platform/upstream/binutils.git
7 years agoFix heap-buffer overflow bugs caused when dumping debug information from a corrupt...
Nick Clifton [Fri, 28 Apr 2017 09:28:04 +0000 (10:28 +0100)]
Fix heap-buffer overflow bugs caused when dumping debug information from a corrupt binary.

PR binutils/21438
* dwarf.c (process_extended_line_op): Do not assume that the
string extracted from the section is NUL terminated.
(fetch_indirect_string): If the string retrieved from the section
is not NUL terminated, return an error message.
(fetch_indirect_line_string): Likewise.
(fetch_indexed_string): Likewise.

7 years agoUse ptid method lwp in mips_linux_new_thread
Yao Qi [Fri, 28 Apr 2017 08:50:51 +0000 (09:50 +0100)]
Use ptid method lwp in mips_linux_new_thread

gdb:

2017-04-28  Yao Qi  <yao.qi@linaro.org>

* mips-linux-nat.c (mips_linux_new_thread): Use ptid method
lwp instead of ptid_get_lwp.

7 years ago[MIPS] Use lwpid from lwp_info instead of inferior_ptid
Yao Qi [Fri, 28 Apr 2017 08:33:16 +0000 (09:33 +0100)]
[MIPS] Use lwpid from lwp_info instead of inferior_ptid

RAJESH reported that GDB gets "Couldn't write debug register: No such
process." on mips64 when GDB attaches to a multi threaded application.

Looks GDB nows PTRACE_GET_WATCH_REGS for inferior_ptid but
PTRACE_SET_WATCH_REGS for lwp->ptid, they may be different.

gdb:

2017-04-28  Yao Qi  <yao.qi@linaro.org>

* mips-linux-nat.c (mips_linux_new_thread): Get lwpid from
lwp_info instead of getting from inferior_ptid.

7 years agoAutomatic date update in version.in
GDB Administrator [Fri, 28 Apr 2017 00:00:34 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoFix overload resolution involving rvalue references and cv qualifiers.
Keith Seitz [Thu, 27 Apr 2017 22:55:26 +0000 (15:55 -0700)]
Fix overload resolution involving rvalue references and cv qualifiers.

The following patch fixes several outstanding overload resolution problems
with rvalue references and cv qualifiers in the test suite. The tests for
these problems typically passed with one compiler version and failed with
another. This behavior occurs because of the ordering of the overloaded
functions in the debug info. So the first best match "won out" over the
a subsequent better match.

One of the bugs addressed by this patch is the failure of rank_one_type to
account for type equality of two overloads based on CV qualifiers.  This was
leading directly to problems evaluating rvalue reference overload quality,
but it is also highlighted in gdb.cp/oranking.exp, where two test KFAIL as
a result of this shortcoming.

I found the overload resolution code committed with the rvalue reference
patch (f9aeb8d49) needlessly over-complicated, and I have greatly simplified
it. This fixes some KFAILing tests in gdb.exp/rvalue-ref-overload.exp.

gdb/ChangeLog

* gdbtypes.c (LVALUE_REFERENCE_TO_RVALUE_BINDING_BADNESS)
DIFFERENT_REFERENCE_TYPE_BADNESS): Remove.
(CV_CONVERSION_BADNESS): Define.
(rank_one_type): Remove overly restrictive rvalue reference
rank checks.
Add cv-qualifier checks and subranks for type equality.
* gdbtypes.h (REFERENCE_CONVERSION_RVALUE,
REFERENCE_CONVERSION_CONST_LVALUE, CV_CONVERSION_BADNESS,
CV_CONVERSION_CONST, CV_CONVERSION_VOLATILE): Declare.

gdb/testsuite/ChangeLog

* gdb.cp/oranking.cc (test15): New function.
(main): Call test15 and declare additional variables for testing.
* gdb.cp/oranking.exp: Remove kfail status for "p foo4(&a)" and
"p foo101('abc')" tests.
* gdb.cp/rvalue-ref-overloads.exp: Remove kfail status for
"lvalue reference overload" test.
* gdb.cp/rvalue-ref-params.exp: Remove kfail status for
"print value of f1 on Child&& in f2" test.

7 years agox86-64: Use "=" instead of "+=" to update 0
H.J. Lu [Thu, 27 Apr 2017 22:24:08 +0000 (15:24 -0700)]
x86-64: Use "=" instead of "+=" to update 0

Use

  if (htab->elf.splt->size == 0)
    htab->elf.splt->size = GET_PLT_ENTRY_SIZE (output_bfd);

instead of

  if (htab->elf.splt->size == 0)
    htab->elf.splt->size += GET_PLT_ENTRY_SIZE (output_bfd);

* elf64-x86-64.c (elf_x86_64_size_dynamic_sections): Use "="
instead of "+=" to update 0.

7 years agoAdd missing incref when creating Inferior Python object
Simon Marchi [Thu, 27 Apr 2017 21:03:20 +0000 (17:03 -0400)]
Add missing incref when creating Inferior Python object

The test py-inferior.exp fails when using a debug build of Python 3.6.  I don't
see it failing with my system's default Python, but it might be related to the
different memory allocation scheme used when doing a build with pydebug.

The issue is that we are missing a Py_INCREF in
inferior_to_inferior_object.  The PyObject_New function initializes the
object with a refcount of 1.  If we assume that this refcount
corresponds to the reference we are returning, then we are missing an
incref for the reference in the inferior data.

The counterpart for the incref that corresponds to the reference in the
inferior data is in py_free_inferior, in the form the gdbpy_ref instance.

Here's how I can get it to crash (with some debug output):

  $ ./gdb -nx -ex "set debug python 1"
  (gdb) add-inferior
  Added inferior 2
  (gdb) python infs = gdb.inferiors()
  Creating Python Inferior object inf = 1
  Creating Python Inferior object inf = 2
  (gdb) remove-inferiors 2
  py_free_inferior inf = 2
  infpy_dealloc inf = <unknown>
  (gdb) python infs = None
  Fatal Python error: Objects/tupleobject.c:243 object at 0x7f9cf1a568d8 has negative ref count -1

  Current thread 0x00007f9cf1b68780 (most recent call first):
    File "<string>", line 1 in <module>
  [1]    408 abort (core dumped)  ./gdb -nx -ex "set debug python 1"

After having created the inferiors object, their refcount is 1 (which
comes from PyObject_New), but it should be two.  The gdb inferior object
has a reference and the "infs" list has a reference.

When invoking remove-inferiors, py_free_inferior gets called.  It does
the decref that corresponds to the reference that the gdb inferior
object kept.  At this moment, the refcount drops to 0 and the object
gets deallocated, even though the "infs" list still has a reference.
When we set "infs" to None, Python tries to decref the already zero
refcount and the assert triggers.

With this patch, it looks better:

  (gdb) add-inferior
  Added inferior 2
  (gdb) python infs = gdb.inferiors()
  Creating Python Inferior object inf = 1
  Creating Python Inferior object inf = 2
  (gdb) remove-inferiors 2
  py_free_inferior inf = 2
  (gdb) python infs = None
  infpy_dealloc inf = <unknown>

gdb/ChangeLog:

* python/py-inferior.c (inferior_to_inferior_object): Increment reference
count when creating the object.

7 years agox86: Create dynamic sections in create_dynamic_sections
H.J. Lu [Thu, 27 Apr 2017 20:55:31 +0000 (13:55 -0700)]
x86: Create dynamic sections in create_dynamic_sections

This patch creates dynamic sections in i386/x86-64 create_dynamic_sections
instead of creating them on demend.  Linker will strip them if they are
empty.  It changes order in x86-64 .eh_frame section.  The extra DW_CFA_nop
paddings is due to

https://sourceware.org/bugzilla/show_bug.cgi?id=21441

bfd/

* elf32-i386.c (elf_i386_create_dynamic_sections): Create the
.plt.got section here.
(elf_i386_check_relocs): Don't create the .plt.got section.
* elf64-x86-64.c (elf_x86_64_create_dynamic_sections): Create
the .plt.got and .plt.bnd sections here.
(elf_x86_64_check_relocs): Don't create the .plt.got nor
.plt.bnd sections.

ld/

* testsuite/ld-x86-64/pr21038a.d: Update DW_CFA_nop paddings
in .eh_frame section.
* testsuite/ld-x86-64/pr21038c.d: Update .eh_frame order.

7 years agoRemove has_bnd_reloc from elf_x86_64_link_hash_entry
H.J. Lu [Thu, 27 Apr 2017 16:32:35 +0000 (09:32 -0700)]
Remove has_bnd_reloc from elf_x86_64_link_hash_entry

has_bnd_reloc was added to elf_x86_64_link_hash_entry track BND
relocations by

commit 0ff2b86e7c14177ec7f9e1257f8e697814794017
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Wed Nov 20 09:01:04 2013 -0800

    Create the second PLT for BND relocations

Since BND relocations have been deprecated by

commit d258b828287a863376af60a1ef7ceafbccc83d93
Author: Igor Zamyatin <igor.zamyatin@intel.com>
Date:   Tue Nov 18 10:52:36 2014 +0300

    Add -z bndplt to generate BND prefix in PLT entries

This patch removes has_bnd_reloc from elf_x86_64_link_hash_entry and
checks bndplt instead of has_bnd_reloc.

* elf64-x86-64.c (elf_x86_64_link_hash_entry): Remove
has_bnd_reloc.
(elf_x86_64_link_hash_newfunc): Don't clear has_bnd_reloc.
(elf_x86_64_copy_indirect_symbol): Don't copy has_bnd_reloc.
(elf_x86_64_check_relocs): Don't set has_bnd_reloc.
(elf_x86_64_finish_dynamic_symbol): Check bndplt instead of
has_bnd_reloc.

Remove has_bnd_reloc

7 years agoChange _bfd_elf_link_setup_gnu_properties to bfd *
H.J. Lu [Thu, 27 Apr 2017 16:07:23 +0000 (09:07 -0700)]
Change _bfd_elf_link_setup_gnu_properties to bfd *

Change setup_gnu_properties to return the first relocatable ELF input
with GNU properties so that a backend can make decision based on GNU
properties.

* elf-bfd.h (elf_backend_data): Change setup_gnu_properties
to return bfd *.
(_bfd_elf_link_setup_gnu_properties): Return bfd *.
* elf-properties.c (_bfd_elf_link_setup_gnu_properties): Return
the first relocatable ELF input with GNU properties.

7 years agoi386: Simplify VxWorks for non-PIC
H.J. Lu [Thu, 27 Apr 2017 15:55:36 +0000 (08:55 -0700)]
i386: Simplify VxWorks for non-PIC

Change

  if (PIC)
    {
      #1
    }
  else
    {
      #2
      if (VxWorks)
        {
          #3
        }
    }
  #4
  if (VxWorks && !PIC)
    {
      #5
    }

to

  #4
  if (PIC)
    {
      #1
    }
  else
    {
      #2
      if (VxWorks)
        {
          #3
          #5
        }
    }

* elf32-i386.c (elf_i386_finish_dynamic_sections): Simplify
VxWorks for non-PIC.

