* configure.in: now h8 is a coff target
* cpu-h8300.c: fix various disassembly problems
* libcoff.h: took out some code which has been #0ed for a long
time.
* targets.c: added h8 coff
* coff-a29k.c, coff-i386.c, coff-i960.c, coff-m68k.c, coff-m88k.c
added new macro RTYPE2HOWTO to take a load of #ifdefs out of
coffcode.h
* coffcode.h: Started to change the way machine dependencies are
handled, from the nest of #ifdefs to macros defined in the
including coff-<foo>.c
from syms.c to new file stab-syms.c.
Renamed to aout_stab_name() and aout_stab_names[].
Also, since GNU extended type codes such as N_SETT are no longer
in ../include/stab.def, include them manually.
rather complicated seeming problem. Eliminate LOGICAL_ versions,
just make N_XXX work by excluding the header from the text segment.
* aoutx.h: Fix comments to match.
just &DEFAULT_VECTOR. This makes executables a lot smaller.
Old behavior can be gotten by defining ALL_TARGETS.
* aoutf1.h, demo64.c, i386aout.c, newsos3.c, sunos.c:
Factored out common code into new file aout-target.h.
Saves a lot of duplicate code for a.out variants.
* Makefile.in: standardize on MINUS_G to disable debugging
* coff-m88k.c: create coff armaps instead of bsd ones.
* opncls.c (bfd_close_all_done), (bfd_close): Mask out random bits
when calling chmod.
of two N_SO stabs, one for directory, and one for filename.
* bfd-in.h (print_vma): Factor out duplicate definition.
Exit a little more gracefully when malloc returns NULL.
* libbfd.c: New function bfd_xmalloc (malloc wrapper).
* opncls.c, syms.c, bout.c, aoutx.h: Replace malloc by bfd_xmalloc.
* libbfd.h: Rre-generated due to libbfd.c update.
* coffcode.h: Defined new macros [PUT|GET]LINENO_LNNO for
manipulation of lnno fields in lineno structs in a coff-<f>.h
independent way. Override it in coff-<f>.h to your favourite
number if you don't have a 16bit lnno field. (coff_swap_lineno_in)
(coff_swap_lineno_out): modified to use the new macros.
* Makefile.in: Added ctor.c
* archures.c (bfd_default_arch_struct) added
bfd_default_reloc_type_lookup.
* coffcode.h: removed a load of #if 0ed code.
(coff_compute_file_section_positions): now works out the section
alignment and size correctly in all cases.
(coff_get_symtab): looks for contructor symbols and calls the
right function when they're found. (coff_canonicalize_reloc):
knows when a section is full of contructors and does the right
thing.
* cpu-m88k.c: Added default_reloc_type lookup to the architecture
description.
* libbfd-in.h: (bfd_default_reloc_type_lookup) defun added.
* libcoff.h: removed #if 0ed code
* reloc.c: (bfd_reloc_code_real_type) added BFD_RELOC_CTOR for
arch independent contructtor relocation type.
(bfd_reloc_type_lookup): removed the comment "this will go away"
since it won't. (bfd_howto_32): stolen from 88k, this is a 32bit
reloc which is used when BFD_RELOC_CTOR falls through to give a
default 32bit reloc for contructors.
(bfd_default_reloc_type_lookup): added.
* aoutf1.h: fixed SEGMENT_SIZE typo.
bfd_print_symbol_nm, for use by nm.
* aoutx.h (NAME(aout,print_symbol)): Add code for new
bfd_print_symbol_nm option, to print in nm format.
* coffcode.h (coff_print_symbol), ieee.c (ieee_print_symbol)
oasys.c (oasys_print_symbol): Provide stub implementations
for bfd_print_symbol_nm.
* syms.c: New function bfd_decode_symclass, used by nm printer.
New function bfd_stab_name to look up string name of stab code.
section size out if there are alignment restrictions so that the
image will be ok on a system where section positions are worked
out by accumulating sizes rather than from the section headers.
* opncls.c (bfd_close_all_done). Added so that generative
programs like gas can close a bfd without causing bfd confusion.
* libbfd.h (changed becuase of protos)
* amdcoff.c: messed with the way that jmp displacements are
calcualated. This may not yet be totally correct.
* archive.c (coff_write_armap): rewrote the way that ranlibs are
written out.
* coffcode.h (fixup_symbol_value): now doesn't core dump if a non
abs symbol has no section (like a register symbol).
