[ARM] Add support for Cortex-R8

2016-03-07  Andre Vieira  <andre.simoesdiasvieira@arm.com>

gas/
    * config/tc-arm.c (arm_cpus): Add cortex-r8.
    * doc/c-arm.texi: Add cortex-r8.

Allow zero length archive elements.

bfd PR binutils/19775
* archive.c (bfd_generic_openr_next_archived_file): Allow zero
length elements in the archive.

binutils PR binutils/19775
* testsuite/binutils-all/ar.exp (proc empty_archive): New proc.
Run the new proc.
* testsuite/binutils-all/empty: New, empty, file.

Support --as-needed in i386/x86-64 tests

Since compiler may pass --as-needed to ld by default, link .o file
before .so file in i386/x86-64 tests.

PR ld/19774
* testsuite/ld-i386/i386.exp: Link tmpdir/pr18900.o before
tmpdir/pr18900.so and test --as-needed.  Link tmpdir/gotpc1.o
before tmpdir/got1d.so and test --as-needed.
* testsuite/ld-x86-64/x86-64.exp: Link tmpdir/pr18900.o before
tmpdir/pr18900.so and test --as-needed.

Fix "set architecture mips:10000" crash

Fix this GDB crash:

  $ gdb -ex "set architecture mips:10000"
  Segmentation fault (core dumped)

Backtrace:

  Program received signal SIGSEGV, Segmentation fault.
  0x0000000000495b1b in mips_gdbarch_init (info=..., arches=0x0) at /home/pedro/gdb/mygit/cxx-convertion/src/gdb/mips-tdep.c:8436
  8436              if (bfd_get_flavour (info.abfd) == bfd_target_elf_flavour
  (top-gdb) bt
  #0  0x0000000000495b1b in mips_gdbarch_init (info=..., arches=0x0) at .../src/gdb/mips-tdep.c:8436
  #1  0x00000000007348a6 in gdbarch_find_by_info (info=...) at .../src/gdb/gdbarch.c:5155
  #2  0x000000000073563c in gdbarch_update_p (info=...) at .../src/gdb/arch-utils.c:522
  #3  0x0000000000735585 in set_architecture (ignore_args=0x0, from_tty=1, c=0x26bc870) at .../src/gdb/arch-utils.c:496
  #4  0x00000000005f29fd in do_sfunc (c=0x26bc870, args=0x0, from_tty=1) at .../src/gdb/cli/cli-decode.c:121
  #5  0x00000000005fd3f3 in do_set_command (arg=0x7fffffffdcdd "mips:10000", from_tty=1, c=0x26bc870) at .../src/gdb/cli/cli-setshow.c:455
  #6  0x0000000000836157 in execute_command (p=0x7fffffffdcdd "mips:10000", from_tty=1) at .../src/gdb/top.c:460
  #7  0x000000000071abfb in catch_command_errors (command=0x835f6b <execute_command>, arg=0x7fffffffdccc "set architecture mips:10000", from_tty=1)
      at .../src/gdb/main.c:368
  #8  0x000000000071bf4f in captured_main (data=0x7fffffffd750) at .../src/gdb/main.c:1132
  #9  0x0000000000716737 in catch_errors (func=0x71af44 <captured_main>, func_args=0x7fffffffd750, errstring=0x106b9a1 "", mask=RETURN_MASK_ALL)
      at .../src/gdb/exceptions.c:240
  #10 0x000000000071bfe6 in gdb_main (args=0x7fffffffd750) at .../src/gdb/main.c:1164
  #11 0x000000000040a6ad in main (argc=4, argv=0x7fffffffd858) at .../src/gdb/gdb.c:32
  (top-gdb)

We already check whether info.abfd is NULL before all other
bfd_get_flavour calls in the same function.  Just this one case was
missing.

(This was exposed by a WIP test that tries all "set architecture ARCH"
values.)

gdb/ChangeLog:
2016-03-07  Pedro Alves  <palves@redhat.com>

* mips-tdep.c (mips_gdbarch_init): Check whether info.abfd is NULL
before calling bfd_get_flavour.

[AArch64] Create .got section if _GLOBAL_OFFSET_TABLE_ referenced

2016-03-07  Jiong Wang  <jiong.wang@arm.com>

bfd/
  * elfnn-aarch64.c (elfNN_aarch64_check_relocs): Always create .got section
  if the symbol "_GLOBAL_OFFSET_TABLE_" referenced.

ld/
  * testsuite/ld-aarch64/implicit_got_section_1.s: New test source file.
  * testsuite/ld-aarch64/implicit_got_section_1.d: New test expected result.
  * testsuite/ld-aarch64/aarch64-elf.exp: Run new test.

Add const qualifiers at various places.

opcodes * mcore-opc.h: Add const qualifiers.
* microblaze-opc.h (struct op_code_struct): Likewise.
* sh-opc.h: Likewise.
* tic4x-dis.c (tic4x_print_indirect): Likewise.
(tic4x_print_op): Likewise.

include * opcode/dlx.h (struct dlx_opcode): Add const qualifiers.
* opcode/h8300.h (struct h8_opcode): Likewise.
* opcode/hppa.h (struct pa_opcode): Likewise.
* opcode/msp430.h: Likewise.
* opcode/spu.h (struct spu_opcode): Likewise.
* opcode/tic30.h (struct _register): Likewise.
* opcode/tic4x.h (struct tic4x_register): Likewise.
(struct tic4x_cond): Likewise.
(struct tic4x_indirect): Likewise.
(struct tic4x_inst): Likewise.
* opcode/visium.h (struct reg_entry): Likewise.

gas * config/tc-arc.c: Add const qualifiers.
* config/tc-h8300.c (md_begin): Likewise.
* config/tc-ia64.c (print_prmask): Likewise.
* config/tc-msp430.c (msp430_operands): Likewise.
* config/tc-nds32.c (struct suffix_name): Likewise.
(struct nds32_parse_option_table): Likewise.
(struct nds32_set_option_table): Likewise.
(do_pseudo_pushpopm): Likewise.
(do_pseudo_pushpop_stack): Likewise.
(nds32_relax_relocs): Likewise.
(nds32_flag): Likewise.
(struct nds32_hint_map): Likewise.
(nds32_find_reloc_table): Likewise.
(nds32_match_hint_insn): Likewise.
* config/tc-s390.c: Likewise.
* config/tc-sh.c (get_specific): Likewise.
* config/tc-tic30.c: Likewise.
* config/tc-tic4x.c (tic4x_inst_add): Likewise.
(tic4x_indirect_parse): Likewise.
* config/tc-vax.c (vax_cons): Likewise.
* config/tc-z80.c (struct reg_entry): Likewise.
* config/tc-epiphany.c (md_assemble): Adjust.
(epiphany_assemble): New function.
(epiphany_elf_section_rtn): Call do_align directly.
(epiphany_elf_section_text): Likewise.
* config/tc-ip2k.c (ip2k_elf_section_rtn): Likewise.
(ip2k_elf_section_text): Likewise.
* read.c (do_align): Make it not static.
* read.h (do_align): New prototype.

Automatic date update in version.in

Set executable bit on analyze-racy-logs.py

I forgot to do it in my previous commit.  This is necessary because we
execute the script directly on gdb/testsuite/Makefile.in.

gdb/testsuite/ChangeLog:
2016-03-06  Sergio Durigan Junior  <sergiodj@redhat.com>

* analyze-racy-logs.py: Set executable bit.

Link tmpdir/copyreloc-main.o before tmpdir/copyreloc-lib.so

Since compiler may pass --as-needed to ld by default, link
tmpdir/copyreloc-main.o before tmpdir/copyreloc-lib.so.

* testsuite/ld-i386/i386.exp: Link tmpdir/copyreloc-main.o
before tmpdir/copyreloc-lib.so and test --as-needed.
* testsuite/ld-x86-64/x86-64.exp: Likewise.

Improve analysis of racy testcases

This is an initial attempt to introduce some mechanisms to identify
racy testcases present in our testsuite.  As can be seen in previous
discussions, racy tests are really bothersome and cause our BuildBot
to pollute the gdb-testers mailing list with hundreds of
false-positives messages every month.  Hopefully, identifying these
racy tests in advance (and automatically) will contribute to the
reduction of noise traffic to gdb-testers, maybe to the point where we
will be able to send the failure messages directly to the authors of
the commits.

I spent some time trying to decide the best way to tackle this
problem, and decided that there is no silver bullet.  Racy tests are
tricky and it is difficult to catch them, so the best solution I could
find (for now?) is to run our testsuite a number of times in a row,
and then compare the results (i.e., the gdb.sum files generated during
each run).  The more times you run the tests, the more racy tests you
are likely to detect (at the expense of waiting longer and longer).
You can also run the tests in parallel, which makes things faster (and
contribute to catching more racy tests, because your machine will have
less resources for each test and some of them are likely to fail when
this happens).  I did some tests in my machine (8-core i7, 16GB RAM),
and running the whole GDB testsuite 5 times using -j6 took 23 minutes.
Not bad.

In order to run the racy test machinery, you need to specify the
RACY_ITER environment variable.  You will assign a number to this
variable, which represents the number of times you want to run the
tests.  So, for example, if you want to run the whole testsuite 3
times in parallel (using 2 cores), you will do:

  make check RACY_ITER=3 -j2

It is also possible to use the TESTS variable and specify which tests
you want to run:

  make check TEST='gdb.base/default.exp' RACY_ITER=3 -j2

And so on.  The output files will be put at the directory
gdb/testsuite/racy_outputs/.

After make invokes the necessary rules to run the tests, it finally
runs a Python script that will analyze the resulting gdb.sum files.
This Python script will read each file, and construct a series of sets
based on the results of the tests (one set for FAIL's, one for
PASS'es, one for KFAIL's, etc.).  It will then do some set operations
and come up with a list of unique, sorted testcases that are racy.
The algorithm behind this is:

  for state in PASS, FAIL, XFAIL, XPASS...; do
    if a test's state in every sumfile is $state; then
      it is not racy
    else
      it is racy

(The algorithm is actually a bit more complex than that, because it
takes into account other things in order to decide whether the test
should be ignored or not).

IOW, a test must have the same state in every sumfile.

After processing everything, the script prints the racy tests it could
identify on stdout.  I am redirecting this to a file named racy.sum.

Something else that I wasn't sure how to deal with was non-unique
messages in our testsuite.  I decided to do the same thing I do in our
BuildBot: include a unique identifier in the end of message, like:

  gdb.base/xyz.exp: non-unique message
  gdb.base/xyz.exp: non-unique message <<2>>

This means that you will have to be careful about them when you use
the racy.sum file.

I ran the script several times here, and it did a good job catching
some well-known racy tests.  Overall, I am satisfied with this
approach and I think it will be helpful to have it upstream'ed.  I
also intend to extend our BuildBot and create new, specialized
builders that will be responsible for detecting the racy tests every X
number of days.