7 years agoRead corrrect auxiliary entry in AIX
Ulrich Weigand [Thu, 27 Apr 2017 13:57:08 +0000 (15:57 +0200)]
Read corrrect auxiliary entry in AIX

Fix handling of XCOFF function auxiliary entries, in particular when
the xlc -qfuncsect or gcc -ffunction-sections compiler option is used
in AIX.  Also handle C_WEAKEXT storage class.

gdb/
2016-10-21  Sangamesh Mallayya  <sangamesh.swamy@in.ibm.com>
    Ulrich Weigand  <uweigand@de.ibm.com>

* xcoffread.c (read_xcoff_symtab): Read correct function auxiliary
entry if xlc -qfuncsect or gcc -ffunction-sections compiler option
is used in AIX.
(read_xcoff_symtab): Handle C_WEAKEXT storage class.
(process_xcoff_symbol): Likewise.
(scan_xcoff_symtab): Likewise.

Signed-off-by: Ulrich Weigand <ulrich.weigand@de.ibm.com>
7 years agoMIPS16/GAS: Factor out duplicate symbol value conversion code
Maciej W. Rozycki [Thu, 27 Apr 2017 01:25:33 +0000 (02:25 +0100)]
MIPS16/GAS: Factor out duplicate symbol value conversion code

Factor out and consolidate duplicate section-relative to PC-relative
symbol value conversion in `mips16_extended_frag' and `md_convert_frag'
used for MIPS16 relaxation, observing that the final calculation in the
latter function implies `stretch == 0'.  Sanitize the formatting of code
moved.

gas/
* config/tc-mips.c (mips16_pcrel_val): New function, factored
out from...
(mips16_extended_frag): ... here.
(md_convert_frag): Use `mips16_pcrel_val' rather than repeated
code in MIPS16 relaxation, with `stretch' hardcoded to 0.

7 years agoMIPS16/GAS: Rename the LONG_BRANCH relaxation flag
Maciej W. Rozycki [Thu, 27 Apr 2017 01:13:21 +0000 (02:13 +0100)]
MIPS16/GAS: Rename the LONG_BRANCH relaxation flag

Following commit 177b4a6ad004 ("infinite loop in mips16 assembler
relaxation"), <https://sourceware.org/ml/binutils/2002-03/msg00345.html>
the LONG_BRANCH flag used in MIPS16 relaxation has lost its use for
branches.  Complement commit 88a7ef168927 ("MIPS16/GAS: Restore
unsupported relocation diagnostics") then, which has removed the remains
of code deactivated by the former commit, and rename the flag to
ALWAYS_EXTENDED, more accurately reflecting its current use to select
the extended form of PC-relative ADDIU, DADDIU, LD and LW instructions.

gas/
* config/tc-mips.c (RELAX_MIPS16_LONG_BRANCH): Rename to...
(RELAX_MIPS16_ALWAYS_EXTENDED): ... this.
(RELAX_MIPS16_MARK_LONG_BRANCH): Rename to...
(RELAX_MIPS16_MARK_ALWAYS_EXTENDED): ... this.
(RELAX_MIPS16_CLEAR_LONG_BRANCH): Rename to...
(RELAX_MIPS16_CLEAR_ALWAYS_EXTENDED): ... this.
(mips16_extended_frag): Adjust accordingly.

7 years agoRemove some MAX_REGISTER_SIZE uses in ia64-tdep.c
Alan Hayward [Thu, 27 Apr 2017 10:38:14 +0000 (11:38 +0100)]
Remove some MAX_REGISTER_SIZE uses in ia64-tdep.c

gdb/
* ia64-tdep.c (examine_prologue): Use get_frame_register_unsigned.
(ia64_sigtramp_frame_prev_register): Use read_memory_unsigned_integer.
(ia64_access_reg): Use get_frame_register_unsigned.
(ia64_access_rse_reg): Likewise.
(ia64_libunwind_frame_prev_register): Likewise.

7 years agoTidy S_FORCE_RELOC
Alan Modra [Thu, 27 Apr 2017 02:50:10 +0000 (12:20 +0930)]
Tidy S_FORCE_RELOC

Separate out symbol flag reasons from section reasons to force a
reloc.  Yes, this adds another section test to the local symbol case
too.

* symbols.c (S_FORCE_RELOC): Separate section and symbol tests.

7 years agoConstify elf_backend_eh_frame_address_size
Alan Modra [Thu, 27 Apr 2017 02:33:39 +0000 (12:03 +0930)]
Constify elf_backend_eh_frame_address_size

* elf-bfd.h (struct elf_backend_data): Make asection param of
elf_backend_eh_frame_address_size const.
(_bfd_elf_eh_frame_address_size): Likewise.
* elf32-m32c.c (_bfd_m32c_elf_eh_frame_address_size): Likewise.
* elf32-msp430.c (elf32_msp430_eh_frame_address_size): Likewise.
* elfxx-mips.c (_bfd_mips_elf_eh_frame_address_size): Likewise.
* elfxx-mips.h (_bfd_mips_elf_eh_frame_address_size): Likewise.
* elf-eh-frame.c (_bfd_elf_eh_frame_address_size): Likewise.
(next_cie_fde_offset): Constify params.
(offset_adjust, adjust_eh_frame_local_symbols): Likewise.

7 years agoEdit .eh_frame symbols
Alan Modra [Thu, 27 Apr 2017 01:38:14 +0000 (11:08 +0930)]
Edit .eh_frame symbols

Experimental support for moving symbols defined in .eh_frame as their
CIEs/FDEs are edited or merged.

* elf-bfd.h (struct eh_cie_fde): Add aug_str_len and aug_data_len.
(_bfd_elf_adjust_eh_frame_global_symbol): Declare.
* elf-eh-frame.c (_bfd_elf_parse_eh_frame): Set aug_str_len and
aug_data_len.
(offset_adjust): New function.
(_bfd_elf_adjust_eh_frame_global_symbol): Likewise.
(adjust_eh_frame_local_symbols): Likewise.
(_bfd_elf_discard_section_eh_frame): Call adjust_eh_frame_local_symbols
after changing anything.  Return true if anything changed.
* elflink.c (bfd_elf_discard_info): If .eh_frame changed, call
_bfd_elf_adjust_eh_frame_global_symbol for globals.

7 years agoClear dynstr_index when forcing symbols local
Alan Modra [Thu, 27 Apr 2017 01:36:33 +0000 (11:06 +0930)]
Clear dynstr_index when forcing symbols local

This is really just cosmetic, but it does protect a little from
accidentally reading a stale value.

* elflink.c (_bfd_elf_link_hash_hide_symbol): Clear dynstr_index
when force_local.

7 years agoPowerPC undefweak handling
Alan Modra [Wed, 26 Apr 2017 23:49:26 +0000 (09:19 +0930)]
PowerPC undefweak handling

This patch fixes a number of cases where -z nodynamic-undefined-weak
was not effective in preventing dynamic relocations or linkage stubs.

* elf32-ppc.c (UNDEFWEAK_NO_DYNAMIC_RELOC): Define.
(ppc_elf_select_plt_layout, ppc_elf_tls_setup): Use it.
(ppc_elf_adjust_dynamic_symbol, allocate_dynrelocs): Likewise.
(ppc_elf_relocate_section): Likewise.  Delete silly optimisation
for undef and undefweak dyn_relocs.
* elf64-ppc.c (UNDEFWEAK_NO_DYNAMIC_RELOC): Define.
(ppc64_elf_adjust_dynamic_symbol, ppc64_elf_tls_setup): Use it.
(allocate_got, allocate_dynrelocs): Likewise.
(ppc64_elf_relocate_section): Likewise.

7 years ago[GOLD] testsuite/plugin_section_order.c fix
Alan Modra [Wed, 26 Apr 2017 23:38:09 +0000 (09:08 +0930)]
[GOLD] testsuite/plugin_section_order.c fix

* testsuite/plugin_section_order.c (onload): Add missing break.

7 years agoAutomatic date update in version.in
GDB Administrator [Thu, 27 Apr 2017 00:00:56 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoMIPS/GAS: Fix `.option picX' handling with relaxation
Maciej W. Rozycki [Wed, 26 Apr 2017 23:47:15 +0000 (00:47 +0100)]
MIPS/GAS: Fix `.option picX' handling with relaxation

Correct the handling of `.option pic0' and `.option pic2' GAS pseudo-ops
in relaxation and use the setting of `mips_pic' (which these directives
control) as at the time a relaxed frag has been created rather than the
final `mips_pic' setting at the end of the source file processed.

To do so record whether `mips_pic' is NO_PIC or not in the frag itself
and use this information throughout relaxation instead of `mips_pic' to
decide which of NO_PIC or SVR4_PIC to produce machine code for, fixing
code generation and removing a possible fatal failure reproducible with:

$ as -32 --relax-branch -o option-pic-relax-3.o option-pic-relax-3.s
option-pic-relax-3.s: Assembler messages:
option-pic-relax-3.s:7: Warning: relaxed out-of-range branch into a jump
option-pic-relax-3.s: Internal error in cvt_frag_to_fill at .../gas/write.c:490.
Please report this bug.
$

using the test source included, due to a buffer overrun in filling the
variable part of a frag.

Likewise use the `fx_tcbit2' flag of a BFD_RELOC_16_PCREL_S2 fixup to
handle the simple case of substituting an out of range unconditional
branch with an equivalent absolute jump in NO_PIC code.

Retain the current way of VXWORKS_PIC use, which commit 41a1578ed17c
("MIPS/GAS: Sanitize `.option picX' pseudo-op") has forbidden the use of
`.option picX' with.

gas/
* config/tc-mips.c (RELAX_ENCODE): Add `PIC' flag.
(RELAX_PIC): New macro.
(RELAX_USE_SECOND, RELAX_SECOND_LONGER, RELAX_NOMACRO)
(RELAX_DELAY_SLOT, RELAX_DELAY_SLOT_16BIT)
(RELAX_DELAY_SLOT_SIZE_FIRST, RELAX_DELAY_SLOT_SIZE_SECOND):
Shift bits.
(RELAX_BRANCH_ENCODE): Add `pic' flag.
(RELAX_BRANCH_UNCOND, RELAX_BRANCH_LIKELY, RELAX_BRANCH_LINK)
(RELAX_BRANCH_TOOFAR): Shift bits.
(RELAX_BRANCH_PIC): New macro.
(RELAX_MICROMIPS_ENCODE): Add `pic' flag.
(RELAX_MICROMIPS_PIC): New macro.
(RELAX_MICROMIPS_UNCOND, RELAX_MICROMIPS_COMPACT)
(RELAX_MICROMIPS_LINK, RELAX_MICROMIPS_NODS)
(RELAX_MICROMIPS_RELAX32): Shift bits.
(relax_close_frag): Pass `mips_pic' setting to RELAX_ENCODE.
(append_insn): Pass `mips_pic' setting to RELAX_BRANCH_ENCODE
and RELAX_MICROMIPS_ENCODE, and record it in `fx_tcbit2' of the
first fixup created.
(md_apply_fix) <BFD_RELOC_16_PCREL_S2>: Use `fx_tcbit2' of the
fixup processed rather than `mips_pic' in choosing to relax an
out of range branch to a jump.
(relaxed_branch_length): Use the `pic' flag of the relaxed frag
rather than `mips_pic'.
(relaxed_micromips_32bit_branch_length): Likewise.
(md_estimate_size_before_relax): Likewise.
(md_convert_frag): Likewise.

