */ | 
 |   for (sub = info->input_bfds; sub != NULL; sub = sub->link.next) | 
 |     sub->output_has_begun = false; | 
 |  | 
 |   /* Mark all sections which are to be included in the link.  This | 
 |      will normally be every section.  We need to do this so that we | 
 |      can identify any sections which the linker has decided to not | 
 |      include.  */ | 
 |   for (o = abfd->sections; o != NULL; o = o->next) | 
 |     { | 
 |       for (p = o->map_head.link_order; p != NULL; p = p->next) | 
 | 	if (p->type == bfd_indirect_link_order) | 
 | 	  p->u.indirect.section->linker_mark = true; | 
 |     } | 
 |  | 
 |   have_link_order_relocs = false; | 
 |   for (o = abfd->sections; o != NULL; o = o->next) | 
 |     { | 
 |       for (p = o->map_head.link_order; | 
 | 	   p != NULL; | 
 | 	   p = p->next) | 
 | 	{ | 
 | 	  if (p->type == bfd_indirect_link_order | 
 | 	      && (bfd_get_flavour (p->u.indirect.section->owner) | 
 | 		  == bfd_target_aout_flavour)) | 
 | 	    { | 
 | 	      bfd *input_bfd; | 
 |  | 
 | 	      input_bfd = p->u.indirect.section->owner; | 
 | 	      if (! input_bfd->output_has_begun) | 
 | 		{ | 
 | 		  if (! aout_link_input_bfd (&aout_info, input_bfd)) | 
 | 		    goto error_return; | 
 | 		  input_bfd->output_has_begun = true; | 
 | 		} | 
 | 	    } | 
 | 	  else if (p->type == bfd_section_reloc_link_order | 
 | 		   || p->type == bfd_symbol_reloc_link_order) | 
 | 	    /* These are handled below.  */ | 
 | 	    have_link_order_relocs = true; | 
 | 	  else | 
 | 	    { | 
 | 	      if (! _bfd_default_link_order (abfd, info, o, p)) | 
 | 		goto error_return; | 
 | 	    } | 
 | 	} | 
 |     } | 
 |  | 
 |   /* Write out any symbols that we have not already written out.  */ | 
 |   bfd_hash_traverse (&info->hash->table, | 
 | 		     aout_link_write_other_symbol, | 
 | 		     &aout_info); | 
 |  | 
 |   /* Now handle any relocs we were asked to create by the linker. | 
 |      These did not come from any input file.  We must do these after | 
 |      we have written out all the symbols, so that we know the symbol | 
 |      indices to use.  */ | 
 |   if (have_link_order_relocs) | 
 |     { | 
 |       for (o = abfd->sections; o != NULL; o = o->next) | 
 | 	{ | 
 | 	  for (p = o->map_head.link_order; | 
 | 	       p != NULL; | 
 | 	       p = p->next) | 
 | 	    { | 
 | 	      if (p->type == bfd_section_reloc_link_order | 
 | 		  || p->type == bfd_symbol_reloc_link_order) | 
 | 		{ | 
 | 		  if (! aout_link_reloc_link_order (&aout_info, o, p)) | 
 | 		    goto error_return; | 
 | 		} | 
 | 	    } | 
 | 	} | 
 |     } | 
 |  | 
 |   free (aout_info.contents); | 
 |   aout_info.contents = NULL; | 
 |   free (aout_info.relocs); | 
 |   aout_info.relocs = NULL; | 
 |   free (aout_info.symbol_map); | 
 |   aout_info.symbol_map = NULL; | 
 |   free (aout_info.output_syms); | 
 |   aout_info.output_syms = NULL; | 
 |   if (includes_hash_initialized) | 
 |     { | 
 |       bfd_hash_table_free (&aout_info.includes.root); | 
 |       includes_hash_initialized = false; | 
 |     } | 
 |  | 
 |   /* Finish up any dynamic linking we may be doing.  */ | 
 |   if (aout_backend_info (abfd)->finish_dynamic_link != NULL) | 
 |     { | 