2016-03-05  Sergio Durigan Junior  <sergiodj@redhat.com>

* Makefile.in (DEFAULT_RACY_ITER): New variable.
(CHECK_TARGET_TMP): Likewise.
(check-single-racy): New rule.
(check-parallel-racy): Likewise.
(TEST_TARGETS): Adjust rule to account for RACY_ITER.
(do-check-parallel-racy): New rule.
(check-racy/%.exp): Likewise.
* README (Racy testcases): New section.
* analyze-racy-logs.py: New file.

Automatic date update in version.in

Fix argument passing for call

When calling function with argument of size more than 8 bytes fails with
an error "That operation is not available on integers of more than 8 bytes.".
avr-gdb considers only 8 bytes (sizeof(long long)) in case of passing the
argument in registers. When the argument is of size more than 8 byte
then the utility function to extract bytes failed with the above error.

    gdb/
* avr-tdep.c (AVR_LAST_ARG_REGNUM): Define.
        (avr_push_dummy_call): Correct last needed argument register.
        Write MSB of argument into register and subsequent bytes into
        other registers in decreasing order.

Automatic date update in version.in

Fix datestamps on ChangeLog entries to read 2015 instead of 2016.

Add missing ChangeLog entries for commit 82838

Fix gold/testsuite/plugin_layout_with_alignment.sh permission.

Fix undefined symbol errors introduced with previous commit.

gold/
PR gold/19019
PR gold/19763
* symtab.cc: Instantiate Sized_symbol::init_constant and
Sized_symbol::init_undefined.

ARM process record: VMOV

ARM process record gets the wrong register number for VMOV (from core
register to single-precision register).  That is, we should record
the D register rather than the S pseudo register.  The patch also
removes the condition "bit (arm_insn_r->arm_insn, 20)" check, which
has been checked above.

It fixes the following internal error,

(gdb) PASS: gdb.reverse/finish-precsave.exp: BP at end of main
continue^M
Continuing.^M
../../binutils-gdb/gdb/regcache.c:649: internal-error: regcache_raw_read: Assertion `regnum >= 0 && regnum < regcache->descr->nr_raw_registers' failed.^M
A problem internal to GDB has been detected,FAIL: gdb.reverse/finish-precsave.exp: run to end of main (GDB internal error)

gdb:

2016-03-04  Yao Qi  <yao.qi@linaro.org>

* arm-tdep.c (arm_record_vdata_transfer_insn): Simplify the
condition check.  Record the right D register number.

Tweak ARM process record

This patch removes the printing "Process record does not support",
and do the print by calling arm_record_unsupported_insn in the
caller.  Also, call arm_record_extension_space only when condition
is 0xf.

gdb:

2016-03-04  Yao Qi  <yao.qi@linaro.org>

* arm-tdep.c (arm_record_extension_space): Remove code
printing "Process record does not support".
(arm_record_data_proc_misc_ld_str): Likewise.
(decode_insn): Call arm_record_extension_space if condition
is 0xf.  Call arm_record_unsupported_insn if ret isn't
ARM_RECORD_SUCCESS.  Use 'ret' instead of 'insn_id' to hold
the value of thumb2_record_decode_insn_handler.

feature_to_c.sh: Print help when passing no arguments

I found that odd that passing no arguments to feature_to_c.sh produces
this:

$ ./feature_to_c.sh
./feature_to_c.sh: 23: shift: can't shift that many

but passing one argument shows the help:

$ ./feature_to_c.sh hello
Usage: ./feature_to_c.sh OUTPUTFILE INPUTFILE...

This patch changes the script to show the help in both cases.

gdb/ChangeLog:

* features/feature_to_c.sh: Print the help when passing no
argument.

Fix a ChangeLog entry

Treat common symbol in executable as definition

Common symbol in executable is a definition, which overrides definition
from shared objects.  When linker sees a new definition from a shared
object, the new dynamic definition should be overridden by the previous
common symbol in executable.

bfd/

PR ld/19579
* elflink.c (_bfd_elf_merge_symbol): Treat common symbol in
executable as definition if the new definition comes from a
shared library.

ld/

PR ld/19579
* testsuite/ld-elf/pr19579a.c: New file.
* testsuite/ld-elf/pr19579b.c: Likewise.
* testsuite/ld-elf/shared.exp: Run PR ld/19579 test.

[ARM] Build attributes for ARMv8.1-A AdvSIMD

binutils/
2016-03-04  Matthew Wahab  <matthew.wahab@arm.com>

* readelf.c (arm_attry_tag_FP_arch): Add "NEON for ARMv8.1".

gas/
2016-03-04  Matthew Wahab  <matthew.wahab@arm.com>

* config/tc-arm.c (aeabi_set_public_attributes): Emit attribute
for ARMv8.1 AdvSIMD use.
* testsuite/gas/arm/attr-march-armv8-a+rdma.d: New.
* testsuite/gas/arm/attr-march-armv8_1-a+simd.d: New.

Change-Id: I3c356e0681b97df2f9c0dabd7c0fd1b441cc2755

[ARM] Add feature check for ARMv8.1 AdvSIMD instructions.

gas/
2016-03-04  Matthew Wahab  <matthew.wahab@arm.com>

* config/gas/tc-arm.c (fpu_neon_ext_v8_1): Restrict to the ARMv8.1 RDMA
feature.
(record_feature_use): New.
(mark_feature_used): Use record_feature_use.
(do_neon_qrdmlah): New.
(insns): Use do_neon_qrdmlah for vqrdmlah and vqrdmlsh and
variants.
(arm_extensions): Put into alphabetical order.  Re-indent "simd"
and "rdma" entries.  Fix the incorrect merge value for "+rdma".
* testsuite/gas/arm/armv8-a+rdma-warning.d: New.
* testsuite/gas/arm/armv8-a+rdma.d: Add assembler command line options.
Make source file explicit.
* testsuite/gas/arm/armv8-a+rdma.l: New.
* testsuite/gas/arm/armv8-a+rdma.s: Remove .arch and .arch_extension
directives.  Fix white-space.
* testsuite/gas/arm/armv8_1-a+simd.d: New.

include/opcode
2016-03-04  Matthew Wahab  <matthew.wahab@arm.com>

* arm.h (ARM_ARCH_V8_1A): Add FPU_NEON_EXT_RDMA.
(ARM_CPU_HAS_FEATURE): Add comment.

Change-Id: Ie19250e8fa50aed44e44ab40ff30b04b38bc1a3d

Add support for STT_SPARC_REGISTER symbols.

gold/
PR gold/19019
* layout.h (Layout::add_target_specific_dynamic_tag): New function.
* layout.cc (Layout::add_target_specific_dynamic_tag): New function.
* mips.cc (Target_mips::make_symbol): Adjust function signature.
* sparc.cc (Target_sparc::Target_sparc): Initialize register_syms_.
(Target_sparc::do_is_defined_by_abi): Remove test for
STT_SPARC_REGISTER.
(Target_sparc::Register_symbol): New struct type.
(Target_sparc::register_syms_): New data member.
(Target_sparc<64, true>::sparc_info): Set has_make_symbol to true.
(Target_sparc::make_symbol): New function.
(Target_sparc::do_finalize_sections): Add register symbols and new
dynamic table entries.
* symtab.h (Sized_symbol::init_undefined): Add value parameter.
(Symbol_table::add_target_global_symbol): New function.
(Symbol_table::target_symbols_): New data member.
* symtab.cc (Sized_symbol::init_undefined): Add value parameter.
(Symbol_table::Symbol_table): Initialize target_symbols_.
(Symbol_table::add_from_object): Pass additional parameters to
Target::make_symbol.
(Symbol_table::define_special_symbol): Likewise.
(Symbol_table::add_undefined_symbol_from_command_line): Pass 0 for
undefined symbol value.
(Symbol_table::set_dynsym_indexes): Process target-specific symbols.
(Symbol_table::sized_finalize): Likewise.
(Symbol_table::sized_write_globals): Likewise.
* target.h (Sized_target::make_symbol): Add name, st_type, object,
st_shndx, and value parameters.

Automatic date update in version.in

Don't handle every plugin symbol as defined.

Gold was fetching --start-lib/--end-lib members if that member defined
or used a currently undefined symbol.

Add new plugin hooks to support querying section alignment and size.

include/
2016-03-03  Than McIntosh <thanm@google.com>

* plugin-api.h: Add new hooks to the plugin transfer vector to
to support querying section alignment and section size.
(ld_plugin_get_input_section_alignment): New hook.
(ld_plugin_get_input_section_size): New hook.
(ld_plugin_tag): Add LDPT_GET_INPUT_SECTION_ALIGNMENT
and LDPT_GET_INPUT_SECTION_SIZE.
(ld_plugin_tv): Add tv_get_input_section_alignment and
tv_get_input_section_size.

gold/
2016-03-03  Than McIntosh  <thanm@google.com>

* plugin.cc (Plugin::load): Include hooks for get_input_section_size
and get_input_section_alignment in transfer vector.
(get_input_section_alignment): New function.
(get_input_section_size): New function.
* testsuite/Makefile.am: Add plugin_layout_with_alignment.sh test.
* testsuite/Makefile.in: [Regenerate.]
* testsuite/plugin_section_alignment.cc: New test file.
* testsuite/plugin_layout_with_alignment.cc: New test file.
* testsuite/plugin_layout_with_alignment.sh: New test file.

get_symbols() plugin API tweak to support --start-lib/--end-lib.

Let the plugin know that a file is not being included in the link by
returning LDPS_NO_SYMS from get_symbols().

include/
* plugin-api.h (enum ld_plugin_tag): Add LDPT_GET_SYMBOLS_V3.

gold/
* plugin.h (Pluginobj::get_symbol_resolution_info): Add version
parameter.
* plugin.cc (get_symbols_v3): New function.
(Plugin::load): Add LDPT_GET_SYMBOLS_V3.
(Pluginobj::get_symbol_resolution_info): Return LDPS_NO_SYMS when using
new version.

gdb.base/skip.exp: Use with_test_prefix.

gdb/testsuite/ChangeLog:

* gdb.base/skip.exp: Use with_test_prefix.

Fix bugs in the simulation of the AArch64's ADDP, FADDP, LD1, CCMP and CCMP instructions.

* simulator.c (set_flags_for_sub32): Correct type of signbit.
(CondCompare): Swap interpretation of bit 30.
(DO_ADDP): Delete macro.
(do_vec_ADDP): Copy source registers before starting to update
destination register.
(do_vec_FADDP): Likewise.
(do_vec_load_store): Fix computation of sizeof_operation.
(rbit64): Fix type of constant.
(aarch64_step): When displaying insn value, display all 32 bits.

Update comments to start_step_over

I happen to see that comments to start_step_over isn't in sync with
code, so this patch is to update the comments.

gdb/gdbserver:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* linux-low.c: Update comments to start_step_over.