* testsuite/gas/mips/option-pic-relax-0.d: New test.
* testsuite/gas/mips/option-pic-relax-1.d: New test.
* testsuite/gas/mips/option-pic-relax-2.d: New test.
* testsuite/gas/mips/option-pic-relax-3.d: New test.
* testsuite/gas/mips/option-pic-relax-3a.d: New test.
* testsuite/gas/mips/option-pic-relax-4.d: New test.
* testsuite/gas/mips/option-pic-relax-5.d: New test.
* testsuite/gas/mips/option-pic-relax-2.l: New stderr output.
* testsuite/gas/mips/option-pic-relax-3.l: New stderr output.
* testsuite/gas/mips/option-pic-relax-4.l: New stderr output.
* testsuite/gas/mips/option-pic-relax-5.l: New stderr output.
* testsuite/gas/mips/option-pic-relax-0.s: New test source.
* testsuite/gas/mips/option-pic-relax-1.s: New test source.
* testsuite/gas/mips/option-pic-relax-2.s: New test source.
* testsuite/gas/mips/option-pic-relax-3.s: New test source.
* testsuite/gas/mips/option-pic-relax-4.s: New test source.
* testsuite/gas/mips/option-pic-relax-5.s: New test source.
* testsuite/gas/mips/mips.exp: Run the new tests.

7 years agox86: Add DT_PLTRELSZ/DT_PLTREL/DT_JMPREL for PLT relocation
H.J. Lu [Wed, 26 Apr 2017 22:34:00 +0000 (15:34 -0700)]
x86: Add DT_PLTRELSZ/DT_PLTREL/DT_JMPREL for PLT relocation

x86, PLT relocation may contain R_386_TLS_DESC or R_X86_64_TLSDESC
even though there is no real PLT.  We need to add DT_PLTRELSZ, DT_PLTREL
and DT_JMPREL if there is a .rel.plt/.rela.plt section.

bfd/

* elf32-i386.c (elf_i386_size_dynamic_sections): Alwasys add
DT_PLTRELSZ, DT_PLTREL and DT_JMPREL for .rel.plt section.
* elf64-x86-64.c (elf_x86_64_size_dynamic_sections): Alwasys
add DT_PLTRELSZ, DT_PLTREL and DT_JMPREL for .rela.plt section.

ld/

* testsuite/ld-i386/tlsdesc2.d: New test.
* testsuite/ld-x86-64/tlsdesc2.d: Likewise.

7 years agoFix a seg-fault when processing a corrupt binary containing reloc(s) with negative...
Nick Clifton [Wed, 26 Apr 2017 15:30:22 +0000 (16:30 +0100)]
Fix a seg-fault when processing a corrupt binary containing reloc(s) with negative addresses.

PR binutils/21434
* reloc.c (bfd_perform_relocation): Check for a negative address
in the reloc.

7 years agoFix PR number in ChangeLog comment.
Nick Clifton [Wed, 26 Apr 2017 14:45:41 +0000 (15:45 +0100)]
Fix PR number in ChangeLog comment.

7 years agoFix invocation of stat() on a NULL pointer.
Nick Clifton [Wed, 26 Apr 2017 14:42:03 +0000 (15:42 +0100)]
Fix invocation of stat() on a NULL pointer.

PR binutils/21407
* bucomm.c (get_file_size): Return -1 if file_name is NULL.
* ar.c (main): Fail with usage() invocation if no file names are
provided.

7 years agoImprove decoding of corrupt/unrecognised gnu build attribute notes.
Nick Clifton [Wed, 26 Apr 2017 13:23:26 +0000 (14:23 +0100)]
Improve decoding of corrupt/unrecognised gnu build attribute notes.

* readelf.c (process_section_headers): Warn about overlarge
sections.
(print_gnu_build_attribute_name): Print the number of unrecognised
note types.  Fix formatting in the presence of errors.
(testsuite/binutils-all/note-2-32.s): Fix encoding of numeric notes.
(testsuite/binutils-all/note-2-64.s): Likewise.

7 years ago[gdbarch] New method "execute_dwarf_cfa_vendor_op" and migrate SPARC to it
Jiong Wang [Wed, 26 Apr 2017 13:05:03 +0000 (14:05 +0100)]
[gdbarch] New method "execute_dwarf_cfa_vendor_op" and migrate SPARC to it

Recently a feature called "return address signing" has been added to GCC to
prevent stack smash stack on AArch64.  For details please refer:

  https://gcc.gnu.org/ml/gcc-patches/2017-01/msg00376.html

GDB needs to be aware of this feature so it can restore the original return
address which is critical for unwinding.

On compiler side, whenever return address, i.e. LR register, is mangled or
restored by hardware instruction, compiler is expected to generate a
DW_CFA_AARCH64_negate_ra_state to toggle return address signing status.

DW_CFA_AARCH64_negate_ra_state is using the same CFI number and
therefore need to be multiplexed with DW_CFA_GNU_window_save which was designed
for SPARC.

A new gdbarch method "execute_dwarf_cfa_vendor_op" is introduced by this patch.
It's parameters has been restricted to those only needed by SPARC and AArch64
for multiplexing DW_CFA_GNU_window_save which is a CFI operation takes none
operand.  Should any further DWARF CFI operation want to be multiplexed in the
future,  the parameter list can be extended.  Below is the current function
prototype.

   typedef int (gdbarch_execute_dwarf_cfa_vendor_op_ftype)
     (struct gdbarch *gdbarch, gdb_byte op, struct dwarf2_frame_state *fs);

DW_CFA_GNU_window_save support for SPARC is migrated to this new gdbarch
method by this patch.

gdb/
* gdbarch.sh: New gdbarch method execute_dwarf_cfa_vendor_op.
* gdbarch.c: Regenerated.
* gdbarch.h: Regenerated.
* dwarf2-frame.c (dwarf2_frame_state_alloc_regs): Made the
visibility external.
(execute_cfa_program): Call execute_dwarf_cfa_vendor_op for CFI
between DW_CFA_lo_user and DW_CFA_high_user inclusive.
(enum cfa_how_kind): Move to ...
(struct dwarf2_frame_state_reg_info): Likewise.
(struct dwarf2_frame_state): Likewise.
* dwarf2-frame.h: ... here.
(dwarf2_frame_state_alloc_regs): New declaration.
* sparc-tdep.c (sparc_execute_dwarf_cfa_vendor_op): New function.
(sparc32_gdbarch_init): Register execute_dwarf_cfa_vendor_op hook.

7 years agoPR ld/21334: Always call `_bfd_elf_link_renumber_dynsyms' if required
Maciej W. Rozycki [Mon, 24 Apr 2017 19:15:31 +0000 (20:15 +0100)]
PR ld/21334: Always call `_bfd_elf_link_renumber_dynsyms' if required

Complement commit e17b0c351f0b ("MIPS/BFD: Respect the ELF gABI dynamic
symbol table sort requirement") and correct an inconsistency in dynamic
symbol accounting data causing an assertion failure in the MIPS backend:

ld: BFD (GNU Binutils) 2.28.51.20170330 assertion fail
../../binutils-gdb/bfd/elfxx-mips.c:3860

in the course of making a GOT entry in a static binary to satisfy a GOT
relocation present in input, due to the local dynamic symbol count not
having been established.

To do so let backends request `_bfd_elf_link_renumber_dynsyms' to be
always called, rather than where a dynamic binary is linked only, and
then make this request in the MIPS backend.

bfd/
PR ld/21334
* elf-bfd.h (elf_backend_data): Add `always_renumber_dynsyms'
member.
* elfxx-target.h [!elf_backend_always_renumber_dynsyms]
(elf_backend_always_renumber_dynsyms): Define.
(elfNN_bed): Initialize `always_renumber_dynsyms' member.
* elfxx-mips.h (elf_backend_always_renumber_dynsyms): Define.
* elflink.c (bfd_elf_size_dynamic_sections): Also call
`_bfd_elf_link_renumber_dynsyms' if the backend has requested
it.
(bfd_elf_size_dynsym_hash_dynstr): Likewise.

ld/
PR ld/21334
* testsuite/ld-mips-elf/pr21334.dd: New test.
* testsuite/ld-mips-elf/pr21334.gd: New test.
* testsuite/ld-mips-elf/pr21334.ld: New test linker script.
* testsuite/ld-mips-elf/pr21334.s: New test source.
* testsuite/ld-mips-elf/mips-elf.exp: Run the new tests.

7 years agoELF/BFD: Limit `_bfd_elf_link_renumber_dynsyms' call in section GC
Maciej W. Rozycki [Mon, 24 Apr 2017 19:13:28 +0000 (20:13 +0100)]
ELF/BFD: Limit `_bfd_elf_link_renumber_dynsyms' call in section GC

Consistently call `_bfd_elf_link_renumber_dynsyms' only if linking a
dynamic binary, complementing code in `bfd_elf_size_dynsym_hash_dynstr'
and commit ccabcbe51e85 ("New attempt at fixing MIPS --gc-sections et
al."), <https://sourceware.org/ml/binutils/2005-08/msg00258.html>.

bfd/
* elflink.c (bfd_elf_size_dynamic_sections): Only call
`_bfd_elf_link_renumber_dynsyms' after section GC if dynamic
sections have been created.

7 years agoMIPS/readelf: With `-A' also dump GOT in static binaries
Maciej W. Rozycki [Mon, 24 Apr 2017 19:41:33 +0000 (20:41 +0100)]
MIPS/readelf: With `-A' also dump GOT in static binaries

A static, non-relocated global offset table will be embedded in static
binaries produced from objects containing any kind of GOT relocations,
generally PIC code.  All symbols will have been resolved in static link
in such binaries making all GOT entries local and their values final as
there is no run-time load processing further performed.

Dump such GOT with `readelf -A' like already done with regular GOT, to
make it easier to examine static code that uses accesses via the GOT
pointer.  There will be no dynamic segment or section in a static binary
to get the GOT pointer (DT_PLTGOT) from, so use section headers to find
a `.got' section instead.

binutils/
* readelf.c (process_mips_specific): Add static GOT support.

7 years agoFix seg-fault attempting to compress a debug section in a corrupt binary.
Nick Clifton [Wed, 26 Apr 2017 12:07:49 +0000 (13:07 +0100)]
Fix seg-fault attempting to compress a debug section in a corrupt binary.

PR binutils/21431
* compress.c (bfd_init_section_compress_status): Check the return
value from bfd_malloc.

7 years agoxtensa_pseudo_register_read/write - Use regcache_raw_read_unsigned
Alan Hayward [Wed, 26 Apr 2017 09:32:35 +0000 (10:32 +0100)]
xtensa_pseudo_register_read/write - Use regcache_raw_read_unsigned

gdb/
* xtensa-tdep.c (xtensa_pseudo_register_read): Use
regcache_raw_read_unsigned.
(xtensa_pseudo_register_write): Likewise.

7 years agonds32: Abort instead of returning REG_UNKNOWN
Alan Hayward [Wed, 26 Apr 2017 08:57:15 +0000 (09:57 +0100)]
nds32: Abort instead of returning REG_UNKNOWN

gdb/
* nds32-tdep.c (nds32_pseudo_register_read): Abort on errors.
(nds32_pseudo_register_write): Likewise.