(coff_write_symbol) now zeros auxent before filling it up to help
with sensitive applications.
* libbfd.c (bfd_write_bigendian_4byte_int): added.
should not include teh size field itself.
* aoutx.h, libaout.h (NAME(aout, sizeof_headers)): Use
adata(abfd)->exec_bytes_size field instead of constant macro,
because aoutx.h compiles to a simple .o file shared by
all 32-bits a.out targets.
and data structures to handle the file, but set a few parameters
from the host's config file (page size, text start addr, etc).
* libaout.h: Define HP BSD machine types.
virtual-memory-address of the .reg section. We can't just make a
section for "all the regs and nothing else" because only GDB knows
exactly where the regs are (scattered around the upage and stack).
Clean up memory allocation. Remove big- and little-endian
vectors, replace with single vector; this only runs on the host,
in host byte order. Replace byteswap routines with aborts in case
anyone calls them.
* targets.c: There's only one trad_core_vec now.
* aoutf1.h (sunos_core_file_p): VMA of .reg* now needs to be 0.
* libbfd.h: Moved ALIGN macro here.
* bout.c (b_out_callback): Replace i960_align by align_power.
* libaout.h (struct aoutdata): Added fields page_size,
segment_size and exec_bytes_size. These help generic code
(in aoutx.h and libaout.h) to figure out where to align
the various segments in a demand paged file.
* libaout.h (WRITE_HEADER): Use new (struct aoutdata) fields
to decide if the exec header counts in a_text (text segment size).
Also, assume D_PAGED flag is set; don't set it here.
* bout.c (b_out_callback): Add initialization of
adata fields page_size, segment_size, and exec_bytes_size.
* host-aout.c (NAME(host_aout, callback) and
NAME(host_aout,mkobject): Ditto.
* i386aout.c (aout386_callback) and new aout386_mkobject): Ditto.
* newos3.c (newsos3_callback) and new newsos_mkobject): Ditto.
* aoutf1.c (sunos4_callback) and new sunos_mkobject): Ditto.
* aoutx.h (NAME(aout,some_aout_object_p)): Don't subtract
EXEC_BYTES_SIZE from the a_test size; this is only appropriate
for some systems (mainly sunos), so do it in the callback instead.
Same routine: Add stuff to the THIS_IS_ONLY_DOCUMENTATION comment.
* aoutx.h (NAME(aout,set_section_contents): Change the way
filepos and size of sections are carried out to support
older styles of demand paged executables.
* bfd.c: Added macros to enable gas to swap coff records.
* coffcode.h: Make the coff_swap_*_out routines return the number
of bytes swapped, and cleaned up their prototypes.
* configure.in: Added ebmon/coff support
* ecoff.c: Changed null definitions of coff_swap_*_out to conform
to new prototypes
* targets.c: Added new entry points for coff swapping
* include/sys/h-*.h: Move to bfd/hosts/h-*.h.
* configure.in: Revise to symlink sysdep.h to hosts/h-xxx.h.
Change some config names to match other dirs.
* *.c: Include bfd.h before sysdep.h, so ansidecl and PROTO()
get defined first.
* Makefile.in: Use -I. to get sysdep.h.
* trad-core.c: Don't disable the whole file if SUN4_SYS.
* config/: Rename some config files to match up h-*.h names.
Remove all the HOST_SYS definitions from the config files.
* hosts/h-amix.h, h-i386v.h: Define POSIX_UTIME, not USG.
* hosts/h-hp9000.h: Remove USG comment.
* hosts/h-rs6000.h: Include <stdlib.h> for malloc.
* hosts/h-news.h, h-sparc-64.h: New config files that had
previously been kludged in the <sysdep.h> file.
* hosts/h-hp9000.h: Remove USG comment.
* hosts/h-rs6000.h: Include <stdlib.h> for malloc.
* hosts/h-news.h, h-sparc-64.h: New config files that had
previously been kludged in the <sysdep.h> file.
* trad-core.c: Document how to use it nowadays.
* i386aout.c, newsos3.c: Clean up.
* i386coff.c: Allow 386 coff files to be used as core files too
(for reading core files from embedded systems).
* elf.c: Remove elf_set_section_contents, use generic one. Lint.
* libbfd-in.h, libbfd.c: Add bfd_generic_set_section_contents.
* libbfd.c (bfd_generic_{get,set}_section_contents): Check that
last byte of transfer, not first byte, is within the section.
