binutils-gdb.git
4 years agold: Add $NOSANTIZE_CFLAGS to more linker tests
H.J. Lu [Wed, 26 Aug 2020 18:28:26 +0000 (11:28 -0700)]
ld: Add $NOSANTIZE_CFLAGS to more linker tests

-fsanitize= can be used to build binutils with

$ CC="gcc -fsanitize=address,undefined" CXX="g++ -fsanitize=address,undefined" .../configure --disable-werror

Since not all linker tests are compatible with -fsanitize=, pass
$NOSANTIZE_CFLAGS to disable -fsanitize= for such tests.

* testsuite/ld-elf/indirect.exp: Append $NOSANTIZE_CFLAGS to CC.
* testsuite/ld-elf/shared.exp: Likewise.
* testsuite/ld-elfcomm/elfcomm.exp: Likewise.
* testsuite/ld-elfvers/vers.exp: Likewise.
* testsuite/ld-elfvsb/elfvsb.exp: Likewise.
* testsuite/ld-elfweak/elfweak.exp: Likewise.
* testsuite/ld-plugin/lto.exp: Likewise.
* testsuite/ld-plugin/plugin.exp: Likewise.
* testsuite/ld-scripts/crossref.exp: Likewise.

4 years agoFix sanitization problems in the BFD library when running the linker testsuite for...
Nick Clifton [Wed, 26 Aug 2020 16:43:39 +0000 (17:43 +0100)]
Fix sanitization problems in the BFD library when running the linker testsuite for the AVR target.

PR 26433
* elf32-avr.c (avr_final_link_relocate):  Fix undefined shift
behaviour.
(avr_elf32_load_records_from_section): Use bfd_get_16 and
bfd_get_32 to load values from potentially unaligned pointers.

4 years agoFix sanitization problems running the linker testsuite for the alpha-elf target.
Nick Clifton [Wed, 26 Aug 2020 15:26:13 +0000 (16:26 +0100)]
Fix sanitization problems running the linker testsuite for the alpha-elf target.

PR 26416
* elf64-alpha.c (elf64_alpha_relax_tls_get_addr): Test for and
ignore local symbols.
(elf64_alpha_relax_got_load): Do not check for local dynamic
symbols.
(OP_LDA, OP_LDAH, OP_LDQ, OP_BR, OP_BSR): Use unsigned constant
values.
(INSN_A) Cast the A parameter to unsigned.
(INSN_AB): Define in terms of INSN_A.
(INSN_ABC): Likewise.
(INSN_ABO): Likewise.
(INSN_AD): Likewise.

4 years agoFix a sanitization problem running the linker testsuite for the AArch64 target.
Nick Clifton [Wed, 26 Aug 2020 14:50:36 +0000 (15:50 +0100)]
Fix a sanitization problem running the linker testsuite for the AArch64 target.

PR 26411
* elfnn-aarch64.c (elfNN_aarch64_relocate_section): Use an
unsigned long constant when creating a mask to test for alignment
issues.

4 years agoopcodes: Add missing entries to ebpf_isa_attr
Jose E. Marchesi [Wed, 26 Aug 2020 14:48:39 +0000 (16:48 +0200)]
opcodes: Add missing entries to ebpf_isa_attr

opcodes/

* disassemble.c (enum epbf_isa_attr): Add ISA_XBPFBE, ISA_EBPFMAX.

4 years agoAdd support to readelf for the OpenBSD segment types.
Nick Clifton [Wed, 26 Aug 2020 14:13:41 +0000 (15:13 +0100)]
Add support to readelf for the OpenBSD segment types.

PR 26405
binutils* readelf.c (get_segment_type): Handle OpenBSD segment types.

include * elf/common.h (PT_OPENBSD_BOOTDATA): Define.
(PT_OPENBSD_RANDOMIZE): Define.
(PT_OPENBSD_WXNEEDED): Define.

4 years agoPR26508 UBSAN: tc-xtensa.c:7764 null pointer bsearch
Alan Modra [Wed, 26 Aug 2020 10:58:40 +0000 (20:28 +0930)]
PR26508 UBSAN: tc-xtensa.c:7764 null pointer bsearch

PR 26508
* config/tc-xtensa.c (xg_get_trampoline_chain): Return early
when n_entries is zero.

4 years agoPR26507 UBSAN: elf32-xtensa.c:6013 null pointer bsearch
Alan Modra [Wed, 26 Aug 2020 10:55:04 +0000 (20:25 +0930)]
PR26507 UBSAN: elf32-xtensa.c:6013 null pointer bsearch

PR 26507
* elf32-xtensa.c (find_removed_literal): Don't bsearch empty map.

4 years agoPR26506 UBSAN: elf32-xtensa.c:3203 null pointer memcpy
Alan Modra [Wed, 26 Aug 2020 10:50:11 +0000 (20:20 +0930)]
PR26506 UBSAN: elf32-xtensa.c:3203 null pointer memcpy

PR 26506
* elf32-xtensa.c (elf_xtensa_combine_prop_entries): Return early
when section is empty.