New test about step over clone syscall

This patch adds a new test for stepping over clone syscall.

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* gdb.base/step-over-syscall.exp (step_over_syscall): Kfail.
Invoke step_over_syscall "clone" and break_cond_on_syscall
"clone".
* gdb.base/step-over-clone.c: New file.

Reformat gdb.base/step-over-syscall.exp

gdb/testsuite:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* gdb.base/step-over-syscall.exp (disp_step_cross_syscall): Fix
code format.

Rename disp-step-syscall.exp to step-over-syscall.exp

disp-step-syscall.exp is extended for stepping over syscall instruction
in different cases, with or without displaced stepping, and stepping
over by GDBserver.

This patch rename disp-step-syscall.exp to step-over-syscall.exp to
reflect this.

gdb/testsuite:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* gdb.base/disp-step-fork.c: Rename to ...
* gdb.base/step-over-fork.c: ... it.  New file.
* gdb.base/disp-step-vfork.c: Rename to ...
* gdb.base/step-over-vfork.c: ... it.  New file.
* gdb.base/disp-step-syscall.exp: Rename to ...
* gdb.base/step-over-syscall.exp: ... it.  New file.
(disp_step_cross_syscall): Rename to ...
(step_over_syscall): ... it.

Step over fork/vfork syscall insn in gdbserver

We can also extend disp-step-syscall.exp to test GDBserver step over
breakpoint on syscall instruction.  That is, we set a breakpoint
with a false condition on syscall instruction, so that GDBserver will
step over it.

This test triggers a GDBserver internal error, which can be fixed by
this series.

(gdb) PASS: gdb.base/disp-step-syscall.exp: fork: break cond on target: break on syscall insns
continue^M
Continuing.^M
Remote connection closed^M
(gdb) FAIL: gdb.base/disp-step-syscall.exp: fork: break cond on target: continue to fork again

In GDBserver, there is an internal error,

/home/yao/SourceCode/gnu/gdb/git/gdb/gdbserver/linux-low.c:1922: A problem internal to GDBserver has been detected.
unsuspend LWP 25554, suspended=-1

the simplified reproducer is like,

$ ./gdb ./testsuite/outputs/gdb.base/disp-step-syscall/disp-step-fork
(gdb) b main
(gdb) c
(gdb) disassemble fork // in order to find the address of insn 'syscall'
....
   0x00007ffff7ad6023 <+179>: syscall
(gdb) b *0x00007ffff7ad6023 if main == 0
(gdb) c

gdb/testsuite:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* gdb.base/disp-step-syscall.exp (break_cond_on_syscall): New.
If target supports condition evaluation on target, invoke
break_cond_on_syscall for fork and vfork.

Step over syscalll insn with disp-step on and off

disp-step-syscall.exp was added to test displaced stepping over syscall
instructions, in which we set breakpoint on syscall instruction, and
step over it.  In fact, we can extend the test to non-displaced-stepping
case.  This patch wraps the test with displaced stepping on and off.
Note that the indentation and format isn't adjusted here to make this
patch easy to read.  The following patch will fix the format separately.

gdb/testsuite:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* gdb.base/disp-step-syscall.exp: Don't invoke
support_displaced_stepping.
(disp_step_cross_syscall): Test with displaced stepping off and
on if supported.

Refactor gdb.base/disp-step-syscall.exp for general step over test

This patch moves some code out of disp_step_cross_syscall to a new proc
check_pc_after_cross_syscall and setup.  Procedure setup is to start a
fresh GDB and compute the syscall instruction address.

gdb/testsuite:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

* gdb.base/disp-step-syscall.exp (check_pc_after_cross_syscall): New
proc.
(setup): New proc.
(disp_step_cross_syscall): Move code to check_pc_after_cross_syscall
and setup.

[GDBserver] Leave child suspended when step over parent

I see the following GDBserver internal error in two cases,

 gdb/gdbserver/linux-low.c:1922: A problem internal to GDBserver has been detected.
 unsuspend LWP 17200, suspended=-1

 1. step over a breakpoint on fork/vfork syscall instruction,
 2. step over a breakpoint on clone syscall instruction and child
    threads hits a breakpoint,

the stack backtrace is

 #0  internal_error (file=file@entry=0x44c4c0 "gdb/gdbserver/linux-low.c", line=line@entry=1922,
    fmt=fmt@entry=0x44c7d0 "unsuspend LWP %ld, suspended=%d\n") at gdb/gdbserver/../common/errors.c:51
 #1  0x0000000000424014 in lwp_suspended_decr (lwp=<optimised out>, lwp=<optimised out>) at gdb/gdbserver/linux-low.c:1922
 #2  0x000000000042403a in unsuspend_one_lwp (entry=<optimised out>, except=0x66e8c0) at gdb/gdbserver/linux-low.c:2885
 #3  0x0000000000405f45 in find_inferior (list=<optimised out>, func=func@entry=0x424020 <unsuspend_one_lwp>, arg=arg@entry=0x66e8c0)
    at gdb/gdbserver/inferiors.c:243
 #4  0x00000000004297de in unsuspend_all_lwps (except=0x66e8c0) at gdb/gdbserver/linux-low.c:2895
 #5  linux_wait_1 (ptid=..., ourstatus=ourstatus@entry=0x665ec0 <last_status>, target_options=target_options@entry=0)
    at gdb/gdbserver/linux-low.c:3632
 #6  0x000000000042a764 in linux_wait (ptid=..., ourstatus=0x665ec0 <last_status>, target_options=0)
    at gdb/gdbserver/linux-low.c:3770
 #7  0x0000000000411163 in mywait (ptid=..., ourstatus=ourstatus@entry=0x665ec0 <last_status>, options=options@entry=0, connected_wait=connected_wait@entry=1)
    at gdb/gdbserver/target.c:214
 #8  0x000000000040b1f2 in resume (actions=0x66f800, num_actions=1) at gdb/gdbserver/server.c:2757
 #9  0x000000000040f660 in handle_v_cont (own_buf=0x66a630 "vCont;c:p45e9.-1") at gdb/gdbserver/server.c:2719

when GDBserver steps over a thread, other threads have been suspended,
the "stepping" thread may create new thread, but GDBserver doesn't set
it suspend count to 1.  When GDBserver unsuspend threads, the child's
suspend count goes to -1, and the assert is triggered.  In fact, GDBserver
has already taken care of suspend count of new thread when GDBserver is
suspending all threads except the one GDBserver wants to step over by
https://sourceware.org/ml/gdb-patches/2015-07/msg00946.html

+   /* If we're suspending all threads, leave this one suspended
+      too.  */
+   if (stopping_threads == STOPPING_AND_SUSPENDING_THREADS)
+     {
+       if (debug_threads)
+ debug_printf ("HEW: leaving child suspended\n");
+       child_lwp->suspended = 1;
+     }

but that is not enough, because new thread is still can be spawned in
the thread which is being stepped over.  This patch extends the
condition that GDBserver set child's suspend count to one if it is
suspending threads or stepping over the thread.

gdb/gdbserver:

2016-03-03  Yao Qi  <yao.qi@linaro.org>

PR server/19736
* linux-low.c (handle_extended_wait): Set child suspended
if event_lwp->bp_reinsert isn't zero.

Skip ld-elf/pr19162.d for hppa-*-*

ld-elf/pr19162.d fails for hppa-*-* since Dave Anglin's fix for PR 12376
makes the data segment always start on a page boundary.

* testsuite/ld-elf/pr19162.d: Skip hppa-*-*.

Correct fr30 comment

* fr30.cpu (f.m4): Replace bogus comment with a better guess
at what is really going on.

Automatic date update in version.in

Xfail mix_pic_and_non_pic on arm and aarch64

* testsuite/ld-elf/shared.exp (mix_pic_and_non_pic): Add xfails.
Xfail mix_pic_and_non_pic on "arm*-*-*" "aarch64*-*-*".

Call enqueue_pending_signal in linux_resume_one_lwp_throw

Replace the code which is exactly what enqueue_pending_signal does.

gdb/gdbserver:

2016-03-02  Yao Qi  <yao.qi@linaro.org>

* linux-low.c (linux_resume_one_lwp_throw): Replace code with
enqueue_pending_signal.

[OBV] gdbserver: Only write ipa_tdesc_idx if agent is actually loaded.

Fixes rather embarassing gdb.trace regressions.

gdb/gdbserver/ChangeLog:

* tracepoint.c (cmd_qtstart): Only set ipa_tdesc_idx if agent
is actually loaded.

testsuite: Remove unnecessary code in fortran vla-history test.

testsuite: Remove unnecessary code in fortran vla-history test.

2016-03-02  Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/testsuite/Changelog:

     * gdb.fortran/vla-history.exp: Remove breakpoint.

testsuite: Fix timeout issues during print of vla-arrays.

Printing and resolving of dynamic array's causes sporadic timeout issues on loaded systems.

2016-03-02  Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/testsuite/Changelog:

     * gdb.fortran/vla-history.exp: Lookup array elements and printing exceeds timeout.

testsuite: Fix run to main issue introduced by GCC 5.x.

Adding a dummy assignment as a new breakpoint anchor because
breakpoint on return statement doesn't work for GCC 5.x.

2016-03-02  Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/testsuite/Changelog:

     * gdb.cp/vla-cxx.cc: Insert dummy assignment as anchor for an breakpoint.

testsuite: Nullify pointers before first usage.

Nullify pointers to avoid an undefined association status.

2016-03-02  Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/testsuite/Changelog:

     * gdb.mi/vla.f90: Nullify pointer after declaration.

Add new maintainer to Write After Approval.

Add new maintainer to Write After Approval.

2016-03-02  Bernhard Heckel  <bernhard.heckel@intel.com>

    * MAINTAINERS (Write After Approval): Add Bernhard Heckel.

Fix detection of gfortran compilers.

Newer gfortran compilers changed naming convention in DW_AT_producer tag.
For example "GNU Fortran 4.6.3" vs "GNU Fortran2008 5.3.0"

2016-03-02  Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/Changelog:

     * dwarf2read.c (new_symbol_full): Fix detection of gfortran compilers.

Regenerate or1k opcodes file

CGEN patch now committed upstream to use hex numbers for large enums.

* or1k-desc.h: Regenerate.

Adjust testsuite/gas/i386/x86_64-intel.d for COFF

* testsuite/gas/i386/x86_64-intel.d: Adjusted for COFF.

Speedup mmo and pe orphan placement for relocatable link

Since there is no need to place output sections in specific order for
relocatable link, we can skip merging flags of other input sections.

PR ld/19739
* emultempl/mmo.em (mmo_place_orphan): Don't merge flags of other
input sections for relocatable link.
* emultempl/pe.em (gld_${EMULATION_NAME}_place_orphan): Likewise.
* emultempl/pep.em (gld_${EMULATION_NAME}_place_orphan): Likewise.

Regenerate rl78 opcodes file

For newer comments.

* rl78-decode.c: Regenerate.

