Default text reordering fix with a flag to turn it off.

2013-01-24  Sriraman Tallam  <tmsriram@google.com>

* layout.cc (Layout::layout): Check for option text_reorder.
(Layout::make_output_section): Ditto.
* options.h (text_reorder): New option.
* output.cc (Input_section_sort_compare): Remove special ordering
of section names.
(Output_section::
 Input_section_sort_section_name_special_ordering_compare::
 operator()): New function.
(Output_section::sort_attached_input_sections): Use new sort function
for .text.
* output.h (Input_section_sort_section_name_special_ordering_compare):
New struct.
* testsuite/Makefile.am (text_section_grouping): Test option
--no-text-reorder
* testsuite/Makefile.in: Regenerate.
* testsuite/text_section_grouping.sh: Check order of functions without
default text reordering.

* ada-lang.h (ada_decode_symbol): Make return type const.
* ada-lang.c (ada_decode_symbol): Likewise.

* v850.h: Add e3v5 support.

Add support for V850E3V5 architecture

2013-01-24  Hafiz Abid Qadeer  <abidh@codesourcery.com>

PR gdb/13443
* gdb.mi/mi-var-block.exp: Make test messages unique.

*** empty log message ***

daily update

* gdb.dwarf2/dw2-error.exp: Pass test name to "file" test.

* gdb.cp/converts.cc (main): Initialize 'a'.

* linespec.c (find_linespec_symbols): Make static.

2013-01-23  Tiago Stürmer Daitx  <tdaitx@linux.vnet.ibm.com>

* ppc-sysv-tdep.c (ppc64_sysv_abi_return_value): Set correct
type on float conversion for complex type.

Add a new class gdb.Architecture which exposes GDB's
internal representation of architecture via GDB Python API.

* Makefile.in: Add entries corresponding to the new file
python/py-arch.c.
* NEWS (Python Scripting): Add entries for the new class
gdb.Architecture and the new method gdb.Frame.architecture.
* python/py-arch.c: Implement gdb.Architecture class.
* python/py-frame.c (frapy_arch): Implement the method
gdb.Frame.architecture().
(frame_object_methods): Add 'architecture' to the method table.
* python/python-internal.h: Add declarations of new utility
functions.
* python/python.c (_initialize_python): Initialize
gdb.Architecture class.

* doc/gdb.texinfo (Architectures In Python): New sub-sub-section
describing the gdb.Architecture class.
(Frames In Python): Add description about the new method
gdb.Frame.architecture().

* testsuite/gdb.python/frame.exp: Add a test for
gdb.Frame.architecture() method.

Work around binutils/15021.
* dwarf2read.c (dwarf2_per_cu_data): Split imported_symtabs and
type_unit_group out of union s.  All uses updated.
(read_index_from_section): Watch for index version 8.
(follow_die_sig): If using .gdb_index version <= 7, record the TU as
an imported symtab.
(write_psymtabs_to_index): Increment version number to 8.

doc/
* gdb.texinfo (Index Section Format): Document .gdb_index version 8.

PR ld/15041
* scripttempl/pep.sc (.pdata): Only accept .pdata sections.
(.xdata): Similarly.
(.debug_frame): Similarly.

* elf32-metag.c: Error on HIADDR16/LOADDR16 in shared link.

PR gas/15039
* config/tc-avr.c: Include dwarf2dbg.h.

PR ld/15037
* scripttempl/avr.sc (.eeprom): Keep it.

* Makefile.am (ALL_64_EMULATION_SOURCES): Add eelf64rdos.c.
(eelf64rdos.c): New rule.
* emulparams/elf64rdos.sh: New file.
* configure.tgt (x86_64-*-rdos*): Use above.
* Makefile.in: Regenerate.

* config.bfd (x86_64-*-rdos*): Remove targ_selvecs.

2013-01-23  Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>

* readelf.c: Add strings for NT_S390_LAST_BREAK and
NT_S390_SYSTEM_CALL.

*** empty log message ***

daily update

ld/
* emultempl/elf32.em (gld${EMULATION_NAME}_before_parse): Set
new_dtags to TRUE for *-*-nacl* targets.

If a breakpoint is not user visible, then there's no point in
bothering the frontend about it...  This is the exact same check MI
does.

I also smoke tested Emacs 23 in gud-gdb mode, both annotations=2
and annotations=3.  I didn't notice anything break.

gdb/
2013-01-22  Pedro Alves  <palves@redhat.com>

* annotate.c (breakpoint_changed): Skip if breakpoint is not
user-visible.

gdb/testsuite/
2013-01-22  Pedro Alves  <palves@redhat.com>

* gdb.base/annota1.exp (signal sent): No longer expect
breakpoints-invalid.
* gdb.cp/annota2.exp (continue until exit)
(watch triggered on a.x): Ditto.