7 years agoAutomatic date update in version.in
GDB Administrator [Wed, 26 Apr 2017 00:00:36 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoChange readonly_p to bool
Yao Qi [Tue, 25 Apr 2017 20:32:05 +0000 (21:32 +0100)]
Change readonly_p to bool

This patch changes readonly_p type to bool.

gdb:

2017-04-25  Yao Qi  <yao.qi@linaro.org>

* regcache.c (struct regcache) <readonly_p>: Change its type
to bool.
(regcache_xmalloc_1): Update parameter type and callers update.

7 years agoMIPS/readelf: Simplify GOT[1] data availability check
Maciej W. Rozycki [Tue, 11 Apr 2017 23:03:41 +0000 (00:03 +0100)]
MIPS/readelf: Simplify GOT[1] data availability check

Unavailable data is handled gracefully in MIPS GOT processing done by
`print_mips_got_entry', so all that is needed in special GOT[1] handling
is to verify whether data can be retrieved for the purpose of the GNU
marker check done with `byte_get'.  Remove the extra error reporting
code then, introduced with commit 75ec1fdbb797 ("Fix runtime seg-fault
in readelf when parsing a corrupt MIPS binary.") in the course of
addressing PR binutils/21344, and defer the error case to regular local
GOT entry processing.

binutils/
* readelf.c (process_mips_specific): Remove error reporting from
GOT[1] processing.

7 years agoMIPS/readelf: Remove extraneous null GOT data check
Maciej W. Rozycki [Tue, 11 Apr 2017 23:02:13 +0000 (00:02 +0100)]
MIPS/readelf: Remove extraneous null GOT data check

Null data is handled gracefully throughout in MIPS GOT processing, with
addresses printed normally and unavailable data shown as `<unknown>' by
`print_mips_got_entry', and special processing code for GOT[1] doing an
explicit check.  Remove an unwanted null GOT data check then, introduced
with commit 592458412fb2 in the course of addressing PR binutils/12855.

binutils/
* readelf.c (process_mips_specific): Remove null GOT data check.

7 years ago[ARC] Enhance enter/leave mnemonics.
Claudiu Zissulescu [Tue, 25 Apr 2017 15:07:00 +0000 (17:07 +0200)]
[ARC] Enhance enter/leave mnemonics.

enter/leave mnemonics are enhanced to not only accept register ranges
but also single register (i.e., r13) or even no GPR register at all.

gas/
2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>

* testsuite/gas/arc/leave_enter.d: Update test.
* testsuite/gas/arc/leave_enter.s: Likewise.

opcodes/
2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>

* arc-dis.c (print_insn_arc): Smartly print enter/leave mnemonics.
* arc-opc.c (insert_r13el): New function.
(R13_EL): Define.
* arc-tbl.h: Add new enter/leave variants.

7 years ago[ARC] Prefer NOP instead of MOV 0,0
Claudiu Zissulescu [Tue, 25 Apr 2017 15:07:00 +0000 (17:07 +0200)]
[ARC] Prefer NOP instead of MOV 0,0

NOP and MOV 0,0 are having the same encoding. As MOV mnemonic is
located before NOP in the instruction table, the disassembler prints
MOV 0,0 for NOP. Reorder the instructions such that NOP is first.

gas/
2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>

* testsuite/gas/arc/b.d: Update test.
* testsuite/gas/arc/noargs_hs.d: Likewise.

opcode/
2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>

* arc-tbl.h: Reorder NOP entry to be before MOV instructions.

7 years agoChange gdbarch_wchar_bit for AArch64 and ARM
Yao Qi [Tue, 25 Apr 2017 14:15:54 +0000 (15:15 +0100)]
Change gdbarch_wchar_bit for AArch64 and ARM

The size of wchar_t on AArch64 and ARM is 4-byte, so we can use the
default value (4*TARGET_CHAR_BIT).

This patch fixes some fails in gdb.cp/wide_char_types.exp on
aarch64-linux.

gdb:

2017-04-25  Yao Qi  <yao.qi@linaro.org>

* aarch64-tdep.c (aarch64_gdbarch_init): Don't call
set_gdbarch_wchar_bit.
* arm-tdep.c (arm_gdbarch_init): Likewise.

7 years agoMIPS16/opcodes: Add `-M no-aliases' disassembler option help text
Maciej W. Rozycki [Tue, 25 Apr 2017 10:49:37 +0000 (11:49 +0100)]
MIPS16/opcodes: Add `-M no-aliases' disassembler option help text

Complement commit 986e18a5a9fd ("Add a second 'pinfo' member to
mips_opcode to extend number of available bits"),
<https://sourceware.org/ml/binutils/2005-01/msg00261.html>, and add a
help text for the `-M no-aliases' disassembler option.

opcodes/
* mips-dis.c (print_mips_disassembler_options): Add
`no-aliases'.

7 years agoMIPS16/opcodes: Annotate instruction aliases
Maciej W. Rozycki [Tue, 25 Apr 2017 10:44:29 +0000 (11:44 +0100)]
MIPS16/opcodes: Annotate instruction aliases

Complement commit 986e18a5a9fd ("Add a second 'pinfo' member to
mips_opcode to extend number of available bits"),
<https://sourceware.org/ml/binutils/2005-01/msg00261.html>, and annotate
MIPS16 NOP, LA, DLA and the synthetic forms of LD and LW instructions as
aliases.  These correspond to MOVE, and the PC-relative ADDIU, DADDIU,
LD and LW hardware instructions respectively.

binutils/
* testsuite/binutils-all/mips/mips16-alias.d: New test.
* testsuite/binutils-all/mips/mips16-noalias.d: New test.
* testsuite/binutils-all/mips/mips16-alias.s: New test source.
* testsuite/binutils-all/mips/mips.exp: Run the new tests.

opcodes/
* mips16-opc.c (AL): New macro.
(mips16_opcodes): Mark "nop", "la", "dla", and synthetic forms
of "ld" and "lw" as aliases.

7 years agoMIPS/GAS: Correct BFD_RELOC_MIPS16_16_PCREL_S1 fixup size
Maciej W. Rozycki [Tue, 25 Apr 2017 09:20:14 +0000 (10:20 +0100)]
MIPS/GAS: Correct BFD_RELOC_MIPS16_16_PCREL_S1 fixup size

Correct the size of a BFD_RELOC_MIPS16_16_PCREL_S1 fixup made in
`md_convert_frag', fixing a bug introduced with commit c9775dde3277
("MIPS16: Add R_MIPS16_PC16_S1 branch relocation support)".  Add test
cases to verify that the overflow of this fixup's in-place addend is
still correctly detected.

gas/
* config/tc-mips.c (md_convert_frag): Correct
BFD_RELOC_MIPS16_16_PCREL_S1 fixup size.
* testsuite/gas/mips/mips16-branch-addend-4.d: New test.
* testsuite/gas/mips/mips16-branch-addend-5.d: New test.
* testsuite/gas/mips/mips16-branch-addend-5.l: New stderr
output.
* testsuite/gas/mips/mips16-branch-addend-4.s: New test source.
* testsuite/gas/mips/mips16-branch-addend-5.s: New test source.
* testsuite/gas/mips/mips.exp: Run the new tests.

7 years agoFix build on gcc < 5 (std::is_trivially_copyable missing)
Pedro Alves [Tue, 25 Apr 2017 09:58:57 +0000 (10:58 +0100)]
Fix build on gcc < 5 (std::is_trivially_copyable missing)

Ref: https://sourceware.org/ml/gdb-patches/2017-04/msg00660.html

Simply skip the poisoning on older compilers.

gdb/ChangeLog:
2017-04-25  Pedro Alves  <palves@redhat.com>

* common/poison.h [!HAVE_IS_TRIVIALLY_COPYABLE] (IsRelocatable)
(BothAreRelocatable, memcopy, memmove): Don't define.
* common/traits.h (__has_feature, HAVE_IS_TRIVIALLY_COPYABLE): New
macros.

7 years agogas: sparc: fix relaxation of CALL instruction into branches in a.out targets
Jose E. Marchesi [Tue, 25 Apr 2017 09:40:43 +0000 (02:40 -0700)]
gas: sparc: fix relaxation of CALL instruction into branches in a.out targets

This patch avoids CALL instructions to be optimized into branches if
the symbols referred to in the CALL instruction are not fully resolved
at the time the assembler writes its output.

Tested in sparc64-linux-gnu and sparc-sun-sunos4.1.3 targets.
No regressions.

gas/ChangeLog:

2017-04-25  Jose E. Marchesi  <jose.marchesi@oracle.com>

PR gas/21407
* config/tc-sparc.c (md_apply_fix): Do not transform `call'
instructions into branch instructions in fixups generating
additional relocations.
* testsuite/gas/sparc/call-relax.s: New file.
* testsuite/gas/sparc/call-relax.d: Likewise.
* testsuite/gas/sparc/call-relax-aout.d: Likewise.
* testsuite/gas/sparc/sparc.exp: Test call-relax and call-relax-aout.

7 years agoPoison non-POD memset & non-trivially-copyable memcpy/memmove
Pedro Alves [Tue, 25 Apr 2017 00:27:41 +0000 (01:27 +0100)]
Poison non-POD memset & non-trivially-copyable memcpy/memmove

This patch catches invalid initialization of non-POD types with
memset, at compile time.

This is what I used to catch the problems fixed by the previous
patches in the series:

  $ make -k 2>&1 | grep "deleted function"
  src/gdb/breakpoint.c:951:53: error: use of deleted function ‘void* memset(T*, int, size_t) [with T = bp_location; <template-parameter-1-2> = void; size_t = long unsigned int]’
  src/gdb/breakpoint.c:7325:32: error: use of deleted function ‘void* memset(T*, int, size_t) [with T = bp_location; <template-parameter-1-2> = void; size_t = long unsigned int]’
  src/gdb/btrace.c:1153:42: error: use of deleted function ‘void* memset(T*, int, size_t) [with T = btrace_insn; <template-parameter-1-2> = void; size_t = long unsigned int]’
...

gdb/ChangeLog:
2017-04-25  Pedro Alves  <palves@redhat.com>

* common/common-defs.h: Include "common/poison.h".
* common/function-view.h: (Not, Or, Requires): Move to traits.h
and adjust.
* common/poison.h: New file.
* common/traits.h: Include <type_traits>.
(Not, Or, Requires): New, moved from common/function-view.h.

7 years agoDon't memset non-POD types: struct breakpoint
Pedro Alves [Tue, 25 Apr 2017 00:27:42 +0000 (01:27 +0100)]
Don't memset non-POD types: struct breakpoint

Eh, struct breakpoint was made non-POD just today, with commit
d28cd78ad820e3 ("Change breakpoint event locations to
event_location_up").  :-)

  src/gdb/breakpoint.c: In function ‘void init_raw_breakpoint_without_location(breakpoint*, gdbarch*, bptype, const breakpoint_ops*)’:
  src/gdb/breakpoint.c:7447:28: error: use of deleted function ‘void* memset(T*, int, size_t) [with T = breakpoint; <template-parameter-1-2> = void; size_t = long unsigned int]’
     memset (b, 0, sizeof (*b));
      ^
  In file included from src/gdb/common/common-defs.h:85:0,
   from src/gdb/defs.h:28,
   from src/gdb/breakpoint.c:20:
  src/gdb/common/poison.h:56:7: note: declared here
   void *memset (T *s, int c, size_t n) = delete;
 ^

gdb/ChangeLog:
2017-04-25  Pedro Alves  <palves@redhat.com>

* breakpoint.h (struct breakpoint): In-class initialize all
fields.  Make boolean fields "bool".
* breakpoint.c (init_raw_breakpoint_without_location): Remove
memset call and initializations no longer necessary.