 |       if (! (*aout_backend_info (abfd)->finish_dynamic_link) (abfd, info)) | 
 | 	goto error_return; | 
 |     } | 
 |  | 
 |   /* Update the header information.  */ | 
 |   abfd->symcount = obj_aout_external_sym_count (abfd); | 
 |   exec_hdr (abfd)->a_syms = abfd->symcount * EXTERNAL_NLIST_SIZE; | 
 |   obj_str_filepos (abfd) = obj_sym_filepos (abfd) + exec_hdr (abfd)->a_syms; | 
 |   obj_textsec (abfd)->reloc_count = | 
 |     exec_hdr (abfd)->a_trsize / obj_reloc_entry_size (abfd); | 
 |   obj_datasec (abfd)->reloc_count = | 
 |     exec_hdr (abfd)->a_drsize / obj_reloc_entry_size (abfd); | 
 |  | 
 |   /* Write out the string table, unless there are no symbols.  */ | 
 |   if (abfd->symcount > 0) | 
 |     { | 
 |       if (bfd_seek (abfd, obj_str_filepos (abfd), SEEK_SET) != 0 | 
 | 	  || ! emit_stringtab (abfd, aout_info.strtab)) | 
 | 	goto error_return; | 
 |     } | 
 |   else if (obj_textsec (abfd)->reloc_count == 0 | 
 | 	   && obj_datasec (abfd)->reloc_count == 0) | 
 |     { | 
 |       /* The layout of a typical a.out file is header, text, data, | 
 | 	 relocs, symbols, string table.  When there are no relocs, | 
 | 	 symbols or string table, the last thing in the file is data | 
 | 	 and a_data may be rounded up.  However we may have a smaller | 
 | 	 sized .data section and thus not written final padding.  The | 
 | 	 same thing can happen with text if there is no data.  Write | 
 | 	 final padding here to extend the file.  */ | 
 |       file_ptr pos = 0; | 
 |  | 
 |       if (exec_hdr (abfd)->a_data > obj_datasec (abfd)->size) | 
 | 	pos = obj_datasec (abfd)->filepos + exec_hdr (abfd)->a_data; | 
 |       else if (obj_datasec (abfd)->size == 0 | 
 | 	       && exec_hdr (abfd)->a_text > obj_textsec (abfd)->size) | 
 | 	pos = obj_textsec (abfd)->filepos + exec_hdr (abfd)->a_text; | 
 |       if (pos != 0) | 
 | 	{ | 
 | 	  bfd_byte b = 0; | 
 |  | 
 | 	  if (bfd_seek (abfd, pos - 1, SEEK_SET) != 0 | 
 | 	      || bfd_bwrite (&b, (bfd_size_type) 1, abfd) != 1) | 
 | 	    goto error_return; | 
 | 	} | 
 |     } | 
 |  | 
 |   return true; | 
 |  | 
 |  error_return: | 
 |   free (aout_info.contents); | 
 |   free (aout_info.relocs); | 
 |   free (aout_info.symbol_map); | 
 |   free (aout_info.output_syms); | 
 |   if (includes_hash_initialized) | 
 |     bfd_hash_table_free (&aout_info.includes.root); | 
 |   return false; | 
 | } | 
 |  | 
 | /* Adjust and write out the symbols for an a.out file.  Set the new | 
 |    symbol indices into a symbol_map.  */ | 
 |  | 
 | static bool | 
 | aout_link_write_symbols (struct aout_final_link_info *flaginfo, bfd *input_bfd) | 
 | { | 
 |   bfd *output_bfd; | 
 |   bfd_size_type sym_count; | 
 |   char *strings; | 
 |   enum bfd_link_strip strip; | 
 |   enum bfd_link_discard discard; | 
 |   struct external_nlist *outsym; | 
 |   bfd_size_type strtab_index; | 
 |   struct external_nlist *sym; | 
 |   struct external_nlist *sym_end; | 
 |   struct aout_link_hash_entry **sym_hash; | 
 |   int *symbol_map; | 
 |   bool pass; | 
 |   bool skip_next; | 
 |  | 
 |   output_bfd = flaginfo->output_bfd; | 
 |   sym_count = obj_aout_external_sym_count (input_bfd); | 
 |   strings = obj_aout_external_strings (input_bfd); | 
