PR 20059 _bfd_elf_copy_link_hash_symbol_type segfault

PR ld/20059
* elfxx-target.h (bfd_elfNN_bfd_copy_link_hash_symbol_type):
Define as _bfd_generic_copy_link_hash_symbol_type when using
generic hash table.

Automatic date update in version.in

Fix for PR gold/19987.

Fix seg fault objdumping a corrupt binary with an invalid sh_link field.

PR binutils/20063
* elf.c (bfd_elf_get_elf_syms): Check for out of range sh_link
field before accessing sections array.

* readelf.c (get_32bit_section_headers): Warn if an out of range
sh_link or sh_info field is encountered.
(get_64bit_section_headers): Likewise.

[ARM/STM32L4XX] PR 20030: --fix-stm32l4xx-629360 fails to create vldm/vpop veneers for double-precision registers

bfd/
PR ld/20030
* elf32-arm.c (is_thumb2_vldm): Account for T1 (DP) encoding.
(stm32l4xx_need_create_replacing_stub): Rename ambiguous nb_regs
to nb_words.
(create_instruction_vldmia): Add is_dp to disambiguate SP/DP
encoding.
(create_instruction_vldmdb): Likewise.
(stm32l4xx_create_replacing_stub_vldm): is_dp detects DP encoding,
uses it to re-encode.

ld/
PR ld/20030
* testsuite/ld-arm/arm-elf.exp: Run new stm32l4xx-fix-vldm-dp
tests. Fix misnamed stm32l4xx-fix-all.
* testsuite/ld-arm/stm32l4xx-fix-vldm-dp.s: New tests for multiple
loads with DP registers.
* testsuite/ld-arm/stm32l4xx-fix-vldm-dp.d: New reference file.
* testsuite/ld-arm/stm32l4xx-fix-vldm.s: Add missing comment.
* testsuite/ld-arm/stm32l4xx-fix-all.s: Add tests for multiple
loads with DP registers.
* testsuite/ld-arm/stm32l4xx-fix-all.d: Update reference.

Revert accidental commit.

PR 19938
* elf32-arm.c (elf32_arm_adjust_dynamic_symbol): Revert accidental
commit.

Update section merge test.

* testsuite/ld-elf/flags1.d (readelf): Dump section header instead
program headers.
(xfail): Remove avr-*-*.
Update regex to check the section flags.
* testsuite/ld-elf/merge.d (xfail): Remove avr-*-*.

opcodes,gas: sparc: fix mnemonic of faligndatai

opcodes/ChangeLog:

2016-05-09  Jose E. Marchesi  <jose.marchesi@oracle.com>

* sparc-opc.c (sparc_opcodes): Fix mnemonic of faligndatai.

gas/ChangeLog:

2016-05-09  Jose E. Marchesi  <jose.marchesi@oracle.com>

* testsuite/gas/sparc/sparc5vis4.s: Fix mnemonic of faligndatai.
* testsuite/gas/sparc/sparc5vis4.d: Likewise.

Regenerate configure

Fix ealphavms.c and eelf64_ia64_vms.c deps

* Makefile.am (ealphavms.c, eelf64_ia64_vms): Correct .em deps.
* Makefile.in: Regenerate.

Redundant hash table check

* elf64-ppc.c (ppc64_elf_init_stub_bfd): Remove redundant NULL check.

Correct script input section pattern

We don't want this to match .rela.text or similar.

* testsuite/ld-scripts/pr14962-2.t: Match .text, not *.text.
* testsuite/ld-scripts/rgn-at5.t: Similarly, .sec not *.sec.
* testsuite/ld-scripts/section-match-1.t: Likewise.

Automatic date update in version.in

remove trivialy unused variables

gdb/ChangeLog:

