Replace -export-dynamic with -Wl,-export-dynamic.

2010-12-04  H.J. Lu  <hongjiu.lu@intel.com>

* ld-elfvers/vers.exp: Replace -export-dynamic with
-Wl,-export-dynamic.

bfd: install plugin-api.h as needed

Since people need plugin-api.h in order to build linker plugins, install
the header with the other library headers when plugin support is enabled.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

PR ld/12277
* elflink.c (elf_link_output_extsym): Set bfd_error on symbol
and section errors.  Allow better translation of error messages.

* ldlang.c (lang_size_sections_1): Align lma using same alignment
as for vma.

   * configure.tgt: Accept microblaze*-*-linux* for linux target
       and microblaze*-*-* for  as target.

* Makefile.in: Use CC_FOR_BUILD to build opc2c.

PR gas/12282
* expr.c (expr_build_dot): Make a clone of the symbol to return if
needed.

*** empty log message ***

daily update

Fix a typo in comments.

2010-12-03  H.J. Lu  <hongjiu.lu@intel.com>

* lexsup.c (ld_options): Fix a typo in comments.

* dwarf.c: #include "bfd_stdint.h".
(do_gdb_index): New global.
(display_gdb_index): New function.
(dwarf_select_sections_by_names) Add "gdb_index".
(dwarf_select_sections_all): Set do_gdb_index.
(debug_displays): Add .gdb_index.
* dwarf.h (do_gdb_index): Declare.
* objdump.c (usage): Add gdb_index.
* readelf.c (usage): Add gdb_index.
(process_section_headers): Process ".gdb_index".
* doc/binutils.texi (readelf): Document gdb_index dump.
(objdump): Ditto.

*** empty log message ***

* mh-cygwin (LDFLAGS): Turn up stack allocation to 12MB.
(BOOT_LDFLAGS): Add matching stack size flag.
* mh-mingw (LDFLAGS): Likewise.
(BOOT_LDFLAGS): Likewise.

daily update

2010-11-29  Michael Snyder  <msnyder@msnyder-server.eng.vmware.com>

* gdb.base/callfuncs.exp: Test for skip_float_tests.
* gdb.base/call-sc.exp: Ditto.
* gdb.base/finish.exp: Ditto.
* gdb.base/return.exp: Ditto.
* gdb.base/return2.exp: Ditto.

* linespec.c (find_method): Move name canonicalization outside
the loop.
Be paranoid and use a cleanup.

2010-12-02  Jeff Johnston  <jjohnstn@redhat.com>

        * COPYING.NEWLIB: Add National Semiconductor notice.

bfd/
* elf32-arm.c (elf32_arm_link_hash_table): Remove sgot, sgotplt,
srelgot, splt, srelplt.
(create_got_section, elf32_arm_link_hash_table_create): Don't set them.
(elf32_arm_create_dynamic_sections): Likewise.  Use htab->root
fields instead.
(arm_type_of_stub): Use the root fields instead of the removed ones.
(cortex_a8_erratum_scan, elf32_arm_size_stubs): Likewise.
(bfd_elf32_arm_process_before_allocation): Likewise.
(elf32_arm_check_relocs, allocate_dynrelocs): Likewise.
(elf32_arm_size_dynamic_sections): Likewise.
(elf32_arm_output_arch_local_syms): Likewise.
(elf32_arm_final_link_relocate): Set sgot, splt and srelgot from the
htab fields instead of looking them up by name.  Consistently use
these local variables instead of htab fields.
(elf32_arm_finish_dynamic_symbol): Likewise sgot, splt and srel.
Use srelbss instead of looking it up by name.
(elf32_arm_finish_dynamic_sections): Use sgotplt, splt and srelplt
instead of looking up the symbols by name.  Use the root fields
instead of the removed ones.

bfd/
* elf32-arm.c (elf32_arm_relocs_copied): Delete.
(elf32_arm_link_hash_entry): Replace relocs_copied with dyn_relocs.
(elf32_arm_link_hash_newfunc): Update accordingly.
(elf32_arm_copy_indirect_symbol, elf32_arm_gc_sweep_hook): Likewise.
(elf32_arm_check_relocs, allocate_dynrelocs): Likewise.
(elf32_arm_readonly_dynrelocs): Likewise.
(elf32_arm_size_dynamic_sections): Likewise.