 |   strip = flaginfo->info->strip; | 
 |   discard = flaginfo->info->discard; | 
 |   outsym = flaginfo->output_syms; | 
 |  | 
 |   /* First write out a symbol for this object file, unless we are | 
 |      discarding such symbols.  */ | 
 |   if (strip != strip_all | 
 |       && (strip != strip_some | 
 | 	  || bfd_hash_lookup (flaginfo->info->keep_hash, | 
 | 			      bfd_get_filename (input_bfd), | 
 | 			      false, false) != NULL) | 
 |       && discard != discard_all) | 
 |     { | 
 |       H_PUT_8 (output_bfd, N_TEXT, outsym->e_type); | 
 |       H_PUT_8 (output_bfd, 0, outsym->e_ovly); | 
 |       H_PUT_16 (output_bfd, 0, outsym->e_desc); | 
 |       strtab_index = add_to_stringtab (output_bfd, flaginfo->strtab, | 
 | 				       bfd_get_filename (input_bfd), false); | 
 |       if (strtab_index == (bfd_size_type) -1) | 
 | 	return false; | 
 |       PUT_WORD (output_bfd, strtab_index, outsym->e_strx); | 
 |       PUT_WORD (output_bfd, | 
 | 		(bfd_section_vma (obj_textsec (input_bfd)->output_section) | 
 | 		 + obj_textsec (input_bfd)->output_offset), | 
 | 		outsym->e_value); | 
 |       ++obj_aout_external_sym_count (output_bfd); | 
 |       ++outsym; | 
 |     } | 
 |  | 
 |   pass = false; | 
 |   skip_next = false; | 
 |   sym = obj_aout_external_syms (input_bfd); | 
 |   sym_end = sym + sym_count; | 
 |   sym_hash = obj_aout_sym_hashes (input_bfd); | 
 |   symbol_map = flaginfo->symbol_map; | 
 |   memset (symbol_map, 0, (size_t) sym_count * sizeof *symbol_map); | 
 |   for (; sym < sym_end; sym++, sym_hash++, symbol_map++) | 
 |     { | 
 |       const char *name; | 
 |       int type; | 
 |       struct aout_link_hash_entry *h; | 
 |       bool skip; | 
 |       asection *symsec; | 
 |       bfd_vma val = 0; | 
 |       bool copy; | 
 |  | 
 |       /* We set *symbol_map to 0 above for all symbols.  If it has | 
 | 	 already been set to -1 for this symbol, it means that we are | 
 | 	 discarding it because it appears in a duplicate header file. | 
 | 	 See the N_BINCL code below.  */ | 
 |       if (*symbol_map == -1) | 
 | 	continue; | 
 |  | 
 |       /* Initialize *symbol_map to -1, which means that the symbol was | 
 | 	 not copied into the output file.  We will change it later if | 
 | 	 we do copy the symbol over.  */ | 
 |       *symbol_map = -1; | 
 |  | 
 |       type = H_GET_8 (input_bfd, sym->e_type); | 
 |       name = strings + GET_WORD (input_bfd, sym->e_strx); | 
 |  | 
 |       h = NULL; | 
 |  | 
 |       if (pass) | 
 | 	{ | 
 | 	  /* Pass this symbol through.  It is the target of an | 
 | 	     indirect or warning symbol.  */ | 
 | 	  val = GET_WORD (input_bfd, sym->e_value); | 
 | 	  pass = false; | 
 | 	} | 
 |       else if (skip_next) | 
 | 	{ | 
 | 	  /* Skip this symbol, which is the target of an indirect | 
 | 	     symbol that we have changed to no longer be an indirect | 
 | 	     symbol.  */ | 
 | 	  skip_next = false; | 
 | 	  continue; | 
 | 	} | 
 |       else | 
 | 	{ | 
 | 	  struct aout_link_hash_entry *hresolve; | 
 |  | 
 | 	  /* We have saved the hash table entry for this symbol, if | 
 | 	     there is one.  Note that we could just look it up again | 
 | 	     in the hash table, provided we first check that it is an | 