2016-05-07  Trevor Saunders  <tbsaunde+binutils@tbsaunde.org>

* aarch64-linux-tdep.c (aarch64_linux_sigframe_init): Remove unused
variables.
* aarch64-tdep.c (aarch64_skip_prologue): Likewise.
(aarch64_scan_prologue): Likewise.
(aarch64_prologue_prev_register): Likewise.
(aarch64_dwarf2_prev_register): Likewise.
(pass_in_v): Likewise.
(aarch64_push_dummy_call): Likewise.
(aarch64_breakpoint_from_pc): Likewise.
(aarch64_return_in_memory): Likewise.
(aarch64_return_value): Likewise.
(aarch64_displaced_step_b_cond): Likewise.
(aarch64_displaced_step_cb): Likewise.
(aarch64_displaced_step_tb): Likewise.
(aarch64_gdbarch_init): Likewise.
(aarch64_process_record): Likewise.
* alpha-mdebug-tdep.c (alpha_mdebug_init_abi): Likewise.
* alpha-tdep.c (_initialize_alpha_tdep): Likewise.
* amd64-dicos-tdep.c (amd64_dicos_init_abi): Likewise.
* amd64-linux-tdep.c (amd64_dtrace_parse_probe_argument): Likewise.
* amd64-tdep.c (fixup_riprel): Likewise.
* amd64-windows-tdep.c (amd64_windows_frame_decode_epilogue): Likewise.
(amd64_windows_frame_decode_insns): Likewise.
(amd64_windows_frame_cache): Likewise.
(amd64_windows_frame_prev_register): Likewise.
(amd64_windows_frame_this_id): Likewise.
(amd64_windows_init_abi): Likewise.
* arm-linux-tdep.c (arm_linux_get_syscall_number): Likewise.
(arm_linux_get_next_pcs_syscall_next_pc): Likewise.
* arm-symbian-tdep.c (arm_symbian_init_abi): Likewise.
* arm-tdep.c (arm_make_epilogue_frame_cache): Likewise.
(arm_epilogue_frame_prev_register): Likewise.
(arm_record_vdata_transfer_insn): Likewise.
(arm_record_exreg_ld_st_insn): Likewise.
* auto-load.c (execute_script_contents): Likewise.
(print_scripts): Likewise.
* avr-tdep.c (avr_frame_prev_register): Likewise.
(avr_push_dummy_call): Likewise.
* bfin-linux-tdep.c (bfin_linux_sigframe_init): Likewise.
* bfin-tdep.c (bfin_gdbarch_init): Likewise.
* blockframe.c (find_pc_partial_function_gnu_ifunc): Likewise.
* break-catch-throw.c (fetch_probe_arguments): Likewise.
* breakpoint.c (breakpoint_xfer_memory): Likewise.
(breakpoint_init_inferior): Likewise.
(breakpoint_inserted_here_p): Likewise.
(software_breakpoint_inserted_here_p): Likewise.
(hardware_breakpoint_inserted_here_p): Likewise.
(bpstat_what): Likewise.
(break_range_command): Likewise.
(save_breakpoints): Likewise.
* coffread.c (coff_symfile_read): Likewise.
* cris-tdep.c (cris_push_dummy_call): Likewise.
(cris_scan_prologue): Likewise.
(cris_register_size): Likewise.
(_initialize_cris_tdep): Likewise.
* d-exp.y: Likewise.
* dbxread.c (dbx_read_symtab): Likewise.
(process_one_symbol): Likewise.
(coffstab_build_psymtabs): Likewise.
(elfstab_build_psymtabs): Likewise.
* dicos-tdep.c (dicos_init_abi): Likewise.
* disasm.c (do_mixed_source_and_assembly): Likewise.
(gdb_disassembly): Likewise.
* dtrace-probe.c (dtrace_process_dof): Likewise.
* dwarf2read.c (error_check_comp_unit_head): Likewise.
(build_type_psymtabs_1): Likewise.
(skip_one_die): Likewise.
(process_imported_unit_die): Likewise.
(dwarf2_physname): Likewise.
(read_file_scope): Likewise.
(setup_type_unit_groups): Likewise.
(create_dwo_cu_reader): Likewise.
(create_dwo_cu): Likewise.
(create_dwo_unit_in_dwp_v1): Likewise.
(create_dwo_unit_in_dwp_v2): Likewise.
(lookup_dwo_unit_in_dwp): Likewise.
(free_dwo_file): Likewise.
(check_producer): Likewise.
(dwarf2_add_typedef): Likewise.
(dwarf2_add_member_fn): Likewise.
(read_unsigned_leb128): Likewise.
(read_signed_leb128): Likewise.
(dwarf2_const_value): Likewise.
(follow_die_sig_1): Likewise.
(dwarf_decode_macro_bytes): Likewise.
* extension.c (restore_active_ext_lang): Likewise.
* frv-linux-tdep.c (frv_linux_sigtramp_frame_cache): Likewise.
* ft32-tdep.c (ft32_analyze_prologue): Likewise.
* gdbtypes.c (lookup_typename): Likewise.
(resolve_dynamic_range): Likewise.
(check_typedef): Likewise.
* h8300-tdep.c (h8300_is_argument_spill): Likewise.
(h8300_gdbarch_init): Likewise.
* hppa-tdep.c (hppa32_push_dummy_call): Likewise.
(hppa_frame_this_id): Likewise.
(_initialize_hppa_tdep): Likewise.
* hppanbsd-tdep.c (hppanbsd_sigtramp_cache_init): Likewise.
* hppaobsd-tdep.c (hppaobsd_supply_fpregset): Likewise.
* i386-dicos-tdep.c (i386_dicos_init_abi): Likewise.
* i386-tdep.c (i386_bnd_type): Likewise.
(i386_gdbarch_init): Likewise.
(i386_mpx_bd_base): Likewise.
* i386nbsd-tdep.c (i386nbsd_sigtramp_cache_init): Likewise.
* i386obsd-tdep.c (i386obsd_elf_init_abi): Likewise.
* ia64-tdep.c (examine_prologue): Likewise.
(ia64_frame_cache): Likewise.
(ia64_push_dummy_call): Likewise.
* infcmd.c (finish_command_fsm_async_reply_reason): Likewise.
(default_print_one_register_info): Likewise.
* infrun.c (infrun_thread_ptid_changed): Likewise.
(thread_still_needs_step_over): Likewise.
(stop_all_threads): Likewise.
(restart_threads): Likewise.
(keep_going_stepped_thread): Likewise.
* iq2000-tdep.c (iq2000_scan_prologue): Likewise.
* language.c (language_init_primitive_type_symbols): Likewise.
* linespec.c (add_sal_to_sals): Likewise.
* linux-nat.c (status_callback): Likewise.
(kill_unfollowed_fork_children): Likewise.
(linux_nat_kill): Likewise.
* linux-tdep.c (linux_fill_prpsinfo): Likewise.
* linux-thread-db.c (thread_db_notice_clone): Likewise.
(record_thread): Likewise.
* location.c (string_to_event_location_basic): Likewise.
* m32c-tdep.c (m32c_prev_register): Likewise.
* m32r-linux-tdep.c (m32r_linux_init_abi): Likewise.
* m32r-tdep.c (decode_prologue): Likewise.
* m68klinux-tdep.c (m68k_linux_sigtramp_frame_cache): Likewise.
* machoread.c (macho_symtab_read): Likewise.
(macho_symfile_read): Likewise.
(macho_symfile_offsets): Likewise.
* maint.c (set_per_command_cmd): Likewise.
* mi/mi-cmd-stack.c (mi_cmd_stack_list_locals): Likewise.
(mi_cmd_stack_list_variables): Likewise.
* mi/mi-main.c (mi_cmd_exec_run): Likewise.
(output_register): Likewise.
(mi_cmd_execute): Likewise.
(mi_cmd_trace_define_variable): Likewise.
(print_variable_or_computed): Likewise.
* minsyms.c (prim_record_minimal_symbol_full): Likewise.
* mn10300-tdep.c (mn10300_frame_prev_register): Likewise.
* msp430-tdep.c (msp430_pseudo_register_write): Likewise.
* mt-tdep.c (mt_registers_info): Likewise.
* nios2-tdep.c (nios2_analyze_prologue): Likewise.
(nios2_push_dummy_call): Likewise.
(nios2_frame_unwind_cache): Likewise.
(nios2_stub_frame_cache): Likewise.
(nios2_stub_frame_sniffer): Likewise.
(nios2_gdbarch_init): Likewise.
* ppc-ravenscar-thread.c: Likewise.
* ppcfbsd-tdep.c (ppcfbsd_sigtramp_frame_cache): Likewise.
* python/py-evts.c (add_new_registry): Likewise.
* python/py-finishbreakpoint.c (bpfinishpy_init): Likewise.
(bpfinishpy_detect_out_scope_cb): Likewise.
* python/py-framefilter.c (py_print_value): Likewise.
* python/py-inferior.c (infpy_write_memory): Likewise.
* python/py-infevents.c (create_inferior_call_event_object): Likewise.
* python/py-infthread.c (thpy_get_ptid): Likewise.
* python/py-linetable.c (ltpy_get_pcs_for_line): Likewise.
(ltpy_get_all_source_lines): Likewise.
(ltpy_is_valid): Likewise.
(ltpy_iternext): Likewise.
* python/py-symtab.c (symtab_and_line_to_sal_object): Likewise.
* python/py-unwind.c (pyuw_object_attribute_to_pointer): Likewise.
(unwind_infopy_str): Likewise.
* python/py-varobj.c (py_varobj_get_iterator): Likewise.
* ravenscar-thread.c (ravenscar_inferior_created): Likewise.
* rs6000-aix-tdep.c (rs6000_push_dummy_call): Likewise.
* rs6000-lynx178-tdep.c (rs6000_lynx178_push_dummy_call): Likewise.
* rs6000-tdep.c (ppc_deal_with_atomic_sequence): Likewise.
* s390-linux-tdep.c (s390_supply_tdb_regset): Likewise.
(s390_frame_prev_register): Likewise.
(s390_dwarf2_frame_init_reg): Likewise.
(s390_record_vr): Likewise.
(s390_process_record): Likewise.
* score-tdep.c (score_push_dummy_call): Likewise.
(score3_analyze_prologue): Likewise.
* sh-tdep.c (sh_extract_return_value_nofpu): Likewise.
* sh64-tdep.c (sh64_analyze_prologue): Likewise.
(sh64_push_dummy_call): Likewise.
(sh64_extract_return_value): Likewise.
(sh64_do_fp_register): Likewise.
* solib-aix.c (solib_aix_get_section_offsets): Likewise.
* solib-darwin.c (darwin_read_exec_load_addr_from_dyld): Likewise.
(darwin_solib_read_all_image_info_addr): Likewise.
* solib-dsbt.c (enable_break): Likewise.
* solib-frv.c (enable_break2): Likewise.
(frv_fdpic_find_canonical_descriptor): Likewise.
* solib-svr4.c (svr4_handle_solib_event): Likewise.
* sparc-tdep.c (sparc_skip_stack_check): Likewise.
* sparc64-linux-tdep.c (sparc64_linux_get_longjmp_target): Likewise.
* sparcobsd-tdep.c (sparc32obsd_init_abi): Likewise.
* spu-tdep.c (info_spu_dma_cmdlist): Likewise.
* stack.c (read_frame_local): Likewise.
* symfile.c (symbol_file_add_separate): Likewise.
(remove_symbol_file_command): Likewise.
* symmisc.c (maintenance_print_one_line_table): Likewise.
* symtab.c (symbol_cache_flush): Likewise.
(basic_lookup_transparent_type): Likewise.
(sort_search_symbols_remove_dups): Likewise.
* target.c (target_memory_map): Likewise.
(target_detach): Likewise.
(target_resume): Likewise.
(acquire_fileio_fd): Likewise.
(target_store_registers): Likewise.
* thread.c (print_thread_info_1): Likewise.
* tic6x-tdep.c (tic6x_analyze_prologue): Likewise.
* tilegx-linux-tdep.c (tilegx_linux_sigframe_init): Likewise.
* tilegx-tdep.c (tilegx_push_dummy_call): Likewise.
(tilegx_analyze_prologue): Likewise.
(tilegx_stack_frame_destroyed_p): Likewise.
(tilegx_frame_cache): Likewise.
* tracefile.c (trace_save): Likewise.
* tracepoint.c (encode_actions_and_make_cleanup): Likewise.
(start_tracing): Likewise.
(print_one_static_tracepoint_marker): Likewise.
* tui/tui.c (tui_enable): Likewise.
* valops.c (value_struct_elt_bitpos): Likewise.
(find_overload_match): Likewise.
(find_oload_champ): Likewise.
* value.c (value_contents_copy_raw): Likewise.
* windows-tdep.c (windows_get_tlb_type): Likewise.
* x86-linux-nat.c (x86_linux_enable_btrace): Likewise.
* xcoffread.c (record_minimal_symbol): Likewise.
(scan_xcoff_symtab): Likewise.
* xtensa-tdep.c (execute_code): Likewise.
(xtensa_gdbarch_init): Likewise.
(_initialize_xtensa_tdep): Likewise.