4 years agoPR26499 UBSAN: eelf32_spu.c:537 member access null pointer
Alan Modra [Wed, 26 Aug 2020 10:42:42 +0000 (20:12 +0930)]
PR26499 UBSAN: eelf32_spu.c:537 member access null pointer

Another &p->field.

PR 26499
* emultempl/spuelf.em (spu_elf_relink): Check for NULL tmp_file_list.

4 years agoPR26498 UBSAN: elf32-spu.c:2292 left shift overflow
Alan Modra [Wed, 26 Aug 2020 10:32:07 +0000 (20:02 +0930)]
PR26498 UBSAN: elf32-spu.c:2292 left shift overflow

PR 26498
* elf32-spu.c (find_function_stack_adjust): Use unsigned vars to
avoid UB left shift.

4 years agoPR 26484-26488, 26490 UBSAN &h->elf null pointer
Alan Modra [Wed, 26 Aug 2020 10:13:18 +0000 (19:43 +0930)]
PR 26484-26488, 26490 UBSAN &h->elf null pointer

PR 26484
PR 26485
PR 26486
PR 26487
PR 26488
PR 26490
* elf64-ppc.c (is_tls_get_addr): Avoid UB &h->elf when h is NULL.
(ppc64_elf_tls_setup): Likewise.
(branch_reloc_hash_match): Likewise.
(build_plt_stub): Likewise.
(ppc64_elf_relocate_section): Likewise.

4 years agoPR26478 UBSAN: mmo.c:2941 null pointer memcpy
Alan Modra [Wed, 26 Aug 2020 09:02:30 +0000 (18:32 +0930)]
PR26478 UBSAN: mmo.c:2941 null pointer memcpy

PR 26478
* mmo.c (mmo_write_symbols_and_terminator): Don't memcpy empty table.

4 years agoPR26475 UBSAN: elfxx-mips.c:12180 null pointer memset
Alan Modra [Wed, 26 Aug 2020 08:52:07 +0000 (18:22 +0930)]
PR26475 UBSAN: elfxx-mips.c:12180 null pointer memset

Another memset(0,0,0)

PR 26475
* elfxx-mips.c (_bfd_mips_elf_finish_dynamic_sections): Check
sstubs->contents != NULL.

4 years agoPR26458 UBSAN: elf32-i386.c:3615 left shift of negative value
Alan Modra [Wed, 26 Aug 2020 01:30:22 +0000 (11:00 +0930)]
PR26458 UBSAN: elf32-i386.c:3615 left shift of negative value

Happens when poking symbol index -2 into r_info.  (The index is
updated before writing out to file.)

PR 26458
* elf/common.h (ELF32_R_INFO): Cast symbol index to unsigned.

4 years agoPR26453 UBSAN: som.c:2885 null pointer memcpy
Alan Modra [Wed, 26 Aug 2020 08:28:40 +0000 (17:58 +0930)]
PR26453 UBSAN: som.c:2885 null pointer memcpy

PR 26453
* som.c (som_prep_for_fixups): Return early when no symbols.

4 years agoPR26448 UBSAN: symbols.c:1586 left shift of negative value
Alan Modra [Wed, 26 Aug 2020 08:09:58 +0000 (17:39 +0930)]
PR26448 UBSAN: symbols.c:1586 left shift of negative value

Besides avoiding the UB, this also makes right shifts inside
expression symbols unsigned, consistent with the way gas evaluates
expressions in source.

PR 26448
* symbols.c: Include limits.h.
(resolve_symbol_value <O_left_shift, O_right_shift>): Do an
unsigned shift.  Warn if shift count larger than valueT size.

4 years agoPR26447 UBSAN: expr.c:1936 left shift of negative value
Alan Modra [Wed, 26 Aug 2020 05:25:39 +0000 (14:55 +0930)]
PR26447 UBSAN: expr.c:1936 left shift of negative value

PR 26447
* expr.c (expr <O_left_shift>): Do an unsigned shift.

4 years agoPR26431 UBSAN: pe-dll.c:568 null pointer bsearch
Alan Modra [Wed, 26 Aug 2020 04:53:42 +0000 (14:23 +0930)]
PR26431 UBSAN: pe-dll.c:568 null pointer bsearch

PR 26431
* pe-dll.c (auto_export): Don't call bsearch with zero count.

4 years agoPR26418 UBSAN: cache.c:386 null pointer fwrite
Alan Modra [Wed, 26 Aug 2020 04:48:53 +0000 (14:18 +0930)]
PR26418 UBSAN: cache.c:386 null pointer fwrite

PR 26418
* ecofflink.c (WRITE): Don't write size 0 chunks.

4 years agoasan: alpha-vms: mmember access within null pointer
Alan Modra [Wed, 26 Aug 2020 02:02:51 +0000 (11:32 +0930)]
asan: alpha-vms: mmember access within null pointer

* bfdio.c (bfd_get_file_size): Don't segv on NULL adata.

