* coff-z8k.c: getting closer
* coffcode.h (coff_get_relocated_section_contents): allow
EXTRA_CASES hooks for different arches to provide different reloc
types.
* seclet.h: #ifndef around it to allow multiple inclusion
* srec.c: minor doc fix
* coffcode.h (coff_renumber_symbols): set the last renumbered
symbol pointer to NULL, as expected by coff_write_linenumbers.
(coff_write_relocs): apparently a non-zero addend is OK for reloc
type R_IHCONST used on the 29k.
update, based on how the underlying file was opened. Obsoletes
FASCIST_FDOPEN.
* hosts/rs6000.h, hosts/tahoe.h, hosts/vaxbsd.h: Remove
all FASCIST_FDOPEN config defines.
* archive.c (bfd_slurp_bsd_armap): if the symdef_count is too
large, assume we're using a swapped byte order and fail with
wrong_format rather than dumping core.
* targets.c, config/i960-bout.mt, hosts/*: added
HOST_BIG_ENDIAN_DEFAULT_VECTOR and
HOST_LITTLE_ENDIAN_DEFAULT_VECTOR because bout archive format
depends on the endianness of the host. Added #define
HOST_BIG_ENDIAN_P and #undef HOST_BIG_ENDIAN_P to a number of
configuration files in hosts/ (it was already in some).
from a coff file (csym is null), dont deref them
* ieee.c (parse_expression): get the answer right when adding an
abs+(sec+off), (ieee_generic_stat_arch_elt): call ieee_object_p
on elts so that filename is filled in.
Tue Aug 18 12:57:45 1992 Ian Lance Taylor (ian@cygnus.com)
* Makefile.in: added FLAGS_TO_PASS, and used it for all recursive
invocations of make. Also, always create installation
directories.
* config/apollov68.mh: removed -g from CC definition.
* libaout.h (struct aout_backend_data): New callback "set_sizes".
* aout-target.h (callback): Don't set page/seg sizes.
(mkobject): Ditto.
(set_sizes): New routine.
(backend_data): Point to it.
* aoutf1.h (sunos4_set_sizes): Heed architecture when setting sizes.
(sunos4_aout_backend): Point to it.
* aoutx.h (set_arch_mach): Call set_sizes callback.
the default. Given that we *write* the suckers (for long
archive member names), we really ought to be able to read them!
* trad-core.c: Don't include <machine/reg.h>. It doesn't
seem to be needed, and many machines don't have it.
pointers to bfd_byte* before performing arithmetic on them. Such
use with bare void* pointers is a gcc extension.
* cache.c (bfd_cache_delete): Forward decl with prototype form.
* archive (normalize): Add to CONST to match actual usages with
CONST.
* elf.c (bfd_section_from_shdr, elf_section_from_bfd_section,
section_from_elf_index): Add SHT_BEPROGBITS case for i860 in
big endian mode.
* elf.c (bfd_section_from_shdr): Test for the possibility that
section_from_elf_index returns NULL and don't dereference it.
* bout.c (bout_swap_exec_header_[in_out]): new field in exec
header e_relaxable, major hackery in callbacks
* libaout.h : add space for a_relaxable.
* seclet.c (rel): don't relax empty sections
Lints for bfd_reloc_code_type->bfd_reloc_code_real_type, and
correct calling of howto special functions and fixes from Raeburn
for gas<>bfdness
* archures.c: nuke bfd_reloc_code_type
* bout.c, cpu-h8300.c, coff-i960.c, coff-m88k.c: special function lint
* coffcode.h (styp_to_sec_flags): STYP_INFO is marked as
SEC_NEVER_LOAD, various other bfdgas newness
* reloc.c->libbfd.h: change protype of bfd_default_reloc_type_lookup
* targets.c: change jump table vector to above
its prototype will be included in libbfd.h. Change result type
to struct * (rather than equivalent typedef) so it can be used in
the prototype, where they typedef won't be known.
* libbfd.h: Updated version.
Correct misconception that there can be only one symbol table.
Only call elf_slurp_symbol_table on the full symbol table, not
the dynamic one which is only a subset of the full one.
(The bfd library was slurping in the first ELF symbol table it
found and then ignoring any others. For dynamically linked
executables, this happened to be the dynamic symbol table, which
is a subset of the full symbol table and only contains the
symbols that need to be used in dynamic linking.)
Various changes to get linker working again for a.out:
* don't set/adjust section file positions or vmas more than once
* use correct page size and segment size when calculating them
* deal with some variations in a.out implementations
Tested on sun4 and sun4->sun3 so far, will be testing further but
needed to get wider exposure&testing. See ChangeLog for details.
Also:
* coffcode.h (coff_write_relocs): Write out swapped reloc, not
pre-swapped version.
* hosts/sparc.h (abort, exit): Hide these names if compiling with
gcc version 2, to avoid warnings.
* don't set/adjust section file positions or vmas more than once
* use correct page size and segment size when calculating them
* deal with some variations in a.out implementations
Tested on sun4 and sun4->sun3 so far, will be testing further but
needed to get wider exposure&testing. See ChangeLog for details.
Also:
* coffcode.h (coff_write_relocs): Write out swapped reloc, not
pre-swapped version.
* hosts/sparc.h (abort, exit): Hide these names if compiling with
gcc version 2, to avoid warnings.