Regenerate bfd files.

* Makefile.in: Regenerate.
* po/SRC-POTFILES.in: Regenerate.

Fix shift left warning at source

cpu/
* fr30.cpu (f-m4): Replace -1 << 4 with -16.
opcodes/
* fr30-ibld.c: Regenerate.

Automatic date update in version.in

Update PR ld/12365 test for GCC 6

PR ld/12365 test is to verify that linker catches the reference to
undefined symbol error caused by a GCC bug.  Since the GCC bug has
been fixed in GCC 6, update the test to verify that there is no
reference to undefined symbol in executable

* testsuite/ld-plugin/lto.exp: Update PR ld/12365 test for GCC 6.

Fix gdb.trace/ftrace-lock.c compilation

Fixes, on F23:

 .../src/gdb/testsuite/gdb.trace/ftrace-lock.c: In function 'gdb_agent_gdb_collect':
 .../src/gdb/testsuite/gdb.trace/ftrace-lock.c:50:3: warning: implicit declaration of function 'sleep' [-Wimplicit-function-declaration]
    sleep (1);
    ^

gdb/testsuite/ChangeLog:
2016-03-01  Pedro Alves  <palves@redhat.com>

* gdb.trace/ftrace-lock.c: Include <unistd.h>.

Fix gdb.threads/watchpoint-fork*.c compilation

This testcase currently fails to compile on Fedora 23:

 .../src/gdb/testsuite/gdb.threads/watchpoint-fork-mt.c: In function 'start':
 .../src/gdb/testsuite/gdb.threads/watchpoint-fork-mt.c:70:11: warning: implicit declaration of function 'pthread_yield' [-Wimplicit-function-declaration]
i = pthread_yield ();
    ^
 .../src/gdb/testsuite/gdb.threads/watchpoint-fork-child.c: In function 'forkoff':
 .../src/gdb/testsuite/gdb.threads/watchpoint-fork-child.c:114:8: warning: implicit declaration of function 'pthread_yield' [-Wimplicit-function-declaratio
 n]
     i = pthread_yield ();
 ^
 /tmp/ccUkNIsI.o: In function `start':
 .../src/gdb/testsuite/gdb.threads/watchpoint-fork-mt.c:70: undefined reference to `pthread_yield'
 (...)
 collect2: error: ld returned 1 exit status

 UNSUPPORTED: gdb.threads/watchpoint-fork.exp: child: multithreaded: Couldn't compile watchpoint-fork-child.c: unrecognized error
 UNTESTED: gdb.threads/watchpoint-fork.exp: child: multithreaded: watchpoint-fork.exp
 testcase .../src/gdb/testsuite/gdb.threads/watchpoint-fork.exp completed i

The glibc manual says, on _GNU_SOURCE:

 "You should define these macros by using ‘#define’ preprocessor
 directives at the top of your source code files. These directives must
 come before any #include of a system header file."

I instead put it in the header all the .c files of the testcase must
include anyway.

gdb/testsuite/ChangeLog:
2016-03-01  Pedro Alves  <palves@redhat.com>

* gdb.threads/watchpoint-fork-child.c: Include "watchpoint-fork.h"
before anything else.
* gdb.threads/watchpoint-fork-mt.c: Likewise.  Don't define
_GNU_SOURCE here.
* gdb.threads/watchpoint-fork-st.c: Include "watchpoint-fork.h"
before anything else.
* gdb.threads/watchpoint-fork.h: Define _GNU_SOURCE.

Fix gdb.base/catch-fork-kill.c compilation

Fixes:

 Running .../src/gdb/testsuite/gdb.base/catch-fork-kill.exp ...
 gdb compile failed, .../src/gdb/testsuite/gdb.base/catch-fork-kill.c: In function 'main':
 .../src/gdb/testsuite/gdb.base/catch-fork-kill.c:81:4: warning: implicit declaration of function 'wait' [-Wimplicit-function-declaration]
     wait (NULL);
     ^

gdb/testsuite/ChangeLog:
2016-03-01  Pedro Alves  <palves@redhat.com>

* gdb.base/catch-fork-kill.c: Include <sys/wait.h>.

Skip ld-elf/pr19162.d for arc target

* testsuite/ld-elf/pr19162.d: Skip arc target.

Fix output path for arm-disp-step.exp

This patch fixes the following error,

ERROR: (/scratch/yao/gdb/build-git/arm-linux-gnueabihf/gdb/testsuite/outputs/gdb.arch/arm-disp-step/arm-disp-step) No such file or directory
FAIL: gdb.arch/arm-disp-step.exp: Can't run to main

gdb/testsuite:

2016-03-01  Yao Qi  <yao.qi@linaro.org>

* gdb.arch/arm-disp-step.exp: Use standard_testfile and
prepare_for_testing.

Compile gdb.arch/arm-neon.c with "quiet"

When we compile gdb.arch/arm-neon.c with options that don't enable NEON,
there are many error/warnings emitted into gdb.sum, which is annoying.
This patch fixes it by passing quiet to prepare_for_testing.

gdb/testsuite:

2016-03-01  Yao Qi  <yao.qi@linaro.org>

* gdb.arch/arm-neon.exp: Pass quiet to prepare_for_testing.

Speedup ELF orphan placement for relocatable link

Since there is no need to place output sections in specific order for
relocatable link, we can skip merging flags of other input sections.

PR ld/19739
* ld/emultempl/elf32.em (gld${EMULATION_NAME}_place_orphan): Don't
merge flags of other input sections for relocatable link.

S390: Fix output path for s390-tdbregs test case

Since test artifacts are always organized in a directory hierarchy, the
s390-tdbregs test case is not executed correctly any more.  This is
because it uses an obsolete way of constructing the executable's path.

This change invokes prepare_for_testing instead.

gdb/testsuite/ChangeLog:

* gdb.arch/s390-tdbregs.exp: Use prepare_for_testing instead of
manually constructing the output path.

S390: Fix internal error with stackless inferior

This fixes a GDB internal error that may occur when the inferior has no
valid stack pointer in r15.

gdb/testsuite/ChangeLog:

* gdb.arch/s390-stackless.S: New.
* gdb.arch/s390-stackless.exp: New.

gdb/ChangeLog:

* s390-linux-tdep.c (s390_backchain_frame_unwind_cache): Avoid
exception when attempting to access the inferior's backchain.

Fix typo in print_insn_rl78_common function.

PR target/19747
* rl78-dis.c (print_insn_rl78_common): Fix typo.

Automatic date update in version.in

Add a testcase for PR ld/19162

PR ld/19162
* testsuite/ld-elf/pr19162.d: New file.
* testsuite/ld-elf/pr19162a.s: Likwise.
* testsuite/ld-elf/pr19162b.s: Likwise.

aarch64-linux process record: more syscalls

The last patch supports several syscalls in linux-record.c, so now
GDB aarch64-linux backend can return these canonicalized syscall numbers
per aarch64 syscall number.

This patch fixes the following fails,

Process record and replay target doesn't support syscall number 59^M
Process record: failed to record execution log.^M
^M
Program stopped.^M
0x00000020000eab28 in pipe () from /lib/aarch64-linux-gnu/libc.so.6^M
(gdb) FAIL: gdb.reverse/pipe-reverse.exp: continue to breakpoint: marker2

Process record and replay target doesn't support syscall number 59^M
Process record: failed to record execution log.^M
^M
Program stopped.^M
0x00000020000eab28 in pipe () from /lib/aarch64-linux-gnu/libc.so.6^M
(gdb) FAIL: gdb.reverse/readv-reverse.exp: continue to breakpoint: marker2

gdb:

2016-02-29  Yao Qi  <yao.qi@linaro.org>

* aarch64-linux-tdep.c (aarch64_canonicalize_syscall): Support
eventfd2, eventfd2, dup3, inotify_init1, fallocate and pipe2.
Return gdb_sys_epoll_create1 instead of gdb_sys_epoll_create
for aarch64_sys_epoll_create1.

Support more syscalls in linux-record: pipe2 epoll_create1 eventfd2 fallocate dup3 and inotify_init1

This patch adds more syscalls in linux-record.

gdb:

2016-02-29  Yao Qi  <yao.qi@linaro.org>

* linux-record.h (enum gdb_syscall) <gdb_sys_fallocate>: New.
<gdb_sys_eventfd2, gdb_sys_epoll_create1, gdb_sys_dup3>: New.
<gdb_sys_pipe2, gdb_sys_inotify_init1>: New.
* linux-record.c (record_linux_system_call): Handle them.

[ARC] Local symbols relocation cleanup

bfd/
2016-02-29  Cupertino Miranda  <cmiranda@synopsys.com>

* elf32-arc.c (elf_arc_relocate_section): Added rules to fix the
relocation addend when sections get merged.

gas/
2016-02-29  Cupertino Miranda  <cmiranda@synopsys.com>
    Claudiu Zissulescu  <Claudiu.Zissulescu@synopsys.com>

        * config/tc-arc.c (arc_extra_reloc): Change size to 0.
        (tc_arc_fix_adjustable): Changed default return value to 1.
        * testsuite/gas/arc/j.d: Updated expected symbol
        * testsuite/gas/arc/jl.d: Likewise
        * testsuite/gas/arc/relax-avoid1.d: Likewise
        * testsuite/gas/arc/st.d: Likewise

ld/
2016-02-29 Cupertino Miranda  <cmiranda@synopsys.com>

* testsuite/ld-elf/merge.d: Removed xfail for ARC.
* testsuite/ld-elf/merge2.d: Likewise.
* testsuite/ld-elf/merge3.d: Likewise.

[ARC] General fixes.

bfd/
2016-02-29  Cupertino Miranda <Cupertino.Miranda@synopsys.com>

* elf32-arc.c (arc_elf_final_write_processing): Add condition to
the flag change.
        (elf_arc_relocate_section): Fixes and conditions to support PIE.
Assert for code sections dynamic relocs.

gas/
2016-02-29  Claudiu Zissulescu  <Claudiu.Zissulescu@synopsys.com>

* config/tc-arc.c: Enable code density instructions for ARC EM.

ld/
2016-02-29  Cupertino Miranda  <Cupertino.Miranda@synopsys.com>

* scripttempl/arclinux.sc: Force .tdata and .tbss to always be
generated.

Automatic date update in version.in

Don't recursively look for a symbol in all imports of imported modules.

Given two or more modules that import each other's scope, the current symbol
lookup routines would go round in circles looking through each import from
each module, possibly checking the same module twice or more until all possible
paths are marked as "searched".

Given enough modules, this causes an exponential slowdown in time taken to find
symbols that do exist, and infinite recursion when they don't.

gdb/ChangeLog:
* d-namespace.c (d_lookup_symbol_imports): Avoid recursive lookups from
cyclic imports.

gdb/testsuite/ChangeLog:
* gdb.dlang/circular.c: New file.
* gdb.dlang/circular.exp: New file.

