Enable powerpc simulator for all powerpc-eabi targets, not just powerpc-eabisim if compiling with GCC. Add --enable-sim-powerpc option to control this.
* config/tc-a29k.c (md_apply_fix): Warn if an attempt is made to
generate a reloc which the linker will not handle correctly. Fix
overflow checking--R_IREL is 18 bits, not 17.
* archive.c (bfd_generic_archive_p): Check the first object file
in an archive even if target_defaulted is set. If the object file
has the wrong xvec, reject it.
* aoutx.h (NAME(aout,set_section_contents)): If a section can not
be represented, report the name via _bfd_error_handler.
(translate_to_native_sym_flags): Likewise.
* elf32-mips.c (mips_elf_final_link): Likewise.
* oasys.c (oasys_write_sections): Likewise.
Turn on INLINES if using GCC to compile simulator; Print more stuff if requests -t trace; If !WITH_ASSERT, do not check whether illegal bits in instruction are set
* coffcode.h (coff_set_alignment_hook): Write RS6000COFF_C version
which checks for STYP_OVRFLO sections.
(coff_compute_section_file_positions): If RS6000COFF_C, handle
reloc and lineno count overflows.
(coff_write_object_contents): Call coff_count_linenumbers before
coff_compute_section_file_positions. If RS6000COFF_C, handle
reloc and lineno count overflows.
* xcofflink.c (_bfd_xcoff_bfd_final_link): Count line numbers and
relocs before dealing with .pad sections. Count overflow section
headers when handling .pad sections.
* emultempl/aix.em (gld${EMULATION_NAME}_before_allocation): Add
new local variable special_sections, and pass it to
size_dynamic_sections. Look through the results, and move the
sections around in the mapping so that they are defined correctly.
* xcofflink.c (XCOFF_DESCRIPTOR): Define.
(struct xcoff_link_hash_table): Add descriptor_section and
special_sections fields.
(_bfd_xcoff_bfd_link_hash_table_create): Initialize new fields.
(xcoff_link_add_symbols): Set linkage section alignment. Create
descriptor section. Check for magic symbol names (_text, etc.),
and record them in special_sections if found. Set
XCOFF_DESCRIPTOR flag for a function descriptor, and set its
descriptor field to point back to the function code symbol.
(xcoff_sweep): Always mark the special descriptor_section.
(bfd_xcoff_export_symbol): Check whether the symbol might be a
function descriptor, and mark it if it is.
(bfd_xcoff_size_dynamic_sections): Add new special_sections
parameter, and fill it in. Allocate space for the descriptor
section.
(xcoff_build_ldsyms): Set XCOFF_DEF_REGULAR flag when defining
global linkage code. If an undefined function descriptor is
exported, arrange to define it. Warn about any other undefined
exported symbol.
(_bfd_xcoff_bfd_final_link): Write out the descriptor section.
(xcoff_write_global_symbol): Create a function descriptor when
necessary.
* bfd-in.h (bfd_xcoff_size_dynamic_sections): Update declaration.
* bfd-in2.h: Rebuild.
Ken Raeburn [Thu, 9 Nov 1995 09:14:01 +0000 (09:14 +0000)]
* configure.in: Added new option --with[out]-mmap. Set USE_MMAP if it is given
and mmep is available.
* acconfig.h: Undef USE_MMAP.
* configure, config.in: Regenerated.
* aoutx.h: If USE_MMAP is not defined, do all symbol and string table handling
the old way.
(aout_get_external_symbols): Don't complain if last byte of string table is
nonzero.
* libbfd.c [HAVE_MADVISE]: Include sys/types.h and sys/mman.h.
(bfd_free_window) [! USE_MMAP]: Don't define.
(bfd_get_file_window, _bfd_generic_get_section_contents_in_window)
[! USE_MMAP]: Abort.
* coff-a29k.c (a29k_reloc): Change handling of R_IREL reloc to be
compatible with AMD generated COFF files. Try to support both AMD
and GNU formats simultaneously.
(coff_a29k_relocate_section): Likewise.
* config/tc-a29k.c (md_operand): Handle $float, $double, and
$extend. Based on code from Eric Freudenthal
<[email protected]>.
* config/tc-a29k.h (LEX_DOLLAR): Define.
* read.c (LEX_DOLLAR): Define if not defined.
(lex_type): Use LEX_DOLLAR.
* libbfd.c (bfd_get_file_window): Change writable parameter from
int to boolean; update all callers. Pass MAP_SHARED if not
writable--it's required on Solaris. Cast fprintf argument to
avoid warning.
* bfd-in.h (bfd_get_file_window): Update declaration.
* bfd-in2.h: Rebuild.
* aoutx.h: Update calls to bfd_get_file_window.
* xcofflink.c (XCOFF_DEF_DYNAMIC): Rename from XCOFF_REF_DYNAMIC.
Change all uses.
(xcoff_swap_ldhdr_in): New static function.
(xcoff_swap_ldsym_in): New static function.
(xcoff_link_add_symbols): If we just created a descriptor, pass it
to _bfd_generic_link_add_one_symbol, to save a hash lookup.
(xcoff_link_add_dynamic_symbols): Rewrite to read .loader symbols
rather than normal symbol table.
(xcoff_mark): When considering called symbols, check whether the
descriptor is from a dynamic object, rather than the symbol
itself.
(xcoff_build_ldsyms): Likewise.
* libbfd.c (bfd_get_file_window): Change return type to boolean.
Cast realloc and malloc return values. If malloc or realloc fail,
set bfd_error_no_memory.
* bfd-in.h (bfd_get_file_window): Change type to boolean.
* bfd-in2.h: Rebuild.