 | 	     external symbol. */ | 
 | 	  h = *sym_hash; | 
 |  | 
 | 	  /* Use the name from the hash table, in case the symbol was | 
 | 	     wrapped.  */ | 
 | 	  if (h != NULL) | 
 | 	    name = h->root.root.string; | 
 |  | 
 | 	  /* If this is an indirect or warning symbol, then change | 
 | 	     hresolve to the base symbol.  We also change *sym_hash so | 
 | 	     that the relocation routines relocate against the real | 
 | 	     symbol.  */ | 
 | 	  hresolve = h; | 
 | 	  if (h != NULL | 
 | 	      && (h->root.type == bfd_link_hash_indirect | 
 | 		  || h->root.type == bfd_link_hash_warning)) | 
 | 	    { | 
 | 	      hresolve = (struct aout_link_hash_entry *) h->root.u.i.link; | 
 | 	      while (hresolve->root.type == bfd_link_hash_indirect | 
 | 		     || hresolve->root.type == bfd_link_hash_warning) | 
 | 		hresolve = ((struct aout_link_hash_entry *) | 
 | 			    hresolve->root.u.i.link); | 
 | 	      *sym_hash = hresolve; | 
 | 	    } | 
 |  | 
 | 	  /* If the symbol has already been written out, skip it.  */ | 
 | 	  if (h != NULL | 
 | 	      && h->root.type != bfd_link_hash_warning | 
 | 	      && h->written) | 
 | 	    { | 
 | 	      if ((type & N_TYPE) == N_INDR | 
 | 		  || type == N_WARNING) | 
 | 		skip_next = true; | 
 | 	      *symbol_map = h->indx; | 
 | 	      continue; | 
 | 	    } | 
 |  | 
 | 	  /* See if we are stripping this symbol.  */ | 
 | 	  skip = false; | 
 | 	  switch (strip) | 
 | 	    { | 
 | 	    case strip_none: | 
 | 	      break; | 
 | 	    case strip_debugger: | 
 | 	      if (is_stab (type, name)) | 
 | 		skip = true; | 
 | 	      break; | 
 | 	    case strip_some: | 
 | 	      if (bfd_hash_lookup (flaginfo->info->keep_hash, name, | 
 | 				   false, false) == NULL) | 
 | 		skip = true; | 
 | 	      break; | 
 | 	    case strip_all: | 
 | 	      skip = true; | 
 | 	      break; | 
 | 	    } | 
 | 	  if (skip) | 
 | 	    { | 
 | 	      if (h != NULL) | 
 | 		h->written = true; | 
 | 	      continue; | 
 | 	    } | 
 |  | 
 | 	  /* Get the value of the symbol.  */ | 
 | 	  if (is_stab (type, name)) | 
 | 	    { | 
 | 	      switch (type) | 
 | 		{ | 
 | 		default: | 
 | 		  symsec = bfd_abs_section_ptr; | 
 | 		  break; | 
 | 		case N_SO: | 
 | 		case N_SOL: | 
 | 		case N_FUN: | 
 | 		case N_ENTRY: | 
 | 		case N_SLINE: | 
 | 		case N_FN: | 
 | 		  symsec = obj_textsec (input_bfd); | 
 | 		  break; | 
 | 		case N_STSYM: | 
 | 		case N_DSLINE: | 
 | 		  symsec = obj_datasec (input_bfd); | 
 | 		  break; | 
 | 		case N_LCSYM: | 
 | 		case N_BSLINE: | 
 | 		  symsec = obj_bsssec (input_bfd); | 
 | 		  break; | 
 | 		} | 
 | 	      val = GET_WORD (input_bfd, sym->e_value); | 
 | 	    } | 
 | 	  else if ((type & N_TYPE) == N_TEXT | 
 | 	      || type == N_WEAKT) | 
 | 	    symsec = obj_textsec (input_bfd); | 
 | 	  else if ((type & N_TYPE) == N_DATA | 
 | 		   || type == N_WEAKD) | 
 | 	    symsec = obj_datasec (input_bfd); | 
 | 	  else if ((type & N_TYPE) == N_BSS | 
 | 		   || type == N_WEAKB) | 
 | 	    symsec = obj_bsssec (input_bfd); | 
 | 	  else if ((type & N_TYPE) == N_ABS | 
 | 		   || type == N_WEAKA) | 
 | 	    symsec = bfd_abs_section_ptr; | 
 | 	  else if (((type & N_TYPE) == N_INDR | 
 | 		    && (hresolve == NULL | 