* host-aout.c: Remove `BSD' archive support. Lint.
* archures.c: Rename `struct bfd_arch_info_struct' to `struct
bfd_arch_info'. Rename `typedef bfd_arch_info_struct_type' to
`bfd_arch_info_type'. All uses changed.
* reloc.c: Rename `bfd_reloc_status_enum_type' to
`bfd_reloc_status_type'. Rename `bfd_reloc_code_enum_real_type'
to `bfd_reloc_code_real_type'. (This seems to be a misnomer,
it needs a better name.) All uses changed.
* targets.c: Rename `enum target_flavour_enum' to `enum
target_flavour', and remove the `_enum' from all of the enum
values themselves. All uses changed.
* configure.in, config/h-i386mach: i386 mach host.
* config/t-i386-aout: Use host-aout.c.
* trad-core.c: Give it its own xvec's to make it independent
of other file formats.
* ecoff.c, host-aout.c: Remove refs to trad-core.
* config/t-dec3100, t-hp300bsd, t-tahoe, t-vax: Define TRAD_CORE.
* targets.c: #ifdef TRAD_CORE, include it in the vector.
Mon Sep 30 15:13:46 1991 Steve Chamberlain (steve at cygnus.com)
* cpu-a29k.c, cpu-i386.c, cpu-m68k.c, cpu-mips.c, cpu-vax.c,
cpu-h8300.c, cpu-i960.c, cpu-m88k.c, cpu-sparc.c: added. These
files will eventually contain processor specific bits for bfd,
like strange relocation information and dis/assembly. So far only
the H8 has been even partially done. This work also ties in with
the change in handling architectures.
* amdcoff.c: (a29k_reloc) fix error message.
* aout-f1.h: (choose_reloc_size) now calls bfd_get_arch to
discover the architecture of the bfd. (sunos4_callback) calls the
function bfd_set_arch_mach rather than stuffing stuff directly
into the bfd. (sunos4_write_object_contents), changed names of
accessor functions.
* aoutx.h: (set_arch_mach) now calls bfd_default_set_arch_mach to
setup the environment.
* archive.c: (bfd_slurp_coff_armap) coff archives always have the
headers in big endian format, regardless of the endianess of the
host or target.
* archures.c: totally changed. Now an architecture is represented
with a pointer to an info structure rather than an enumerated type
and a long. The old info is available as two elements in the
structure. Future enhancements to architecure support will
involve pointers to methods being placed into the info structure.
* bfd.c: changed the definition of the bfd structure for the new
architecture stuff.
* bout.c: (b_out_set_arch_mach) changed to use the new
architecture mechanism.
* coffcode.h: (coff_set_arch_mach, coff_set_flags) changed to use
the new architecture mechanism.
* configure.in: added h8 stuff.
* ieee.c: too many changes to note. Now ieee files written with
bfd gas and ld can be read by gld and ieee only linkers and
simulators.
* libbfd.c, libbfd.h: changed prototype of bfd_write.
* newsos3.c: (newos3_callback) now calls bfd_set_arch_mach rather
than fixing the structure directly.
* oasys.c: (oasys_object_p) now calls bfd_default_set_arch_mach rather
than fixing the structure directly.
* opncls.c: (new_bfd) makes sure that bfd_init has been called
before opening a bfd.
* srec.c: (srec_set_arch_mach) now calls bfd_default_set_arch_mach
rather than fixing the structure directly.
* targets.c: (target_vector) now by defining SELECT_VECS (perhaps
in the t/hmake file) a user can select which backends they want
linked with bfd without changing the source.
* init.c: new, looks after initializing modules.
* howto.c: for future use, will allow an application to work out
what cookie to use as a handle on a relcoatio howto.
* cpu-a29k.c, cpu-i386.c, cpu-m68k.c, cpu-mips.c, cpu-vax.c,
cpu-h8300.c, cpu-i960.c, cpu-m88k.c, cpu-sparc.c: added. These
files will eventually contain processor specific bits for bfd,
like strange relocation information and dis/assembly. So far only
the H8 has been even partially done. This work also ties in with
the change in handling architectures.
* aout-f1.h: (choose_reloc_size) now calls bfd_get_arch to
discover the architecture of the bfd. (sunos4_callback) calls the
function bfd_set_arch_mach rather than stuffing stuff directly
into the bfd. (sunos4_write_object_contents), changed names of
accessor functions.
