* elf-hppa.c (elf_hppa_final_write_processing): New function.
* elf32-hppa.c: Remove everything related to symbol extension
sections & records. Use the common elf_hppa_final_write_processing.
* elf32-ppc.c (ppc_elf_size_dynamic_sections): Set unused dynindx
field to 0, not -1.
(ppc_elf_finish_dynamic_sections): Check for positive section
dynindx field, rather than comparing against -1.
(ppc_elf_relocate_section): Only return false if undefined_symbol
or reloc_overflow fail, not always upon encountering an error.
* elflink.c (_bfd_elf_link_adjust_dynindx): Define it.
* elflink.h (elf_link_remove_section_and_adjust_dynindices): New
function.
(bfd_elf_size_dynamic_sections): Use it.
* elf32-ppc.c (ppc_elf_relocate_section): Don't barf on out of
range undefweak symbols.
* hash.c: Add missing comma after @xref{}
* linker.c: Likewise.
* dwarf2.c (struct line_info): Add member END_SEQUENCE to keep
track of end_sequence markers.
(add_line_info): Add END_SEQUENCE arg.
(decode_line_info): Don't try to infer lo_pc and hi_pc from the
debug-line info---it doesn't work right if a compilation unit
consists of multiple discontiguous code-sequences. It would be
worthwhile to optimize for the common case where a compilation
unit results in a contiguous set of code sequences, but this is
quite tricky to get right for arbitrary DWARF2 files.
(lookup_address_in_line_info_table): Don't use the last line entry
for a compilation unit for anything with an address higher than
this line entry. Also, check for end_sequence markers to
correctly handle discontinuities.
(_bfd_dwarf2_find_nearest_line): When checking previously loaded
compilation units, check all compilation units with each->high==0
just like when reading compilation units.
* dwarf2.c (decode_line_info): Initialize table->files and
table->last_line to NULL to avoid segfaults due to random
values in these members.
(concat_filename): Check for out-of-range file number before
indexing filename table. Segfaults suck.
* dwarf2.c (decode_line_info): Don't truncate address to least
significant 32 bits (breaks 64 bit targets).
(lookup_address_in_function_table): Ditto.
(comp_unit_contains_address): Ditto.
* elf64-mips.c (mips_elf64_howto_table_rel): Likewise.
(mips_elf64_howto_table_rela): Likewise.
* elfcode.h (elf_slurp_reloc_table_from_section): New function,
split out from ...
(elf_slurp_reloc_table): Here. Use it to handle the case where a
single section has two associated relocation sections.
may_use_rel_p, may_use_rela_p, default_use_rela_p.
(bfd_elf_section_data): Add use_rela_p.
* elf.c (bfd_section_from_shdr): Set use_rela_p appropriately.
(_bfd_elf_new_section_hook): Likewise.
(elf_fake_sections): Use may_use_rela_p, etc., instead of
use_rela_p.
(_bfd_elf_copy_private_section_data): Copy use_rela_p.
* elfcode.h (write_relocs): Determine whether or not use rela
relocs based on the relocation section header.
* elflink.c (_bfd_elf_create_dynamic_sections): Use default_use_rela_p
instead of use_rela_p.
* elfxx-target.h (elf_backend_may_use_relp): New macro.
(elf_backend_may_use_rela_p): Likewise.
(elf_backend_default_use_rela_p): Likewise.
(elfNN_bed): Use them.
(BFD_RELOC_MIPS_GOT_PAGE): Likewise.
(BFD_RELOC_MIPS_GOT_OFST): Likewise.
(BFD_RELOC_MIPS_GOT_DISP): Likewise.
* bfd-in2.h: Regenerated.
* libbfd.h: Likewise.
* elf32-mips.c (mips_info_to_howto_rela): New function.
(USE_REL): Adjust for new conventions.
(MINUS_ONE): New macro.
(elf_mips_howto_table): Add R_MIPS_SUB.
(mips_r): Add entries for MIPS_SUB, MIPS_GOT_PAGE, MIPS_GOT_OFST,
and MIPS_GOT_DISP.
(mips_elf_final_write_processing): Set sh_link, not sh_info, for a
.MIPS.content section.
(_bfd_mips_elf_fake_sections): Treat all sections that begin
with .MIPS.content as .MIPS.content sections. Set
SHF_MNIPS_NOSTRIP for such section.
(elf_info_to_howto): Define to mips_info_to_howto_rela.
* elf64-mips.c (mips_r): Add entries for MIPS_SUB, MIPS_GOT_PAGE,
MIPS_GOT_OFST, and MIPS_GOT_DISP.
* elf32-arm.h (elf32_arm_relocate_section): Reinstate change of
1993-03-25 (!!). Take into account the bitmasks for the reloc so
the addend does not overflow into the rest of the word
* archive.c: Define getuid and getgid as macros if HAVE_GETUID or
HAVE_GETGID are not defined, respectively.
(bfd_write_armap): Don't special case on _WIN32 for getuid and
getgid.
* configure.host: Set ac_cv_func_get{ug}id for *-*-windows*.
* configure, config.in: Rebuild.
* config.bfd (i[3456]86*-*-rtems*, m68k*-*-rtems*): Added to
list of target formats (targ_selvecs).
(i[3456]86*-*-rtemself*, mips*el-*-rtems*): New targets.
(powerpcle*-*rtems*, sh-*-rtemself*): New targets.
reasonable mode until "ar x" is smart enough to skip it (fixes
gcc/libgcc.a builds on mips-ecoff targets
* coffcode.h (styp_to_sec_flags): Explain how COMDATs are supposed
to work. Hack to support MS import libraries, which use different
COMDAT types than GNU.
(coff_slurp_symbol_table): C_SECTION symbols are local; they refer
to implied zero-length sections (see peicode below)
* coffgen.c (coff_get_normalized_symtab): Properly read long MS
filename symbols, which use one *or more* auxents.
* coffswap.h (coff_swap_aux_in): ditto
* peicode.h (coff_swap_sym_in): Build the implied zero-length
sections
* dwarf2.c (read_abbrevs): Change cast of dwarf_abbrev_buffer
assignment from "unsigned char *" to "char *".
(decode_line_info): Likewise for dwarf_line_buffer assignment.