4 years agoPR26415 UBSAN: vms-misc.c:636 left shift cannot be represented
Alan Modra [Wed, 26 Aug 2020 01:56:35 +0000 (11:26 +0930)]
PR26415 UBSAN: vms-misc.c:636 left shift cannot be represented

An unsigned short value is promoted to int, thus triggering UB on a
left shift of a positive value that results in a negative int.

PR 26415
* vms-misc.c (vms_time_t_to_vms_time): Don't use unsigned short vars.

4 years agoPR26412 UBSAN: objcopy.c:3026 null pointer fwrite
Alan Modra [Wed, 26 Aug 2020 00:41:38 +0000 (10:11 +0930)]
PR26412 UBSAN: objcopy.c:3026 null pointer fwrite

PR 26412
* objcopy.c (copy_object): Don't fwrite NULL contents.

4 years agobpf: add xBPF ISA
David Faust [Wed, 26 Aug 2020 13:39:00 +0000 (15:39 +0200)]
bpf: add xBPF ISA

This patch adds support for xBPF, another ISA targetting the BPF
virtual architecture. For now, the primary difference between eBPF
and xBPF is that xBPF supports indirect calls through the
'call %reg' form of the call instruction.

bfd/
* archures.c (bfd_mach_xbpf): Define.
* bfd-in2.h: Regenerate.
* cpu-bpf.c (bfd_xbpf_arch) New.
(bfd_bpf_arch) Update next in list field to point to xbpf arch.

cpu/
* bpf.cpu (arch bpf): Add xbpf mach and isas.
(define-xbpf-isa) New pmacro.
(all-isas) Add xbpfle,xbpfbe.
(endian-isas): New pmacro.
(mach xbpf): New.
(model xbpf-def): Likewise.
(h-gpr): Add xbpf mach.
(f-dstle, f-srcle, dstle, srcle): Add xbpfle isa.
(f-dstbe, f-srcbe, dstbe, srcbe): Add xbpfbe isa.
(define-alu-insn-un): Use new endian-isas pmacro.
(define-alu-insn-bin, define-alu-insn-mov): Likewise.
(define-endian-insn, define-lddw): Likewise.
(dlind, dxli, dxsi, dsti): Likewise.
(define-cond-jump-insn, define-call-insn): Likewise.
(define-atomic-insns): Likewise.

gas/
* config/tc-bpf.c: Add option -mxbpf to select xbpf isa.
* testsuite/gas/bpf/indcall-1.d: New file.
* testsuite/gas/bpf/indcall-1.s: Likewise.
* testsuite/gas/bpf/indcall-bad-1.l: Likewise.
* testsuite/gas/bpf/indcall-bad-1.s: Likewise.
* testsuite/gas/bpf/bpf.exp: Run new tests.

opcodes/
* bpf-desc.c: Regenerate.
* bpf-desc.h: Likewise.
* bpf-opc.c: Likewise.
* bpf-opc.h: Likewise.
* disassemble.c (disassemble_init_for_target): Set bits for xBPF
ISA when appropriate.

4 years agowindres: AUTOCHECKBOX default style must have WS_TABSTOP style
Katayama Hirofumi [Wed, 26 Aug 2020 10:57:02 +0000 (11:57 +0100)]
windres: AUTOCHECKBOX default style must have WS_TABSTOP style

PR 26340
* rcparse.y (AUTOCHECKBOX): Add WS_TABSTOP to the base style.

4 years agogdb: Add ARC target and maintainer to MAINTAINERS
Shahab Vahedi [Wed, 5 Aug 2020 15:58:45 +0000 (17:58 +0200)]
gdb: Add ARC target and maintainer to MAINTAINERS

This patch updates gdb/MAINTAINERS with ARC as a target and
myself as the maintainer.  There is no mention of "-Werror"
because that is enabled by default for gdb/ targets now.

gdb/ChangeLog:

* MAINTAINERS: Add ARC target and maintainer.

4 years agoRe: CSKY: Add new arch CK860
Alan Modra [Wed, 26 Aug 2020 02:10:28 +0000 (11:40 +0930)]
Re: CSKY: Add new arch CK860

bfd-in2.h is a generated file.  Put the new machine where it belongs.

* archures.c (bfd_mach_ck860): Define.