* configure.in: Rewrite target parsing to use simple format.
Handle sparc-sun-solaris2 configuration.
* aout64.h: Avoid ANSI C brain death warning.
* elf.c: Avoid trigraph (???) or /* in comments.
elf_core_tdata_struct into a single common struct. Core files
wouldn't have worked at all without this.
(bfd_elf_find_section): New function for GDB's undercover use
to find string sections that BFD hides from it.
(elf_get_str_section): Avoid multiple alloc&reads for same data; lint.
(elf_object_p, elf_core_file_p): Allocate internal file header
storage dynamically.
* bfd.c (union {...} tdata): Remove elf_core_tdata_struct.
* demo64.c: Prevent "empty translation unit" warnings from idiots.
* working reloc lookup support
* better support for creation of COFF files from scratch
* misc doc changes, macro tweaks, "const" additions, typo fixes, etc.
* one or two ugly kludges that should get cleaned up (but what's one
or two more, in BFD?)
See ChangeLog for details.
* working reloc lookup support
* better support for creation of COFF files from scratch
* misc doc changes, macro tweaks, "const" additions, typo fixes, etc.
* one or two ugly kludges that should get cleaned up (but what's one
or two more, in BFD?)
See ChangeLog for details.
* coff-a29k.c (perform_reloc): fix bug in jmp/call evaluation
* coff-h8300.c (reloc_processing): all relocs are relative to
section start.
* opncls.c: don't use fdopen on DOS systems
short patches from Glenn Kasten (glenn@ready.com)
* target.c: enabled oasys support
* oasys.c: Insert . when UNDERSCORE_HACK is enabled,
fix problem where a relocation which crossed a modification byte
boundary did not work. Fix problem where a relocation near the end
of a data record did not work.
Major change; changed calling convention for
bfd_get_relocated_section_contents so that caller allocates
memory for section data.
* coffcode.h (bfd_coff_get_relocated_section_contents), reloc.c,
seclet.c, targets.c, bfd.c: reflect new convention.
* coffcode.h (styp_to_sec_flags): if styp_flags is not a special
case, then use reasonable default values for SEC_* flags.
* ieee.c: if some places where null pointers were used instead of &bfd_abs_section.
* configure.in: tandem target is st2000
* coff-m68k.c: rename static howto_table to global
m68kcoff_howto_table.
* bout.c: remove unnecessary abort
bout.c (b_out_squirt_out_relocs): treat abs sumbols the right way.
* reloc.c (bfd_perform-relocation): don't relocate refs to
absolute symbols if doing a partial link.
structure before initializing the fields that are used, so that the unused
fields are in a known state. Two of these structures are memcmp'd so
any fields not explicitly initialized must be set to a known state.
* bfd.c, Makefile.in: added tekhex
* aoutx.h (set_section_contents): pages should be padded to the
size of a page, not the size of a segment.
* configure.in: added go32 host and i386-aout target.
* i386aout.c, libaout.h: now works for go32 target
* ieee.c: fix bit rot.
* seclet.c: support for padding seclet type.
if they make sense as independent sections; symbols and relocs are
turned into BFD symbols and relocs; objects are now created.
objdump and copy mostly work on .o files.
this check-in will be followed by gdb testing.
the code has *not* yet been split up the way coff is to
support multiple architectures and bit widths.
no ChangeLog entries yet; too big... nearly everything changed.
* configure.in: Add new host and target configs.
* elf.c (bfd_prstatus, bfd_fpregset): Avoid typename of register
structs, which vary; just use member name.
(elf_object_p, elf_core_file_p): Lint, fix comments.
* config/sparc-elf.mt, hosts/sysv4.h: New config files.
symbol addresss for undefined symbols.
* coff-rs6000.c: Various updates (due to target-
independent changes).
* libbfd.h: Ditto (tdata field is now a union).
* hosts/h-sparc.h: Removed prototype for bogus function
'emset' (presumably should have been 'memset' - which is there).
* bfd.c: lint
* coff-h8300.c: fix stupid reloc subtraction bug
* coffcode.h (coff_swap_aux_out): only swap a tvndx by 2 bytes,
since that's how wide it is. (coff_slurp_symbol_table): always zero out
the symbol.flags and done_lineno fields.
out tagndx too.
* aoutx.h (swap_std_reloc_out, swap_ext_reloc_out): use the output
section of a symbol rather than the section of a reffed symbol to
determine the r_index of an emmitted reloc.
"done_lineno" so that a symbol which appears twice in the symbol
table only gets it's linenumbers relocated once. Modifed
(coff_write_native_symbol) and (coff_make_empty_symbol) to make
* aout-target.h: Use TARGET_BIG_ENDIAN_P to set up vector.
* host-aout.c: Don't use aout-params.h; configure these systems
in their xm files, not in a randomly generated file.
* aoutx.h: cast an enum
* coff-a29k.c: many ints in reloc structure turned to enum
* coffcode.h: more enums
* cpu-h8300.c: more enums
* opncls.c: more enums
elf.c: Remove extraneous paren from core_prpsinfo and core_prpstatus
macros, rename core_prpstatus to core_prstatus. Replace references to
old section "size" member with new "_raw_size" member. Implement
elf_make_empty_symbol, which is now used.