Automatic date update in version.in

Automatic date update in version.in

Fix CXX invalid cast from void *.

This is an obvious patch to fix the following build error seen with
--enable-build-with-cxx:

../../src/gdb/rs6000-tdep.c: In function ‘rs6000_frame_cache* rs6000_frame_cache(frame_info*, void**)’:
../../src/gdb/rs6000-tdep.c:3242:15: error: invalid conversion from ‘void*’ to ‘rs6000_frame_cache*’ [-fpermissive]
       return (*this_cache);
              ~^~~~~~~~~~~~

gdb/ChangeLog

* rs6000-tdep.c (rs6000_frame_cache): Explicitly cast return result
to avoid invalid conversion from void *.

Add aarch64-*-rtems* target

Add x86_64-*-rtems* target

ld/configure.tgt: Also add stanza for x86_64-*-elf.

[BFD][AARCH64]Fix MOVW_SABS_G(0,1,2) relocation overflow check.

For these three relocations, 17 bit signed value should be used, instead of
16 bit. The bitsize field is changed from 16 to 17, this field in aarch64
backend is used for overflow check only.

bfd/

2016-02-26  Renlin Li  <renlin.li@arm.com>

    * elfnn-aarch64.c (elfNN_aarch64_howto_table): Fix signed overflow
    check for MOVW_SABS_G0, MOVW_SABS_G1, MOVW_SABS_G2.

ld/

2016-02-26  Renlin Li  <renlin.li@arm.com>

    * testsuite/ld-aarch64/aarch64-elf.exp: Run new testcases.
    * testsuite/ld-aarch64/emit-relocs-270.d: Update to use new boundary.
    * testsuite/ld-aarch64/emit-relocs-271.d: Likewise.
    * testsuite/ld-aarch64/emit-relocs-272.d: Likewise.
    * testsuite/ld-aarch64/emit-relocs-270-overflow.d: New.
    * testsuite/ld-aarch64/emit-relocs-270-overflow.s: New.
    * testsuite/ld-aarch64/emit-relocs-271-overflow.d: New.
    * testsuite/ld-aarch64/emit-relocs-271-overflow.s: New.
    * testsuite/ld-aarch64/emit-relocs-272-overflow.d: New.
    * testsuite/ld-aarch64/emit-relocs-272-overflow.s: New.

Optimize x86 GOT32X/GOTPCRELX relocations

R_386_GOT32X, R_X86_64_GOTPCRELX and R_X86_64_REX_GOTPCRELX relocations
retrieve the symbol address via its GOT slot.  If the symbol address is
known at the link-time, we can use it directly by changing instruction
encoding.  Indirect branch can only be converted to PC relative direct
branch.  MOV can be changed to LEA or encoded differently with signed
address.  The subset of binary operations can be encoded only with
signed address.

If undefined weak symbol is resolved to zero link-time, we can use it
as address.  Zero addresss can't used with PC relative direct branch
when PIC is true since the current PC is unknown.  In 64-bit, 32-bit
relocation for PC relatiave direct branch to zero may also overflow.

If this optimization causes relocation overflow, --no-relax can be used
to work around it.

bfd/

PR ld/19609
* elf32-i386.c (elf_i386_convert_load): Convert to R_386_32 for
load with locally bound symbols if PIC is false or there is no
base register.  Optimize branch to 0 if PIC is false.
(elf_i386_relocate_section): Don't generate dynamic relocations
against undefined weak symbols if PIC is false.
* elf64-x86-64.c (elf_x86_64_convert_load): Disable optimization
if we can't estimate relocation overflow with --no-relax.
Convert to R_X86_64_32S/R_X86_64_32 for load with locally bound
symbols if PIC is false.  Optimize branch to 0 if PIC is false.
(elf_x86_64_relocate_section): Don't generate dynamic relocations
against undefined weak symbols if PIC is false.

ld/

PR ld/19609
* testsuite/ld-i386/got1.dd: Updated.
* testsuite/ld-i386/lea1c.d: Likewise.
* testsuite/ld-i386/load1-nacl.d: Likewise.
* testsuite/ld-i386/load1.d: Likewise.
* testsuite/ld-i386/load4b.d: Likewise.
* testsuite/ld-i386/load5b.d: Likewise.
* testsuite/ld-i386/mov1b.d: Likewise.
* testsuite/ld-x86-64/mov1b.d: Likewise.
* testsuite/ld-x86-64/mov1d.d: Likewise.
* testsuite/ld-ifunc/ifunc-21-i386.d: Likewise.
* testsuite/ld-ifunc/ifunc-21-x86-64.d: Likewise.
* testsuite/ld-ifunc/ifunc-22-i386.d: Likewise.
* testsuite/ld-ifunc/ifunc-22-x86-64.d: Likewise.
* testsuite/ld-x86-64/gotpcrel1.dd: Likewise.
* testsuite/ld-x86-64/lea1a.d: Likewise.
* testsuite/ld-x86-64/lea1b.d: Likewise.
* testsuite/ld-x86-64/lea1c.d: Likewise.
* testsuite/ld-x86-64/lea1d.d: Likewise.
* testsuite/ld-x86-64/lea1e.d: Likewise.
* testsuite/ld-x86-64/lea1f.d: Likewise.
* testsuite/ld-x86-64/mov1b.d: Likewise.
* testsuite/ld-x86-64/mov1d.d: Likewise.
* testsuite/ld-x86-64/pr13082-3b.d: Likewise.
* testsuite/ld-x86-64/pr13082-4b.d: Likewise.
* testsuite/ld-x86-64/lea1.s: Add tests for 32-bit registers.
* testsuite/ld-i386/pr19609-1.s: New file.
* testsuite/ld-i386/pr19609-1a.d: Likewise.
* testsuite/ld-i386/pr19609-1b.d: Likewise.
* testsuite/ld-i386/pr19609-1c.d: Likewise.
* testsuite/ld-i386/pr19609-1d.d: Likewise.
* testsuite/ld-i386/pr19609-1e.d: Likewise.
* testsuite/ld-i386/pr19609-1f.d: Likewise.
* testsuite/ld-i386/pr19609-1g.d: Likewise.
* testsuite/ld-i386/pr19609-1h.d: Likewise.
* testsuite/ld-i386/pr19609-1i.d: Likewise.
* testsuite/ld-i386/pr19609-2.s: Likewise.
* testsuite/ld-i386/pr19609-2a.d: Likewise.
* testsuite/ld-i386/pr19609-2b.d: Likewise.
* testsuite/ld-i386/pr19609-2c.d: Likewise.
* testsuite/ld-i386/undefweak.s: Likewise.
* testsuite/ld-i386/undefweaka.d: Likewise.
* testsuite/ld-i386/undefweakb.d: Likewise.
* testsuite/ld-x86-64/pr13082-3c.d: Likewise.
* testsuite/ld-x86-64/pr13082-3d.d: Likewise.
* testsuite/ld-x86-64/pr19609-1.s: Likewise.
* testsuite/ld-x86-64/pr19609-1a.d: Likewise.
* testsuite/ld-x86-64/pr19609-1b.d: Likewise.
* testsuite/ld-x86-64/pr19609-1c.d: Likewise.
* testsuite/ld-x86-64/pr19609-1d.d: Likewise.
* testsuite/ld-x86-64/pr19609-1e.d: Likewise.
* testsuite/ld-x86-64/pr19609-1f.d: Likewise.
* testsuite/ld-x86-64/pr19609-1g.d: Likewise.
* testsuite/ld-x86-64/pr19609-1h.d: Likewise.
* testsuite/ld-x86-64/pr19609-1i.d: Likewise.
* testsuite/ld-x86-64/pr19609-1j.d: Likewise.
* testsuite/ld-x86-64/pr19609-1k.d: Likewise.
* testsuite/ld-x86-64/pr19609-1l.d: Likewise.
* testsuite/ld-x86-64/pr19609-1m.d: Likewise.
* testsuite/ld-x86-64/pr19609-2.s: Likewise.
* testsuite/ld-x86-64/pr19609-2a.d: Likewise.
* testsuite/ld-x86-64/pr19609-2b.d: Likewise.
* testsuite/ld-x86-64/pr19609-2c.d: Likewise.
* testsuite/ld-x86-64/pr19609-2d.d: Likewise.
* testsuite/ld-x86-64/pr19609-3.s: Likewise.
* testsuite/ld-x86-64/pr19609-3a.d: Likewise.
* testsuite/ld-x86-64/pr19609-3b.d: Likewise.
* testsuite/ld-x86-64/pr19609-4.s: Likewise.
* testsuite/ld-x86-64/pr19609-4a.d: Likewise.
* testsuite/ld-x86-64/pr19609-4b.d: Likewise.
* testsuite/ld-x86-64/pr19609-4c.d: Likewise.
* testsuite/ld-x86-64/pr19609-4d.d: Likewise.
* testsuite/ld-x86-64/pr19609-4e.d: Likewise.
* testsuite/ld-x86-64/pr19609-5.s: Likewise.
* testsuite/ld-x86-64/pr19609-5a.d: Likewise.
* testsuite/ld-x86-64/pr19609-5b.d: Likewise.
* testsuite/ld-x86-64/pr19609-5c.d: Likewise.
* testsuite/ld-x86-64/pr19609-5d.d: Likewise.
* testsuite/ld-x86-64/pr19609-5e.d: Likewise.
* testsuite/ld-x86-64/pr19609-6.s: Likewise.
* testsuite/ld-x86-64/pr19609-6a.d: Likewise.
* testsuite/ld-x86-64/pr19609-6b.d: Likewise.
* testsuite/ld-x86-64/pr19609-6c.d: Likewise.
* testsuite/ld-x86-64/pr19609-6d.d: Likewise.
* testsuite/ld-x86-64/pr19609-7.s: Likewise.
* testsuite/ld-x86-64/pr19609-7a.d: Likewise.
* testsuite/ld-x86-64/pr19609-7b.d: Likewise.
* testsuite/ld-x86-64/pr19609-7c.d: Likewise.
* testsuite/ld-x86-64/pr19609-7d.d: Likewise.
* testsuite/ld-i386/i386.exp: Run undefweak tests and tests for
PR ld/19609.
* testsuite/ld-x86-64/x86-64.exp: Run pr13082-3c, pr13082-3d
and tests for PR ld/19609.

Discard FDEs for zero-length address ranges.

2016-02-26  Egor Kochetov  <egor.kochetov@intel.com>
    Cary Coutant  <ccoutant@gmail.com>

gold/
PR gold/19735
* ehframe.h (Cie::fde_encoding): New method.
* ehframe.cc (Eh_frame::read_fde): Discard FDEs for zero-length
address ranges.

Fix various bugs in arm_record_exreg_ld_st_insn

This patch fixes various bugs in arm_record_exreg_ld_st_insn, and use
gdb.reverse/insn-reverse.c to test more arm instructions.