Automatic date update in version.in

Automatic date update in version.in

Strip global symbol defined in discarded section

When a global symbol is defined in COMDAT group, we shouldn't leave an
undefined symbol in symbol table when the symbol section is discarded
unless there is a reference to the symbol outside of COMDAT group.

bfd/

PR ld/17550
* elf-bfd.h (elf_link_hash_entry): Update comments for indx,
documenting that indx == -3 if symbol is defined in a discarded
section.
* elflink.c (elf_link_add_object_symbols): Set indx to -3 if
symbol is defined in a discarded section.
(elf_link_output_extsym): Strip a global symbol defined in a
discarded section.

ld/

PR ld/17550
* testsuite/ld-elf/pr17550-1.s: New file.
* testsuite/ld-elf/pr17550-2.s: Likewise.
* testsuite/ld-elf/pr17550-3.s: Likewise.
* testsuite/ld-elf/pr17550-4.s: Likewise.
* testsuite/ld-elf/pr17550a.d: Likewise.
* testsuite/ld-elf/pr17550b.d: Likewise.
* testsuite/ld-elf/pr17550c.d: Likewise.
* testsuite/ld-elf/pr17550d.d: Likewise.

Define elf_backend_add_symbol_hook for Intel MCU

elf_backend_add_symbol_hook is undefined for FreeBSD.  Define it for
Intel MCU to support STB_GNU_UNIQUE for Intel MCU and NaCl.

* elf32-i386.c (elf_backend_add_symbol_hook): Defined for Intel
MCU.

* ld/testsuite/ld-srec/srec.exp: Mark test as XFAIL for AVR.

[ARM][gas] Fix warnings about uninitialised uses and unused const variables

* config/tc-arm.c (fpu_arch_vfp_v1): Mark with ATTRIBUTE_UNUSED.
(fpu_arch_vfp_v3): Likewise.
(fpu_arch_neon_v1): Likewise.
(arm_arch_full): Likewise.
(parse_neon_el_struct_list): Initialize fields of firsttype.

Test load conversion with symbol versioning for x86

* testsuite/ld-i386/i386.exp: Run load7.
* testsuite/ld-i386/load7.d: New file.
* testsuite/ld-i386/load7.map: Likewise.
* testsuite/ld-i386/load7.s: Likewise.
* testsuite/ld-x86-64/load2.d: Likewise.
* testsuite/ld-x86-64/load2.map: Likewise.
* testsuite/ld-x86-64/load2.s: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Run load2.

Add support for FMLA (by element) to AArch64 sim.

* simulator.c (do_FMLA_by_element): New function.
(do_vec_op2): Call it.

Automatic date update in version.in

More setting CHECK_RELOCS_AFTER_OPEN_INPUT to yes

* emulparams/elf_iamcu.sh (CHECK_RELOCS_AFTER_OPEN_INPUT): New.
* emulparams/elf_k1om.sh (CHECK_RELOCS_AFTER_OPEN_INPUT):
Likewise.
* emulparams/elf_l1om.sh (CHECK_RELOCS_AFTER_OPEN_INPUT):
Likewise.

Pass --no-relax to ld in ld-x86-64/pr18591.d

Pass --no-relax to ld so that GOTPCREL relocation will be kept.

* testsuite/ld-x86-64/pr18591.d: Pass --no-relax to ld.

Fix ChangeLog entry format

Extract convert_load_reloc from x86 convert_load

* elf32-i386.c (elf_i386_convert_load): Extract the GOT load
conversion to ...
(elf_i386_convert_load_reloc): This.  New function.
* elf64-x86-64.c (elf_x86_64_convert_load): Extract the GOT load
conversion to ...
(elf_x86_64_convert_load_reloc): This.  New function.

Cache the section contents in x86 check_relocs

Cache the section contents in x86 check_relocs for sections without
TLS relocations.

* elf32-i386.c (elf_i386_check_tls_transition): Remove abfd.
Don't check if contents == NULL.
(elf_i386_tls_transition): Add from_relocate_section.  Check
from_relocate_section instead of contents != NULL.  Update
elf_i386_check_tls_transition call.
(elf_i386_check_relocs): Cache the section contents if
keep_memory is FALSE.  Pass FALSE as from_relocate_section to
elf_i386_tls_transition.
(elf_i386_relocate_section): Pass TRUE as from_relocate_section
to elf_i386_tls_transition.
(elf_backend_caches_rawsize): New.
* elf64-x86-64.c (elf_x86_64_check_tls_transition): Don't check
if contents == NULL.
(elf_x86_64_tls_transition): Add from_relocate_section.  Check
from_relocate_section instead of contents != NULL.
(elf_x86_64_check_relocs): Cache the section contents if
keep_memory is FALSE.  Pass FALSE as from_relocate_section to
elf_x86_64_tls_transition.
(elf_x86_64_relocate_section): Pass TRUE as from_relocate_section
to elf_x86_64_tls_transition.
(elf_backend_caches_rawsize): New.

Initialize res in get_next_pcs_read_memory_unsigned_integer

This patch initialize res to zero, otherwise, it may have some garbage
bits after the *the_target->read_memory call.

gdb/gdbserver:

2016-05-05  Yao Qi  <yao.qi@linaro.org>

* linux-arm-low.c (get_next_pcs_read_memory_unsigned_integer):
Initialize res to zero.

Change type of cpsr in arm_sigreturn_next_pc

Variable cpsr holds the value of cpsr register, which is 32-bit.  It
is better to explicitly use uint32_t.

gdb/gdbserver:

2016-05-05  Yao Qi  <yao.qi@linaro.org>

* linux-arm-low.c (arm_sigreturn_next_pc): Change type of cpsr
to uint32_t.

Pass %E to einfo when bfd_section == NULL

Report BFD error When bfd_section == NULL in init_os.

* ldlang.c (init_os): Pass %E to einfo when bfd_section == NULL.

Automatic date update in version.in

[spu] Fix C++ build problems

ChangeLog:

* spu-linux-nat.c (spu_bfd_iovec_pread): Add pointer cast for C++.
(spu_bfd_open): Likewise.

gdbserver/ChangeLog:

* spu-low.c (fetch_ppc_register): Cast PowerPC-Linux-specific value
used as first ptrace argument to PTRACE_TYPE_ARG1 for C++.
(fetch_ppc_memory_1, store_ppc_memory_1): Likewise.

LD/testsuite: Add STB_GNU_UNIQUE cross-linker test

Some targets are only really, or at least regularly, regression-tested
in a crossed configuration.  Currently we only have native compiled test
cases for the STB_GNU_UNIQUE feature in the linker test suite.  This is
nice, covering run-time semantics even, but quite often not run at all.
Consequently a regression may remain unnoticed for long.

Add a simple test case then to provide basic linker coverage with no
need for a compiler or a native toolchain.

ld/
* testsuite/ld-unique/unique.d: New test.
* testsuite/ld-unique/unique.exp: Run the new test.  Adjust
messages for compiled tests.

[ARC] Add SYNTAX_NOP and SYNTAX_1OP for extension instructions

gas/
2016-05-03  Claudiu Zissulescu  <claziss@synopsys.com>

* config/tc-arc.c (syntaxclass): Add SYNTAX_NOP and SYNTAX_1OP.
(arc_extinsn): Handle new introduced syntax.
* testsuite/gas/arc/textinsn1op.d: New file.
* testsuite/gas/arc/textinsn1op.s: Likewise.
* doc/c-arc.texi: Document SYNTAX_NOP and SYNTAX_1OP.

opcodes/
2016-05-03  Claudiu Zissulescu  <claziss@synopsys.com>

* arc-ext.c (dump_ARC_extmap): Handle SYNATX_NOP and SYNTAX_1OP.
(arcExtMap_genOpcode): Likewise.
* arc-opc.c (arg_32bit_rc): Define new variable.
(arg_32bit_u6): Likewise.
(arg_32bit_limm): Likewise.

include/
2016-05-03  Claudiu Zissulescu  <claziss@synopsys.com>

* opcode/arc.h (ARC_SYNTAX_1OP): Declare
(ARC_SYNTAX_NOP): Likewsie.
(ARC_OP1_MUST_BE_IMM): Update defined value.
(ARC_OP1_IMM_IMPLIED): Likewise.
(arg_32bit_rc, arg_32bit_u6, arg_32bit_limm): Declare.

Throw NOT_AVAILABLE_ERROR in read_stack and read_code

Nowadays, read_memory may throw NOT_AVAILABLE_ERROR (it is done by
patch http://sourceware.org/ml/gdb-patches/2013-08/msg00625.html)
however, read_stack and read_code still throws MEMORY_ERROR only.  This
causes PR 19947, that is prologue unwinder is unable unwind because
code memory isn't available, but MEMORY_ERROR is thrown, while unwinder
catches NOT_AVAILABLE_ERROR.