All annotate_breakpoints_changed calls are along-side
observer_notify_breakpoints_changed calls.  All, except the
init_raw_breakpoint one.  But that one is actually wrong.  The
breakpoint is being constructed at that point, and hasn't been placed
on the breakpoint chain yet.  It would be better placed in
install_breakpoint, and I actually started out that way.  But once the
annotate_breakpoints_changed are parallel to the observer calls, we
can fully move annotations to observers too.

One issue is that this changes the order of annotations a bit.
Before, we'd emit the annotation, and after call "mention()" on the
breakpoint (which prints the breakpoint number, etc.).  But, we call
the observers _after_ mention is called, so the annotation output will
change a little:

void
install_breakpoint (int internal, struct breakpoint *b, int update_gll)
{
  add_to_breakpoint_chain (b);
  set_breakpoint_number (internal, b);
  if (is_tracepoint (b))
    set_tracepoint_count (breakpoint_count);
  if (!internal)
    mention (b);
  observer_notify_breakpoint_created (b);

  if (update_gll)
    update_global_location_list (1);
}

I believe this order doesn't really matter (the frontend needs to wait
for the prompt anyway), so I just adjust the expected output in the
tests.  Emacs in annotations mode doesn't seem to complain.  Couple
that with the previous patch that suppressed duplicated annotations,
and, the fact that some annotations calls were actually missing (were
we do have observer calls), more changes to the tests are needed
anyway.

Tested on x86_64 Fedora 17.

gdb/
2013-01-22  Pedro Alves  <palves@redhat.com>

* annotate.c (annotate_breakpoints_changed): Rename to ...
(annotate_breakpoints_invalid): ... this.  Make static.
(breakpoint_changed): Adjust.
(_initialize_annotate): Always install the observers.  Install a
"breakpoint_created" observer.
* annotate.h (annotate_breakpoints_changed): Delete declaration.
* breakpoint.c (set_breakpoint_condition)
(breakpoint_set_commands, do_map_commands_command)
(init_raw_breakpoint, clear_command, set_ignore_count)
(enable_breakpoint_disp): No longer call
annotate_breakpoints_changed.

gdb/testsuite/
2013-01-22  Pedro Alves  <palves@redhat.com>

* gdb.base/annota1.exp (breakpoints_invalid): New variable.
Adjust tests to breakpoints-invalid changes.
* gdb.cp/annota2.exp (breakpoints_invalid, frames_invalid): New
variables.
Adjust tests to breakpoints-invalid changes.

With some changes to how software single-step (SSS) breakpoints are
handled, one of those being to place SSS breakpoints on the breakpoint
chain as all other breakpoints, annota1.exp times out with lots and
lots of breakpoint-invalid and frame-changed annotations.  All those
extra annotations are actually unnecessary.  For one, SSS breakpoints
are internal breakpoints, so the frontend shouldn't care if they were
added, removed or changed.  Then, there's really no point in emitting
"breakpoints-invalid" or "frames-invalid" more than once between times
the frontend/user can actually issues GDB commands; the frontend will
have to wait for the GDB prompt to refresh its state, so emitting
those annotations at most once between prompts is enough.  Non-stop or
async would complicate this, but no frontend will be using annotations
in those modes (one of goes of emacs switching to MI was non-stop mode
support, AFAIK).  The previous patch reveals there has been an
intention in the past to suppress multiple breakpoints-invalid
annotations caused by ignore count changes.  As the previous patch
shows, that's always been broken, but in any case, this patch actually
makes it work.  The next patch will remove several annotation-specific
calls in breakpoint.c in favor of always using the breakpoint modified
& friends observers, and that causes yet more of these annotations,
because several calls to the corresponding annotate_* functions in
breakpoint.c are missing, particularly in newer code.

So all in all, here's a simple mechanism that avoids sending the same
annotation to the frontend more than once until gdb is ready to accept
further commands.

Tested on x86_64 Fedora 17.

2013-01-22  Pedro Alves  <palves@redhat.com>

* annotate.c: Include "inferior.h".
(frames_invalid_emitted)
(breakpoints_invalid_emitted): New globals.
(async_background_execution_p): New function.
(annotate_breakpoints_changed, annotate_frames_invalid): Skip
emitting the annotation if it has already been emitted.
(annotate_display_prompt): New function.
* annotate.h (annotate_display_prompt): New declaration.
* event-top.c: Include annotate.h.
(display_gdb_prompt): Call annotate_display_prompt.