 - Set flag SINGLE_REG correctly.  In the arch reference manual,
   SING_REG is true when the bit 8 of instruction is zero.
 - Record the right D registers for instructions changing S registers.
 - Fix the order of length and address in record_buf_mem array.
 - Shift the offset by 2 instead of by 24.

This patch also fixes one internal error,

(gdb) PASS: gdb.reverse/finish-precsave.exp: BP at end of main
continue^M
Continuing.^M
../../binutils-gdb/gdb/utils.c:1072: internal-error: virtual memory exhausted.^M
A problem internal to GDB has been detected,FAIL: gdb.reverse/finish-precsave.exp: run to end of main (GDB internal error)

gdb:

2016-02-26  Yao Qi  <yao.qi@linaro.org>

* arm-tdep.c (arm_record_exreg_ld_st_insn): Set 'single_reg'
per bit 8.  Check bit 20 instead of bit 4 for VMOV
instruction.  Record D registers for instructions changing
S registers.  Change of the order of length and address
in record_buf_mem array.

gdb/testsuite:

2016-02-26  Yao Qi  <yao.qi@linaro.org>

* gdb.reverse/insn-reverse.c [__arm__] (ext_reg_load): New.
[__arm__] (ext_reg_mov, ext_reg_push_pop): New.
(testcases): Update.

Record right reg num of thumb special data instructions

When GDB decodes these thumb special data instructions, such as 'mov sp, r7'
the Rd is got incorrectly.  According to the arch reference manual, the Rd
is DN:Rdn, in which DN is bit 7 and Rdn is bits 0 to 2.  This patch fixes it.

gdb:

2016-02-26  Yao Qi  <yao.qi@linaro.org>

* arm-tdep.c (thumb_record_ld_st_reg_offset): Fix the register
number of Rd.

Rename gdb.reverse/aarch64.{exp,c} to gdb.reverse/insn-reverse.{exp,c}

gdb/testsuite:

2016-02-26  Yao Qi  <yao.qi@linaro.org>

* gdb.reverse/aarch64.c: Rename to ...
* gdb.reverse/insn-reverse.c: ... it.
* gdb.reverse/aarch64.exp: Rename to ...
* gdb.reverse/insn-reverse.exp: ... it.

Generalize gdb.reverse/aarch64.exp

I said we can generialize gdb.reverse/aarch64.exp for other
architectures https://sourceware.org/ml/gdb-patches/2015-05/msg00482.html
and here is the patch to change aarch64.exp so that it can be used to
test for other architectures as well.

gdb/testsuite:

2016-02-26  Yao Qi  <yao.qi@linaro.org>

* gdb.reverse/aarch64.c: [__aarch64__] Include arm_neon.h.
(testcase_ftype): New.
(testcases): New array.
(n_testcases): New.
(main): Call each element in testcases.
* gdb.reverse/aarch64.exp: Remove is_aarch64_target check.
(read_testcase): New.
Do the tests in a loop.

Add ChangeLog entries for PR ld/19645

Properly implement STT_COMMON

The BFD configure option, --enable-elf-stt-common, can't be to used to
verify STT_COMMON implementation with the normal binutils build.  Instead,
this patch removes it from BFD.  It adds --elf-stt-common=[no|yes] to ELF
assembler/objcopy and adds -z common/-z nocommon to ld.

A configure option, --enable-elf-stt-common, is added to gas to specify
whether ELF assembler should generate common symbols with the STT_COMMON
type by default.

Since BSF_KEEP_G is never used, it is renamed to BSF_ELF_COMMON for ELF
common symbols.

bfd/

PR ld/19645
* bfd.c (bfd): Change flags to 20 bits.
(BFD_CONVERT_ELF_COMMON): New.
(BFD_USE_ELF_STT_COMMON): Likewise.
(BFD_FLAGS_SAVED): Add BFD_CONVERT_ELF_COMMON and
BFD_USE_ELF_STT_COMMON.
(BFD_FLAGS_FOR_BFD_USE_MASK): Likewise.
* configure.ac: Remove --enable-elf-stt-common.
* elf.c (swap_out_syms): Choose STT_COMMON or STT_OBJECT for
common symbol depending on BFD_CONVERT_ELF_COMMON and
BFD_USE_ELF_STT_COMMON.
* elfcode.h (elf_slurp_symbol_table): Set BSF_ELF_COMMON for
STT_COMMON.
* elflink.c (bfd_elf_link_mark_dynamic_symbol): Also check
STT_COMMON.
(elf_link_convert_common_type): New function.
(elf_link_output_extsym): Choose STT_COMMON or STT_OBJECT for
common symbol depending on BFD_CONVERT_ELF_COMMON and
BFD_USE_ELF_STT_COMMON.  Set sym.st_info after sym.st_shndx.
* elfxx-target.h (TARGET_BIG_SYM): Add BFD_CONVERT_ELF_COMMON
and BFD_USE_ELF_STT_COMMON to object_flags.
(TARGET_LITTLE_SYM): Likewise.
* syms.c (BSF_KEEP_G): Renamed to ...
(BSF_ELF_COMMON): This.
* bfd-in2.h: Regenerated.
* config.in: Likewise.
* configure: Likewise.

binutils/

PR ld/19645
* NEWS: Mention --elf-stt-common= for objcopy.
* doc/binutils.texi: Document --elf-stt-common= for objcopy.
* objcopy.c (do_elf_stt_common): New.
(command_line_switch): Add OPTION_ELF_STT_COMMON.
(copy_options): Add --elf-stt-common=.
(copy_usage): Add --elf-stt-common=.
(copy_object): Also check do_elf_stt_common for ELF targets.
(copy_file): Handle do_elf_stt_common.
(copy_main): Handle OPTION_ELF_STT_COMMON.
* readelf.c (apply_relocations): Support STT_COMMON.
* testsuite/binutils-all/common-1.s: New file.
* testsuite/binutils-all/common-1a.d: Likewise.
* testsuite/binutils-all/common-1b.d: Likewise.
* testsuite/binutils-all/common-1c.d: Likewise.
* testsuite/binutils-all/common-1d.d: Likewise.
* testsuite/binutils-all/common-1e.d: Likewise.
* testsuite/binutils-all/common-1f.d: Likewise.
* testsuite/binutils-all/common-2.s: Likewise.
* testsuite/binutils-all/common-2a.d: Likewise.
* testsuite/binutils-all/common-2b.d: Likewise.
* testsuite/binutils-all/common-2c.d: Likewise.
* testsuite/binutils-all/common-2d.d: Likewise.
* testsuite/binutils-all/common-2e.d: Likewise.
* testsuite/binutils-all/common-2f.d: Likewise.
* testsuite/binutils-all/objcopy.exp
(objcopy_test_elf_common_symbols): New proc.
Run objcopy_test_elf_common_symbols for ELF targets

gas/

PR ld/19645
* NEWS: Mention --enable-elf-stt-common and --elf-stt-common=
for ELF assemblers.
* as.c (flag_use_elf_stt_common): New.
(show_usage): Add --elf-stt-common=.
(option_values): Add OPTION_ELF_STT_COMMON.
(std_longopts): Add --elf-stt-common=.
(parse_args): Handle --elf-stt-common=.
* as.h (flag_use_elf_stt_common): New.
* config.in: Regenerated.
* configure: Likewise.
* configure.ac: Add --enable-elf-stt-common and define
DEFAULT_GENERATE_ELF_STT_COMMON.
* gas/write.c (write_object_file): Set BFD_CONVERT_ELF_COMMON
and BFD_USE_ELF_STT_COMMON if flag_use_elf_stt_common is set.
* doc/as.texinfo: Document --elf-stt-common=.
* testsuite/gas/elf/common3.s: New file.
* testsuite/gas/elf/common3a.d: Likewise.
* testsuite/gas/elf/common3b.d: Likewise.
* testsuite/gas/elf/common4.s: Likewise.
* testsuite/gas/elf/common4a.d: Likewise.
* testsuite/gas/elf/common4b.d: Likewise.
* testsuite/gas/i386/dw2-compress-3b.d: Likewise.
* testsuite/gas/i386/dw2-compressed-3b.d: Likewise.
* testsuite/gas/elf/elf.exp: Run common3a, common3b, common4a
and common4b.
* testsuite/gas/i386/dw2-compress-3.d: Renamed to ...
* testsuite/gas/i386/dw2-compress-3a.d: This.  Pass
--elf-stt-common=no to as.
* testsuite/gas/i386/dw2-compressed-3.d: Renamed to ...
* testsuite/gas/i386/dw2-compressed-3a.d: This.  Pass
--elf-stt-common=no to as.
* testsuite/gas/i386/i386.exp: Run dw2-compress-3a,
dw2-compress-3b, dw2-compressed-3a and dw2-compressed-3b instead
of dw2-compress-3 and dw2-compressed-3.

include/

PR ld/19645
* bfdlink.h (bfd_link_elf_stt_common): New enum.
(bfd_link_info): Add elf_stt_common.

ld/

PR ld/19645
* NEWS: Mention -z common/-z nocommon for ELF targets.
* emultempl/elf32.em (gld${EMULATION_NAME}_handle_option): Handle
-z common and -z nocommon.
* ld.texinfo: Document -z common/-z nocommon.
* lexsup.c (elf_shlib_list_options): Add -z common/-z nocommon.
* testsuite/ld-elf/tls_common.exp: Test --elf-stt-common=no and
--elf-stt-common=yes with assembler.
* testsuite/ld-elfcomm/common-1.s: New file.
* testsuite/ld-elfcomm/common-1a.d: Likewise.
* testsuite/ld-elfcomm/common-1b.d: Likewise.
* testsuite/ld-elfcomm/common-1c.d: Likewise.
* testsuite/ld-elfcomm/common-1d.d: Likewise.
* testsuite/ld-elfcomm/common-1e.d: Likewise.
* testsuite/ld-elfcomm/common-1f.d: Likewise.
* testsuite/ld-elfcomm/common-2.s: Likewise.
* testsuite/ld-elfcomm/common-2a.d: Likewise.
* testsuite/ld-elfcomm/common-2b.d: Likewise.
* testsuite/ld-elfcomm/common-2c.d: Likewise.
* testsuite/ld-elfcomm/common-2d.d: Likewise.
* testsuite/ld-elfcomm/common-2e.d: Likewise.
* testsuite/ld-elfcomm/common-2f.d: Likewise.
* testsuite/ld-elfcomm/common-3a.rd: Likewise.
* testsuite/ld-elfcomm/common-3b.rd: Likewise.
* testsuite/ld-i386/pr19645.d: Likewise.
* testsuite/ld-i386/pr19645.s: Likewise.
* testsuite/ld-x86-64/largecomm-1.s: Likewise.
* testsuite/ld-x86-64/largecomm-1a.d: Likewise.
* testsuite/ld-x86-64/largecomm-1b.d: Likewise.
* testsuite/ld-x86-64/largecomm-1c.d: Likewise.
* testsuite/ld-x86-64/largecomm-1d.d: Likewise.
* testsuite/ld-x86-64/largecomm-1e.d: Likewise.
* testsuite/ld-x86-64/largecomm-1f.d: Likewise.
* testsuite/ld-x86-64/pr19645.d: Likewise.
* testsuite/ld-x86-64/pr19645.s: Likewise.
* testsuite/ld-elfcomm/elfcomm.exp: Test --elf-stt-common=yes
with assembler.
(assembler_generates_commons): Removed.
Run -z common/-z nocommon tests.  Run *.d tests.
* testsuite/ld-i386/i386.exp: Run pr19645.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-x86-64/dwarfreloc.exp: Test --elf-stt-common with
assembler.  Test STT_COMMON with readelf.