 | 			|| (hresolve->root.type != bfd_link_hash_defined | 
 | 			    && hresolve->root.type != bfd_link_hash_defweak | 
 | 			    && hresolve->root.type != bfd_link_hash_common))) | 
 | 		   || type == N_WARNING) | 
 | 	    { | 
 | 	      /* Pass the next symbol through unchanged.  The | 
 | 		 condition above for indirect symbols is so that if | 
 | 		 the indirect symbol was defined, we output it with | 
 | 		 the correct definition so the debugger will | 
 | 		 understand it.  */ | 
 | 	      pass = true; | 
 | 	      val = GET_WORD (input_bfd, sym->e_value); | 
 | 	      symsec = NULL; | 
 | 	    } | 
 | 	  else | 
 | 	    { | 
 | 	      /* If we get here with an indirect symbol, it means that | 
 | 		 we are outputting it with a real definition.  In such | 
 | 		 a case we do not want to output the next symbol, | 
 | 		 which is the target of the indirection.  */ | 
 | 	      if ((type & N_TYPE) == N_INDR) | 
 | 		skip_next = true; | 
 |  | 
 | 	      symsec = NULL; | 
 |  | 
 | 	      /* We need to get the value from the hash table.  We use | 
 | 		 hresolve so that if we have defined an indirect | 
 | 		 symbol we output the final definition.  */ | 
 | 	      if (h == NULL) | 
 | 		{ | 
 | 		  switch (type & N_TYPE) | 
 | 		    { | 
 | 		    case N_SETT: | 
 | 		      symsec = obj_textsec (input_bfd); | 
 | 		      break; | 
 | 		    case N_SETD: | 
 | 		      symsec = obj_datasec (input_bfd); | 
 | 		      break; | 
 | 		    case N_SETB: | 
 | 		      symsec = obj_bsssec (input_bfd); | 
 | 		      break; | 
 | 		    case N_SETA: | 
 | 		      symsec = bfd_abs_section_ptr; | 
 | 		      break; | 
 | 		    default: | 
 | 		      val = 0; | 
 | 		      break; | 
 | 		    } | 
 | 		} | 
 | 	      else if (hresolve->root.type == bfd_link_hash_defined | 
 | 		       || hresolve->root.type == bfd_link_hash_defweak) | 
 | 		{ | 
 | 		  asection *input_section; | 
 | 		  asection *output_section; | 
 |  | 
 | 		  /* This case usually means a common symbol which was | 
 | 		     turned into a defined symbol.  */ | 
 | 		  input_section = hresolve->root.u.def.section; | 
 | 		  output_section = input_section->output_section; | 
 | 		  BFD_ASSERT (bfd_is_abs_section (output_section) | 
 | 			      || output_section->owner == output_bfd); | 
 | 		  val = (hresolve->root.u.def.value | 
 | 			 + bfd_section_vma (output_section) | 
 | 			 + input_section->output_offset); | 
 |  | 
 | 		  /* Get the correct type based on the section.  If | 
 | 		     this is a constructed set, force it to be | 
 | 		     globally visible.  */ | 
 | 		  if (type == N_SETT | 
 | 		      || type == N_SETD | 
 | 		      || type == N_SETB | 
 | 		      || type == N_SETA) | 
 | 		    type |= N_EXT; | 
 |  | 
 | 		  type &=~ N_TYPE; | 
 |  | 
 | 		  if (output_section == obj_textsec (output_bfd)) | 
 | 		    type |= (hresolve->root.type == bfd_link_hash_defined | 
 | 			     ? N_TEXT | 
 | 			     : N_WEAKT); | 
 | 		  else if (output_section == obj_datasec (output_bfd)) | 
 | 		    type |= (hresolve->root.type == bfd_link_hash_defined | 
 | 			     ? N_DATA | 
 | 			     : N_WEAKD); | 
 | 		  else if (output_section == obj_bsssec (output_bfd)) | 
 | 		    type |= (hresolve->root.type == bfd_link_hash_defined | 