 #0  memory_error (err=err@entry=TARGET_XFER_E_IO, memaddr=memaddr@entry=140737349781158) at /home/yao/SourceCode/gnu/gdb/git/gdb/corefile.c:217
 #1  0x000000000065f5ba in read_code (memaddr=memaddr@entry=140737349781158, myaddr=myaddr@entry=0x7fffffffd7b0 "\340\023<\001", len=len@entry=1)
     at /home/yao/SourceCode/gnu/gdb/git/gdb/corefile.c:288
 #2  0x000000000065f7b5 in read_code_unsigned_integer (memaddr=memaddr@entry=140737349781158, len=len@entry=1, byte_order=byte_order@entry=BFD_ENDIAN_LITTLE)
     at /home/yao/SourceCode/gnu/gdb/git/gdb/corefile.c:363
 #3  0x00000000004717e0 in amd64_analyze_prologue (gdbarch=gdbarch@entry=0x13c13e0, pc=140737349781158, current_pc=140737349781165, cache=cache@entry=0xda0cb0)
     at /home/yao/SourceCode/gnu/gdb/git/gdb/amd64-tdep.c:2267
 #4  0x0000000000471f6d in amd64_frame_cache_1 (cache=0xda0cb0, this_frame=0xda0bf0) at /home/yao/SourceCode/gnu/gdb/git/gdb/amd64-tdep.c:2437
 #5  amd64_frame_cache (this_frame=0xda0bf0, this_cache=<optimised out>) at /home/yao/SourceCode/gnu/gdb/git/gdb/amd64-tdep.c:2508
 #6  0x000000000047214d in amd64_frame_this_id (this_frame=<optimised out>, this_cache=<optimised out>, this_id=0xda0c50)
     at /home/yao/SourceCode/gnu/gdb/git/gdb/amd64-tdep.c:2541
 #7  0x00000000006b94c4 in compute_frame_id (fi=0xda0bf0) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:481
 #8  get_prev_frame_if_no_cycle (this_frame=this_frame@entry=0xda0b20) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1809
 #9  0x00000000006bb6c9 in get_prev_frame_always_1 (this_frame=0xda0b20) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1983
 #10 get_prev_frame_always (this_frame=this_frame@entry=0xda0b20) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1999
 #11 0x00000000006bbe11 in get_prev_frame (this_frame=this_frame@entry=0xda0b20) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:2241
 #12 0x00000000006bc13c in unwind_to_current_frame (ui_out=<optimised out>, args=args@entry=0xda0b20) at /home/yao/SourceCode/gnu/gdb/git/gdb/frame.c:1485

The fix is to let read_stack and read_code throw NOT_AVAILABLE_ERROR too,
in order to align with read_memory.

gdb:

2016-05-04  Yao Qi  <yao.qi@linaro.org>

PR gdb/19947
* corefile.c (read_memory): Rename it to ...
(read_memory_object): ... it.  Add parameter object.
(read_memory): Call read_memory_object.
(read_stack): Likewise.
(read_code): Likewise.

Fix solib-display.exp remote check

This test currently uses [is_remote target] to check if the test is
supported.  This is not quite correct, as the limitation is actually
that it requires support for "running", ruling out stub-like targets.
Therefore, it should check for use_gdb_stub.

This has no visible effect right now, but it will once we make the
native-gdbserver board non-dejagnu-remote.

gdb/testsuite/ChangeLog:

* gdb.base/solib-display.exp: Check for [use_gdb_stub] instead
of [is_remote target],

Introduce procedure use_gdb_stub

This patch introduces the use_gdb_stub procedure, which allows getting
the right value of the use_gdb_stub variable/property in any all
situations.

When calling it before the $use_gdb_stub global variable has been set,
it will return the value of the use_gdb_stub property from the board
file.  This happens when tests want to bail out early (even before gdb
has been started) when the current test setup is a stub.

Otherwise, it returns the value of the $use_gdb_stub global.

It's possible for these two to differ when a test file overrides the
value of the global.

gdb/testsuite/ChangeLog:

* lib/gdb.exp (use_gdb_stub): New procedure.

Fix some AVR test failures.

binutils* testsuite/lib/binutils-common.exp (is_elf_format): Add avr-*-*.

ld * testsuite/ld-elf/pr18735.d: Allow other symbols.
* testsuite/ld-elf/sec64k.exp: Skip 64ksec for avr.
* testsuite/ld-gc/pr14265.d: Allow other symbols.
* testsuite/ld-plugin/plugin.exp: Add PR ld/17973 to
plugin_tests only if check_shared_lib_support is true.
* testsuite/ld-selective/selective.exp: Add --section-start
flag for avr.

Automatic date update in version.in

PR symtab/19914 fix handling of dwp + split debug

PR symtab/19914
* dwarf2read.c (open_and_init_dwp_file): Look at backlink if objfile
is separate debug file.

testsuite/
* gdb.dwarf2/dwp-sepdebug.c: New file.
* gdb.dwarf2/dwp-sepdebug.exp: New file.

Fix typos in gdb_pipe function comment

gdb/ChangeLog:
2016-05-03  Don Breazeal <donb@codesourcery.com>

* serial.h (gdb_pipe): Fix argument names in comment.

PR 10549: MIPS/LD: Handle OSABI setting for STB_GNU_UNIQUE

Add missing OSABI handling for MIPS/LD, complementing:

commit f64b2e8d60f277b705c96d4b4e585f5f7c847cc4
Author: Nick Clifton <nickc@redhat.com>
Date:   Mon Apr 11 08:13:22 2011 +0000

Adjust test coverage accordingly.

bfd/
PR 10549
* elfxx-mips.c (_bfd_mips_elf_add_symbol_hook): Handle
STB_GNU_UNIQUE.

ld/
PR 10549
* testsuite/ld-unique/unique.exp: Also run for `mips*-*-*'.

Updated Swedish translation for the binutils.

Fix gdb/python/python.c use-after-free

Valgrind shows:

 ==26964== Invalid read of size 1
 ==26964==    at 0x6E14100: __GI_strcmp (strcmp.S:180)
 ==26964==    by 0x6DB55AA: setlocale (setlocale.c:238)
 ==26964==    by 0x4E0455: _initialize_python() (python.c:1731)
 ==26964==    by 0x786731: initialize_all_files() (init.c:319)
 ==26964==    by 0x72EF0A: gdb_init(char*) (top.c:1929)
 ==26964==    by 0x60BCAC: captured_main(void*) (main.c:863)
 ==26964==    by 0x606AD5: catch_errors(int (*)(void*), void*, char*, return_mask) (exceptions.c:234)
 ==26964==    by 0x60C608: gdb_main(captured_main_args*) (main.c:1165)
 ==26964==    by 0x40CAEC: main (gdb.c:32)
 ==26964==  Address 0x81d30a0 is 0 bytes inside a block of size 181 free'd
 ==26964==    at 0x4C29CF0: free (vg_replace_malloc.c:530)
 ==26964==    by 0x6DB5B65: setname (setlocale.c:201)
 ==26964==    by 0x6DB5B65: setlocale (setlocale.c:388)
 ==26964==    by 0x4E037F: _initialize_python() (python.c:1712)
 ==26964==    by 0x786731: initialize_all_files() (init.c:319)
 ==26964==    by 0x72EF0A: gdb_init(char*) (top.c:1929)
 ==26964==    by 0x60BCAC: captured_main(void*) (main.c:863)
 ==26964==    by 0x606AD5: catch_errors(int (*)(void*), void*, char*, return_mask) (exceptions.c:234)
 ==26964==    by 0x60C608: gdb_main(captured_main_args*) (main.c:1165)
 ==26964==    by 0x40CAEC: main (gdb.c:32)

The problem is doing this:

  oldloc = setlocale (LC_ALL, NULL);
  setlocale (LC_ALL, "");
  ...
  setlocale (LC_ALL, oldloc);

I.e., the second setlocale call frees 'oldloc'.

From http://pubs.opengroup.org/onlinepubs/9699919799/functions/setlocale.html :

 "The returned string pointer might be invalidated or the string
 content might be overwritten by a subsequent call to setlocale()."

gdb/ChangeLog:
2016-05-03  Pedro Alves <palves@redhat.com>

PR python/20037
* python/python.c (_initialize_python) [IS_PY3K]: xstrdup/xfree
oldloc.

Remove gdb/python/python.c code that handles strlen failing with -1

This makes no sense -- strlen doesn't really ever fail with -1.

gdb/ChangeLog:
2016-05-03  Pedro Alves <palves@redhat.com>

* python/python.c (_initialize_python) [IS_PY3K]: Remove dead
code.

[AArch64] Also puts value in place for R_AARCH64_RELATIVE

When handling absolute relocations for global symbols bind within the
shared object, AArch64 will generate one dynamic RELATIVE relocation,
but won't apply the value for this absolution relocations at static
linking stage. This is different from AArch64 gold linker and x86-64.