[x86] Resolve non-PIC undefweak symbols in executable

For i386 and x86-64, non-PIC references to undefined weak symbols are
resolved without dynamic relocation when creating executable.  Resolved
undefined weak symbols are removed from the dynamic symbol table in
executable.  One exception is on i386, we need resolved undefined weak
symbols in the dynamic symbol table in PIE if input relocatable files
contain branchs without PLT so that we can branch to 0 with dynamic
relocation in text section.

This makes behaviors of dynamic executable and position independent
executable predictable with mixed PIC and non-PIC references to undefined
weak symbols.  If all references to undefined weak symbols are PIC,
dynamic relocations against undefined weak symbols will be generated
in executable unless -z nodynamic-undefined-weak is passed to linker.

bfd/

PR ld/19636
PR ld/19704
PR ld/19719
* elf32-i386.c (UNDEFINED_WEAK_RESOLVED_TO_ZERO): New.
(elf_i386_link_hash_entry): Add has_got_reloc and
has_non_got_reloc.
(elf_i386_link_hash_table): Add interp.
(elf_i386_link_hash_newfunc): Initialize has_got_reloc and
has_non_got_reloc.
(elf_i386_copy_indirect_symbol): Copy has_got_reloc and
has_non_got_reloc.
(elf_i386_check_relocs): Set has_got_reloc and has_non_got_reloc.
(elf_i386_fixup_symbol): New function.
(elf_i386_pie_finish_undefweak_symbol): Likewise.
(elf_i386_allocate_dynrelocs): Don't allocate space for dynamic
relocations and discard relocations against resolved undefined
weak symbols in executable.  Don't make resolved undefined weak
symbols in executable dynamic.  Keep dynamic non-GOT/non-PLT
relocation against undefined weak symbols in PIE.
(elf_i386_size_dynamic_sections): Set interp to .interp section.
(elf_i386_relocate_section): Don't generate dynamic relocations
against resolved undefined weak symbols in PIE, except for
R_386_PC32.
(elf_i386_finish_dynamic_symbol): Keep PLT/GOT entries without
dynamic PLT/GOT relocations for resolved undefined weak symbols.
Don't generate dynamic relocation against resolved undefined weak
symbol in executable.
(elf_i386_finish_dynamic_sections): Call
elf_i386_pie_finish_undefweak_symbol on all symbols in PIE.
(elf_backend_fixup_symbol): New.
* elf64-x86-64.c (UNDEFINED_WEAK_RESOLVED_TO_ZERO): New.
(elf_x86_64_link_hash_entry): Add has_got_reloc and
has_non_got_reloc.
(elf_x86_64_link_hash_table): Add interp.
(elf_x86_64_link_hash_newfunc): Initialize has_got_reloc and
has_non_got_reloc.
(elf_x86_64_copy_indirect_symbol): Copy has_got_reloc and
has_non_got_reloc.
(elf_x86_64_check_relocs): Set has_got_reloc and
has_non_got_reloc.
(elf_x86_64_fixup_symbol): New function.
(elf_x86_64_pie_finish_undefweak_symbol): Likewise.
(elf_x86_64_allocate_dynrelocs): Don't allocate space for dynamic
relocations and discard relocations against resolved undefined
weak symbols in executable.  Don't make resolved undefined weak
symbols in executable dynamic.
(elf_x86_64_size_dynamic_sections): Set interp to .interp section.
(elf_x86_64_relocate_section): Check relocation overflow for
dynamic relocations against unresolved weak undefined symbols.
Don't generate dynamic relocations against resolved weak
undefined symbols in PIE.
(elf_x86_64_finish_dynamic_symbol): Keep PLT/GOT entries without
dynamic PLT/GOT relocations for resolved undefined weak symbols.
Don't generate dynamic relocation against resolved undefined weak
symbol in executable.
(elf_x86_64_finish_dynamic_sections): Call
elf_x86_64_pie_finish_undefweak_symbol on all symbols in PIE.
(elf_backend_fixup_symbol): New.

include/

PR ld/19636
PR ld/19704
PR ld/19719
* bfdlink.h (bfd_link_info): Add dynamic_undefined_weak.

ld/

PR ld/19636
PR ld/19704
PR ld/19719
* Makefile.am (ELF_X86_DEPS): Add dynamic_undefined_weak.sh.
* Makefile.in: Regenerated.
* NEWS: Mention -z nodynamic-undefined-weak.
* ld.texinfo: Document -z nodynamic-undefined-weak.
* ldmain.c (main): Initialize dynamic_undefined_weak to -1.
* emulparams/dynamic_undefined_weak.sh: New file.
* emulparams/elf32_x86_64.sh: Source dynamic_undefined_weak.sh.
* emulparams/elf_i386.sh: Likewise.
* emulparams/elf_i386_be.sh: Likewise.
* emulparams/elf_i386_chaos.sh: Likewise.
* emulparams/elf_i386_ldso.sh: Likewise.
* emulparams/elf_i386_vxworks.sh: Likewise.
* emulparams/elf_iamcu.sh: Likewise.
* emulparams/elf_k1om.sh: Likewise.
* emulparams/elf_l1om.sh: Likewise.
* emulparams/elf_x86_64.sh: Likewise.
* emulparams/extern_protected_data.sh (PARSE_AND_LIST_OPTIONS):
Append.
(PARSE_AND_LIST_ARGS_CASE_Z): Likewise.
* testsuite/ld-elf/pr19719a.c: New file.
* testsuite/ld-elf/pr19719b.c: Likewise.
* testsuite/ld-elf/pr19719c.c: Likewise.
* testsuite/ld-elf/pr19719d.c: Likewise.
* testsuite/ld-i386/pr19636-1.s: Likewise.
* testsuite/ld-i386/pr19636-1a.d: Likewise.
* testsuite/ld-i386/pr19636-1b.d: Likewise.
* testsuite/ld-i386/pr19636-1c.d: Likewise.
* testsuite/ld-i386/pr19636-1d-nacl.d: Likewise.
* testsuite/ld-i386/pr19636-1d.d: Likewise.
* testsuite/ld-i386/pr19636-1e.d: Likewise.
* testsuite/ld-i386/pr19636-1f.d: Likewise.
* testsuite/ld-i386/pr19636-1g.d: Likewise.
* testsuite/ld-i386/pr19636-1h.d: Likewise.
* testsuite/ld-i386/pr19636-1i.d: Likewise.
* testsuite/ld-i386/pr19636-2.s: Likewise.
* testsuite/ld-i386/pr19636-2a.d: Likewise.
* testsuite/ld-i386/pr19636-2b.d: Likewise.
* testsuite/ld-i386/pr19636-2c-nacl.d: Likewise.
* testsuite/ld-i386/pr19636-2c.d: Likewise.
* testsuite/ld-i386/pr19636-2d-nacl.d: Likewise.
* testsuite/ld-i386/pr19636-2d.d: Likewise.
* testsuite/ld-i386/pr19636-2e-nacl.d: Likewise.
* testsuite/ld-i386/pr19636-2e.d: Likewise.
* testsuite/ld-i386/pr19636-3.s: Likewise.
* testsuite/ld-i386/pr19636-3a.d: Likewise.
* testsuite/ld-i386/pr19636-3b.d: Likewise.
* testsuite/ld-i386/pr19636-3c.d: Likewise.
* testsuite/ld-i386/pr19636-3d.d: Likewise.
* testsuite/ld-i386/pr19636-3e.d: Likewise.
* testsuite/ld-i386/pr19636-3f.d: Likewise.
* testsuite/ld-i386/pr19636-3g.d: Likewise.
* testsuite/ld-i386/pr19636-4.s: Likewise.
* testsuite/ld-i386/pr19636-4a.d: Likewise.
* testsuite/ld-i386/pr19636-4b.d: Likewise.
* testsuite/ld-i386/pr19636-4c.d: Likewise.
* testsuite/ld-i386/pr19636-4d.d: Likewise.
* testsuite/ld-i386/pr19704.out: Likewise.
* testsuite/ld-i386/pr19704a.c: Likewise.
* testsuite/ld-i386/pr19704b.c: Likewise.
* testsuite/ld-x86-64/pr19636-1.s: Likewise.
* testsuite/ld-x86-64/pr19636-1a.d: Likewise.
* testsuite/ld-x86-64/pr19636-1b.d: Likewise.
* testsuite/ld-x86-64/pr19636-1c.d: Likewise.
* testsuite/ld-x86-64/pr19636-1d.d: Likewise.
* testsuite/ld-x86-64/pr19636-1e.d: Likewise.
* testsuite/ld-x86-64/pr19636-1f.d: Likewise.
* testsuite/ld-x86-64/pr19636-1g.d: Likewise.
* testsuite/ld-x86-64/pr19636-2.s: Likewise.
* testsuite/ld-x86-64/pr19636-2a.d: Likewise.
* testsuite/ld-x86-64/pr19636-2b.d: Likewise.
* testsuite/ld-x86-64/pr19636-2c.d: Likewise.
* testsuite/ld-x86-64/pr19636-2d-nacl.d: Likewise.
* testsuite/ld-x86-64/pr19636-2d.d: Likewise.
* testsuite/ld-x86-64/pr19636-2e.d: Likewise.
* testsuite/ld-x86-64/pr19636-2f.d: Likewise.
* testsuite/ld-x86-64/pr19636-2g.d: Likewise.
* testsuite/ld-x86-64/pr19636-2h.d: Likewise.
* testsuite/ld-x86-64/pr19636-2i.d: Likewise.
* testsuite/ld-x86-64/pr19636-3.s: Likewise.
* testsuite/ld-x86-64/pr19636-3a.d: Likewise.
* testsuite/ld-x86-64/pr19636-3b.d: Likewise.
* testsuite/ld-x86-64/pr19636-3c.d: Likewise.
* testsuite/ld-x86-64/pr19636-3d.d: Likewise.
* testsuite/ld-x86-64/pr19704.out: Likewise.
* testsuite/ld-x86-64/pr19704a.c: Likewise.
* testsuite/ld-x86-64/pr19704b.c: Likewise.
* testsuite/ld-elf/shared.exp (mix_pic_and_non_pic): New.
Run mix_pic_and_non_pic.
* testsuite/ld-i386/i386.exp (undefined_weak): New.
Run undefined_weak and PR ld/19636 tests.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-x86-64/pr13082-3b.d: Updated.
* testsuite/ld-x86-64/pr13082-4b.d: Likewise.