4 years agoAutomatic date update in version.in
GDB Administrator [Wed, 26 Aug 2020 00:00:10 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agoarc: Add GNU/Linux support for ARC
Anton Kolesov [Wed, 6 Jul 2016 17:36:49 +0000 (20:36 +0300)]
arc: Add GNU/Linux support for ARC

ARC Linux targets differences from baremetal:

- No support for hardware single instruction stepping.
- Different access rules to registers.
- Use of another instruction for breakpoints.

v2: Changes after Tom's remarks [1]
 arc-linux-tdep.c
  - Use true/false instead of TRUE/FALSE.
  - arc_linux_sw_breakpoint_from_kind (): Break long lines into two.
  - arc_linux_sw_breakpoint_from_kind (): Remove starting blank line.
  - Use explicit number evaluation, e.g: if (a & b) -> if ((a & b) != 0)
 arc-tdep.c
  - Use explicit number evaluation, e.g: if (a & b) -> if ((a & b) != 0)
 gdb/configure.tgt
  - arc*-*-linux*): Remove "build_gdbserver=yes".

v3: Changes after Simon's remarks [2]
  arc-linux-tdep.c
  - Use "return trap_size" instead of cryptic "return 2".
  - Removed unnecessary curly braces.
  - Removed "void" from "_initialize_arc_linux_tdep (void)".

v5: Changes after Simon's remarks [3]
- Remove unnecessary empty lines.
- Replace "breakpoint uses" with "breakpoints use" in a comment.
- "return condition;" i.s.o. "if (condition) return true; else return false;"

[1] Tom's remarks
https://sourceware.org/pipermail/gdb-patches/2020-April/167887.html

[2] Simon's remarks on v2
https://sourceware.org/pipermail/gdb-patches/2020-May/168513.html

[3] Simon's remarks on v4
https://sourceware.org/pipermail/gdb-patches/2020-August/170994.html

gdb/ChangeLog:

2020-08-25  Anton Kolesov  <anton.kolesov@synopsys.com>

* configure.tgt: ARC support for GNU/Linux.
* Makefile.in (ALL_TARGET_OBJS): Likewise.
* arc-linux-tdep.c: New file.
* arc-tdep.h (ARC_STATUS32_L_MASK, ARC_STATUS32_DE_MASK): Declare.
* arc-tdep.c (arc_write_pc): Use it.

4 years agoarc: Add hardware loop detection
Shahab Vahedi [Thu, 9 Jul 2020 15:44:01 +0000 (17:44 +0200)]
arc: Add hardware loop detection

For ARC there are registers that are not part of a required set in XML
target descriptions by default, but are almost always present on ARC
targets and are universally exposed by the ptrace interface.  Hardware
loop registers being one of them.

LP_START and LP_END auxiliary registers are hardware loop start and end.
Formally, they are optional, but it is hard to find an ARC configuration
that doesn't have them.  They are always present in processors that can
run GNU/Linux.  GDB needs to know about those registers to implement
proper software single stepping, since they affect  what instruction
will be next.

This commit adds the code to check for the existance of "lp_start" and
"lp_end" in XML target descriptions. If they exist, then the function
reports that the target supports hardware loops.

gdb/ChangeLog:

* arc-tdep.c (arc_check_for_hardware_loop): New.
* arc-tdep.h (gdbarch_tdep): New field has_hw_loops.

gdb/doc/ChangeLog:

* gdb.texinfo (Synopsys ARC): Document LP_START, LP_END and BTA.

4 years agoarc: Add inclusion of "gdbarch.h" in "arc-tdep.h"
Shahab Vahedi [Mon, 3 Aug 2020 18:41:42 +0000 (20:41 +0200)]
arc: Add inclusion of "gdbarch.h" in "arc-tdep.h"

The "arc-tdep.h" makes use of definitions in "gdbarch.h", but it
does not include it explicitly.  I have piggy backed this fix
in another commit [1], but I was asked to do it separately [2].

[1] arc: Add hardware loop detection
https://sourceware.org/pipermail/gdb-patches/2020-July/170800.html

[2] Simon's remarks to "arc: Add hardware loop detection"
https://sourceware.org/pipermail/gdb-patches/2020-August/170993.html

gdb/ChangeLog:

* arc-tdep.h: Include "gdbarch.h".

4 years agoarc: Add ARCv2 XML target along with refactoring
Shahab Vahedi [Thu, 9 Jul 2020 15:43:13 +0000 (17:43 +0200)]
arc: Add ARCv2 XML target along with refactoring

A few changes have been made to make the register support simpler,
more flexible and extendible.  The trigger for most of these changes
are the remarks [1] made earlier for v2 of this patch.  The noticeable
improvements are:

- The arc XML target features are placed under gdb/features/arc
- There are two cores (based on ISA) and one auxiliary feature:
  v1-core: ARC600, ARC601, ARC700
  v2-core: ARC EM, ARC HS
  aux: common in both
- The XML target features represent a minimalistic sane set of
  registers irrespective of application (baremetal or linux).
- A concept of "feature" class has been introduced in the code.
  The "feature" object is constructed from BFD and GDBARCH data.
  It contains necessary information (ISA and register size) to
  determine which XML target feature to use.
- A new structure (ARC_REGISTER_FEATURE) is added that allows
  providing index, names, and the necessity of registers. This
  simplifies the sanity checks and future extendibility.
- Documnetation has been updated to reflect ARC features better.
- Although the feature names has changed, there still exists
  backward compatibility with older names through
  find_obsolete_[core,aux]_names() functions.

The last two points were inspired from RiscV port.