This is not a bug as AArch64 is RELA, there is only guarantee that
relocation addend is placed in the relocation entry.  But some
system softwares originally writen for x86-64 might assume AArch64
bfd linker gets the same behavior as x86-64, then they could take
advantage of this buy skipping those RELATIVE dynamic relocations
if the load address is the same as the static linking address.

This patch makes AArch64 BFD linker applies absolution relocations at
static linking stage for scenario described above.  Meanwhile old AArch64
android loader has a bug (PR19163) which relies on current linker behavior
as a workaround, so the same option --no-apply-dynamic-relocs added.

Fix generation of AArhc64 instruction table.

* aarch64-gen.c (VERIFIER): Define.
* aarch64-opc.c (VERIFIER): Define.
(verify_ldpsw): Use static linkage.
* aarch64-opc.h (verify_ldpsw): Remove.
* aarch64-tbl.h: Use VERIFIER for verifiers.

Assume that the GAS testsuite for the AVR is being run without -mlink-relax specified.

* testsuite/gas/lns/lns.exp: Add avr to list of targets using
DW_LNS_fixed_advance_pc.

Skip ELF STT_COMMON linker tests if shared libraries are not supported.

* testsuite/ld-elfcomm/elfcomm.exp: Check for shared lib support
before running STT_COMMON tests.

Fix PR gdb/16818, workaround Python's forcing of -export-dynamic

GDB's use of --dynamic-list to only export the proc-service symbols is
broken due to Python's "python-config --ldflags" saying we should link
with -export-dynamic, causing us to export _all_ extern symbols
anyway.  On Fedora 23:

 $ python-config --ldflags
 -lpython2.7 -lpthread -ldl -lutil -lm -Xlinker -export-dynamic
 $ python3.4-config --ldflags
  -L/usr/lib64 -lpython3.4m -lpthread -ldl  -lutil -lm  -Xlinker -export-dynamic

Having GDB export all its symbols leads to issues such as PR gdb/16818
(GDB crashes when using name for target remote hostname:port), where a
GDB symbol unintentionally preempts a symbol in one of the NSS modules
glibc loads into the process.  NSS modules should not define symbols
outside the implementation namespace or the relevant standards, but,
alas, that's a longstanding and hard to fix issue.  See libc-alpha
discussion at:

  [symbol name space issues with NSS modules]
  https://sourceware.org/ml/libc-alpha/2016-04/msg00130.html

Python should instead be either using GCC's symbol visibility feature
or -Wl,--dynamic-list as well, to only export Python API symbols, but,
it doesn't.  There are bugs open upstream for that:

  [Use -Wl,--dynamic-list=x.list, not -Xlinker -export-dynamic]
  http://bugs.python.org/issue10112

  [Use GCC visibility attrs in PyAPI_*]
  http://bugs.python.org/issue11410

But that's taking a long while to resolve.

I thought of working around this Python issue by making GDB build with
-fvisibility=hidden, as Jan suggests in Python issue 10112, as then
Python's "-Xlinker -export-dynamic" has no effect.  However, that
would need to be done in the whole source tree (bfd, libiberty, etc.),
and I think that would break GCC plugins, as I believe those have
access to all of GCCs symbols, by "design".  So we'd need a new
configure switch, or have the libraries in the tree detect which of
GCC or GDB is being built, but that doesn't work, because the answer
can be "both" with combined builds...

So this patch instead works around Python's bug, by simply sed'ing
away "-Xlinker -export-dynamic" from the result of python-config.py
--ldflags, making -Wl,--dynamic-list work again as it used to.  It's
ugly, but so is the bug...

Note that if -Wl,--dynamic-list doesn't work, we always link with
-rdynamic, so static Python should still work.

Tested on F23 with --python=python (Python 2.7) and
--python=python3.4.

gdb/ChangeLog:y
2016-05-03  Pedro Alves  <palves@redhat.com>

* configure.ac (PYTHON_LIBS): Sed away "-Xlinker -export-dynamic".
* configure: Regenerate.

Fix "-Wl,--dynamic-list" gdb/configure test

The -Wl,--dynamic-list test is currently broken on Fedora 23, when you
configure with --with-python=python3.4.  We see:

 configure:13741: checking for the dynamic export flag
 configure:13796: gcc -o conftest -g3 -O0  -fno-strict-aliasing -DNDEBUG -fwrapv    -Wl,--dynamic-list=/home/pedro/gdb/mygit/src/gdb/proc-service.list conftest.c -ldl -lncurses -lm -ldl  -lpthread -ldl -lutil -lm -lpython3.4m -Xlinker -export-dynamic >&5
 conftest.c:182:30: fatal error: python3.4/Python.h: No such file or directory
 compilation terminated.
 configure:13796: $? = 1

The correct -I path is in PYTHON_CPPFLAGS:

 PYTHON_CPPFLAGS='-I/usr/include/python3.4m -I/usr/include/python3.4m'

(Other Python-related tests in the file are already doing this.)

gdb/ChangeLog:
2016-05-03  Pedro Alves  <palves@redhat.com>

* configure.ac (checking for the dynamic export flag): Add
$PYTHON_CPPFLAGS to CPPFLAGS.
* configure: Regenerate.

[gdb] Fix -Wparentheses warnings

2016-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>

* symfile.c (find_pc_overlay): Add braces to avoid -Wparentheses
warning.
(find_pc_mapped_section): Likewise.
(list_overlays_command): Likewise.

Automatic date update in version.in

Fix detach.exp remote check

This test seems to work with both native-gdbserver and
native-extended-gdbserver, so I removed the remote check.

When running with native-gdbserver (a stub-like target), detach makes
gdbserver stop and gdb disconnect.  runto_main just spawns a brand new
gdbserver.  So it tests the exact same thing twice.  It doesn't hurt
though.

With native-extended-gdbserver, the test is probably a bit more useful
(and similar to native).  It tests running/detaching twice using the
same gdb/gdbserver instances, since with extended-remote, you can
detach/attach/run all you want, unlike with remote.

gdb/testsuite/ChangeLog:

* gdb.base/detach.exp: Remove is_remote check.

Fix annota-input-while-running.exp remote check

The comment says that we can't use runto_main here becore it doesn't
know how to handle annotation.  Instead, the test puts a breakpoint at
main and calls run by hand.  Therefore, it can't work with stub targets,
since they can't "run".  The check should be then changed to check the
use_gdb_stub variable instead of [is_remote target].

But as an alternative, we can just use runto_main and enable annotations
after, since the "run to main" part is not really part of what we want
to test.

I also removed the "set test..." line that is unused.

gdb/testsuite/ChangeLog:

* gdb.base/annota-input-while-running.exp: Don't check for
[is_remote target].  Enable annotations after running to main.
Remove unused "set test..." line.

Fix startup on MS-Windows when 'gdb.ini' is found in $HOME

* windows-nat.c (_initialize_check_for_gdb_ini): Fix off-by-one
error in allocation of space for "$HOME/.gdbinit" string.  This
caused GDB to abort on startup whenever a '~/gdb.ini' file was
actually found, because xsnprintf would hit an assertion
violation.

Automatic date update in version.in

Automatic date update in version.in

Automatic date update in version.in

Set interpreter in x86 create_dynamic_sections

Set interpreter in x86 create_dynamic_sections to make this information
available to x86 check_relocs.

* elf32-i386.c (elf_i386_size_dynamic_sections): Move interp
setting to ...
(elf_i386_create_dynamic_sections): Here.
* elf64-x86-64.c (elf_x86_64_size_dynamic_sections): Move
interp setting to ...
(elf_x86_64_create_dynamic_sections): Here.

Pass GOT_RELOC to UNDEFINED_WEAK_RESOLVED_TO_ZERO

When UNDEFINED_WEAK_RESOLVED_TO_ZERO is checked to convert load via
GOT, has_got_reloc is always TRUE.  This patch adds GOT_RELOC, which
is TRUE in x86 convert_load, to UNDEFINED_WEAK_RESOLVED_TO_ZERO.

* elf32-i386.c (UNDEFINED_WEAK_RESOLVED_TO_ZERO): Take GOT_RELOC
and replace (EH)->has_got_reloc with GOT_RELOC.
(elf_i386_fixup_symbol): Pass has_got_reloc to
UNDEFINED_WEAK_RESOLVED_TO_ZERO.
(elf_i386_allocate_dynrelocs): Likewise.
(elf_i386_relocate_section): Likewise.
(elf_i386_finish_dynamic_symbol): Likewise.
(elf_i386_convert_load): Pass TRUE to
UNDEFINED_WEAK_RESOLVED_TO_ZERO.
* elf64-x86-64.c (UNDEFINED_WEAK_RESOLVED_TO_ZERO): Take
GOT_RELOC and replace (EH)->has_got_reloc with GOT_RELOC.
(elf_x86_64_fixup_symbol): Pass has_got_reloc to
UNDEFINED_WEAK_RESOLVED_TO_ZERO.
(elf_x86_64_allocate_dynrelocs): Likewise.
(elf_x86_64_relocate_section): Likewise.
(elf_x86_64_finish_dynamic_symbol): Likewise.
(elf_x86_64_convert_load): Pass TRUE to
UNDEFINED_WEAK_RESOLVED_TO_ZERO.