7 years agoDon't memset non-POD types: struct btrace_insn
Pedro Alves [Tue, 25 Apr 2017 00:27:42 +0000 (01:27 +0100)]
Don't memset non-POD types: struct btrace_insn

struct btrace_insn is not a POD [1] so we shouldn't be using memset to
initialize it [2].

Use list-initialization instead, wrapped in a "pt insn to btrace insn"
function, which looks like just begging to be added next to the
existing pt_reclassify_insn/pt_btrace_insn_flags functions.

[1] - because its field "flags" is not POD, because enum_flags has a
non-trivial default ctor.

gdb/ChangeLog:
2017-04-25  Pedro Alves  <palves@redhat.com>

* btrace.c (pt_btrace_insn_flags): Change parameter type to
reference.
(pt_btrace_insn): New function.
(ftrace_add_pt): Remove memset call and use pt_btrace_insn.

7 years agoDon't memset non-POD types: struct bp_location
Pedro Alves [Tue, 25 Apr 2017 00:27:42 +0000 (01:27 +0100)]
Don't memset non-POD types: struct bp_location

struct bp_location is not a POD, so we shouldn't be using memset to
initialize it.

Caught like this:

  src/gdb/breakpoint.c: In function ‘bp_location** get_first_locp_gte_addr(CORE_ADDR)’:
  src/gdb/breakpoint.c:950:53: error: use of deleted function ‘void* memset(T*, int, size_t) [with T = bp_location; <template-parameter-1-2> = void; size_t = long unsigned int]’
     memset (&dummy_loc, 0, sizeof (struct bp_location));
       ^
  In file included from src/gdb/defs.h:28:0,
   from src/gdb/breakpoint.c:20:
  src/gdb/common/common-defs.h:126:7: note: declared here
   void *memset (T *s, int c, size_t n) = delete;
 ^

gdb/ChangeLog:
2017-04-25  Pedro Alves  <palves@redhat.com>

* ada-lang.c (ada_catchpoint_location): Now a "class".  Remove
"base" field and inherit from "bp_location" instead.  Add
non-default ctor.
(allocate_location_exception): Use new non-default ctor.
* breakpoint.c (get_first_locp_gte_addr): Remove memset call.
(init_bp_location): Convert to ...
(bp_location::bp_location): ... this new ctor, and remove memset
call.
(base_breakpoint_allocate_location): Use the new non-default ctor.
* breakpoint.h (bp_location): Now a class.  Declare default and
non-default ctors.  In-class initialize all members.
(init_bp_location): Remove declaration.

7 years agoDon't memcpy non-trivially-copyable types: Make enum_flags triv. copyable
Pedro Alves [Tue, 25 Apr 2017 00:27:42 +0000 (01:27 +0100)]
Don't memcpy non-trivially-copyable types: Make enum_flags triv. copyable

The delete-memcpy-with-non-trivial-types patch exposed many instances
of this problem:

  src/gdb/btrace.h: In function ‘btrace_insn_s* VEC_btrace_insn_s_quick_insert(VEC_btrace_insn_s*, unsigned int, const btrace_insn_s*, const char*, unsigned int)’:
  src/gdb/common/vec.h:948:62: error: use of deleted function ‘void* memmove(T*, const U*, size_t) [with T = btrace_insn; U = btrace_insn; <template-parameter-1-3> = void; size_t = long unsigned int]’
     memmove (slot_ + 1, slot_, (vec_->num++ - ix_) * sizeof (T));    \
^
  src/gdb/common/vec.h:436:1: note: in expansion of macro ‘DEF_VEC_FUNC_O’
   DEF_VEC_FUNC_O(T)         \
   ^
  src/gdb/btrace.h:84:1: note: in expansion of macro ‘DEF_VEC_O’
   DEF_VEC_O (btrace_insn_s);
   ^
[...]
  src/gdb/common/vec.h:1060:31: error: use of deleted function ‘void* memcpy(T*, const U*, size_t) [with T = btrace_insn; U = btrace_insn; <template-parameter-1-3> = void; size_t = long unsigned int]’
  sizeof (T) * vec2_->num);       \
 ^
  src/gdb/common/vec.h:437:1: note: in expansion of macro ‘DEF_VEC_ALLOC_FUNC_O’
   DEF_VEC_ALLOC_FUNC_O(T)         \
   ^
  src/gdb/btrace.h:84:1: note: in expansion of macro ‘DEF_VEC_O’
   DEF_VEC_O (btrace_insn_s);
   ^

So, VECs (given it's C roots) rely on memcpy/memcpy of VEC elements to
be well defined, in order to grow/reallocate its internal elements
array.  This means that we can only put trivially copyable types in
VECs.  E.g., if a type requires using a custom copy/move ctor to
relocate, then we can't put it in a VEC (so we use std::vector
instead).  But, as shown above, we're violating that requirement.

btrace_insn is currently not trivially copyable, because it contains
an enum_flags field, and that is itself not trivially copyable.  This
patch corrects that, by simply removing the user-provided copy
constructor and assignment operator.  The compiler-generated versions
work just fine.

Note that std::vector relies on std::is_trivially_copyable too to know
whether it can reallocate its elements with memcpy/memmove instead of
having to call copy/move ctors and dtors, so if we have types in
std::vectors that weren't trivially copyable because of enum_flags,
this will make such vectors more efficient.

gdb/ChangeLog:
2017-04-25  Pedro Alves  <palves@redhat.com>

* common/enum-flags.h (enum_flags): Don't implement copy ctor and
assignment operator.

7 years agoAutomatic date update in version.in
GDB Administrator [Tue, 25 Apr 2017 00:00:36 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agox86-64: Force symbol dynamic if it isn't undefined weak
H.J. Lu [Mon, 24 Apr 2017 21:37:12 +0000 (14:37 -0700)]
x86-64: Force symbol dynamic if it isn't undefined weak

Force symbol dynamic if it isn't undefined weak.  Generate relative
relocation for GOT reference against non-dynamic symbol in PIC to
avoid unnecessary dynamic symbols.

bfd/

* elf64-x86-64.c (elf_x86_64_link_hash_entry): Add
no_finish_dynamic_symbol.
(elf_x86_64_link_hash_newfunc): Set no_finish_dynamic_symbol to
0.
(elf_x86_64_allocate_dynrelocs): If a symbol isn't undefined
weak symbol, don't make it dynamic.
(elf_x86_64_relocate_section): If a symbol isn't dynamic in PIC,
set no_finish_dynamic_symbol and generate R_X86_64_RELATIVE
relocation for GOT reference.
(elf_x86_64_finish_dynamic_symbol): Abort if
no_finish_dynamic_symbol isn't 0.

ld/

* testsuite/ld-x86-64/no-plt.exp: Also check no-plt-1e.nd.
* testsuite/ld-x86-64/no-plt-1e.nd: New file.

7 years agoi386: Force symbol dynamic if it isn't undefined weak
H.J. Lu [Mon, 24 Apr 2017 20:42:33 +0000 (13:42 -0700)]
i386: Force symbol dynamic if it isn't undefined weak

Force symbol dynamic if it isn't undefined weak.  Generate R_386_RELATIVE
relocation for R_386_GOT32 relocation against non-dynamic symbol in PIC.

PR ld/21402
* elf32-i386.c (elf_i386_allocate_dynrelocs): If a symbol isn't
undefined weak symbol, don't make it dynamic.
(elf_i386_relocate_section): If a symbol isn't dynamic in PIC,
set no_finish_dynamic_symbol and generate R_386_RELATIVE
relocation for R_386_GOT32.

7 years agoUse floatformat_totalsize_bytes
Yao Qi [Mon, 24 Apr 2017 20:29:12 +0000 (21:29 +0100)]
Use floatformat_totalsize_bytes

The code can be replaced by floatformat_totalsize_bytes.

gdb:

2017-04-24  Yao Qi  <yao.qi@linaro.org>

* doublest.c (convert_doublest_to_floatformat): Call
floatformat_totalsize_bytes.

7 years agoi386: Skip pr12570 tests for nacl targets
H.J. Lu [Mon, 24 Apr 2017 16:46:04 +0000 (09:46 -0700)]
i386: Skip pr12570 tests for nacl targets

commit f129e49f4d07f4d36319ac757fdcf3a8ce7d605b
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Tue Jan 10 11:30:25 2017 -0800

    Don't use elf_i386_eh_frame_plt directly

fixed i386 PLT eh_frame generation.  Skip pr12570 tests since they are
for non-nacl targets.

* testsuite/ld-i386/pr12570a.d: Skip for nacl targets.
* testsuite/ld-i386/pr12570b.d: Likewise.

7 years agoi386: Set ELF_MAXPAGESIZE to 0x1000 for VxWorks
H.J. Lu [Mon, 24 Apr 2017 16:37:10 +0000 (09:37 -0700)]
i386: Set ELF_MAXPAGESIZE to 0x1000 for VxWorks

commit a27e437177412e5b52999723f3c5d5d0d37b9087
Author: Roland McGrath <roland@gnu.org>
Date:   Thu Jul 28 22:35:15 2011 +0000

    BFD vector for elf32-i386-nacl:

changed ELF_MAXPAGESIZE to 0x10000 for VxWorks.  This patch fixes it
and updated testsuite/ld-i386/vxworks2.sd to add space for program
headers.

bfd/

PR ld/21425
* elf32-i386.c (ELF_MAXPAGESIZE): Set to 0x1000 for VxWorks.

ld/

PR ld/20815
* testsuite/ld-i386/vxworks2.sd: Add space for program headers.

7 years ago[GAS/ARM] Fix expansion of ldr pseudo instruction
Thomas Preud'homme [Mon, 24 Apr 2017 13:49:48 +0000 (14:49 +0100)]
[GAS/ARM] Fix expansion of ldr pseudo instruction

The LDR rX, =cst pseudo-instruction suffers from two issues for loading
integer constants in Thumb mode:

- movs is used if the constant and register can be encoded using that
  instruction which leads to unexpected behavior due to its flag-setting
  behavior
- mov.w, movw and mvn are used for r13 (sp) and r15 (pc) but these
  encoding are marked as UNPREDICTABLE

This patch fixes those issues and update testing accordingly.

2017-04-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>

gas/
* config/tc-arm.c (move_or_literal_pool): Remove code generating MOVS.
Forbid MOV.W and MOVW if destination is SP or PC.
* testsuite/gas/arm/thumb2_ldr_immediate_highregs_armv6t2.s: Explain
expectation of LDR not generating a MOVS for low registers and small
constants.  Add tests of MOVW generation.
* testsuite/gas/arm/thumb2_ldr_immediate_highregs_armv6t2.d: Update
expected disassembly.