[1]
https://sourceware.org/pipermail/gdb-patches/2020-May/168511.html

gdb/ChangeLog:

* arch/arc.h
  (arc_gdbarch_features): New class to stir the selection of target XML.
  (arc_create_target_description): Use FEATURES to choose XML target.
  (arc_lookup_target_description): Use arc_create_target_description
  to create _new_ target descriptions or return the already created
  ones if the FEATURES is the same.
* arch/arc.c: Implementation of prototypes described above.
* gdb/arc-tdep.h (arc_regnum enum): Add more registers.
  (arc_gdbarch_features_init): Initialize the FEATURES struct.
* arc-tdep.c (*_feature_name): Make feature names consistent.
  (arc_register_feature): A new struct to hold information about
  registers of a particular target/feature.
  (arc_check_tdesc_feature): Check if XML provides registers in
  compliance with ARC_REGISTER_FEATURE structs.
  (arc_update_acc_reg_names): Add aliases for r58 and r59.
  (determine_*_reg_feature_set): Which feature name to look for.
  (arc_gdbarch_features_init): Given MACH and ABFD, initialize FEATURES.
  (mach_type_to_arc_isa): Convert from a set of binutils machine types
  to expected ISA enums to be used in arc_gdbarch_features structs.
* features/Makefile (FEATURE_XMLFILES): Add new files.
* gdb/features/arc/v1-aux.c: New file.
* gdb/features/arc/v1-aux.xml: Likewise.
* gdb/features/arc/v1-core.c: Likewise.
* gdb/features/arc/v1-core.xml: Likewise.
* gdb/features/arc/v2-aux.c: Likewise.
* gdb/features/arc/v2-aux.xml: Likewise.
* gdb/features/arc/v2-core.c: Likewise.
* gdb/features/arc/v2-core.xml: Likewise.
* NEWS (Changes since GDB 9): Announce obsolence of old feature names.

gdb/doc/ChangeLog:

* gdb.texinfo (Synopsys ARC): Update the documentation for ARC
Features.

gdb/testsuite/ChangeLog:

* gdb.arch/arc-tdesc-cpu.xml: Use new feature names.

4 years agogdb/testsuite: fix gdb.threads/stepi-random-signal.exp pattern (gdb/26532)
Simon Marchi [Tue, 25 Aug 2020 15:06:43 +0000 (11:06 -0400)]
gdb/testsuite: fix gdb.threads/stepi-random-signal.exp pattern (gdb/26532)

Commit 1eb8556f5a8b ("gdb: add infrun_debug_printf macro") changed the
debug output format for `set debug infrun 1`.  The test
gdb.threads/stepi-random-signal.exp uses that debug output, and was
updated, but not correctly.  It results in this failure:

    FAIL: gdb.threads/stepi-random-signal.exp: stepi (no random signal)

Fix it by adjusting the pattern in the test.

gdb/testsuite/ChangeLog:

PR gdb/26532
* gdb.threads/stepi-random-signal.exp: Update pattern.

Change-Id: If5fa525e9545e32a286effe6a6184358374bd37c

4 years agogdb/testsuite: fix gdb.base/ui-redirect.exp pattern (gdb/26532)
Simon Marchi [Tue, 25 Aug 2020 15:06:27 +0000 (11:06 -0400)]
gdb/testsuite: fix gdb.base/ui-redirect.exp pattern (gdb/26532)

Commit 1eb8556f5a8b ("gdb: add infrun_debug_printf macro") changed the
debug output format for `set debug infrun 1`.  It broke test
gdb.base/ui-redirect.exp, which I missed:

    FAIL: gdb.base/ui-redirect.exp: debugging: continue

Fix it by adjusting the pattern in the test to the new reality.

gdb/testsuite/ChangeLog:

PR gdb/26532
* gdb.base/ui-redirect.exp: Update pattern.

Change-Id: Ie8a8f6675e35a0cab55109b1534b44eb51baec9d

4 years agoFix ChangeLog entry for commit b04aa1fc8c9d4a79e6293a3a1df7507052afedf3
Gary Benson [Tue, 25 Aug 2020 14:53:33 +0000 (15:53 +0100)]
Fix ChangeLog entry for commit b04aa1fc8c9d4a79e6293a3a1df7507052afedf3

4 years agoFix the linker's handling of DWARF-5 line number tables.
Mark Wielaard [Tue, 25 Aug 2020 14:33:00 +0000 (15:33 +0100)]
Fix the linker's handling of DWARF-5 line number tables.

When building with gcc with -gdwarf-5 ld tests (including ld-elf/dwarf.exp)
fail because they try to read the .debug_ranges section. But DWARF5
introduces a new .debug_rnglists section that encodes the address ranges
more efficiently. Implement reading the debug_rnglists in bfd/dwarf2.c.
Which makes all tests pass again and fixes several gcc testsuite tests
when defaulting to DWARF5.

*  dwarf2.c (struct dwarf2_debug_file): Add dwarf_rnglists_buffer
and dwarf_rnglists_size fields.
(dwarf_debug_sections): Add debug_rnglists.
(dwarf_debug_section_enum): Likewise.
(read_debug_rnglists): New function.
(read_rangelist): New function to call either read_ranges or
read_rnglists. Rename original function to...
(read_ranges): ...this.
(read_rnglists): New function.