gas/
* symbols.c (S_FORCE_RELOC): Return true for indirect functions
even if !strict.
* expr.c (operand): Don't convert absolute symbols to constants
if S_FORCE_RELOC is true.
(expr): Only reduce subtractions between different symbols if
S_FORCE_RELOC is false for both of them.
* write.c (fixup_segment): Don't remove symbols if S_FORCE_RELOC
is true for them, regardless of their segment.

gas/testsuite/
* gas/i386/ifunc-2.s, gas/i386/ifunc-2.l: New test.
* gas/i386/ifunc-3.s, gas/i386/ifunc-3.d: Likeise.
* gas/i386/i386.exp: Run them.

* pdp11.c (aout_link_add_symbols): Ignore debug symbols.

Break long lines.

Re-indent plugin.h.

2010-12-01  H.J. Lu  <hongjiu.lu@intel.com>

* plugin.h: Re-indent.

*** empty log message ***

daily update

* symbols.h (dot_symbol): New declaration.
(dot_symbol_init): New prototype.
* symbols.c (dot_symbol): New variable.
(symbol_clone): Assert it's not dot_symbol being cloned.
(dot_symbol_init): New function.
(symbol_clone_if_forward_ref): Create a new temporary symbol
when trying to clone dot_symbol.
* expr.c (current_location): Refer to dot_symbol instead of
making a new temporary symbol.
* read.c (read_a_source_file): Update dot_symbol as we go.
* as.c (main): Call dot_symbol_init.

* symbols.c (symbol_clone_if_forward_ref): Don't limit cloning
to expr_section symbols; clone all equated symbols.  Clear
sy_resolving of the cloned copy.
* expr.c (operand): Only clone equated symbols on a final
(i.e. non-equated) reference.

* config/tc-mips.c (md_convert_frag): Remove a call to
S_GET_VALUE and use the result of resolve_symbol_value as the
value of the symbol processed in MIPS16 relaxation.

* dwarf_reader.h (class Sized_dwarf_line_info): Add
track_relocs_type_ field.
* dwarf_reader.cc (Sized_dwarf_line_info::Sized_dwarf_line_info):
Set track_relocs_type_.
(Sized_dwarf_line_info::process_one_opcode): Ignore the section
contents when using RELA relocs.
(Sized_dwarf_line_info::read_relocs): Add the reloc addend to
reloc_map_.
* reloc.cc (Track_relocs::next_addend): New function.
* reloc.h (class Track_relocs): Declare next_addend.

* dwarf2read.c (save_gdb_index_command): Add comment.

* testsuite/icf_virtual_function_folding_test.cc (class Bar): Add
virtual destructor.

* README: Update compilers known to work and fail.

* valops.c (value_assign): Returned value is never lazy.  If a
C++ class type is returned, fix incorrect enclosing type / embedded
offset.  If internal variable is returned, allocate new internalvar
value using value_of_internalvar.

* NEWS: Document changes in behavior of "print x = 0" and similar
expressions.

2010-12-01  Kai Tietz  <kai.tietz@onevision.com>

       PR binutils/11065
       * dlltool.c (dll_name_set_by_exp_name): New variable.
       (def_name): Allow setting of dll_name by .def file.
       (def_library): Likewise.
       (main): Set dll_name_set_by_exp_name, if dll_name is
       set indirect by exp_name.

*** empty log message ***

daily update

* lib/gdb.exp (gdb_test_sequence): Return result of gdb_expect_list.
(gdb_expect_list): Fix spelling errors in comments.
* gdb.arch/system-gcore.exp: Use gdb_test_sequence instead of
send_gdb/gdb_expect_list where applicable.
* gdb.base/call-ar-st.exp: Ditto.
* gdb.base/funcargs.exp: Ditto.
* gdb.base/gcore.exp: Ditto.
* gdb.base/page.exp: Ditto.
* gdb.base/sigaltstack.exp: Ditto.
* gdb.base/siginfo.exp: Ditto.
* gdb.base/sigstep.exp: Ditto.
* gdb.base/trace-commands.exp: Ditto.