Download source only on remote host

There is no need to download source if we aren't on remote host.
Otherwise, each ld test run on local host leaves behind a couple
test files.

* config/default.exp (NOPIE_CFLAGS): Download source only on
remote host.
(NOPIE_LDFLAGS): Likewise.
* testsuite/lib/ld-lib.exp (check_lto_available): Likewise.
(check_lto_fat_available): Likewise.
(check_lto_shared_available): Likewise.
(check_ifunc_available): Likewise.
(check_ifunc_attribute_available): Likewise.

Enhance readelf's recognition of AVR relocation types.

* readelf.c (is_32bit_pcrel_reloc): Return true if reloc is 32-bit
PC relocation for AVR target.
(is_none_reloc): Return true if reloc is any of AVR diff
relocations.

i386: Don't relocate section when check_relocs failed

No need to relocate section when check_relocs failed.

* elf32-i386.c (check_relocs_failed): New.
(elf_i386_check_relocs): Set check_relocs_failed on error.
(elf_i386_relocate_section): Skip if check_relocs failed.

X86-64: Set check_relocs_failed on error

When checking relocations, set check_relocs_failed on error.

* elf64-x86-64.c (elf_x86_64_check_relocs): Set
check_relocs_failed on error.

Enhance support for copying and stripping Solaris and ARM binaries.

PR 19938
bfd * elf-bfd.h (struct elf_backend_data): Rename
elf_backend_set_special_section_info_and_link to
elf_backend_copy_special_section_fields.
* elfxx-target.h: Likewise.
* elf.c (section_match): Ignore the SHF_INFO_LINK flag when
comparing section flags.
(copy_special_section_fields): New function.
(_bfd_elf_copy_private_bfd_data): Copy the EI_ABIVERSION field.
Perform two scans over special sections.  The first one looks for
a direct mapping between the output section and an input section.
The second scan looks for a possible match based upon section
characteristics.
* elf32-arm.c (elf32_arm_copy_special_section_fields): New
function.  Handle setting the sh_link field of SHT_ARM_EXIDX
sections.
* elf32-i386.c (elf32_i386_set_special_info_link): Rename to
elf32_i386_copy_solaris_special_section_fields.
* elf32-sparc.c (elf32_sparc_set_special_section_info_link):
Rename to elf32_sparc_copy_solaris_special_section_fields.
* elf64-x86-64.c (elf64_x86_64_set_special_info_link): Rename to
elf64_x86_64_copy_solaris_special_section_fields.

binutils* readelf.c (get_solaris_segment_type): New function.
(get_segment_type): Call it.

Automatic date update in version.in

Don't show deprecated commands in help

Just like completion doesn't show deprecated commands, I think that help
should not list them, so that we don't incite users to use them.

gdb/ChangeLog:

* cli/cli-decode.c (help_cmd_list): Do not list commands that
are deprecated.

Add test for tracepoint enable/disable

This patch adds a test for tracepoints enabling/disabling, which
didn't work properly for fast tracepoints on big endian systems.

gdb/testsuite/ChangeLog:

* gdb.trace/trace-enable-disable.exp: New file.
* gdb.trace/trace-enable-disable.c: New file.

Fix write endianness/size problem for fast tracepoint enabled flag

I am sending this fix on behalf of Par Olsson, as a follow-up of this
one:

https://www.sourceware.org/ml/gdb-patches/2015-10/msg00196.html

This problem is exposed when enabling/disabling fast tracepoints on big
endian machines.  The flag is defined as an int8_t, but is written from
gdbserver as an integer (usually 32 bits).  When the agent code reads it
as an int8_t, it only considers the most significant byte, which is
always 0.

Also, we were writing 32 bits in an 8 bits field, so the write would
overflow, but since the following bytes are padding (the next field is
an uint64_t), it luckily didn't cause any issue on little endian
systems.

The fix was originally tested on ARM big endian systems, but I don't
have access to such a system.  However, thanks to Marcin's PowerPC fast
tracepoint patches and gcc110 (big endian Power7) on the gcc compile
farm, I was able to reproduce the problem, test the fix and write a
test (the following patch).

gdb/gdbserver/ChangeLog:

YYYY-MM-DD  Par Olsson  <par.olsson@windriver.com>

* tracepoint.c (write_inferior_int8): New function.
(cmd_qtenable_disable): Write enable flag using
write_inferior_int8.

ftrace tests: Use gdb_load_shlib result to lookup IPA in info sharedlibrary

Some fast tracepoints tests make sure that the in-process agent library
is properly loaded, by searching for the library name in "info
sharedlibrary".

Originally, it would search for the full path.  Since patch "Make ftrace
tests work with remote targets" [1], the "runtime" location of the IPA,
in the standard output directory, is not the same as the original
location, in the gdbserver build directory.  Therefore, the patch
changed the checks:

  gdb_test "info sharedlibrary" ".*${libipa}.*" "IPA loaded"

to

  gdb_test "info sharedlibrary" ".*[file tail ${libipa}].*" "IPA loaded"

so that only the "libinproctrace.so" part would be searched for.
Antoine (in CC) pointed out that I missed some, so I have to update
them.  In the mean time, I noticed that I missed a few test failures:
adding the SONAME to the IPA makes it possible for the test executable
to erroneously pick up libinproctrace.so from /usr/lib if the test
harness failed to put the libinproctrace.so we want to test in the right
place.  To mitigate that kind of error in the future, we can use the
return value of gdb_load_shlib (the path of the "runtime" version of the
library) and use that to search in the output of info sharedlibrary.

When testing locally, gdb_load_shlib returns the full normalized path of
the destination library, which the test executable should use e.g.:

  /path/to/gdb/testsuite/outputs/gdb.trace/thetest/libinproctrace.so

My testing showed that it was the same path that gdb displayed in info
sharedlibrary.  If the test executable picks up another
libinproctrace.so, the test will fail.

When testing remotely, gdb_load_shlib/gdb_remote_download only returns
us "libinproctrace.so", so the situation doesn't really change.  If
there is a rogue libinproctrace.so in /usr/lib on the target and we fail
to download ours, it might cover up a test failure.  But that situation
is probably still better than the original one, where it wasn't possible
to test remotely using the IPA at all.

[1] https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;a=commit;h=6e774b13c3b81ac2599812adf058796948ce7e95

gdb/testsuite/ChangeLog:

* gdb.arch/ftrace-insn-reloc.exp: Save gdb_load_shlib result,
use it in info sharedlibrary test.
* gdb.trace/ftrace-lock.exp: Likewise.
* gdb.trace/ftrace.exp: Likewise.
* gdb.trace/range-stepping.exp: Likewise.
* gdb.trace/trace-break.exp: Likewise.
* gdb.trace/trace-condition.exp: Likewise.
* gdb.trace/trace-mt.exp: Likewise.

Updated Chinese (simplified) translations for bfd, binutils and gold.

Remove need_step_over from struct lwp_info

Hi,
I happen to see that field need_step_over in struct lwp_info is only
used to print a debug info.  need_step_over is set in linux_wait_1
when breakpoint_here is true, however, we check breakpoint_here too in
need_step_over_p and do the step over.  I think we don't need field
need_step_over, and check breakpoint_here directly in need_step_over_p.

This field was added in this patch
https://sourceware.org/ml/gdb-patches/2010-03/msg00605.html and the code
wasn't changed much since then.

This patch is to remove it.

gdb/gdbserver:

2016-04-28  Yao Qi  <yao.qi@linaro.org>

* linux-low.h (struct lwp_info) <need_step_over>: Remove.
* linux-low.c (linux_wait_1): Update.
(need_step_over_p): Likewise.

Add support to AArch64 disassembler for verifying instructions.  Add verifier for LDPSW.

PR target/19722
opcodes * aarch64-dis.c (aarch64_opcode_decode): Run verifier if present.
* aarch64-opc.c (verify_ldpsw): New function.
* aarch64-opc.h (verify_ldpsw): New prototype.
* aarch64-tbl.h: Add initialiser for verifier field.
(LDPSW): Set verifier to verify_ldpsw.

binutils* testsuite/binutils-all/aarch64/illegal.s: New test.
* testsuite/binutils-all/aarch64/illegal.d: New test driver.

include * opcode/aarch64.h (struct aarch64_opcode): Add verifier field.