7 years agold: fix regressions with rett instructions in sparc tests.
Jose E. Marchesi [Mon, 24 Apr 2017 12:31:31 +0000 (05:31 -0700)]
ld: fix regressions with rett instructions in sparc tests.

A few tests in the ld testsuite were expecting the disassembler to
emit `rett' instructions in V9.  This patch updates the tests to
expect `return' instead.

ld/ChangeLog:

2017-04-24  Jose E. Marchesi  <jose.marchesi@oracle.com>

* testsuite/ld-sparc/tlssunbin64.dd: Expect `return' instructions
instead of `rett' in V9.
* testsuite/ld-sparc/tlssunnopic64.dd: Likewise.
* testsuite/ld-sparc/tlssunpic64.dd: Likewise.

7 years agoFix snafu in aarch64 opcodes debugging statement.
Tamar Christina [Mon, 24 Apr 2017 10:55:44 +0000 (11:55 +0100)]
Fix snafu in aarch64 opcodes debugging statement.

* aarch64-opc.c (aarch64_logical_immediate_p): Update DEBUG_TRACE
arguments.

7 years agoAutomatic date update in version.in
GDB Administrator [Mon, 24 Apr 2017 00:00:39 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoPR 21418, ar -N lacks arg check
Alan Modra [Sun, 23 Apr 2017 08:09:48 +0000 (17:39 +0930)]
PR 21418, ar -N lacks arg check

PR 21418
* ar.c (main): Check -a, -b, -i and -N args are given.

7 years agoPR 21417, ar arg check failure
Alan Modra [Sun, 23 Apr 2017 07:58:14 +0000 (17:28 +0930)]
PR 21417, ar arg check failure

PR 21417
* ar.c (main): Check that an archive file is given after options.

7 years agoPR 21415, objdump fails to check bfd_get_section_contents status
Alan Modra [Sun, 23 Apr 2017 09:01:40 +0000 (18:31 +0930)]
PR 21415, objdump fails to check bfd_get_section_contents status

PR 21415
* objdump.c (disassemble_section): Check bfd_get_section_contents
status.

7 years agoPR 21414, null pointer deref of _bfd_elf_large_com_section sym
Alan Modra [Sun, 23 Apr 2017 05:51:11 +0000 (15:21 +0930)]
PR 21414, null pointer deref of _bfd_elf_large_com_section sym

PR 21414
* section.c (GLOBAL_SYM_INIT): Make available in bfd.h.
* elf.c (lcomm_sym): New.
(_bfd_elf_large_com_section): Use lcomm_sym section symbol.
* bfd-in2.h: Regenerate.

7 years agoPR 21412, get_reloc_section assumes .rel/.rela name for SHT_REL/RELA.
Alan Modra [Sun, 23 Apr 2017 01:33:34 +0000 (11:03 +0930)]
PR 21412, get_reloc_section assumes .rel/.rela name for SHT_REL/RELA.

This patch fixes an assumption made by code that runs for objcopy and
strip, that SHT_REL/SHR_RELA sections are always named starting with a
.rel/.rela prefix.  I'm also modifying the interface for
elf_backend_get_reloc_section, so any backend function just needs to
handle name mapping.

PR 21412
* elf-bfd.h (struct elf_backend_data <get_reloc_section>): Change
parameters and comment.
(_bfd_elf_get_reloc_section): Delete.
(_bfd_elf_plt_get_reloc_section): Declare.
* elf.c (_bfd_elf_plt_get_reloc_section, elf_get_reloc_section):
New functions.  Don't blindly skip over assumed .rel/.rela prefix.
Extracted from..
(_bfd_elf_get_reloc_section): ..here.  Delete.
(assign_section_numbers): Call elf_get_reloc_section.
* elf64-ppc.c (elf_backend_get_reloc_section): Define.
* elfxx-target.h (elf_backend_get_reloc_section): Update.

7 years agoPR 21409, segfault in _bfd_dwarf2_find_nearest_line
Alan Modra [Sun, 23 Apr 2017 04:25:49 +0000 (13:55 +0930)]
PR 21409, segfault in _bfd_dwarf2_find_nearest_line

PR 21409
* dwarf2.c (_bfd_dwarf2_find_nearest_line): Don't segfault when
no symbols.

7 years agoPR 21408, segfault in display_debug_lines_decoded
Alan Modra [Sun, 23 Apr 2017 04:19:50 +0000 (13:49 +0930)]
PR 21408, segfault in display_debug_lines_decoded

PR 21408
* dwarf.c (display_debug_lines_decoded): Don't segfault on NULL
file_table.

7 years agoAutomatic date update in version.in
GDB Administrator [Sun, 23 Apr 2017 00:00:35 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoFix ldn/stn multiple instructions. Fix testcases with unaligned data.
Jim Wilson [Sat, 22 Apr 2017 23:36:01 +0000 (16:36 -0700)]
Fix ldn/stn multiple instructions.  Fix testcases with unaligned data.

sim/aarch64/
* simulator.c (vec_load): Add M argument.  Rewrite to iterate over
registers based on structure size.
(LD4, LD3, LD2, LD1_2, LD1_3, LD1_4): Pass new arg to vec_load.
(LD1_1): Replace with call to vec_load.
(vec_store): Add new M argument.  Rewrite to iterate over registers
based on structure size.
(ST4, ST3, ST2, ST1_2, ST1_3, ST1_4): Pass new arg to vec_store.
(ST1_1): Replace with call to vec_store.

sim/testsuite/sim/aarch64/
* fcvtz.s, fstur.s, ldn_single.s, ldnr.s, mla.s, mls.s, uzp.s: Align
data.
* sumulh.s: Delete unnecessary data alignment.
* stn_single.s: Align data.  Fix unaligned ldr insns.  Adjust cmp
arguments to match change.
* ldn_multiple.s, stn_multiple.s: New.

7 years agoUse ui_out_emit_list
Tom Tromey [Wed, 12 Apr 2017 22:30:21 +0000 (16:30 -0600)]
Use ui_out_emit_list

This changes some spots to use ui_out_emit_list.  This only touches
"easy" cases, where the cleanup was used in a block-structured way.
There's also one more use of ui_out_emit_tuple in here.

ChangeLog
2017-04-22  Tom Tromey  <tom@tromey.com>

* mi/mi-cmd-file.c (mi_cmd_file_list_shared_libraries): Use
ui_out_emit_list.
* stack.c (print_frame): Use ui_out_emit_list.
* mi/mi-symbol-cmds.c (mi_cmd_symbol_list_lines): Use
ui_out_emit_list.
* mi/mi-main.c (print_one_inferior)
(mi_cmd_data_list_register_names)
(mi_cmd_data_list_register_values, mi_cmd_list_features)
(mi_cmd_list_target_features, mi_cmd_trace_frame_collected): Use
ui_out_emit_list.
* mi/mi-interp.c (mi_on_normal_stop_1): Use ui_out_emit_list.
(mi_output_solib_attribs): Use ui_out_emit_list,
ui_out_emit_tuple.
* mi/mi-cmd-var.c (varobj_update_one): Use ui_out_emit_list.
* mi/mi-cmd-stack.c (mi_cmd_stack_list_frames)
(mi_cmd_stack_list_args, list_args_or_locals): Use
ui_out_emit_list.
* disasm.c (do_assembly_only): Use ui_out_emit_list.
* breakpoint.c (print_solib_event, output_thread_groups): Use
ui_out_emit_list.

7 years agoUse ui_out_emit_tuple in more places in MI
Tom Tromey [Wed, 12 Apr 2017 22:16:10 +0000 (16:16 -0600)]
Use ui_out_emit_tuple in more places in MI

This patch changes a few more spots in MI to use ui_out_emit_tuple.
These changes required the use of gdb::optional.

ChangeLog
2017-04-22  Tom Tromey  <tom@tromey.com>

* mi/mi-main.c (print_variable_or_computed): Use ui_out_emit_tuple.
* mi/mi-cmd-var.c (varobj_update_one): Use ui_out_emit_tuple.
* mi/mi-cmd-stack.c (list_arg_or_local): Use ui_out_emit_tuple.

7 years agoUse ui_out_emit_tuple in tracepoint.c
Tom Tromey [Wed, 12 Apr 2017 22:15:14 +0000 (16:15 -0600)]
Use ui_out_emit_tuple in tracepoint.c

This changes some code in tracepoint.c to use ui_out_emit_tuple.  One
of these involved removing an otherwise unrelated cleanup (changing
type to std::string) and the other involved introducing a new block.

ChangeLog
2017-04-22  Tom Tromey  <tom@tromey.com>

* tracepoint.c (tvariables_info_1)
(print_one_static_tracepoint_marker): Use ui_out_emit_tuple.

7 years agoMore uses of ui_out_emit_tuple
Tom Tromey [Wed, 12 Apr 2017 22:10:02 +0000 (16:10 -0600)]
More uses of ui_out_emit_tuple

This patch adds a few more uses of ui_out_emit_tuple.  In these cases
a slightly more complicated change was needed.  This also adds
annotate_arg_emitter, for use in stack.c, to avoid having to introduce
a new scope and reindent the code for a single call.

ChangeLog
2017-04-22  Tom Tromey  <tom@tromey.com>

* stack.c (print_frame_arg): Use ui_out_emit_tuple,
annotate_arg_emitter.
* breakpoint.c (print_mention_watchpoint)
(print_mention_masked_watchpoint): Use ui_out_emit_tuple.
* annotate.h (struct annotate_arg_emitter): New.

7 years agoUse ui_out_emit_tuple
Tom Tromey [Wed, 12 Apr 2017 14:23:08 +0000 (08:23 -0600)]
Use ui_out_emit_tuple

This patch changes various places to use ui_out_emit_tuple,
eliminating a number of cleanups.  This patch only tackles "easy"
cases, which are ones where the cleanups in question were
block-structured and did not involve any changes other than the
obvious replacement.

ChangeLog
2017-04-22  Tom Tromey  <tom@tromey.com>

* record-btrace.c (record_btrace_insn_history)
(record_btrace_insn_history_range, record_btrace_call_history)
(record_btrace_call_history_range): Use ui_out_emit_tuple.
* thread.c (do_captured_list_thread_ids, print_thread_info_1): Use
ui_out_emit_tuple.
* stack.c (print_frame_info): Use ui_out_emit_tuple.
* solib.c (info_sharedlibrary_command): Use ui_out_emit_tuple.
* skip.c (skip_info): Use ui_out_emit_tuple.
* remote.c (show_remote_cmd): Use ui_out_emit_tuple.
* progspace.c (print_program_space): Use ui_out_emit_tuple.
* probe.c (info_probes_for_ops): Use ui_out_emit_tuple.
* osdata.c (info_osdata): Use ui_out_emit_tuple.
* mi/mi-symbol-cmds.c (mi_cmd_symbol_list_lines): Use
ui_out_emit_tuple.
* mi/mi-main.c (print_one_inferior, list_available_thread_groups)
(output_register, mi_cmd_data_read_memory)
(mi_cmd_data_read_memory_bytes, mi_load_progress)
(mi_cmd_trace_frame_collected): Use ui_out_emit_tuple.
* mi/mi-cmd-var.c (mi_cmd_var_list_children, varobj_update_one):
Use ui_out_emit_tuple.
* mi/mi-cmd-stack.c (mi_cmd_stack_list_args): Use
ui_out_emit_tuple.
* mi/mi-cmd-info.c (mi_cmd_info_ada_exceptions)
(mi_cmd_info_gdb_mi_command): Use ui_out_emit_tuple.
* linux-thread-db.c (info_auto_load_libthread_db): Use
ui_out_emit_tuple.
* inferior.c (print_inferior): Use ui_out_emit_tuple.
* gdb_bfd.c (print_one_bfd): Use ui_out_emit_tuple.
* disasm.c (do_mixed_source_and_assembly_deprecated)
(do_mixed_source_and_assembly): Use ui_out_emit_tuple.
* cp-abi.c (list_cp_abis): Use ui_out_emit_tuple.
* cli/cli-setshow.c (cmd_show_list): Use ui_out_emit_tuple.
* breakpoint.c (print_one_breakpoint_location)
(print_one_breakpoint): Use ui_out_emit_tuple.
* auto-load.c (print_script, info_auto_load_cmd): Use
ui_out_emit_tuple.
* ada-tasks.c (print_ada_task_info): Use ui_out_emit_tuple.