 | 			     ? N_BSS | 
 | 			     : N_WEAKB); | 
 | 		  else | 
 | 		    type |= (hresolve->root.type == bfd_link_hash_defined | 
 | 			     ? N_ABS | 
 | 			     : N_WEAKA); | 
 | 		} | 
 | 	      else if (hresolve->root.type == bfd_link_hash_common) | 
 | 		val = hresolve->root.u.c.size; | 
 | 	      else if (hresolve->root.type == bfd_link_hash_undefweak) | 
 | 		{ | 
 | 		  val = 0; | 
 | 		  type = N_WEAKU; | 
 | 		} | 
 | 	      else | 
 | 		val = 0; | 
 | 	    } | 
 | 	  if (symsec != NULL) | 
 | 	    val = (symsec->output_section->vma | 
 | 		   + symsec->output_offset | 
 | 		   + (GET_WORD (input_bfd, sym->e_value) | 
 | 		      - symsec->vma)); | 
 |  | 
 | 	  /* If this is a global symbol set the written flag, and if | 
 | 	     it is a local symbol see if we should discard it.  */ | 
 | 	  if (h != NULL) | 
 | 	    { | 
 | 	      h->written = true; | 
 | 	      h->indx = obj_aout_external_sym_count (output_bfd); | 
 | 	    } | 
 | 	  else if ((type & N_TYPE) != N_SETT | 
 | 		   && (type & N_TYPE) != N_SETD | 
 | 		   && (type & N_TYPE) != N_SETB | 
 | 		   && (type & N_TYPE) != N_SETA) | 
 | 	    { | 
 | 	      switch (discard) | 
 | 		{ | 
 | 		case discard_none: | 
 | 		case discard_sec_merge: | 
 | 		  break; | 
 | 		case discard_l: | 
 | 		  if (!is_stab (type, name) | 
 | 		      && bfd_is_local_label_name (input_bfd, name)) | 
 | 		    skip = true; | 
 | 		  break; | 
 | 		case discard_all: | 
 | 		  skip = true; | 
 | 		  break; | 
 | 		} | 
 | 	      if (skip) | 
 | 		{ | 
 | 		  pass = false; | 
 | 		  continue; | 
 | 		} | 
 | 	    } | 
 |  | 
 | 	  /* An N_BINCL symbol indicates the start of the stabs | 
 | 	     entries for a header file.  We need to scan ahead to the | 
 | 	     next N_EINCL symbol, ignoring nesting, adding up all the | 
 | 	     characters in the symbol names, not including the file | 
 | 	     numbers in types (the first number after an open | 
 | 	     parenthesis).  */ | 
 | 	  if (type == N_BINCL) | 
 | 	    { | 
 | 	      struct external_nlist *incl_sym; | 
 | 	      int nest; | 
 | 	      struct aout_link_includes_entry *incl_entry; | 
 | 	      struct aout_link_includes_totals *t; | 
 |  | 
 | 	      val = 0; | 
 | 	      nest = 0; | 
 | 	      for (incl_sym = sym + 1; incl_sym < sym_end; incl_sym++) | 
 | 		{ | 
 | 		  int incl_type; | 
 |  | 
 | 		  incl_type = H_GET_8 (input_bfd, incl_sym->e_type); | 
 | 		  if (incl_type == N_EINCL) | 
 | 		    { | 
 | 		      if (nest == 0) | 
 | 			break; | 
 | 		      --nest; | 
 | 		    } | 
 | 		  else if (incl_type == N_BINCL) | 
 | 		    ++nest; | 
 | 		  else if (nest == 0) | 
 | 		    { | 
 | 		      const char *s; | 
 |  | 
 | 		      s = strings + GET_WORD (input_bfd, incl_sym->e_strx); | 
 | 		      for (; *s != '\0'; s++) | 
 | 			{ | 
 | 			  val += *s; | 
 | 			  if (*s == '(') | 
 | 			    { | 
 | 			      /* Skip the file number.  */ | 
 | 			      ++s; | 
 | 			      while (ISDIGIT (*s)) | 
 | 				++s; | 
 | 			      --s; | 
 | 			    } | 
 | 			} | 
 | 		    } | 
 | 		} | 
 |  | 
 | 	      /* If we have already included a header file with the | 
 | 		 same value, then replace this one with an N_EXCL | 
 | 		 symbol.  */ | 
 | 	      copy = ! flaginfo->info->keep_memory; | 