Automatic date update in version.in

Limit ld-elf/compressed1b.d to Linux/GNU targets

Since not all ELF targets use the elf.em emulation to support ld option:
--compress-debug-sections=zlib-gnu, limit compressed1b.d to Linux/GNU
targets.

* testsuite/ld-elf/compressed1b.d: Only run for Linux/GNU targets.

Rename gdb_load_shlibs to gdb_load_shlib

Rename gdb_load_shlibs to gdb_load_shlib to reflect that it can only
load a single shlib at the time.

gdb/testsuite/ChangeLog:

* lib/gdb.exp (gdb_load_shlibs): Rename to...
(gdb_load_shlib): ... this.
* gdb.arch/ftrace-insn-reloc.exp: Adjust gdb_load_shlibs ->
gdb_load_shlib.
* gdb.base/catch-load.exp (one_catch_load_test): Likewise.
* gdb.base/ctxobj.exp: Likewise.
* gdb.base/dprintf-pending.exp: Likewise.
* gdb.base/dso2dso.exp: Likewise.
* gdb.base/fixsection.exp: Likewise.
* gdb.base/gcore-relro.exp: Likewise.
* gdb.base/gdb1555.exp: Likewise.
* gdb.base/global-var-nested-by-dso.exp: Likewise.
* gdb.base/gnu-ifunc.exp: Likewise.
* gdb.base/hbreak-in-shr-unsupported.exp: Likewise.
* gdb.base/jit-so.exp (one_jit_test): Likewise.
* gdb.base/pending.exp: Likewise.
* gdb.base/print-file-var.exp: Likewise.
* gdb.base/print-symbol-loading.exp: Likewise.
* gdb.base/shlib-call.exp: Likewise.
* gdb.base/shreloc.exp: Likewise.
* gdb.base/so-impl-ld.exp: Likewise.
* gdb.base/solib-disc.exp: Likewise.
* gdb.base/solib-nodir.exp: Likewise.
* gdb.base/solib-overlap.exp: Likewise.
* gdb.base/solib-symbol.exp: Likewise.
* gdb.base/solib-weak.exp (do_test): Likewise.
* gdb.base/sym-file.exp: Likewise.
* gdb.base/symtab-search-order.exp: Likewise.
* gdb.base/type-opaque.exp: Likewise.
* gdb.base/unload.exp: Likewise.
* gdb.base/watchpoint-solib.exp: Likewise.
* gdb.compile/compile.exp: Likewise.
* gdb.cp/gdb2384.exp: Likewise.
* gdb.cp/infcall-dlopen.exp: Likewise.
* gdb.cp/re-set-overloaded.exp: Likewise.
* gdb.fortran/library-module.exp: Likewise.
* gdb.opt/solib-intra-step.exp: Likewise.
* gdb.python/py-finish-breakpoint.exp: Likewise.
* gdb.python/py-shared.exp: Likewise.
* gdb.reverse/solib-precsave.exp: Likewise.
* gdb.reverse/solib-reverse.exp: Likewise.
* gdb.server/solib-list.exp: Likewise.
* gdb.threads/dlopen-libpthread.exp: Likewise.
* gdb.threads/tls-shared.exp: Likewise.
* gdb.threads/tls-so_extern.exp: Likewise.
* gdb.trace/change-loc.exp: Likewise.
* gdb.trace/ftrace-lock.exp: Likewise.
* gdb.trace/ftrace.exp: Likewise.
* gdb.trace/mi-tracepoint-changed.exp (test_reconnect): Likewise.
* gdb.trace/pending.exp: Likewise.
* gdb.trace/range-stepping.exp: Likewise.
* gdb.trace/strace.exp (strace_remove_socket): Likewise.
(strace_info_marker): Likewise.
(strace_probe_marker): Likewise.
(strace_trace_on_same_addr): Likewise.
(strace_trace_on_diff_addr): Likewise.
* gdb.trace/trace-break.exp: Likewise.
* gdb.trace/trace-condition.exp: Likewise.
* gdb.trace/trace-mt.exp: Likewise.

Make gdb_load_shlibs return the destination path of the library

This patch makes gdb_load_shlibs return the destination path of the
copied library.  To make the procedure implementation and interface more
straightforward, it also changes it so that it accepts a single shared
library path at the time.  Therefore, calls that are passed multiple
libraries:

  gdb_load_shlibs $lib1 $lib2

must be changed to separate calls:

  gdb_load_shlibs $lib1
  gdb_load_shlibs $lib2

A subtle impact is the solib-search-path handling.  In the former
version, solib-search-path is set using the directory of the first
passed lib (further calls overwrite the value).  In the later version,
the directory of the library passed to the last call to gdb_load_shlibs
remnains.  I don't think that's a problem in practice, since if we had
tests that needed multiple different paths in solib-search-path, they
wouldn't work in the first place.

Changed in v2:

* Split behavioural and rename changes in two separate patches.

gdb/testsuite/ChangeLog:

* lib/gdb.exp (gdb_load_shlibs):  Accept a single argument.  Return
result of gdb_remote_download.
* gdb.base/ctxobj.exp: Split gdb_load_shlibs call.
* gdb.base/dso2dso.exp: Likewise.
* gdb.base/global-var-nested-by-dso.exp: Likewise.
* gdb.base/print-file-var.exp: Likewise.
* gdb.base/shlib-call.exp: Likewise.
* gdb.base/shreloc.exp: Likewise.
* gdb.base/solib-overlap.exp: Likewise.
* gdb.base/solib-weak.exp (do_test): Likewise.
* gdb.base/unload.exp: Likewise.

Workaround gdbserver<7.7 for setfs

With current FSF GDB HEAD and old FSF gdbserver I expected I could do:
gdb -ex 'file target:/root/redhat/threadit' -ex 'target remote :1234'
(supplying that unsupported qXfer:exec-file:read by "file")
But that does not work because:
Sending packet: $vFile:setfs:0#bf...Packet received: OK
Packet vFile:setfs (hostio-setfs) is supported
...
Sending packet: $vFile:setfs:104#24...Packet received: OK
"target:/root/redhat/threadit": could not open as an executable file: Invalid argument

GDB documentation says:
The valid responses to Host I/O packets are:
An empty response indicates that this operation is not recognized.

This "empty response" vs. "OK" was a bug in gdbserver < 7.7.  It was fixed by:
commit e7f0d979dd5cc4f8b658df892e93db69d6d660b7
Author: Yao Qi <yao@codesourcery.com>
Date:   Tue Dec 10 21:59:20 2013 +0800
    Fix a bug in matching notifications.
Message-ID: <1386684626-11415-1-git-send-email-yao@codesourcery.com>
https://sourceware.org/ml/gdb-patches/2013-12/msg00373.html
2013-12-10  Yao Qi  <yao@codesourcery.com>
* notif.c (handle_notif_ack): Return 0 if no notification
matches.

with unpatched old FSF gdbserver and patched FSF GDB HEAD:
gdb -ex 'file target:/root/redhat/threadit' -ex 'target remote :1234'
Sending packet: $vFile:setfs:0#bf...Packet received: OK
Packet vFile:setfs (hostio-setfs) is NOT supported
...
(gdb) info sharedlibrary
From                To                  Syms Read   Shared Object Library
0x00007ffff7ddbae0  0x00007ffff7df627a  Yes (*)     target:/lib64/ld-linux-x86-64.so.2
0x00007ffff7bc48a0  0x00007ffff7bcf514  Yes (*)     target:/lib64/libpthread.so.0

gdb/ChangeLog
2016-04-26  Jan Kratochvil  <jan.kratochvil@redhat.com>

* remote.c (remote_start_remote): Detect PACKET_vFile_setfs.support.

Skip debug sections when estimating distances

Skip debug sections when estimating distances between output sections
since compressed_size is used to compress debug sections and debug
sections aren't excluded from distances between output sections.

bfd/

PR ld/20006
* elf64-x86-64.c (elf_x86_64_convert_load): Skip debug sections
when estimating distances between output sections.

ld/

PR ld/20006
* testsuite/ld-elfvsb/elfvsb.exp (COMPRESS_LDFLAG): New.
(visibility_run): Pass COMPRESS_LDFLAG to visibility_test on
ELF targets.

Skip gdb.base/branch-to-self.exp if gdb,nosignals exists

I get a timeout fail in branch-to-self.exp when it is compiled by a
bare-mental target running qemu, which doesn't have signal.