There's code in annotate.c and breakpoint.c that is supposed to
suppress multiple breakpoints-invalid annotations when the ignore
count of a breakpoint changes, up until the target actually stops.

But, the code is bogus:

void
annotate_breakpoints_changed (void)
{
  if (annotation_level == 2)
    {
      target_terminal_ours ();
      printf_unfiltered (("\n\032\032breakpoints-invalid\n"));
      if (ignore_count_changed)
ignore_count_changed = 0;   /* Avoid multiple break annotations.  */
    }
}

The "ignore_count_changed" flag isn't actually guarding the output of
the annotation at all.  It would have been better written something
like:

void
annotate_breakpoints_changed (void)
{
  if (annotation_level == 2 && !ignore_count_changed)
    {
      target_terminal_ours ();
      printf_unfiltered (("\n\032\032breakpoints-invalid\n"));
      ignore_count_changed = 0;   /* Avoid multiple break annotations.  */
    }
}

but, it wasn't.  AFAICS, that goes all the way back to the original
patch'es submission and check in, at
<http://sourceware.org/ml/gdb-patches/1999-q4/msg00106.html>.  I
looked a tar of HP's wdb from 1999, and even though that contains
local changes in the annotate code, this suppression seems borked
there too to me.

The original patch added a test to supposedly exercise this
suppression, but, it actually doesn't.  It merely tests that
"breakpoints-invalid" is output after "stopped", but doesn't check
whether the duplicates supression actually works (IOW, check that only
_one_ annotation is seen).  I was going to simply delete the tests
too, but a following patch will eliminate the duplicates in a
different way (which I needed for a different reason), so instead, I'm
making the tests actually fail if a duplicate annotation is seen.

Worry not, the test doesn't actually fail!  The reason is that
breakpoint.c does:

      else if (b->ignore_count > 0)
{
  b->ignore_count--;
  annotate_ignore_count_change ();
  bs->stop = 0;
  /* Increase the hit count even though we don't stop.  */
  ++(b->hit_count);
  observer_notify_breakpoint_modified (b);
}

where the annotate_ignore_count_change call is meant to inform the
"breakpoint_modified" annotation observer to ignore the notification.
All sounds good.  But, the trouble is that nowadays annotate.c only
installs the observers if GDB is started with annotations enabled with
a command line option (gdb --annotate=2):

void
_initialize_annotate (void)
{
  if (annotation_level == 2)
    {
      observer_attach_breakpoint_deleted (breakpoint_changed);
      observer_attach_breakpoint_modified (breakpoint_changed);
    }
}

and annota1.exp, to enable annotations, starts GDB normally, and
afterwards does "set annotate 2", so the observers aren't installed
when annota1.exp is run, and therefore changing the ignore count isn't
triggering any annotation at all...

gdb/
2013-01-22  Pedro Alves  <palves@redhat.com>

* annotate.c (ignore_count_changed): Delete.
(annotate_breakpoints_changed): Don't clear ignore_count_changed.
(annotate_ignore_count_change): Delete.
(annotate_stopped): Don't emit a delayed breakpoints-changed
annotation.
* annotate.h (annotate_ignore_count_change): Delete.
* breakpoint.c (bpstat_check_breakpoint_conditions): Don't call
annotate_ignore_count_change.

gdb/testsuite/
2013-01-22  Pedro Alves  <palves@redhat.com>

* gdb.base/annota1.exp (annotate ignore count change): Add
expected output for failure case.

* dwarf2loc.c (dwarf2_compile_expr_to_ax) <DW_OP_fbreg>: Only
require_rvalue for a register location.

* gdb.gdb/selftest.exp (do_steps_and_nexts): Handle bfd_init
call.

Updated Changelog and testsuite/Changelog because of bad formatting.

*** empty log message ***

2013-01-21  Marc Khouzam  <marc.khouzam@ericsson.com>

        * breakpoint.c (print_one_breakpoint_location): Add MI
        field 'thread-groups' when printing a breakpoint.
        (output_thread_groups): New function.

2013-01-21  Marc Khouzam  <marc.khouzam@ericsson.com>

        * gdb.texinfo (GDB/MI Breakpoint Commands): Document new
        'thread-groups' field when printing a breakpoint in MI.