* ld/testsuite/ld-scripts/script.exp: Remove full path from REGION_ALIAS test
names.

Add sparc64-rtems.

*** empty log message ***

* python/lib/gdb/printing.py (register_pretty_printer): Change
printer-name:subprinter-name to printer-name;subprinter-name.
* python/lib/gdb/command/pretty_printers.py (parse_printer_regexps):
Ditto.
(InfoPrettyPrinter, EnablePrettyPrinter, DisablePrettyPrinter): Ditto.

doc/
* gdb.texinfo (Pretty-Printer Introduction): Change
printer-name:subprinter-name to printer-name;subprinter-name.

testsuite/
* gdb.python/py-pp-maint.exp: Change printer-name:subprinter-name to
printer-name;subprinter-name.

daily update

gdb
* opencl-lang.c (lval_func_check_synthetic_pointer): New
function.
* value.h (struct lval_funcs) <indirect, check_synthetic_pointer>:
New fields.
(value_bits_synthetic_pointer): Declare.
* value.c (value_bits_synthetic_pointer): New function.
* valprint.c (valprint_check_validity): Handle synthetic
pointers.
* valops.c (value_ind): Use new 'indirect' lval_funcs method.
* valarith.c (value_ptradd): Use set_value_component_location.
* p-valprint.c (pascal_object_print_value_fields): Handle
synthetic pointers.
* jv-valprint.c (java_print_value_fields): Handle synthetic
pointers.
* dwarf2read.c (dwarf_stack_op_name): Add
DW_OP_GNU_implicit_pointer.
(dwarf2_fetch_die_location_block): Add get_frame_pc, baton
arguments.  Handle location lists.
(fill_in_loclist_baton): New function.
(dwarf2_symbol_mark_computed): Use it.
* dwarf2loc.h (dwarf2_find_location_expression): Declare.
(dwarf2_fetch_die_location_block): Add get_frame_pc, baton
arguments.
* dwarf2loc.c (dwarf2_find_location_expression): Rename from
find_location_expression.  No longer static.  Update all callers.
(dwarf_expr_frame_pc): New function.
(per_cu_dwarf_call): Add get_frame_pc, baton arguments.  Update
all callers.
(struct piece_closure) <per_cu>: New field.
(allocate_piece_closure): Add per_cu argument.
(read_pieced_value): Handle DWARF_VALUE_IMPLICIT_POINTER.
(check_pieced_value_bits): Remove validity argument, add check_for
argument.  Handle DWARF_VALUE_IMPLICIT_POINTER.
(check_pieced_value_validity, check_pieced_value_invalid):
Update.
(check_pieced_synthetic_pointer): New function.
(get_frame_address_in_block_wrapper): New function.
(indirect_pieced_value): New function.
(pieced_value_funcs): Update.
(invalid_synthetic_pointer): New function.
(dwarf2_evaluate_loc_desc_full): Rename from
dwarf2_evaluate_loc_desc.  Add byte_offset argument.
(dwarf2_evaluate_loc_desc): Rewrite.
(dwarf2_loc_desc_needs_frame): Set new field on context.
(get_ax_pc): New function.
(disassemble_dwarf_expression): Handle
DW_OP_GNU_implicit_pointer.
* dwarf2expr.h (enum dwarf_value_location)
<DWARF_VALUE_IMPLICIT_POINTER>: New constant.
(struct dwarf_expr_context) <get_frame_pc>: New field.
(struct dwarf_expr_piece) <v.ptr>: New field.
* dwarf2expr.c (add_piece): Handle DWARF_VALUE_IMPLICIT_POINTER.
(execute_stack_op): Handle DW_OP_GNU_implicit_pointer.
* dwarf2-frame.c (no_get_frame_pc): New function.
(execute_stack_op): Set new field on context.
* cp-valprint.c (cp_print_value_fields): Handle synthetic
pointers.
gdb/testsuite
* gdb.dwarf2/implptr.exp: New file.
* gdb.dwarf2/implptr.c: New file.
* gdb.dwarf2/implptr.S: New file.

merge from gcc

2010-11-29  Phil Muldoon  <pmuldoon@redhat.com>

        PR python/12199

* python/py-breakpoint.c (bppy_delete_breakpoint): New function.