4 years agoDisable Clang's integrated assembler for two testcases
Gary Benson [Tue, 25 Aug 2020 14:25:04 +0000 (15:25 +0100)]
Disable Clang's integrated assembler for two testcases

gdb.dwarf2/dw2-dir-file-name.exp fails to build using Clang because
the generated assembly language contains .ascii directives with more
than one string literal.  gdb.dwarf2/dw2-restore.exp fails to build
using Clang because it contains .func and .endfunc directives.
This commit causes Clang to invoke the system assembler to assemble
the relevant files.

gdb/testsuite/ChangeLog:

* gdb.dwarf2/dw2-dir-file-name.exp: Use system assembler
when compiling with clang.
* gdb.dwarf2/dw2-restore.exp: Likewise

4 years agoEnable gdb.cp/ambiguous.exp with GCC and clang
Gary Benson [Tue, 25 Aug 2020 14:14:46 +0000 (15:14 +0100)]
Enable gdb.cp/ambiguous.exp with GCC and clang

gdb.cp/ambiguous.exp failed to build using clang with the following
error:

 gdb compile failed, /gdbtest/src/gdb/testsuite/gdb.cp/ambiguous.cc:70:36:
   warning: direct base 'A1' is inaccessible due to ambiguity:
     class JVA1 -> class KV -> class A1
     class JVA1 -> class A1 [-Winaccessible-base]
 class JVA1 : public KV, public LV, public A1 {
                                   ^~~~~~~~~

This commit builds this testcase with -Wno-inaccessible-base when
using clang, to avoid this failure.

Furthermore, gdb.cp/ambiguous.exp has been disabled when using GCC
since 1998.  This commit enables this testcase, building with
-Wno-inaccessible-base when using GCC >= 10.1, and -w otherwise.

gdb/testsuite/ChangeLog:

* gdb.cp/ambiguous.exp: Enable test when compiling with GCC.
Add additional_flags=-Wno-inaccessible-base when compiling
with GCC >= 10.1 or clang.  Add additional_flags=-w when
compiling with GCC < 10.

4 years agoPR26505, ASAN: xstormy16_elf_relax_section elf32-xstormy16.c:595
Alan Modra [Tue, 25 Aug 2020 08:26:08 +0000 (17:56 +0930)]
PR26505, ASAN: xstormy16_elf_relax_section elf32-xstormy16.c:595

PR 26505
* elf32-xstormy16.c (xstormy16_elf_relax_section): Check
is_elf_hash_table before accessing elf fields.

4 years agoPR26504, ASAN: parse_disassembler_options vax-dis.c:142
Alan Modra [Tue, 25 Aug 2020 08:16:59 +0000 (17:46 +0930)]
PR26504, ASAN: parse_disassembler_options vax-dis.c:142

PR 26504
* vax-dis.c (parse_disassembler_options): Always add at least one
to entry_addr_total_slots.

4 years agoPR26501, ASAN: tic54x_undefined_symbol tc-tic54x.c:5015
Alan Modra [Tue, 25 Aug 2020 07:46:28 +0000 (17:16 +0930)]
PR26501, ASAN: tic54x_undefined_symbol tc-tic54x.c:5015

PR26501
* gas/config/tc-tic54x.c (tic54x_undefined_symbol): Properly treat
misc_symbol_hash entries without values.

4 years agoPR26500, ASAN: tic4x_inst_make tc-tic4x.c:1247
Alan Modra [Tue, 25 Aug 2020 07:26:07 +0000 (16:56 +0930)]
PR26500, ASAN: tic4x_inst_make tc-tic4x.c:1247

PR 26500
* tc-tic4x.c (tic4x_inst_make): Don't die on terminating insn
with name = "".

4 years agoPR26482, ASAN: _bfd_xcoff_sizeof_headers coff-rs6000.c:2585
Alan Modra [Tue, 25 Aug 2020 06:45:27 +0000 (16:15 +0930)]
PR26482, ASAN: _bfd_xcoff_sizeof_headers coff-rs6000.c:2585

PR 26482
* coff-rs6000.c (_bfd_xcoff_sizeof_headers): Ignore sections that
won't be output.

4 years agoPR26463, ASAN: m32c_elf_relax_section elf32-m32c.c:1448
Alan Modra [Tue, 25 Aug 2020 08:53:38 +0000 (18:23 +0930)]
PR26463, ASAN: m32c_elf_relax_section elf32-m32c.c:1448

PR 26463
* elf32-m32c.c (m32c_elf_relax_section): Check is_elf_hash_table
before accessing elf fields.

4 years agoPR26452, ASAN: som_compute_checksum som.c:4293
Alan Modra [Tue, 25 Aug 2020 06:16:02 +0000 (15:46 +0930)]
PR26452, ASAN: som_compute_checksum som.c:4293

PR 26452
* som.c (som_compute_checksum): XOR 32-bit words in header,
not unsigned long sized words.