 | 	      incl_entry = aout_link_includes_lookup (&flaginfo->includes, | 
 | 						      name, true, copy); | 
 | 	      if (incl_entry == NULL) | 
 | 		return false; | 
 | 	      for (t = incl_entry->totals; t != NULL; t = t->next) | 
 | 		if (t->total == val) | 
 | 		  break; | 
 | 	      if (t == NULL) | 
 | 		{ | 
 | 		  /* This is the first time we have seen this header | 
 | 		     file with this set of stabs strings.  */ | 
 | 		  t = bfd_hash_allocate (&flaginfo->includes.root, | 
 | 					 sizeof *t); | 
 | 		  if (t == NULL) | 
 | 		    return false; | 
 | 		  t->total = val; | 
 | 		  t->next = incl_entry->totals; | 
 | 		  incl_entry->totals = t; | 
 | 		} | 
 | 	      else | 
 | 		{ | 
 | 		  int *incl_map; | 
 |  | 
 | 		  /* This is a duplicate header file.  We must change | 
 | 		     it to be an N_EXCL entry, and mark all the | 
 | 		     included symbols to prevent outputting them.  */ | 
 | 		  type = N_EXCL; | 
 |  | 
 | 		  nest = 0; | 
 | 		  for (incl_sym = sym + 1, incl_map = symbol_map + 1; | 
 | 		       incl_sym < sym_end; | 
 | 		       incl_sym++, incl_map++) | 
 | 		    { | 
 | 		      int incl_type; | 
 |  | 
 | 		      incl_type = H_GET_8 (input_bfd, incl_sym->e_type); | 
 | 		      if (incl_type == N_EINCL) | 
 | 			{ | 
 | 			  if (nest == 0) | 
 | 			    { | 
 | 			      *incl_map = -1; | 
 | 			      break; | 
 | 			    } | 
 | 			  --nest; | 
 | 			} | 
 | 		      else if (incl_type == N_BINCL) | 
 | 			++nest; | 
 | 		      else if (nest == 0) | 
 | 			*incl_map = -1; | 
 | 		    } | 
 | 		} | 
 | 	    } | 
 | 	} | 
 |  | 
 |       /* Copy this symbol into the list of symbols we are going to | 
 | 	 write out.  */ | 
 |       H_PUT_8 (output_bfd, type, outsym->e_type); | 
 |       H_PUT_8 (output_bfd, H_GET_8 (input_bfd, sym->e_ovly), outsym->e_ovly); | 
 |       H_PUT_16 (output_bfd, H_GET_16 (input_bfd, sym->e_desc), outsym->e_desc); | 
 |       copy = false; | 
 |       if (! flaginfo->info->keep_memory) | 
 | 	{ | 
 | 	  /* name points into a string table which we are going to | 
 | 	     free.  If there is a hash table entry, use that string. | 
 | 	     Otherwise, copy name into memory.  */ | 
 | 	  if (h != NULL) | 
 | 	    name = h->root.root.string; | 
 | 	  else | 
 | 	    copy = true; | 
 | 	} | 
 |       strtab_index = add_to_stringtab (output_bfd, flaginfo->strtab, | 
 | 				       name, copy); | 
 |       if (strtab_index == (bfd_size_type) -1) | 
 | 	return false; | 
 |       PUT_WORD (output_bfd, strtab_index, outsym->e_strx); | 
 |       PUT_WORD (output_bfd, val, outsym->e_value); | 
 |       *symbol_map = obj_aout_external_sym_count (output_bfd); | 
 |       ++obj_aout_external_sym_count (output_bfd); | 
 |       ++outsym; | 
 |     } | 
 |  | 
 |   /* Write out the output symbols we have just constructed.  */ | 
 |   if (outsym > flaginfo->output_syms) | 
 |     { | 
 |       bfd_size_type size; | 
 |  | 
 |       if (bfd_seek (output_bfd, flaginfo->symoff, SEEK_SET) != 0) | 
 | 	return false; | 
 |       size = outsym - flaginfo->output_syms; | 
 |       size *= EXTERNAL_NLIST_SIZE; | 
 |       if (bfd_bwrite ((void *) flaginfo->output_syms, size, output_bfd) != size) | 
 | 	return false; | 