7 years agoPowerPC VLE insn set additions
Alan Modra [Sat, 22 Apr 2017 04:39:21 +0000 (14:09 +0930)]
PowerPC VLE insn set additions

opcodes/
* ppc-opc.c (ELEV): Define.
(vle_opcodes): Add se_rfgi and e_sc.
(powerpc_opcodes): Enable lbdx, lhdx, lwdx, stbdx, sthdx, stwdx
for E200Z4.
gas/
* testsuite/gas/ppc/vle.s: Format.  Add se_rfgi and e_sc.
* testsuite/gas/ppc/vle.d: Update.

7 years agodoc: Improve documentation about MI thread output
Simon Marchi [Sat, 22 Apr 2017 01:50:22 +0000 (21:50 -0400)]
doc: Improve documentation about MI thread output

I noticed that the documentation on how the info about threads is output
in MI is duplicated and not up to date.  The duplication is between the
"GDB/MI Thread Information" page and the -thread-info result
description.

I improved the "GDB/MI Thread Information" page a bit and referred to it
in the -thread-info doc.  This way, the -thread-info doc is more precise
(it did not mention the "threads" and "current-thread-id" attributes)
and concise.

gdb/doc/ChangeLog:

* gdb.texinfo (GDB/MI Thread Information): Add missing
fields, re-word some things.
(GDB/MI Thread Commands): Describe fields found in the output of
-thread-info, remove description of fields in the
thread output tuple, replace with a cross-reference to "GDB/MI
Thread Information".

7 years agoRemove dead code and "current" field from MI thread output doc
Simon Marchi [Sat, 22 Apr 2017 01:45:15 +0000 (21:45 -0400)]
Remove dead code and "current" field from MI thread output doc

The MI documentation says that -thread-info output contains a "current"
field in the current thread tuple, with the value "*".  Current GDB
master does not do this, and I couldn't find any GDB version that did.
I suspect that it was never the case.

The code that would correspond to this in print_thread_info_1 is
essentially dead code.  The calls to uiout->text end up in
mi_out::do_text, which is empty.

This patch removes the documentation bit and the dead code.  This
"current" field is not necessary, since -thread-info outputs a
"current-thread-id" field.

gdb/ChangeLog:

* thread.c (print_thread_info_1): Remove dead code.

gdb/doc/ChangeLog:

* gdb.texinfo (GDB/MI Thread Commands): Remove "current" field
from -thread-info output.

7 years agoAutomatic date update in version.in
GDB Administrator [Sat, 22 Apr 2017 00:00:47 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoi386: Avoid dynamic symbol with GOT reference in PIE
H.J. Lu [Fri, 21 Apr 2017 19:03:14 +0000 (12:03 -0700)]
i386: Avoid dynamic symbol with GOT reference in PIE

GOT reference to global symbol in PIE will lead to dynamic symbol.  It
becomes a problem when "time" or "times" is defined as a variable in
an executable, clashing with functions of the same name in libc.  If
a symbol isn't undefined weak symbol, don't make it dynamic in PIE and
generate R_386_RELATIVE relocation.

bfd/

PR ld/21402
* elf32-i386.c (elf_i386_link_hash_entry): Add
no_finish_dynamic_symbol.
(elf_i386_link_hash_newfunc): Set no_finish_dynamic_symbol to 0.
(elf_i386_allocate_dynrelocs): If a symbol isn't undefined weak
symbol, don't make it dynamic in PIE.
(elf_i386_relocate_section): If a symbol isn't dynamic in PIE,
set no_finish_dynamic_symbol and generate R_386_RELATIVE
relocation for R_386_GOT32
(elf_i386_finish_dynamic_symbol): Abort if no_finish_dynamic_symbol
isn't 0.

ld/

PR ld/21402
* testsuite/ld-elf/indirect.exp: Don't skip PIE indirect5 and
indirect6 tests on i386.

7 years agoRequire --no-dynamic-linker with -static -E/--dynamic-list
H.J. Lu [Fri, 21 Apr 2017 19:00:55 +0000 (12:00 -0700)]
Require --no-dynamic-linker with -static -E/--dynamic-list

When -static -E/--dynamic-list are passed to linker, linker may create
executable with dynamic sections which aren't supported by run-time.
We require --no-dynamic-linker together with -static -E/--dynamic-list
before adding dynamic symbol table to static executable.

bfd/

PR ld/19617
PR ld/21086
* elflink.c (elf_link_add_object_symbols): Require
--no-dynamic-linker with -E/--dynamic-list when creating
dynamic sections.

ld/

PR ld/19617
PR ld/21086
* testsuite/ld-elf/pr19617a.d: Pass --no-dynamic-linker to ld.
* testsuite/ld-elf/pr19617b.d: Likewise.
* testsuite/ld-elf/pr19617c.d: Likewise.
*testsuite/ld-i386/pr19636-4d.d: Likewise.
* testsuite/ld-elf/readelf.exp: Pass --no-dynamic-linker to ld
with --export-dynamic.
* testsuite/ld-elf/shared.exp: Pass --no-dynamic-linker to ld
with -E.

7 years agorelease branch: Fix: --enable-werror
Jan Kratochvil [Fri, 21 Apr 2017 14:14:37 +0000 (16:14 +0200)]
release branch: Fix: --enable-werror

gdb-8.0-branch
./configure --enable-werror --enable-targets=all
aarch64-tdep.c:3045:13: error: ‘void selftests::aarch64_process_record_test()’ declared ‘static’ but never defined [-Werror=unused-function]
arm-tdep.c:9601:13: error: ‘void selftests::arm_record_test()’ declared ‘static’ but never defined [-Werror=unused-function]

gdb/ChangeLog
2017-04-21  Jan Kratochvil  <jan.kratochvil@redhat.com>

* aarch64-tdep.c (selftests::aarch64_process_record_test): Make it #if
GDB_SELF_TEST.
* arm-tdep.c (selftests::arm_record_test): Likewise.

7 years agoSimplify regcache_restore
Yao Qi [Fri, 21 Apr 2017 13:59:27 +0000 (14:59 +0100)]
Simplify regcache_restore

This patches removes the 2nd argument of regcache_restore, because it
is only called by regcache_cpy.  In regcache_cpy, if regcache_restore
is called, dst is not readonly, but src is readonly.  So this patch
adds an assert that src is readonly in regcache_restore.
regcache_cook_read read everything from a readonly regcache cache
(src)'s register_buffer, and register status is from ->register_status.

gdb:

2017-04-21  Yao Qi  <yao.qi@linaro.org>

* regcache.c (regcache_restore): Remove argument 2.  Replace
argument 3 with regcache.  Get register status from
src->register_status and get register contents from
register_buffer (src, regnum).
(regcache_cpy): Update.

7 years agoopcodes: mark SPARC RETT instructions as v6notv9.
Jose E. Marchesi [Fri, 21 Apr 2017 12:19:06 +0000 (05:19 -0700)]
opcodes: mark SPARC RETT instructions as v6notv9.

This prevents the disassembler to show `return' instructions as
`rett' in V9 and later architectures.

opcodes/ChangeLog:

2017-04-21  Jose E. Marchesi  <jose.marchesi@oracle.com>

* sparc-opc.c (sparc_opcodes): Mark RETT instructions as v6notv9.

7 years agoFix shift overflow when parsing an overlarge note value.
Nick Clifton [Fri, 21 Apr 2017 11:31:59 +0000 (12:31 +0100)]
Fix shift overflow when parsing an overlarge note value.

PR binutils/21378
* readelf.c (print_gnu_build_attribute_name): Check for an
overlarge name field.

7 years agoFix detection of illegal AArch64 opcodes that resemble LD1R, LD2R, LD3R and LD4R.
Nick Clifton [Fri, 21 Apr 2017 11:18:06 +0000 (12:18 +0100)]
Fix detection of illegal AArch64 opcodes that resemble LD1R, LD2R, LD3R and LD4R.

PR binutils/21380
opcodes * aarch64-tbl.h (aarch64_opcode_table): Fix masks for LD1R, LD2R,
LD3R and LD4R.

gas * testsuite/gas/aarch64/illegal-3.s: New file.
* testsuite/gas/aarch64/illegal-3.d: New file.

7 years agoAutomatic date update in version.in
GDB Administrator [Fri, 21 Apr 2017 00:00:18 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agoSkip PIE indirect5 and indirect6 tests on i386
H.J. Lu [Thu, 20 Apr 2017 19:33:55 +0000 (12:33 -0700)]
Skip PIE indirect5 and indirect6 tests on i386

On i386, since GOT reference is needed to access global symbols in PIE,
those symbols are made dynamic.  Crash happens when there is a reference
to the same global symbol with a different symbol type in a shared
object.  Since mixing different types of the same symbol doesn't work in
general, this patch skips those tests on i386 as well as compiles non-PIE
indirect5 and indirect6 tests with $NOPIE_LDFLAGS and $NOPIE_CFLAGS.

PR ld/21402
* testsuite/ld-elf/indirect.exp: Pass $NOPIE_LDFLAGS and
$NOPIE_CFLAGS to non-PIE indirect5 and indirect6 tests.  Skip
PIE indirect5 and indirect6 tests on i386.

7 years agoELF/LD: Always consider STB_LOCAL symbols local
Maciej W. Rozycki [Thu, 13 Apr 2017 09:26:05 +0000 (10:26 +0100)]
ELF/LD: Always consider STB_LOCAL symbols local

Do not require forced local (STB_LOCAL) symbols to have a definition in
a regular file to be considered to resolve local to the current module,
matching `elf_link_renumber_local_hash_table_dynsyms'.  In the absence
of a regular definition any reference to a STB_LOCAL symbol will have to
be garbage collected along with the undefined symbol itself, or the link
will eventually fail.  Either way the symbol concerned is not going to
be external.

bfd/
* elflink.c (_bfd_elf_symbol_refs_local_p): Always return TRUE
if forced local.