4 years agoPR26441, ASAN: get_b_cc tc-cr16.c:1521
Alan Modra [Tue, 25 Aug 2020 06:04:10 +0000 (15:34 +0930)]
PR26441, ASAN: get_b_cc tc-cr16.c:1521

PR 26441
* config/tc-cr16.c (get_b_cc): Return NULL early if op isn't
two or three chars, and don't bother copying.

4 years agoPR26430, ASAN: nacl_modify_segment_map elf-nacl.c:164
Alan Modra [Tue, 25 Aug 2020 05:17:50 +0000 (14:47 +0930)]
PR26430, ASAN: nacl_modify_segment_map elf-nacl.c:164

PR 26430
* elf-nacl.c (nacl_modify_segment_map): Correct alloc size and
amount copied for elf_segment_map defined with one element
sections array.

4 years agoPR26426, ASAN: neon_quad tc-arm.c:15175
Alan Modra [Tue, 25 Aug 2020 04:42:49 +0000 (14:12 +0930)]
PR26426, ASAN: neon_quad tc-arm.c:15175

PR 26426
* config/tc-arm.c (do_neon_mvn, do_neon_swp): Bail out on
NS_NULL shape.

4 years agoPR26422, ASAN: elf32_arm_final_link_relocate elf32-arm.c:10351
Alan Modra [Tue, 25 Aug 2020 04:35:10 +0000 (14:05 +0930)]
PR26422, ASAN: elf32_arm_final_link_relocate elf32-arm.c:10351

Always reading 32 bits in order to extract addends from instruction
fields is wrong when the field size is smaller.  It also leads to
reading past the end of the section.  This patch tidies that by
reading the proper field size, which allows some later refetching of
addends to disappear.

PR 26422
* elf32-arm.c (elf32_arm_final_link_relocate): Use the appropriate
bfd_get_x size function to read addends out of fields.  Apply
rightshift adjustment too.  Don't apply the now unnecessary
howto->size shift to branch REL addends.  Don't refetch R_ARM_ABS8
and R_ARM_ABS16 addends.  Don't refetch thumb branch addends.
Correct R_ARM_THM_JUMP6 addend.

4 years agoPR26419, ASAN: mn10300_elf_relax_section elf-m10300.c:3943
Alan Modra [Tue, 25 Aug 2020 00:07:23 +0000 (09:37 +0930)]
PR26419, ASAN: mn10300_elf_relax_section elf-m10300.c:3943

PR 26419
* elf-m10300.c (mn10300_elf_relax_section): Don't attempt access
before start of section.

4 years agoPR26410, UBSAN: symbols.c:1818
Alan Modra [Tue, 25 Aug 2020 05:21:21 +0000 (14:51 +0930)]
PR26410, UBSAN: symbols.c:1818

Don't memset(0,0,0).

PR 26410
* symbols.c (dollar_label_count, dollar_label_max): Make size_t.
(dollar_label_clear): Don't call memset with NULL pointer.

4 years agogas warning fixes
Alan Modra [Tue, 25 Aug 2020 00:27:58 +0000 (09:57 +0930)]
gas warning fixes

Some versions of gcc with -Werror=format-overflow complain about using
a perfectly good 7 char buffer for "r%dr%d" when the int is between 0
and 64, apparently not seeing the value range.

note: __builtin___sprintf_chk output between 5 and 24 bytes into a destination of size 7

* config/tc-arc.c (declare_register_set): Avoid false positive
format-overflow warning.
* config/tc-epiphany.c (md_assemble): Likewise.
* config/tc-mips.c (md_begin): Likewise.
* config/tc-mmix.c (mmix_md_begin): Likewise.
* config/tc-nds32.c (nds32_elf_append_relax_relocs): Avoid false
positive "may be used uninitialized" warning.

4 years agotic54x ld testsuite
Alan Modra [Mon, 24 Aug 2020 16:29:55 +0000 (01:59 +0930)]
tic54x ld testsuite

* testsuite/ld-libs/libs.exp: Don't run on sh-pe, tic30 or tic54x.
* testsuite/ld-scripts/data.d: xfail tic4x and tic54x.
* testsuite/ld-scripts/empty-address-2a.d: xfail tic54x.
* testsuite/ld-scripts/empty-address-2b.d: Likewise.
* testsuite/ld-scripts/include-1.d: xfail tic4x and tic54x.
* testsuite/ld-scripts/provide-1.d: xfail tic54x.
* testsuite/ld-scripts/provide-2.d: Likewise.
* testsuite/ld-scripts/provide-4.d: Likewise.

4 years agogdb/modula-2: parsing of multi-subscript arrays
Gaius Mulley [Tue, 25 Aug 2020 08:39:27 +0000 (09:39 +0100)]
gdb/modula-2: parsing of multi-subscript arrays

Fix bug PR m2/26372, GDB's inability to parse multi-dimensional
modula-2 arrays.