2010-11-29  Phil Muldoon  <pmuldoon@redhat.com>

        PR python/12199

* gdb.python/py-breakpoint.exp: Test the delete method.

2010-11-29  Phil Muldoon  <pmuldoon@redhat.com>

        PR python/12199

* gdb.texinfo (Breakpoints In Python): Document "delete" method.

* configure.ac: Move comment to remove extra space in last argument
of GCC_TARGET_TOOL.

*** empty log message ***

daily update

gdb/
Fix step_resume_breakpoint unsaved during an infcall.
* gdbthread.h (struct thread_control_state): Move here field
step_resume_breakpoint ...
(struct thread_info): ... from here.
* infrun.c (save_infcall_control_state): Reset
control.step_resume_breakpoint to NULL.
(restore_infcall_control_state, discard_infcall_control_state): Delete
control.step_resume_breakpoint.
* arm-linux-tdep.c, infrun.c, thread.c: Update all the references to
the moved field.

gdb/testsuite/
Fix step_resume_breakpoint unsaved during an infcall.
* gdb.base/step-resume-infcall.exp: New file.
* gdb.base/step-resume-infcall.c: New file.

gdb/
Rename and move inferior_thread_state and inferior_status.
* gdbthread.h (struct thread_control_state): New struct, move fields
step_range_start, step_range_end, step_frame_id, step_stack_frame_id,
trap_expected, proceed_to_finish, in_infcall, step_over_calls,
stop_step and stop_bpstat here from struct thread_info.
(struct thread_suspend_state): New struct, move field stop_signal here
from struct thread_info.
(struct thread_info): Move the fields above from this struct.
* inferior.h: Move the inferior_thread_state and inferior_status
declarations comment to their definitions at infrun.c.
(struct inferior_control_state): New struct, move field stop_soon from
struct inferior here.
(struct inferior_suspend_state): New empty struct.
(struct inferior): New fields control and suspend.  Move out field
stop_soon.
* infrun.c (struct inferior_thread_state): Rename to ...
(infcall_suspend_state): ... here.  Replace field stop_signal by
fields thread_suspend and inferior_suspend.
(save_inferior_thread_state): Rename to ...
(save_infcall_suspend_state): ... here.  New variable inf.  Update the
code for new fields.
(restore_inferior_thread_state): Rename to ...
(restore_infcall_suspend_state): ... here.  New variable inf.  Update
the code for new fields.
(do_restore_inferior_thread_state_cleanup): Rename to ...
(do_restore_infcall_suspend_state_cleanup): ... here.
(make_cleanup_restore_inferior_thread_state): Rename to ...
(make_cleanup_restore_infcall_suspend_state): ... here.
(discard_inferior_thread_state): Rename to ...
(discard_infcall_suspend_state): ... here.
(get_inferior_thread_state_regcache): Rename to ...
(get_infcall_suspend_state_regcache): ... here.
(struct inferior_status): Rename to ...
(struct infcall_control_state): ... here.  Replace fields
step_range_start, step_range_end, step_frame_id, step_stack_frame_id,
trap_expected, proceed_to_finish, in_infcall, step_over_calls,
stop_step, stop_bpstat and stop_soon by fields thread_control and
inferior_control.
(save_inferior_status): Rename to ...
(save_infcall_control_state): ... here.  Update the code for new
fields.
(restore_inferior_status): Rename to ...
(restore_infcall_control_state): ... here.  Update the code for new
fields.
(do_restore_inferior_status_cleanup): Rename to ...
(do_restore_infcall_control_state_cleanup): ... here.
(make_cleanup_restore_inferior_status): Rename to ...
(make_cleanup_restore_infcall_control_state): ... here.
(discard_inferior_status): Rename to ...
(discard_infcall_control_state): ... here.
* alpha-tdep.c, breakpoint.c, dummy-frame.c, dummy-frame.h,
exceptions.c, fbsd-nat.c, gdbthread.h, infcall.c, infcmd.c,
inferior.c, inferior.h, infrun.c, linux-nat.c, mi/mi-interp.c,
mips-tdep.c, procfs.c, solib-irix.c, solib-osf.c, solib-spu.c,
solib-sunos.c, solib-svr4.c, thread.c, windows-nat.c: Update all the
references to the moved fields and renamed functions.