Fix powerpc64 -r --save-restore-funcs

* elf64-ppc.c (create_linkage_sections): Create sfpr when
save_restore_funcs, rest of sections when not relocatable.
(ppc64_elf_init_stub_bfd): Always call create_linkage_sections.
(sfpr_define): Define all symbols on emitted code.
(ppc64_elf_func_desc_adjust): Adjust for sfpr now being created
when relocatable.  Move sfpr_define loop earlier.

Automatic date update in version.in

gdbserver/s390: Enable high GPRs, VX, TDB with 31-bit gdbserver.

Currently, 31-bit gdbserver doesn't support collecting/supplying high
GPRs, VX registers, and TDB data.  This is not much of a problem now,
since machines that have them usually have a 64-bit gdbserver that can
be used to debug 31-bit targets just fine.  However, with fast
tracepoints, it's not possible to use a 64-bit gdbserver with a 31-bit
IPA (and thus a 31-bit target), so 31-bit gdbserver has to be used
for 31-bit targets.  Thus, this patch is needed to allow collecting
high GPRs and VX registers on 31-bit targets via fast tracepoints.

gdb/gdbserver/ChangeLog:

* linux-s390-low.c (s390_num_regs_3264): Define on 31-bit too.
(s390_regmap_3264) [!__s390x__]: New global.
(s390_collect_ptrace_register): Skip map entries containing -1.
(s390_supply_ptrace_register): Ditto.
(s390_fill_gprs_high): New function.
(s390_store_gprs_high): New function.
(s390_regsets): Add NT_S390_HIGH_GPRS.
(s390_get_hwcap): Enable on 31-bit.
(have_hwcap_s390_high_gprs): Enable on 31-bit.
(s390_arch_setup): Enable detection of high GPRs, TDB, VX on 31-bit.
Detect NT_S390_HIGH_GPRS.
(s390_usrregs_info_3264): Enable on 31-bit.
(s390_regs_info): Enable regs_info_3264 on 31-bit.
(initialize_low_arch): Initialize s390_regsets_info_3264 on 31-bit.

avoid compiler warnings in remote-m32r-sdi.c

gdb/ChangeLog:

* remote-m32r-sdi.c (recv_char_data): Initialize val to avoid
compiler warning.
(recv_long_data): Ditto.

Remove gdb.base/branches.c

This patch removes gdb.base/branches.c which was added by the following
commit, but it is not used at all.

commit ea8122af1432abdeb256b2c669eb3d0cf8cb97bf
Author: John Metzler <jmetzler@cygnus>
Date:   Thu Apr 16 17:56:11 1998 +0000

    Thu Apr 16 10:52:34 1998  John Metzler  <jmetzler@cygnus.com>

        * gdb.base/branches.c: Code with lots of loops and
        subroutines. Used to test gdbs ability to single step through PC
        changes, especially to test mips-tdep.c:mips_next_pc

gdb/testsuite:

2016-02-25  Yao Qi  <yao.qi@linaro.org>

* gdb.base/branches.c: Remove.

[PR gdb/13808] gdb.trace: Pass tdesc selected in gdbserver to IPA.

If gdbserver and IPA are using different tdesc, they will disagree
about 'R' trace packet size.  This results in mangled traces.

To make sure they pick the same tdesc, gdbserver pokes the tdesc
(specified as an index in a target-specific list) into a global
variable in IPA.  In theory, IPA could find out the tdesc on its
own, but that may be complex (in particular, I don't know how to
tell whether we have LAST_BREAK on s390 without messing with ptrace),
and we'd have to duplicate the logic.

Tested on i386 and x86_64.  On i386, it fixes two FAILs in ftrace.exp.
On x86_64, these failures have been KFAILed - one of them works now,
but the other now fails due to an unrelated reason (ugh).

gdb/gdbserver/ChangeLog:

PR gdb/13808
* Makefile.in: Add i386-*-linux-ipa.o and amd64-*-linux-ipa.o.
* configure.srv: Ditto.
* linux-aarch64-ipa.c (get_ipa_tdesc): New function.
(initialize_low_tracepoint): Remove ipa_tdesc assignment.
* linux-amd64-ipa.c: Add "linux-x86-tdesc.h" include.
(init_registers_amd64_linux): Remove prototype.
(tdesc_amd64_linux): Remove declaration.
(get_ipa_tdesc): New function.
(initialize_low_tracepoint): Remove ipa_tdesc assignment,
initialize remaining tdescs.
* linux-i386-ipa.c: Add "linux-x86-tdesc.h" include.
(init_registers_i386_linux): Remove prototype.
(tdesc_i386_linux): Remove declaration.
(get_ipa_tdesc): New function.
(initialize_low_tracepoint): Remove ipa_tdesc assignment,
initialize remaining tdescs.
* linux-low.c (linux_get_ipa_tdesc_idx): New function.
(linux_target_ops): wire in linux_get_ipa_tdesc_idx.
* linux-low.h (struct linux_target_ops): Add get_ipa_tdesc_idx.
* linux-x86-low.c: Move tdesc declarations to linux-x86-tdesc.h.
(x86_get_ipa_tdesc_idx): New function.
(the_low_target): Wire in x86_get_ipa_tdesc_idx.
* linux-x86-tdesc.h: New file.
* target.h (struct target_ops): Add get_ipa_tdesc_idx.
(target_get_ipa_tdesc_idx): New macro.
* tracepoint.c (ipa_tdesc_idx): New macro.
(struct ipa_sym_addresses): Add addr_ipa_tdesc_idx.
(symbol_list): Add ipa_tdesc_idx.
(cmd_qtstart): Write ipa_tdesc_idx in the target.
(ipa_tdesc): Remove.
(ipa_tdesc_idx): New variable.
(get_context_regcache): Use get_ipa_tdesc.
(gdb_collect): Ditto.
(gdb_probe): Ditto.
* tracepoint.h (get_ipa_tdesc): New prototype.
(ipa_tdesc): Remove.

gdb/testsuite/ChangeLog:

PR gdb/13808
* gdb.trace/ftrace.exp (test_fast_tracepoints): Remove kfail.

Convert more variables to a constant form.

* as.c (select_emulation_mode): Add const qualifiers.
* as.h: Likewise.
* config/bfin-defs.h: Likewise.
* config/bfin-parse.y: Likewise.
* config/rx-parse.y: Likewise.
* config/tc-aarch64.c (struct aarch64_option_table): Likewise.
(struct aarch64_cpu_option_table): Likewise.
(struct aarch64_arch_option_table): Likewise.
(struct aarch64_option_cpu_value_table): Likewise.
(struct aarch64_long_option_table): Likewise.
(struct aarch64_option_abi_value_table): Likewise.
* config/tc-arm.c (struct reloc_entry): Likewise.
(tc_gen_reloc): Likewise.
(struct arm_option_table): Likewise.
(struct arm_legacy_option_table): Likewise.
(struct arm_cpu_option_table): Likewise.
(struct arm_arch_option_table): Likewise.
(struct arm_option_extension_value_table): Likewise.
(struct arm_option_fpu_value_table): Likewise.
(struct arm_option_value_table): Likewise.
(struct arm_long_option_table): Likewise.
* config/tc-avr.c (struct avr_opcodes_s): Likewise.
(struct mcu_type_s): Likewise.
(struct exp_mod_s): Likewise.
(avr_operand): Likewise.
(avr_operands): Likewise.
* config/tc-d10v.c (md_begin): Likewise.
* config/tc-dlx.c: Likewise.
* config/tc-fr30.c (fr30_is_colon_insn): Likewise.
* config/tc-ft32.c (parse_condition): Likewise.
* config/tc-h8300.c (do_a_fix_imm): Likewise.
* config/tc-hppa.c (pa_ip): Likewise.
(hppa_regname_to_dw2regnum): Likewise.
* config/tc-i370.c (i370_elf_suffix): Likewise.
* config/tc-i960.c (struct tabentry): Likewise.
* config/tc-m32r.c: Likewise.
* config/tc-m68k.c: Likewise.
* config/tc-m68k.h: Likewise.
* config/tc-mcore.c (parse_psrmod): Likewise.
* config/tc-metag.c (struct metag_core_option): Likewise.
(struct metag_long_option): Likewise.
* config/tc-microblaze.c: Likewise.
* config/tc-mips.c (macro): Likewise.
* config/tc-mn10200.c: Likewise.
* config/tc-mn10300.c: Likewise.
* config/tc-msp430.c (struct rcodes_s): Likewise.
(struct hcodes_s): Likewise.
(md_parse_option): Likewise.
* config/tc-ns32k.c (struct ns32k_option): Likewise.
(optlist): Likewise.
* config/tc-ppc.c (ppc_elf_suffix): Likewise.
(tc_ppc_regname_to_dw2regnum): Likewise.
* config/tc-ppc.h: Likewise.
* config/tc-rl78.c: Likewise.
* config/tc-rx.c (struct cpu_type): Likewise.
* config/tc-sh.c (sh_regname_to_dw2regnum): Likewise.
* config/tc-sparc.c (struct priv_reg_entry): Likewise.
(sparc_ip): Likewise.
* config/tc-spu.c (insn_fmt_string): Likewise.
* config/tc-tic54x.c (tic54x_set_default_include): Likewise.
* config/tc-v850.c: Likewise.
* config/tc-visium.c (struct visium_arch_option_table): Likewise.
(struct visium_long_option_table): Likewise.
* config/tc-xgate.c: Likewise.
* config/tc-z8k.c: Likewise.
* read.c (add_include_dir): Likewise.
* read.h: Likewise.

Initialize variables in i386_linux_handle_segmentation_fault

We see this error when building with gcc 4.3.

../../gdb/i386-linux-tdep.c: In function ‘i386_linux_handle_segmentation_fault’:
../../gdb/i386-linux-tdep.c:399: error: ‘access’ may be used uninitialized in this function
../../gdb/i386-linux-tdep.c:399: error: ‘upper_bound’ may be used uninitialized in this function
../../gdb/i386-linux-tdep.c:399: error: ‘lower_bound’ may be used uninitialized in this function

It's a false positive, since the variables will always get initialized
in the TRY clause, and the CATCH returns.

gdb/ChangeLog:

* i386-linux-tdep.c (i386_linux_handle_segmentation_fault):
Initialize variables.

gdb.trace: Remove unnecessary target check from ftrace.exp.

The check used hardcoded targets and wasn't doing anything useful anyway,
since unsupported architectures blow up on link due to missing the IPA
library before they ever get to that check.

gdb/testsuite/ChangeLog:

* gdb.trace/ftrace.exp: Remove unnecessary target check.