2013-01-21  Marc Khouzam  <marc.khouzam@ericsson.com>

        * gdb.mi/mi-break.exp: Expect new 'thread-groups' field.
        * gdb.mi/mi-nsmoribund.exp: Expect new 'thread-groups' field.
        Also handle 'thread' field.
        * gdb.mi/mi-simplerun.exp: Expect new 'thread-groups' field.
        * gdb.mi/mi-watch.exp: Ditto.
        * lib/mi-support.exp: Ditto.

daily update

* python/lib/gdb/commands/explore.py
(CompoundExplorer.explore_expr): Correct the name of a method
being invoked.
(ExploreTypeCommand.invoke): Add a missing 'return'.
* testsuite/gdb.python/py-explore.exp: Improve a test

* gdb_obstack.h (obconcat): Move declaration here, from...
* symfile.h (obconcat): ... here.
* gdb_obstack.c: New file.
(obconcat): Move from...
* symfile.c (obconcat): ... here.
* Makefile.in (SFILES): Add gdb_obstack.c.
(COMMON_OBS): Add gdb_obstack.o.

* symfile.h (obsavestring): Don't declare.
* symfile.c (obsavestring): Remove.
* ada-exp.y: Use obstack_copy0, not obsavestring.
* ada-lang.c: Use obstack_copy0, not obsavestring.
* coffread.c: Use obstack_copy0, not obsavestring.
* cp-namespace.c: Use obstack_copy0, not obsavestring.
* dbxread.c: Use obstack_copy0, not obsavestring.
* dwarf2read.c: Use obstack_copy0, not obsavestring.
* jit.c: Use obstack_copy0, not obsavestring.
* mdebugread.c: Use obstack_copy0, not obsavestring.
* psymtab.c: Use obstack_copy0, not obsavestring.
* stabsread.c: Use obstack_copy0, not obsavestring.
* xcoffread.c: Use obstack_copy0, not obsavestring.

* dwarf2read.c (fixup_go_packaging): Save package name
on objfile obstack.
* gdbtypes.c (init_type): Don't copy name.

* dwarf2read.c (struct partial_die_info) <name, scope>: Now
const.
(struct attribute) <u.str>: Now const.
(struct fnfieldlist) <name>: Now const.
(dw2_get_file_names_reader, init_cutu_and_read_dies): Update.
(partial_die_parent_scope): Make return type const.
(partial_die_full_name, add_partial_symbol): Update.
(dwarf2_compute_name, dwarf2_full_name, dwarf2_physname): Make
'name' const.
(find_file_and_directory): Make 'name' and 'comp_dir' const.
(read_file_scope, read_func_scope, dwarf2_add_field)
(dwarf2_add_member_fn, read_structure_type)
(process_enumeration_scope, read_array_type, read_module_type)
(read_base_type, read_subrange_type): Update.
(dwarf2_start_symtab): Make 'name' and 'comp_dir' const.
(new_symbol_full, guess_full_die_structure_name): Update.
(dwarf2_canonicalize_name): Return const type.  Make 'name' const.
(dwarf2_name): Return const type.
(dwarf_decode_macro_bytes, dwarf_decode_macros): Make 'comp_dir'
const.

* gdbtypes.c (init_type): Make 'name' const.
* gdbtypes.h (init_type): Update.

* buildsym.c (patch_subfile_names): Use set_last_source_file.
(start_symtab): Make 'name' and 'dirname' const.  Use
set_last_source_file.
(restart_symtab, reset_symtab_globals): Use set_last_source_file.
(last_source_file): Define.  Now static.
(set_last_source_file, get_last_source_file): New functions.
* buildsym.h (last_source_file): Don't declare.
(start_symtab): Update.
(set_last_source_file, get_last_source_file): Declare.
* coffread.c (complete_symtab): Use set_last_source_file.
(coff_end_symtab): Likewise.
(coff_symtab_read): Use set_last_source_file, get_last_source_file.
* dbxread.c (read_dbx_symtab, read_ofile_symtab): Use
set_last_source_file.
(process_one_symbol): Use get_last_source_file.
* mdebugread.c (parse_partial_symbols): Use set_last_source_file.
(psymtab_to_symtab_1): Use get_last_source_file.
* xcoffread.c (process_linenos): Use get_last_source_file.
(complete_symtab): Use set_last_source_file.
(read_xcoff_symtab): Use set_last_source_file, get_last_source_file.
(scan_xcoff_symtab): Use set_last_source_file.

* symtab.c (struct demangled_name_entry) <mangled>: Now const.
(symbol_set_names): Remove casts.  Handle field const-ness.

* dwarf2read.c (new_symbol_full): Remove cast.
* symtab.c (symbol_set_demangled_name): Make 'name' const.
* symtab.h (symbol_set_demangled_name): Update.

* main.c (captured_main): Call bfd_init.