*** empty log message ***

daily update

* bootstrap-lto.mk (BOOT_ADAFLAGS): Delete.

*** empty log message ***

daily update

* dwarf.c (decode_location_expression, display_debug_frames): Move
code handling DW_EH_PE_pcrel into..
(get_encoded_value): ..here.

PR other/46020
* configure.ac (CXX_FOR_TARGET): Add -funconfigured-libstdc++-v3.
* Makefile.def (CXX_FOR_TARGET): Removed from flags_to_pass.
* Makefile.tpl (CXX_FOR_TARGET_FLAG_TO_PASS): New.
(BASE_FLAGS_TO_PASS): Use it.
* configure: Rebuilt.
* Makefile.in: Rebuilt.

gdb/
* infcall.c (struct inferior_thread_state) <siginfo_gdbarch>: Fix up
the comment.
(save_inferior_thread_state): Reset tp->stop_signal.

Fix ChangeLog entry.

gdb/
* infcall.c (caller_state_cleanup): Remove variable caller_state_cleanup.

update japanese translation

*** empty log message ***

2010-11-25  Marc Khouzam  <marc.khouzam@ericsson.com>

PR breakpoints/12217

* breakpoint.c (is_marker_spec): Check for NULL.

daily update

* breakpoint.h: No longer include python.h or python-internal.h.
(struct breakpoint_object): Forward declare.
* defs.h (PyObject) [!HAVE_PYTHON]: Don't define.
* varobj.c (PyObject) [!HAVE_PYTHON]: Define.
* python/py-breakpoint.c (build_bp_list): Cast py_bp_object to
PyObject pointer.
(gdbpy_breakpoint_created): Remove casts around py_bp_object
accesses.

* ld/testsuite/ld-arm/arm-elf.exp (armeabitests): Replace --defsym argument in
jump-reloc-veneers* tests with --section-start .text=0x8000.

2010-11-25  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* config/tc-s390.c (current_cpu): Initialize with latest CPU.
(init_default_arch): Default to z/Architecture mode if CPU provides it.
Remove the check setting the CPU default.

2010-11-25  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* opcode/s390.h (enum s390_opcode_cpu_val): Add S390_OPCODE_MAXCPU.

Update translations

* scripttempl/elf.sc: Group .text.exit, text.startup and .text.hot
sections.

PR gas/12264
* compress-debug.c: Include config.h first.

Add rs6000-core.c/.o to the list of optional backend files...

This makes sure that rs6000-core.c gets recompiled whenever any of its
dependencies get changed.

bfd/ChangeLog:

        * Makefile.am (OPTIONAL_BACKENDS): Add rs6000-core.lo.
        (OPTIONAL_BACKENDS_CFILES): Add rs6000-core.c.
        * Makefile.in: Regenerate.

Build failure on ppc-aix (bfd_elf_get_obj_attr_int is undefined)

bfd_uses_spe_extensions should only be used when BFD has been built
with ELF support.  The typical way of checking that in GDB is to use
the HAVE_ELF macro.

gdb/ChangeLog:

        * rs6000-tdep.c (bfd_uses_spe_extensions): Use bfd_elf_get_obj_attr_int
        only if HAVE_ELF is defined.