7 years agoMIPS/LD/testsuite: Join `__ehdr_start' tests
Maciej W. Rozycki [Thu, 20 Apr 2017 00:39:30 +0000 (01:39 +0100)]
MIPS/LD/testsuite: Join `__ehdr_start' tests

Complement commit 3807734dbe48 ("PR ld/15428: MIPS/LD/testsuite:
Un-KFAIL `__ehdr_start' test 2") and join tests that do not need to be
split anymore.

ld/
* testsuite/ld-mips-elf/mips-elf.exp: Join `__ehdr_start' tests.

7 years agoMIPS/BFD: Reorder comment about dynamic symbol sorting
Maciej W. Rozycki [Wed, 19 Apr 2017 15:51:17 +0000 (16:51 +0100)]
MIPS/BFD: Reorder comment about dynamic symbol sorting

Move the comment about dynamic symbol sorting next to where it happens.

bfd/
* elfxx-mips.c (_bfd_mips_elf_final_link): Reorder comment about
dynamic symbol sorting.

7 years agoHandle symbol defined in IR and referenced in DSO
H.J. Lu [Thu, 20 Apr 2017 14:48:24 +0000 (07:48 -0700)]
Handle symbol defined in IR and referenced in DSO

We need to make an IR symbol visible if it is defined in an IR object
and referenced in a dynamic object.  When --as-needed is used, since
linker removes the IR symbol reference of the dynamic object if the
dynamic object isn't needed in the first pass, the IR definition isn't
visible to the dynamic object even if the dynamic object becomes needed
in the second pass.  Add dynamic_ref_after_ir_def to bfd_link_hash_entry
to track IR symbol which is defined in an IR object and later referenced
in a dynamic object.  dynamic_ref_after_ir_def is preserved when restoring
the symbol table for unneeded dynamic object.

bfd/

PR ld/21382
* elflink.c (elf_link_add_object_symbols): Preserve
dynamic_ref_after_ir_def when restoring the symbol table for
unneeded dynamic object.

include/

PR ld/21382
* bfdlink.h (bfd_link_hash_entry): Add dynamic_ref_after_ir_def.

ld/

PR ld/21382
* plugin.c (is_visible_from_outside): Symbol may be visible
from outside if dynamic_ref_after_ir_def is set.
(plugin_notice): Set dynamic_ref_after_ir_def if the symbol is
defined in an IR object and referenced in a dynamic object.
* testsuite/ld-plugin/lto.exp: Run PR ld/21382 tests.
* testsuite/ld-plugin/pr21382a.c: New file.
* testsuite/ld-plugin/pr21382b.c: Likewise.

7 years agoAutomatic date update in version.in
GDB Administrator [Thu, 20 Apr 2017 00:00:41 +0000 (00:00 +0000)]
Automatic date update in version.in

7 years agox86-64: Handle undefined IFUNC symbol with -z now
H.J. Lu [Wed, 19 Apr 2017 16:59:14 +0000 (09:59 -0700)]
x86-64: Handle undefined IFUNC symbol with -z now

Since undefined IFUNC symbol is treated as normal FUNC symbol, don't
abort on undefined IFUNC symbol in the second PLT.

bfd/

PR ld/21401
* elf64-x86-64.c (elf_x86_64_finish_dynamic_symbol): Don't abort
on on undefined IFUNC symbol in the second PLT.

ld/

PR ld/21401
* testsuite/ld-ifunc/ifunc.exp: Add a libtest-2-now.so test with
-z now.

7 years agogdbthread.h: Fix comment typo
Pedro Alves [Wed, 19 Apr 2017 12:16:05 +0000 (13:16 +0100)]
gdbthread.h: Fix comment typo

gdb/ChangeLog:
2017-04-19  Pedro Alves  <palves@redhat.com>

* gdbthread.h (thread): Add missing closing parenthesis in
comment.

7 years agoFix removing inferiors from within "thread apply" commands
Pedro Alves [Wed, 19 Apr 2017 12:12:23 +0000 (13:12 +0100)]
Fix removing inferiors from within "thread apply" commands

This patch fixes an internal error exposed by a test that does
something like:

  define kill-and-remove
    kill inferiors 2
    remove-inferiors 2
  end

  # Start one inferior.
  start

  # Start another inferior.
  add-inferior 2
  inferior 2
  start

  # Kill and remove inferior 1 while inferior 2 is selected.
  thread apply 1.1 kill-and-remove

The internal error looks like this:

 Thread 1.1 (Thread 0x7ffff7fc2700 (LWP 20677)):
 [Switching to inferior 1 [process 20677] (gdb/testsuite/outputs/gdb.threads/threadapply/threadapply)]
 [Switching to thread 1.1 (Thread 0x7ffff7fc2700 (LWP 20677))]
 #0  main () at src/gdb/testsuite/gdb.threads/threadapply.c:38
 38          for (i = 0; i < NUM; i++)
 src/gdb/inferior.c:66: internal-error: void set_current_inferior(inferior*): Assertion `inf != NULL' failed.
 A problem internal to GDB has been detected,
 further debugging may prove unreliable.
 Quit this debugging session? (y or n) FAIL: gdb.threads/threadapply.exp: kill_and_remove_inferior: try kill-and-remove: thread apply 1.1 kill-and-remove (GDB internal error)

There are several problems around this area of the code.  One is that
in do_restore_current_thread_cleanup, we do a look up of inferior by
ptid, which can find the wrong inferior if the previously selected
inferior exited and some other inferior was started with a reused pid
(rare, but still...).

The other problem is that the "remove-inferiors" command rejects
attempts to remove the current inferior, but when we get to
"remove-inferiors" in a "thread apply THR remove-inferiors 2" command,
the current inferior is the inferior of thread THR, not the previously
selected inferior, so if the previously selected inferior was inferior
2, that command still manages to wipe it, and then gdb restores the
old selected inferior, which is now a dangling pointer...

So the fix here is:

- Make make_cleanup_restore_current_thread store a pointer to the
  previously selected inferior directly, and use it directly instead
  of doing ptid look ups.

- Add a refcount to inferiors, very similar to thread_info's refcount,
  that is incremented/decremented by
  make_cleanup_restore_current_thread, and checked before deleting an
  inferior.  To avoid duplication, a new refcounted_object type is
  added, that both thread_info and inferior inherit from.

gdb/ChangeLog:
2017-04-19  Pedro Alves  <palves@redhat.com>

* common/refcounted-object.h: New file.
* gdbthread.h: Include "common/refcounted-object.h".
(thread_info): Inherit from refcounted_object and add comments.
(thread_info::incref, thread_info::decref)
(thread_info::m_refcount): Delete.
(thread_info::deletable): Use the refcounted_object::refcount()
method.
* inferior.c (current_inferior_): Add comment.
(set_current_inferior): Increment/decrement refcounts.
(prune_inferiors, remove_inferior_command): Skip inferiors marked
not-deletable instead of comparing with the current inferior.
(initialize_inferiors): Increment the initial inferior's refcount.
* inferior.h (struct inferior): Forward declare.
Include "common/refcounted-object.h".
(current_inferior, set_current_inferior): Move declaration to
before struct inferior's definition, and fix comment.
(inferior): Inherit from refcounted_object.  Add comments.
* thread.c (switch_to_thread_no_regs): Reference the thread's
inferior pointer directly instead of doing a ptid lookup.
(switch_to_no_thread): New function.
(switch_to_thread(thread_info *)): New function, factored out
from ...
(switch_to_thread(ptid_t)): ... this.
(restore_current_thread): Delete.
(current_thread_cleanup): Remove 'inf_id' and 'was_removable'
fields, and add 'inf' field.
(do_restore_current_thread_cleanup): Check whether old->inf is
alive instead of looking up an inferior by ptid.  Use
switch_to_thread and switch_to_no_thread.
(restore_current_thread_cleanup_dtor): Use old->inf directly
instead of lookup up an inferior by id.  Decref the inferior.
Don't restore 'removable'.
(make_cleanup_restore_current_thread): Same the inferior pointer
in old, instead of the inferior number.  Incref the inferior.
Don't save/clear 'removable'.

gdb/testsuite/ChangeLog:
2017-04-19  Pedro Alves  <palves@redhat.com>

* gdb.threads/threadapply.exp (kill_and_remove_inferior): New
procedure.
(top level): Call it.
* lib/gdb.exp (gdb_define_cmd): New procedure.

7 years agoMake inferior::detaching a bool, and introduce scoped_restore::release()
Pedro Alves [Wed, 19 Apr 2017 12:12:23 +0000 (13:12 +0100)]
Make inferior::detaching a bool, and introduce scoped_restore::release()

I left making inferior::detaching a bool to a separate patch, because
doing that makes a make_cleanup_restore_integer call in
infrun.c:prepare_for_detach no longer compile (passing a 'bool *' when
an 'int *' is expected).  Since we want to get rid of cleanups anyway,
I looked at converting that to a scoped_restore.  However,
prepare_for_detach wants to discard the cleanup on success, and
scoped_restore doesn't have an equivalent for that.  So I added one --
I called it "release()" because it seems like a natural fit in the way
standard components call similarly-spirited methods, and, it's also
what the proposal for a generic scope guard calls it too, AFAICS:

  http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n4189.pdf

I've added some scoped_guard unit tests, while at it.

gdb/ChangeLog:
2017-04-19  Pedro Alves  <palves@redhat.com>

* Makefile.in (SUBDIR_UNITTESTS_SRCS): Add
unittests/scoped_restore-selftests.c.
(SUBDIR_UNITTESTS_OBS): Add scoped_restore-selftests.o.
* common/scoped_restore.h (scoped_restore_base): Make "class".
(scoped_restore_base::release): New public method.
(scoped_restore_base::scoped_restore_base): New protected ctor.
(scoped_restore_base::m_saved_var): New protected field.
(scoped_restore_tmpl::scoped_restore_tmpl(T*)): Initialize the
scoped_restore_base base class instead of m_saved_var directly.
(scoped_restore_tmpl::scoped_restore_tmpl(T*, T2)): Likewise.
(scoped_restore_tmpl::scoped_restore_tmpl(const
scoped_restore_tmpl<T>&)): Likewise.
(scoped_restore_tmpl::~scoped_restore_tmpl): Use the saved_var
method.
(scoped_restore_tmpl::saved_var): New method.
(scoped_restore_tmpl::m_saved_var): Delete.
* inferior.h (inferior::detaching): Now a bool.
* infrun.c (prepare_for_detach): Use a scoped_restore instead of a
cleanup.
* unittests/scoped_restore-selftests.c: New file.

7 years agogdb/Makefile.in: Re-sort SUBDIR_UNITTESTS_SRCS/SUBDIR_UNITTESTS_OBS
Pedro Alves [Wed, 19 Apr 2017 12:06:20 +0000 (13:06 +0100)]
gdb/Makefile.in: Re-sort SUBDIR_UNITTESTS_SRCS/SUBDIR_UNITTESTS_OBS

Note to self: 'o' before 'p'.

gdb/ChangeLog:
2017-04-19  Pedro Alves  <palves@redhat.com>

* Makefile.in (SUBDIR_UNITTESTS_SRCS, SUBDIR_UNITTESTS_OBS):
Re-sort in alphabetic order.

7 years agoCorrect PE reloc display
Wedson Almeida Filho [Wed, 19 Apr 2017 08:32:53 +0000 (18:02 +0930)]
Correct PE reloc display

* peXXigen.c (pe_print_reloc): Correct chunk_end.