* gnu-v2-abi.c (_initialize_gnu_v2_abi): Don't set default ABI.
* gnu-v3-abi.c (_initialize_gnu_v3_abi): Set default ABI.
* minsyms.c (install_minimal_symbols): Don't check for _Z symbols.
* NEWS: Update.

gdb/
* symmisc.c (maintenance_print_msymbols): Check also ST_DEV.

gdb/
Fix gdb.fortran/common-block.exp crash in PIE mode.
* dwarf2read.c (new_symbol_full) <DW_TAG_common_block>: Use
LOC_COMMON_BLOCK.
* f-valprint.c (info_common_command_for_block): Expect
LOC_COMMON_BLOCK in gdb_assert.
* symtab.h (struct general_symbol_info): Update comment for the
common_block member.
(domain_enum): Extend comment for the COMMON_BLOCK_DOMAIN member.
(enum address_class): New member LOC_COMMON_BLOCK.

ld: enable new dtags by default for linux/gnu targets

The "new" dtags options have been around for 14+ years now, so for Linux
and GNU targets, enable them by default.

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

* ld-size/size.exp (build_tests <size-7, size-8>): Pass
--no-as-needed in cflags.

*** empty log message ***

daily update

*** empty log message ***

daily update

Add HOSTING_SCRT0 for PIE test

ld/

* Makefile.am (HOSTING_SCRT0): New.

* configure.host (HOSTING_SCRT0): New.  Used for PIE.

* configure.in (HOSTING_SCRT0): New AC_SUBST.

* Makefile.in: Regenerated.
* configure: Likewise.

ld/testsuite/

* config/default.exp (get_target_emul): Also set HOSTING_SCRT0.

* lib/ld-lib.exp (default_ld_link): Use HOSTING_SCRT0 for -pie.

*** empty log message ***

daily update

Resolve size relocation against non-zero TLS symbol

bfd/

* elf32-i386.c (elf_i386_allocate_dynrelocs): Clear pc_count for
non-zero TLS symbol.
(elf_i386_relocate_section): Resolve size relocation against
non-zero TLS symbol.
* elf64-x86-64.c (elf_x86_64_allocate_dynrelocs): Clear pc_count
for non-zero TLS symbol.
(elf_x86_64_relocate_section): Resolve size relocation against
non-zero TLS symbol.

ld/testsuite/

* ld-size/size-10.rd: Updated.
* ld-size/size-8.rd: Likewise.
* ld-size/size32-2-i386.d: Likewise.
* ld-size/size32-2-x32.d: Likewise.
* ld-size/size32-2-x86-64.d: Likewise.
* ld-size/size64-2-x32.d: Likewise.
* ld-size/size64-2-x86-64.d: Likewise.

gdb
* ChangeLog: Fix errors in my previous commit: whitespace->tabs, date,
and trailing blank line. (from review by Sergio Durigan Junior)

gdb
* MAINTAINERS (Write After Approval): Add "David Blaikie".

gdb/testsuite
* gdb.base/label.c (main): Correct the type of the second
parameter.

PR c++/14999:
* dwarf2loc.c (dwarf2_compile_expr_to_ax) <DW_OP_fbreg>:
Call require_rvalue.
gdb/testsuite
* gdb.dwarf2/trace-crash.s: New file.
* gdb.dwarf2/trace-crash.exp: New file.

gold: enable new dtags by default

The "new" dtags options have been around for 14+ years, and for all the
targets that gold supports, these flags have always existed.  So enable
them by default.

Having behavior be different from ld.bfd isn't new, and this behavior
is the "better" one, so there shouldn't be a problem based on that.

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

ld: change --enable-new-dtags to only generate new dtags

The "new" dtags options have been around for 14+ years, so there
shouldn't be a need to generate both new & old tags anymore.

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

Support size relocation only for ELF

* config/tc-i386.c (reloc): Support size relocation only for ELF.
(tc_i386_fix_adjustable): Likewise.
(lex_got): Likewise.
(tc_gen_reloc): Likewise.

PR binutils/15026
* addr2line.c (translate_addresses): When pretty printing, print
unknown function names on the same line as unknown symbol names.