*** empty log message ***

daily update

gas: bfin: convert to new doc style

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

gdb/
Code cleanup.
* infrun.c (struct inferior_status) <stepping_over_breakpoint>
(save_inferior_status) <stepping_over_breakpoint>
(restore_inferior_status) <stepping_over_breakpoint>: Rename to
trap_expected.
(struct inferior_status) <step_resume_break_address>: Remove.

gdb/
* infrun.c (struct inferior_status, save_inferior_status)
(restore_inferior_status): Group and reorder the fields, comment their
groups.

gdb/
Code cleanup.
* dictionary.c
(struct dict_vector) <iter_match_first, iter_match_next>
(iter_match_first_hashed, iter_match_next_hashed)
(iter_match_first_linear, iter_match_next_linear)
(dict_iter_match_first, dict_iter_match_next, iter_match_first_hashed)
(iter_match_next_hashed, iter_match_first_linear)
(iter_match_next_linear): Use symbol_compare_ftype.
* dictionary.h: Include symfile.h.
(dict_iter_match_first, dict_iter_match_next): Use
symbol_compare_ftype.
* dwarf2read.c (dw2_map_matching_symbols): Likewise.
* psymtab.c (match_partial_symbol, match_partial_symbol, map_block)
(map_matching_symbols_psymtab): Likewise.
* symfile.h (symbol_compare_ftype): New typedef.
(struct quick_symbol_functions) <map_matching_symbols): Use
symbol_compare_ftype.

2010-11-24  Edjunior Machado  <emachado@br.ibm.com>

* gdb.asm/powerpc.inc: Use 'sys_exit' on gdbasm_exit0 macro.

2010-11-24  Edjunior Machado  <emachado@br.ibm.com>

* gdb.asm/asm-source.exp: Add new target "powerpc64"; expect leading `.'
on ppc64's symbols; "finish" and "return" commands must accept to
show the caller line again as well as the line after.
* gdb.asm/powerpc64.inc: New file.

*** empty log message ***

PR ld/12253
* elf-eh-frame.c (_bfd_elf_write_section_eh_frame): Correct
DW_EH_PE_datarel handling.  Truncate .eh_frame_hdr address to
ptr_size.

daily update

* lib/gdb.exp (gdb_test_sequence): New function.
(gdb_expect_list): Add verbose -log call for each pattern.
* gdb.base/signals.exp (test_handle_all_print): Call it.
Reduce timeout increment from 6 minutes to 1 minute.
* gdb.server/ext-run.exp: Call it.

include/opcode/
* mips.h: Fix previous commit.

2010-11-23  Sterling Augustine  <sterling@tensilica.com>

* doc/as.texinfo: Refer to and include c-xtensa.texi for Xtensa
options.  Move Xtensa options to proper alphabetical location.
* doc/c-xtensa.texi: Add markup for use in manpage generation.

Properly check default linker.

2010-11-23  H.J. Lu  <hongjiu.lu@intel.com>

PR binutils/12258
* configure.ac: Correct comments for --enable-gold/--enable-ld.
Properly check default linker.
* configure: Regnerated.

* mips.h (INSN_CHIP_MASK): Update according to INSN_LOONGSON_3A.
(INSN_LOONGSON_3A): Clear bit 31.

* elfxx-mips.c (mips_set_isa_flags): Move bfd_mach_loongson_3a
after bfd_mach_mips_sb1.

* config/tc-mips.c (mips_cpu_info_table): Move loongson3a after sb1.

Refer to and include c-i386.texi for i386 options.

2010-11-23  H.J. Lu  <hongjiu.lu@intel.com>

* doc/as.texinfo: Refer to and include c-i386.texi for i386
options.

* doc/c-i386.texi: Add markup for use in manpage generation.

gdb
* configure: Rebuild.
* configure.ac (READLINE_TEXI_INCFLAG): New subst.
gdb/doc
* gdb.texinfo (Top): Check SYSTEM_READLINE.
(Editing): Likewise.
(Command History): Likewise.
(TUI Keys): Likewise.
(Bug Reporting): Conditionally include rluser.texi and
inc-history.texinfo.
* Makefile.in (READLINE_TEXI_INCFLAG): New variable.
(GDB_DOC_SOURCE_INCLUDES): Add comment.
(GDBvn.texi): Set SYSTEM_READLINE when appropriate.
(gdb.dvi): Use READLINE_TEXI_INCFLAG.
(gdb.pdf): Likewise.
(gdb.info): Likewise.
(gdb/index.html): Likewise.

gdb
* Makefile.in (SUBDIRS): Add doc.
* configure: Rebuild.
* configure.ac: Don't configure in doc.  Create doc/Makefile.
gdb/doc
* Makefile.in (Makefile): Run ../config.status.
(distclean): Update.
* configure: Remove.
* configure.ac: Remove.