* aoutx.h: (set_arch_mach) now calls bfd_default_set_arch_mach to
setup the environment.
* archive.c: (bfd_slurp_coff_armap) coff archives always have the
headers in big endian format, regardless of the endianess of the
host or target.
* archures.c: totally changed. Now an architecture is represented
with a pointer to an info structure rather than an enumerated type
and a long. The old info is available as two elements in the
structure. Future enhancements to architecure support will
involve pointers to methods being placed into the info structure.
* bfd.c: changed the definition of the bfd structure for the new
architecture stuff.
* bout.c: (b_out_set_arch_mach) changed to use the new
architecture mechanism.
* coffcode.h: (coff_set_arch_mach, coff_set_flags) changed to use
the new architecture mechanism.
* configure.in: added h8 stuff.
* ieee.c: too many changes to note. Now ieee files written with
bfd gas and ld can be read by gld and ieee only linkers and
simulators.
* libbfd.c, libbfd.h: changed prototype of bfd_write.
* newsos3.c: (newos3_callback) now calls bfd_set_arch_mach rather
than fixing the structure directly.
* oasys.c: (oasys_object_p) now calls bfd_default_set_arch_mach rather
than fixing the structure directly.
* opncls.c: (new_bfd) makes sure that bfd_init has been called
before opening a bfd.
* srec.c: (srec_set_arch_mach) now calls bfd_default_set_arch_mach
rather than fixing the structure directly.
* targets.c: (target_vector) now by defining SELECT_VECS (perhaps
in the t/hmake file) a user can select which backends they want
linked with bfd without changing the source.
* init.c: new, looks after initializing modules.
* howto.c: for future use, will allow an application to work out
what cookie to use as a handle on a relcoatio howto.
(srec_object_p): Avoid assuming any file that starts with 'S' is
an S-record file! (Check for 3 uppercase-hex digits after it.)
(ISHEX): Define new macro for checking supposedly hex characters.
(pass_over): Treat as EOF any S-record line that:
* doesn't have hex in either size character
* produces a size larger than 0x7F
* has an unrecognized type number
* ecoff.c: If compiling for DEC3100, use trad_unix core files,
else dummy out the core file support.
* trad-core.c (trad_unix_core_file_p): If HOST_DATA_START_ADDR is
specified, use it to locate the data section.
* coffcode.h (coff_write_symbols): Declare buffer as bfd_bytes
rather than as chars (lint).
* Makefile.in: use targets "bfd.dvi" and "bfd.ps" instead of
texdoc and psdoc; make these depend on generated .texi's
* scanit, (new) unPROTO: turn PROTO macros into ANSI declarations in
doc
* bfd.c, reloc.c, syms.c: minor spelling/wording fixes in doc
portions
* Makefile.in, awkscan-ip, awkscan-p, intobfd,
libbfd.h, libcoff.h, tolibcoff; new bfd-in.h, libbfd-in.h,
libcoff-in.h: (a) use separate files for invariant parts of bfd.h,
libbfd.h, and libcoff.h; (b) in generated parts of same, use less
obtrusive marks indicating .c origins.
* bfd.texinfo: generalize most references to linker
* aoutx.h, archive.c, archures.c, bfd.c, bfd.texinfo, cache.c,
coffcode.h, core.c, format.c, libbfd.c, libbfd.h, libcoff.h,
opncls.c, reloc.c, section.c, syms.c, targets.c (documentation
segments): used BFD (caps) more consistently as a name in
discourse, fixed a few other minor typos and uses of fonts
The stuff after the comment /* THE FOLLOWING IS EXTRACTED ... */ is
removed from the header files, so the header files broke themselves.
This is now fixed because the description is split over two lines. Not
a great way. I'll fix it better soon.
bfd.doc bfd.texinfo. Merged what was usefull into bfd.texinfo and
deleted bfd.doc.
Install header comments and attribute most modules to their
authors. BFD's name is now officially "Binary File Descriptor",
so remove any conflicting perjoratives.
bfd.texinfo oasys.c opncls.c reloc.c srec.c: More documentation on
lint fixes.
* amdcoff.c bfd.texinfo Makefile.in : Folded in changes for amd
29k coff by David Wood (wood@lab.ultra.nyu.edu).
choose_reloc_size. (sunos4_write_object_contents) now calls
choose_reloc_size so outputs relocs correctly, also calculates the
size of the sections correctly.
* aout.c: (look in the rrgs I hate VI)
~
~
~
~