2013-01-18  Hafiz Abid Qadeer<abidh@codesourcery.com>

        PR gdb/13443
        * gdb.base/checkpoint.exp: Update test messages to make them
        unique.

gdb/gdbserver:

* ax.h (struct eval_agent_expr_context): New.
(gdb_eval_agent_expr): Update declaration.
* ax.c (gdb_eval_agent_expr): Remove argument REGCACHE and
TFRAME.  Add new argument CTX.
* server.h (struct eval_agent_expr_context): Declare.
(agent_mem_read, agent_tsv_read): Update declaration.
(agent_mem_read_string): Likewise.
* tracepoint.c (eval_tracepoint_agent_expr): Remove.
(add_traceframe_block): Add new argument TPOINT.
Increase TPOINT->traceframe_usage.
(do_action_at_tracepoint): Call gdb_eval_agent_expr instead of
eval_tracepoint_agent_expr.
(condition_true_at_tracepoint): Likewise.
(agent_mem_read): Remove argument TFRAME.  Add argument CTX.
(agent_mem_read_string, agent_tsv_read): Likewise.
Callers update.

gdb/testsuite:

* gdb.trace/infotrace.exp: Check 'traceframe usage' in the
output of 'info tracepoints'.
* gdb.trace/disconnected-tracing.exp (disconnected_tracing):
Likewise.
* gdb.trace/tstatus.exp (run_trace_experiment): Likewise.
* gdb.trace/disconnected-tracing.c (struct foo): New.

gdb/
* dbxread.c (dbx_psymtab_to_symtab): Delete the declaration.
(dbx_read_symtab): New declaration.
(dbx_psymtab_to_symtab): Delete.
(dbx_read_symtab): Rename from dbx_psymtab_to_symtab.
Rename parameter PST to SELF.  Exchanged two parameters.
(start_psymtab): Caller update.
* dwarf2read.c (dwarf2_psymtab_to_symtab): Delete the declaration.
(dwarf2_read_symtab): New declaration.
(dwarf2_psymtab_to_symtab): Delete.
(dwarf2_read_symtab): Rename from dwarf2_psymtab_to_symtab.
Rename parameter PST to SELF.  Exchanged two parameters.
(create_partial_symtab): Caller update.
* mdebugread.c (mdebug_psymtab_to_symtab): Delete.
(mdebug_read_symtab): Rename from mdebug_psymtab_to_symtab.
Rename parameter PST to SELF.  Exchanged two parameters.
(parse_partial_symbols, new_psymtab): Caller update.
* psympriv.h (struct partial_symtab) <read_symtab>: Exchange
two parameters.
* psymtab.c (psymtab_to_symtab): Caller update.
* xcoffread.c (xcoff_psymtab_to_symtab): Delete.
(xcoff_read_symtab): Rename from xcoff_psymtab_to_symtab.
Rename parameter PST to SELF.  Exchanged two parameters.
(xcoff_start_psymtab): Caller update.

gdb/
* infrun.c (proceed): Rename local variable 'oneproc' to
'force_step'.

Verify run-time size relocations if supported

* ld-size/size-7.out: New file.
* ld-size/size-8.out: Likewise.
* ld-size/size-9.out: Likewise.
* ld-size/size-9.rd: Likewise.
* ld-size/size-9a.c: Likewise.
* ld-size/size-9b.c: Likewise.
* ld-size/size-10.out: Likewise.
* ld-size/size-10.rd: Likewise.
* ld-size/size-10a.c: Likewise.
* ld-size/size-10b.c: Likewise.

* ld-size/size.exp (build_tests): Build libsize-9.so and
libsize-10.so.
Run-time size relocation tests if supported.
(run_time_tests): New.

Add size-1 and size-2 tests

* ld-size/size-1.c: New file.
* ld-size/size-1.out: Likewise.
* ld-size/size-1a.c: Likewise.
* ld-size/size-1b.c: Likewise.
* ld-size/size-2.c: Likewise.
* ld-size/size-2.out: Likewise.
* ld-size/size-2a.c: Likewise.
* ld-size/size-2b.c: Likewise.

* ld-size/size.exp (build_tests): Build libsize-1.so and
libsize-2.so.
(run_tests): Run size-1 and size-2.

*** empty log message ***

Add missing files

Count size relocation as PC-relative relocation

bfd/

* elf32-i386.c (elf_i386_check_relocs): Count size relocation as
PC-relative relocation.
* elf64-x86-64.c (elf_x86_64_check_relocs): Count size relocation
as PC-relative relocation.

ld/testsuite/

* ld-size/size32-3-i386.d: New file.
* ld-size/size32-3-x32.d: Likewise.
* ld-size/size32-3-x86-64.d: Likewise.
* ld-size/size32-3.s: Likewise.

* powerpc.cc (Stub_table::find_plt_call_entry): Make types
used in declaration and definition consistent.
(Target_powerpc::symval_for_branch): Ditto.

* dwarf2read.c (dw2_build_type_unit_groups_reader): Delete.
(dw2_build_type_unit_groups): Delete.  All uses updated.