The test should be skipped if gdb,nosignals exists, and that is
what this patch does.

gdb/testsuite:

2016-04-27  Yao Qi  <yao.qi@linaro.org>

* gdb.base/branch-to-self.exp: Skip it if gdb,nosignals
exists.

c_value_print: Revert 'val' to a reference for TYPE_CODE_STRUCT

Currently c_value_print will turn struct reference values into pointers before
doing a set of RTTI checks.  This was introduced as a fix to PR c++/15401.
If there's RTTI the pointer will be adjusted and converted back to a reference.
However, if there's no RTTI the value will still be treated as a pointer during
the remainder of the function.
This patch moves the conversion down so that it's always performed when needed.

Notice this currently has not user-visible effects, so can be seen as a small
code cleanup.  However, it'll be necessary for the bug-fix for handling
synthetic C++ references.  It causes no testsuite regressions.

gdb/ChangeLog:
2016-04-26  Martin Galvan  <martin.galvan@tallertechnologies.com>

* c-valprint.c (c_value_print): Always convert val back to reference
type if we converted it to a pointer type.

Tweak doc on command tfind

Hi,
command "tfind" will find the first trace snapshot if no trace
snapshot is selected, but this behavior isn't documented.

This patch completes the doc of command "tfind" without argument.

gdb/doc:

2016-04-27  Yao Qi  <yao.qi@linaro.org>

* gdb.texinfo (tfind): Complete doc about tfind without
argument.

Avoid non-C++-enabled babeltrace versions

In some babeltrace versions before 1.2.0, the header file iterator.h
declares the enum values `BT_SEEK_*' within the struct declaration of
bt_iter_pos.  The enum values are supposed to be globally-scoped, which
works for C, but not for C++.  Later babeltrace versions declare the
enum outside the struct:

  https://lists.lttng.org/pipermail/lttng-dev/2013-September/021411.html

Now that GDB is compiled with C++, the GDB build fails on a system with
an affected babeltrace version: the compiler complains about a missing
declaration of BT_SEEK_BEGIN in ctf.c.

This patch enhances the configure check to recognize such babeltrace
versions as unusable for GDB.

gdb/ChangeLog:

* configure.ac: Enhance configure check for babeltrace to reject
non-C++-enabled versions.
* configure: Regenerate.

Pass --compress-debug-sections=none to ld

Since ld may generate compressed debug sections by default, pass
--compress-debug-sections=none to ld to avoid compressed debug
sections.

* ld-elf/compressed1b.d: Pass --compress-debug-sections=none
to ld.
* ld-elf/compressed1c.d: Likewise.

Fix a typo in the check for SNANs in the RX simulator.

PR target/20000
* fpu.c (check_exceptions): Fix typo checking for signalling
NANs.

Add support for the --trace-decode option to the AArch64 simulator.

* simulator.c: Add TRACE_DECODE statements to all emulation
functions.

Fix powerpc64 ld testsuite fail

PR target/19985
* configure.tgt: Don't use var+=.

Provide xmemdup0

and some tidies in as.h

* as.h (inline, __PTR_TO_INT, __INT_TO_PTR): Don't define.
(xmemdup0): New inline function.

Cache result of scan for __start_* and __stop_* sections

include/
* bfdlink.h (struct bfd_link_hash_entry): Add "section" field to
undef.  Formatting.
bfd/
* elflink.c (_bfd_elf_is_start_stop): New function.
(_bfd_elf_gc_mark_rsec): Use it.
* elf-bfd.h (_bfd_elf_is_start_stop): Declare.

add casts to avoid arithmetic on void *

arithmetic on void * is undefined in ISO C, so we should avoid it.  In
GNU C sizeof void * is defined as 1, and that is pretty clearly what
this code wants, so change it to do arithmetic on bfd_byte *.
Unfortunately most of the argument types come from virtual function
interfaces so changing the types to bfd_byte * isn't trivial though it
might make the code clearer.  So for the moment its easiest to leave the
variable types as void * and cast before doing arithmetic.

bfd/ChangeLog:

2016-04-26  Trevor Saunders  <tbsaunde+binutils@tbsaunde.org>

* elf32-rx.c (rx_set_section_contents): Avoid arithmetic on void *.
* mmo.c (mmo_get_section_contents): Likewise.
(mmo_set_section_contents): Likewise.

Automatic date update in version.in

fort_dyn_array: Use value constructor instead of raw-buffer manipulation.

Instead of pre-computing indices into a fortran array re-use
the value_* interfaces to subscript a fortran array.
The benefit of using the new interface is that it takes care of
dynamic types and resolve them when needed.
This fixes issues when printing structures with dynamic arrays from toplevel.

Before:
(gdb) p twov
$1 = ( (( ( 6352320, 0, -66, -1, 267) ( 343476, 1, -15, 1, 0) ( 5, 0, 5, 0, 1) ...

After:
(gdb) p twov
$1 = ( (( ( 1, 1, 1, 1, 1) ( 1, 1, 321, 1, 1) ( 1, 1, 1, 1, 1) ...

2016-04-26  Sanimir Agovic  <sanimir.agovic@intel.com>
            Keven Boell  <keven.boell@intel.com>
            Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/Changelog:
* f-valprint.c (f77_create_arrayprint_offset_tbl): Remove
function.
(F77_DIM_SIZE, F77_DIM_OFFSET): Remove macro.
(f77_print_array_1): Use value_subscript to subscript a
value array.
(f77_print_array): Remove call to f77_create_arrayprint_offset_tbl.
(f_val_print): Use value_field to construct a field value.

gdb/testsuite/Changelog:
* vla-type.exp: Print structure from toplevel.

fort_dyn_array: Support evaluation of dynamic elements inside arrays.

Resolve type of an array's element to be printed in case it is dynamic.
Otherwise we don't use the correct boundaries nor the right location.

Before:
  ptype fivearr(1)
  type = Type five
      Type one
          integer(kind=4) :: ivla(34196784:34196832,34197072:34197120,34197360:34197408)
      End Type one :: tone
  End Type five

After:
  ptype fivearr(1)
  type = Type five
      Type one
          integer(kind=4) :: ivla(2,4,6)
      End Type one :: tone
  End Type five

2016-04-26  Bernhard Heckel  <bernhard.heckel@intel.com>

gdb/Changelog:
* valarith.c (value_address): Resolve dynamic types.

gdb/testsuite/Changelog:
* gdb.fortran/vla-type.f90: Add test for static and dynamic arrays
of dynamic types.
* gdb.fortran/vla-type.exp: Add test for static and dynamic arrays
of dynamic types.

fort_dyn_array: Enable dynamic member types inside a structure.

Fortran supports dynamic types for which bounds, size and location
can vary during their lifetime. As a result of the dynamic
behaviour, they have to be resolved at every query.
This patch will resolve the type of a structure field when it
is dynamic.

2016-04-26  Bernhard Heckel  <bernhard.heckel@intel.com>
2016-04-26  Keven Boell  <keven.boell@intel.com>

Before:
(gdb) print threev%ivla(1)
Cannot access memory at address 0x3
(gdb) print threev%ivla(5)
no such vector element

After:
(gdb) print threev%ivla(1)
$9 = 1
(gdb) print threev%ivla(5)
$10 = 42

gdb/Changelog:

* NEWS: Add new supported features for fortran.
* gdbtypes.c (remove_dyn_prop): New.
(resolve_dynamic_struct): Keep type length for fortran structs.
* gdbtypes.h: Forward declaration of new function.
* value.c (value_address): Return dynamic resolved location of a value.
(set_value_component_location): Adjust the value address
for single value prints.
(value_primitive_field): Support value types with a dynamic location.
(set_internalvar): Remove dynamic location property of
internal variables.

gdb/testsuite/Changelog:

* gdb.fortran/vla-type.f90: New file.
* gdb.fortran/vla-type.exp: New file.

Always count the NULL entry in dynamic symbol table

There is an unused NULL entry at the head of dynamic symbol table which
we must account for in our count even if the table is empty or unused
since it is intended for the mandatory DT_SYMTAB tag (.dynsym section)
in .dynamic section.

* elf-bfd.h (elf_link_hash_table): Update comments for
dynsymcount.
* elflink.c (_bfd_elf_link_renumber_dynsyms): Always count for
the unused NULL entry at the head of dynamic symbol table.
(bfd_elf_size_dynsym_hash_dynstr): Remove dynsymcount != 0
checks.

Automatic date update in version.in

Allow powerpc64le-linux-gnu toolchain to support big endian targets as well.

PR target/19985
* configure.tgt: Include big endian PPC64 emulations with little
endian PPC64 targets.

AVR: Allow the start address of the .text section to be set by --section-start.

* scripttempl/avrtiny.sc (.text): Do not set LMA to zero.