* configure.ac: For --enable-gold, handle value `default' instead of
`both*'.   New configure option --{en,dis}able-ld.

ld, gold/
* configure.ac: For --enable-gold, handle value `default' instead of
`both*'. Always install ld as ld.gold, install as ld if gold is
the default.

2010-11-23  Phil Muldoon  <pmuldoon@redhat.com>

        PR python/12212

* python/py-inferior.c (find_thread_object): Check if PIDGET
returns 0.

2010-11-23  Phil Muldoon  <pmuldoon@redhat.com>

        PR python/12212

* gdb.python/python.exp: Check that selected_thread raises an
error when no inferior is loaded.

sim: profile: fix building with --disable-sim-profile

When the sim is configured with profile support disabled, the build fails:
./../common/sim-profile.c: In function 'profile_option_handler':
./../common/sim-profile.c:337:6: warning: implicit declaration of function 'PROFILE_PC_FREQ'
./../common/sim-profile.c:337:6: error: lvalue required as left operand of assignment
./../common/sim-profile.c:351:6: warning: implicit declaration of function 'PROFILE_PC_NR_BUCKETS'
./../common/sim-profile.c:351:6: error: lvalue required as left operand of assignment
./../common/sim-profile.c:381:6: warning: implicit declaration of function 'PROFILE_PC_SHIFT'
./../common/sim-profile.c:381:6: error: lvalue required as left operand of assignment
./../common/sim-profile.c:405:8: warning: implicit declaration of function 'PROFILE_PC_START'
./../common/sim-profile.c:405:8: error: lvalue required as left operand of assignment
./../common/sim-profile.c:406:8: warning: implicit declaration of function 'PROFILE_PC_END'
./../common/sim-profile.c:406:8: error: lvalue required as left operand of assignment
./../common/sim-profile.c: In function 'profile_uninstall':
./../common/sim-profile.c:1299:7: warning: implicit declaration of function 'PROFILE_INSN_COUNT'
./../common/sim-profile.c:1299:37: warning: comparison between pointer and integer
./../common/sim-profile.c:1300:2: warning: passing argument 1 of 'zfree' makes pointer from integer without a cast
../common/sim-utils.h:30:6: note: expected 'void *' but argument is of type 'int'
make[2]: *** [sim-profile.o] Error 1

So add some stubs similar to how some of the other subsystems are
stubbed out so things build correctly.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sim: cast away hw/device differences

When building with device and hw support, the sim-core code generates a
lot of build time warnings such as:

./../common/sim-core.c: In function 'sim_core_map_attach':
./../common/sim-core.c:198:7: warning: passing argument 1 of 'device_error' from incompatible pointer type
../common/sim-core.h:347:6: note: expected 'struct device *' but argument is of type 'struct hw *'
./../common/sim-core.c:235:7: warning: passing argument 1 of 'device_error' from incompatible pointer type
../common/sim-core.h:347:6: note: expected 'struct device *' but argument is of type 'struct hw *'

In reality, these two structures get cast back and forth in the core
code already and so are "compatible".  So tweak the three functions
that generate all of these warnings to include the casts automatically.
I know this isn't exactly clean, but the current device/hw ifdef
approach is full of landmines itself and I'm not entirely sure how
to unscrew it.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

Delete redundant variable in ada_template_to_fixed_record_type_1.

gdb/ChangeLog:

        * ada-lang.c (ada_template_to_fixed_record_type_1): Delete bit_incr.