* symtab.h (struct symbol_search): Add comment.

daily update

gdb/
* symtab.c (compare_filenames_for_search): New comment for
HAS_DRIVE_SPEC.

gdb/testsuite/
* gdb.dwarf2/dw2-dos-drive.S: New file.
* gdb.dwarf2/dw2-dos-drive.exp: New file.

* cp-abi.c (cp_abi_completer): Fix typo in assignment.

* boards/dwarf4-gdb-index.exp: Use any existing CC_FOR_TARGET,
CXX_FOR_TARGET.

gdb/
* symtab.c (iterate_over_some_symtabs): New variable cleanups,
initialize it by existing make_cleanup.  Call new do_cleanups.

* boards/cc-with-tweaks.exp: New file.

* cp-abi.c (cp_abi_completer): New function.
(_initialize_cp_abi): Set completer for "set cp-abi".
gdb/testsuite
* gdb.base/completion.exp: Add "set cp-abi" completion test.

include/opcode/

2013-01-17  Yufeng Zhang  <yufeng.zhang@arm.com>

* aarch64.h (aarch64_op): Remove OP_V_MOVI_B.

opcodes/

2013-01-17  Yufeng Zhang  <yufeng.zhang@arm.com>

* aarch64-asm.c (aarch64_ins_advsimd_imm_modified): Handle 8-bit MOVI.
* aarch64-dis.c (aarch64_ext_advsimd_imm_modified): Likewise.
* aarch64-opc.c (operand_general_constraint_met_p): For
AARCH64_MOD_LSL, move the range check on the shift amount before the
alignment check; change to call set_sft_amount_out_of_range_error
instead of set_imm_out_of_range_error.
* aarch64-tbl.h (QL_SIMD_IMM_B): Replace NIL with LSL.
(aarch64_opcode_table): Remove the OP enumerator from the asimdimm
8-bit MOVI entry; change the 2nd operand from SIMD_IMM to
SIMD_IMM_SFT.

gas/

2013-01-17  Yufeng Zhang  <yufeng.zhang@arm.com>

* config/tc-aarch64.c (output_operand_error_record): Change to output
the out-of-range error message as value-expected message if there is
only one single value in the expected range.
(programmer_friendly_fixup): Remove the handling of 8-bit MOVI with
LSL #0 as a programmer-friendly feature.

gas/testsuite/

2013-01-17  Yufeng Zhang  <yufeng.zhang@arm.com>

* gas/aarch64/diagnostic.l: Update.
* gas/aarch64/movi.s: Add tests.
* gas/aarch64/movi.d: Update.
* gas/aarch64/programmer-friendly.s: Add comment.

* mem-break.c: Remove obsolete comment.
* bfin-tdep.c (bfin_breakpoint_from_pc): Fix comment.

Merge dg-extract-results.sh from gcc upstream (svn 195224).

gdb/testsuite/
2013-01-17  Pedro Alves  <palves@redhat.com>

Merge dg-extract-results.sh from upstream (svn 195224).

2013-01-15  David Blaikie <dblaikie@gmail.com>
     * dg-extract-results.sh: Fix order of summary counts.

2013-01-15  David Blaikie <dblaikie@gmail.com>
     * dg-extract-results.sh: Constrain the start-of-log pattern.

2013-01-15  David Blaikie <dblaikie@gmail.com>
     * dg-extract-results.sh: Handle KPASSes.

2010-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
* dg-extract-results.sh: Redirect grep output to /dev/null instead
of grep -q.

gdb/testsuite

 * gdb.base/jit-reader.exp: New file. Test case for the jit-reader
   interface.
 * gdb.base/jithost.c: New file.
 * gdb.base/jithost.h: New file.
 * gdb.base/jitreader.c : New file.
 * gdb.base/jit-protocol.h: New file.

* jit.c (jit_reader_load_command): Interpret the jit reader name as an
  absolute path if it begins with a forward slash.

PR gdb/14550

 * jit.c (finalize_symtab): Ensure that only the global block has a
   NULL superblock.

2013-01-17  Pedro Alves  <palves@redhat.com>

* acinclude.m4: Include ../config/plugins.m4,
../config/largefile.m4 and ../config/lead-dot.m4.  Add comments.
* Makefile.in (aclocal_m4_deps): Update.
* aclocal.m4: Renegerate.

Remove debug output

Add x86 size relocation support to gas

gas/

* config/tc-i386.c (reloc): Support BFD_RELOC_SIZE32.
(tc_i386_fix_adjustable): Keep symbol for BFD_RELOC_32_SIZE and
BFD_RELOC_64_SIZE relocations.
(lex_got): Support "symbol@SIZE" and don't create GOT symbol
for it.
(tc_gen_reloc): Resolve BFD_RELOC_SIZE32 and BFD_RELOC_SIZE64
relocations against local symbols.

gas/testsuite/

* gas/i386/i386.exp: Run size-1, size-2, size-3, size-4,
x86-64-size-1, x86-64-size-2, x86-64-size-3, x86-64-size-4,
x86-64-size-5 and x86-64-size-inval-1.