* config/mips/vr4300.mt: Added simulator to default VR4300 build.
NOTE: Only the BigEndian build has the simulator included, because a
LittleEndian simulator is not constructed at the moment (it should
work, but has not been tested fully).
Initial check-in of the MIPS simulator. Work still needs to be done on
the run-time support code (interp.c) to provide better tracing, and
also to add profiling and architecture specific support. At the moment
the simulator has a fixed size, fixed address memory area, and
simulates a subset of the IDT monitor calls (enough to execute test
programs).
The other major feature (could even be a bug) is that the simulator
makes use of the GCC "long long" extension. Work has been started to
make this a build configuration option... but there is still a lot of
this to be done.
Stu Grossman [Wed, 8 Nov 1995 00:33:16 +0000 (00:33 +0000)]
* remote-mips.c (mips_initialize): Fix brain damage found by
Jamie. Basically had case statement in the wrong place...
* (mips_load): Remove unnecessary `db tty0' command. It's all
handled by mips_initialize now.
* configure: Default ${build} correctly. Avoid picking up extra
spaces when reading CC and CXX from Makefile. When doing a
Canadian Cross, use plausible default values for numerous
variables.
* configure.in: When doing a Canadian Cross, don't try to
configure tools whose configure script can't handle it.
* ecofflink.c (_bfd_ecoff_locate_line): Keep looking through stabs
entries until both the line number address and the function name
address are too large.
PR 8364.
* remote-mips.c (mips_initialize): Updated to talk to VR4300 RISQ
monitor board.
The mips_initialize() code seemed to have a "funny" switch
statement. I also removed an unnecessary switch into debug monitor
mode just before the call to mips_initialize() in mips_load().
* configure.in: Add test for whether free must be declared.
* sysdep.h: Declare free if necessary. Include "ansidecl.h".
* acconfig.h: Explain NEED_DECLARATION_FREE.
* configure, config.in: Rebuild.
* ld-shared/shared.exp: Run tests on rs6000*-*-aix* and
powerpc*-*-aix*. Add code to create appropriate exports files,
and pass appropriate compilation flags, and test against
appropriate expected output.
* ld-shared/xcoff.dat: New file.
* ld-shared/main.c: Put #ifndef XCOFF_TEST around tests that can
not be linked using XCOFF. Use shlib_shlibvar1 as a sample
function pointer, rather than shlib_mainvar.
* ld-shared/sh1.c: Likewise.
* ld-shared/shared.dat: Update for change from using shlib_mainvar
to using shlib_shlibvar1.
* ld-shared/sun4.dat: Likewise.
* xcofflink.c (xcoff_link_add_symbols): Set the alignment power of
the created .tc section to 2.
(xcoff_mark): Don't keep a .loader reloc for a call to an
undefined symbol when creating a shared library.
(xcoff_build_ldsyms): When creating a shared library, generate
global linkage code for a call to an undefined symbol.
Moved the "sim_open" call to after the callback initialisation. This
is to allow the simulator initialisation code to make use of the host
callback facilities. Had discussed this with sac, and it should be a
harmless change since none of the other simulators really make use of
the call, and it does not return a result.
Ken Raeburn [Mon, 6 Nov 1995 10:08:03 +0000 (10:08 +0000)]
Changes for mmap; details in change log.
Added some new interfaces, and a new entry in the target vector. Under the new
interfaces, mmap will be used if available, otherwise malloc/seek/read, as
before. Old interfaces all still intact.
Most configurations (including all used by "--enable-targets=all") simply
changed to call the default routine for that entry in the target vector. I
might've missed some targets only included in special configurations.
Support for a.out symbol and string table reading now goes through new
interfaces, and will use mmap when available.
Linker hooks (e.g., avoiding reallocation under malloc) not ready yet.
Ken Raeburn [Mon, 6 Nov 1995 02:51:26 +0000 (02:51 +0000)]
* Makefile.in (CFILES): Rebuild list from BFD_LIBS, ALL_MACHINES,
BFD32_BACKENDS, BFD64_BACKENDS.
(.dep1): Put mkdep output in a separate file.
Rebuilt dependencies.
Fred Fish [Sun, 5 Nov 1995 05:40:15 +0000 (05:40 +0000)]
* core.c: Rename to corefile.c
* core.h: Rename to corefile.h
* inline.c: Include corefile.h, renamed from core.h.
* cpu.h: Include corefile.h, renamed from core.h
* vm.c: Include corefile.h, renamed from core.h
* Makefile.in (CPU_H): Change core.h to corefile.h
(vm.o): Change dependency to corefile.h
(LIB_SRC): Change core.c to corefile.c.
(LIB_OBJ): Change core.o to corefile.o.
(corefile.o): Change dependencies to corefile.c, corefile.h.
* corefile.c: Include corefile.h rather than core.h
* README.psim (KNOWN PROBLEMS): Change core.* references to corefile.*
references.
Jason Molenda [Sun, 5 Nov 1995 03:12:15 +0000 (03:12 +0000)]
* configure.in: add AC_HEADER_STDC check.
* mmalloc.h: check if STDC_HEADERS instead of __STDC__.
* mmprivate.h: check if STDC_HEADERS instead of __STDC__.
This change is necessary to build under AIX 3.2.5 w/ cc after Stan's
Oct 24 change. I'm not convinced that changing mmalloc.h in this way
is such a good thing--if a directory that doesn't use autoconf (or one
that DOES use autoconf but doesn't put all the defines on the command
line), and it includes mmalloc.h, the right thing won't happen.
I believe gdb is the only directory which uses mmalloc, though, so
it should be OK.
projects / binutils.git / log