 |       flaginfo->symoff += size; | 
 |     } | 
 |  | 
 |   return true; | 
 | } | 
 |  | 
 | /* Write out a symbol that was not associated with an a.out input | 
 |    object.  */ | 
 |  | 
 | static bfd_vma | 
 | bfd_getp32 (const void *p) | 
 | { | 
 |   const bfd_byte *addr = p; | 
 |   unsigned long v; | 
 |  | 
 |   v = (unsigned long) addr[1] << 24; | 
 |   v |= (unsigned long) addr[0] << 16; | 
 |   v |= (unsigned long) addr[3] << 8; | 
 |   v |= (unsigned long) addr[2]; | 
 |   return v; | 
 | } | 
 |  | 
 | #define COERCE32(x) (((bfd_signed_vma) (x) ^ 0x80000000) - 0x80000000) | 
 |  | 
 | static bfd_signed_vma | 
 | bfd_getp_signed_32 (const void *p) | 
 | { | 
 |   const bfd_byte *addr = p; | 
 |   unsigned long v; | 
 |  | 
 |   v = (unsigned long) addr[1] << 24; | 
 |   v |= (unsigned long) addr[0] << 16; | 
 |   v |= (unsigned long) addr[3] << 8; | 
 |   v |= (unsigned long) addr[2]; | 
 |   return COERCE32 (v); | 
 | } | 
 |  | 
 | static void | 
 | bfd_putp32 (bfd_vma data, void *p) | 
 | { | 
 |   bfd_byte *addr = p; | 
 |  | 
 |   addr[0] = (data >> 16) & 0xff; | 
 |   addr[1] = (data >> 24) & 0xff; | 
 |   addr[2] = (data >> 0) & 0xff; | 
 |   addr[3] = (data >> 8) & 0xff; | 
 | } | 
 |  | 
 | const bfd_target MY (vec) = | 
 | { | 
 |   TARGETNAME,			/* Name.  */ | 
 |   bfd_target_aout_flavour, | 
 |   BFD_ENDIAN_LITTLE,		/* Target byte order (little).  */ | 
 |   BFD_ENDIAN_LITTLE,		/* Target headers byte order (little).  */ | 
 |   (HAS_RELOC | EXEC_P |		/* Object flags.  */ | 
 |    HAS_LINENO | HAS_DEBUG | | 
 |    HAS_SYMS | HAS_LOCALS | WP_TEXT), | 
 |   (SEC_HAS_CONTENTS | SEC_ALLOC | SEC_LOAD | SEC_RELOC | SEC_CODE | SEC_DATA), | 
 |   MY_symbol_leading_char, | 
 |   AR_PAD_CHAR,			/* AR_pad_char.  */ | 
 |   15,				/* AR_max_namelen.  */ | 
 |   0,				/* match priority.  */ | 
 |   TARGET_KEEP_UNUSED_SECTION_SYMBOLS, /* keep unused section symbols.  */ | 
 |   bfd_getl64, bfd_getl_signed_64, bfd_putl64, | 
 |      bfd_getp32, bfd_getp_signed_32, bfd_putp32, | 
 |      bfd_getl16, bfd_getl_signed_16, bfd_putl16, /* Data.  */ | 
 |   bfd_getl64, bfd_getl_signed_64, bfd_putl64, | 
 |      bfd_getp32, bfd_getp_signed_32, bfd_putp32, | 
 |      bfd_getl16, bfd_getl_signed_16, bfd_putl16, /* Headers.  */ | 
 |   {				/* bfd_check_format.  */ | 
 |     _bfd_dummy_target, | 
 |     MY_object_p, | 
 |     bfd_generic_archive_p, | 
 |     MY_core_file_p | 
 |   }, | 
 |   {				/* bfd_set_format.  */ | 
 |     _bfd_bool_bfd_false_error, | 
 |     MY_mkobject, | 
 |     _bfd_generic_mkarchive, | 
 |     _bfd_bool_bfd_false_error | 
 |   }, | 
 |   {			/* bfd_write_contents.  */ | 
 |     _bfd_bool_bfd_false_error, | 
 |     MY_write_object_contents, | 
 |     _bfd_write_archive_contents, | 
 |     _bfd_bool_bfd_false_error | 
 |   }, | 
 |  | 
 |   BFD_JUMP_TABLE_GENERIC (MY), | 
 |   BFD_JUMP_TABLE_COPY (MY), | 
 |   BFD_JUMP_TABLE_CORE (MY), | 
 |   BFD_JUMP_TABLE_ARCHIVE (MY), | 
 |   BFD_JUMP_TABLE_SYMBOLS (MY), | 
 |   BFD_JUMP_TABLE_RELOCS (MY), | 
 |   BFD_JUMP_TABLE_WRITE (MY), | 
 |   BFD_JUMP_TABLE_LINK (MY), | 
 |   BFD_JUMP_TABLE_DYNAMIC (MY), | 
 |  | 
 |   /* Alternative_target.  */ | 
 |   NULL, | 
 |  | 
 |   (void *) MY_backend_data | 
 | }; |