* gas/i386/size-1.d: New file.
* gas/i386/size-1.s: Likewise.
* gas/i386/size-2.d: Likewise.
* gas/i386/size-2.s: Likewise.
* gas/i386/size-3.d: Likewise.
* gas/i386/size-3.s: Likewise.
* gas/i386/size-4.d: Likewise.
* gas/i386/size-4.s: Likewise.
* gas/i386/x86-64-size-1.d: Likewise.
* gas/i386/x86-64-size-2.d: Likewise.
* gas/i386/x86-64-size-3.d: Likewise.
* gas/i386/x86-64-size-4.d: Likewise.
* gas/i386/x86-64-size-5.d: Likewise.
* gas/i386/x86-64-size-5.s: Likewise.
* gas/i386/x86-64-size-inval-1.l: Likewise.
* gas/i386/x86-64-size-inval-1.s: Likewise.
* gas/i386/ilp32/x86-64-size-1.d: Likewise.
* gas/i386/ilp32/x86-64-size-2.d: Likewise.
* gas/i386/ilp32/x86-64-size-3.d: Likewise.
* gas/i386/ilp32/x86-64-size-4.d: Likewise.
* gas/i386/ilp32/x86-64-size-5.d: Likewise.

ld/testsuite/

* ld-size/size.exp: New file.
* ld-size/size32-1-i386.d: Likewise.
* ld-size/size32-1-x32.d: Likewise.
* ld-size/size32-1-x86-64.d: Likewise.
* ld-size/size32-1.s: Likewise.
* ld-size/size32-2-i386.d: Likewise.
* ld-size/size32-2-x32.d: Likewise.
* ld-size/size32-2-x86-64.d: Likewise.
* ld-size/size32-2.s: Likewise.
* ld-size/size64-1-x32.d: Likewise.
* ld-size/size64-1-x86-64.d: Likewise.
* ld-size/size64-1.s: Likewise.
* ld-size/size64-2-x32.d: Likewise.
* ld-size/size64-2-x86-64.d: Likewise.
* ld-size/size64-2.s: Likewise.
* ld-size/size-3.c: Likewise.
* ld-size/size-3.out: Likewise.
* ld-size/size-3a.c: Likewise.
* ld-size/size-3b.c: Likewise.
* ld-size/size-3c.c: Likewise.
* ld-size/size-4.out: Likewise.
* ld-size/size-4a.c: Likewise.
* ld-size/size-4b.c: Likewise.
* ld-size/size-5.out: Likewise.
* ld-size/size-5a.c: Likewise.
* ld-size/size-5b.c: Likewise.
* ld-size/size-6.out: Likewise.
* ld-size/size-6a.c: Likewise.
* ld-size/size-6b.c: Likewise.
* ld-size/size-7.rd: Likewise.
* ld-size/size-7a.c: Likewise.
* ld-size/size-7b.c: Likewise.
* ld-size/size-8.rd: Likewise.
* ld-size/size-8a.c: Likewise.
* ld-size/size-8b.c: Likewise.

* objdump.c (dump_target_specific): Fix NULL pointer test.

2013-01-16  Sriraman Tallam  <tmsriram@google.com>

* testsuite/plugin_final_layout.cc: Fix comment.

*** empty log message ***

daily update

2013-01-16  Sriraman Tallam  <tmsriram@google.com>

* layout.cc (Layout::layout): Do not do default sorting for
text sections when section ordering is specified.
(make_output_section): Ditto.
* testsuite/plugin_final_layout.cc: Name the function sections
to catch reordering issues.

Update x86 size relocation check

* elf32-i386.c (elf_i386_check_relocs): Update R_386_SIZE32
check.
(elf_i386_relocate_section): Don't check TLS for R_386_SIZE32.

* elf64-x86-64.c (elf_x86_64_check_relocs): Update R_X86_64_SIZE32
and R_X86_64_SIZE64 check.
(elf_x86_64_relocate_section): Don't check TLS for R_X86_64_SIZE32
nor R_X86_64_SIZE64.

Add OPERAND_TYPE_IMM32_64

* i386-gen.c (operand_type_init): Add OPERAND_TYPE_IMM32_64.

* i386-init.h: Regenerated.
* i386-tbl.h: Likewise.

* contrib/cc-with-tweaks.sh: Add references to Fission docs.