[Ada] array pointers encoded as typedef to fat pointer

A recent change in check_typedef caused the following regression,
considering:

        type String_Access is access String;
        S1 : String_Access := null;

Trying to print S1, we get:

        (gdb) print s1
        $1 = (string) (null)

The type name is wrong.  We were expecting:

        (gdb) print s1
        $1 = (string_bug.string_access) (null)

The extensive comment in this patch explains how pointers to arrays
are encoded when the array is a "fat pointer".  What happened is
that the change in check_typedef broke a type-equality test that
we were performing.   The test really was on the fringe, but it
turns out that, for what we're doing, we're really only interested
in the main-type portion of the type.

The patch adjust the check accordingly.

gdb/ChangeLog:

        * ada-lang.c (ada_to_fixed_type): Expand function documentation.
        Return the original type if the main type portions match rather
         than when the type themselves match.

GDB SEGV while trying to print uninitialize variant record

We have a variant record whose value is defined as follow:

   type Discriminant_Record (Num1, Num2,
                             Num3, Num4 : Natural) is record
      Field1 : My_Record_Array (1 .. Num2);
      Field2 : My_Record_Array (Num1 .. 10);
      Field3 : My_Record_Array (Num1 .. Num2);
      Field4 : My_Record_Array (Num3 .. Num2);
      Field5 : My_Record_Array (Num4 .. Num2);
   end record;
   Dire : Discriminant_Record (1, 7, 3, 0);

However, we're trying to print "Dire" before it is elaborated.
This is common if one breaks on a function and then starts doing
"info locals" for instance.

What happens is that GDB reads bogus values for fields Num1 to
Num4, and deduces a bogus (ginormouos) size for component
"Field1".  The length is so large that it then later causes
an overflow in the overall record length computation. Things
go downhill from there, because length(field1) > length(record).
So, when after we've fetched the value of the record based on
the computed size, we crash trying to access unallocated memory
when accessing field1...

The first fix we can do is to check the size of the field
against the maximum object size.  If it exceeds that size,
then we know the record will also exceed that size...

gdb/ChangeLog:

        * ada-lang.c (ada_template_to_fixed_record_type_1):
        For dynamic fields, check the field size against the maximum
        object size.

support for mips-irix on-stack trampolines

On mips-irix, the debugger has trouble stepping over the following
line of code:

    S: Shape'Class := R;   <<<<---- STOP here

Here is what happens:

    (gdb) n
    warning: GDB can't find the start of the function at 0x7fff2bd8.

        GDB is unable to find the start of the function at 0x7fff2bd8
    and thus can't determine the size of that function's stack frame.
    This means that GDB may be unable to access that stack frame, or
    the frames below it.
        This problem is most likely caused by an invalid program counter or
    stack pointer.
        However, if you think GDB should simply search farther back
    from 0x7fff2bd8 for code which looks like the beginning of a
    function, you can increase the range of the search using the `set
    heuristic-fence-post' command.
    0x7fff2bd8 in ?? ()

The program does in fact jump to this code location, which is a trampoline
located on the stack (there is an implicit call to a routine internally
generated by the Ada expander). As it is on the stack, GDB is naturally
 unable to find the bounds of the current function, or any debugging
information, and is thus unable to continue.

This patch adds support for this sort of trampoline.

gdb/ChangeLog:

        * mips-irix-tdep.c (mips_irix_n32_stack_tramp_frame_init): New
        function.
        (mips_irix_n32_stack_tramp_frame): New static global.
        (mips_irix_init_abi): Add mips_irix_n32_stack_tramp_frame to
        list of unwinder.

[Ada/tasks] Optimize ATCB lookups

Type symbols that are used to decode the Ada task control block may be
duplicated in any unit that uses a tasking feature; an Ada lookup of
these symbols would collect all the duplicated instances and could
provoke the load of a huge amount of debug symbols in the process.

However, when decoding the task information, we are only interested in
one instance; and we have the full name of the symbol anyway. So use
a C lookup instead of an Ada lookup.

2010-11-05  Jerome Guitton  <guitton@adacore.com>

* ada-tasks.c (get_tcb_types_info): Use C lookups to get
ATCB symbols.