We previously had two rules for handling the parsing of array
sub-scripts.  I have reproduced them here with the actual handler
blocks removed to make the bug clearer:

  exp     :    exp '[' non_empty_arglist ']'
          ;

  exp     :    exp '[' exp ']'
          ;

  non_empty_arglist
          :       exp
          ;

  non_empty_arglist
          :       non_empty_arglist ',' exp
          ;

This is ambiguous as the pattern "exp '[' exp" could match either of
the 'exp' rules.  Currently it just so happens that the parser picks
the second 'exp' rule which means we can only handle a single array
index.

As the handler code for the first 'exp' pattern will correctly handle
and number of array indexes then lets just remove the second pattern.

gdb/ChangeLog:

PR m2/26372
        * m2-exp.y (exp): Improve comment for non_empty_arglist case, add
an assert.  Remove single element array indexing pattern as the
MULTI_SUBSCRIPT support will handle this case too.

gdb/testsuite/ChangeLog:

PR m2/26372
        * gdb.modula2/multidim.c: New file.
        * gdb.modula2/multidim.exp: New file.

4 years agoRevert "Fix for Bug 26372 [Modula-2] Parsing of multi-subscript arrays"
Andrew Burgess [Tue, 25 Aug 2020 09:25:32 +0000 (10:25 +0100)]
Revert "Fix for Bug 26372 [Modula-2] Parsing of multi-subscript arrays"

This reverts commit 07758bdfa9e5a762f2ec0deeb51b11d6ad5fe376.

4 years agoFix for Bug 26372 [Modula-2] Parsing of multi-subscript arrays
Gaius Mulley [Tue, 25 Aug 2020 09:16:56 +0000 (11:16 +0200)]
Fix for Bug 26372 [Modula-2] Parsing of multi-subscript arrays

Here is a bugfix for Pr 26372 [Modula-2] Parsing of multi-subscript arrays.
Also included is a dejagnu testcase.  No extra regressions are caused on
Debian GNU/Linux Buster amd64.

gdb/ChangeLog:

2020-08-25  Gaius Mulley  <gaiusmod2@gmail.com>

PR m2/26372
* m2-exp.y: Rewrite array subscript rules to support multidimension
array access.  (ArgumentList) replaces non_empty_arglist.

gdb/testsuite/ChangeLog:

2020-08-25  Gaius Mulley  <gaiusmod2@gmail.com>

PR m2/26372
* testsuite/gdb.modula2/multidim.exp: New file.
* testsuite/gdb.modula2/multidim.c: New file.

4 years agogdb: move declaration of valprint_check_validity to valprint.h
Simon Marchi [Tue, 25 Aug 2020 02:51:50 +0000 (22:51 -0400)]
gdb: move declaration of valprint_check_validity to valprint.h

The implementation is in valprint.c, so the declaration belongs in
valprint.h.

gdb/ChangeLog:

* value.h (valprint_check_validity): Move declaration from
here...
* valprint.h (valprint_check_validity): ... to here.

Change-Id: Ibe577d3696720099e6d79888d4ee8e3c1bf05a26

4 years agoAutomatic date update in version.in
GDB Administrator [Tue, 25 Aug 2020 00:00:09 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agogdb/testsuite: make runto always emit a FAIL on internal error
Simon Marchi [Mon, 24 Aug 2020 23:44:53 +0000 (19:44 -0400)]
gdb/testsuite: make runto always emit a FAIL on internal error

I noticed that when a test uses `runto_main` and a GDB internal error
happens while running to main, no error or fail is emitted.  This is
because `runto_main` uses the `no-message` option of `runto`.

As a result, if a test fails to run to main and exits, no sign that
something went wrong is emitted.  For example, add this always-false
assertion to compute_frame_id:

    --- a/gdb/frame.c
    +++ b/gdb/frame.c
    @@ -545,6 +545,7 @@ static void
     compute_frame_id (struct frame_info *fi)
     {
       gdb_assert (!fi->this_id.p);
    +  gdb_assert (false);

       if (frame_debug)
         fprintf_unfiltered (gdb_stdlog, "{ compute_frame_id (fi=%d) ",

... and run gdb.dwarf2/dw2-align.exp.  No fail or sign that something
went wrong is shown.  It just appears as if the test gets skipped.

A developer introducing such a regression in this test today would
likely notice it, because we are used to diff-ing test results.  So we
would see some PASSes dispappear for no good reason and look into it.

But I find it worrysome for two reasons:

1. Scripts that analyze regressions (such as the one on the buildbot)
   may only look for new FAILs or new ERRORs.  It would probably miss
   this.
2. Imagine that we one day have a testsuite that runs cleanly (some
   people might already run subsets of the testsuite and expect it to
   all pass), we would just run the testsuite and check that there are
   no fails.  It would be easy to miss something like this.

In case of internal error, I suggest making `runto` emit a FAIL even if
`no-message` was passed.  This is different from other failure modes
that might be expected (whchi rightfully cause the test to simply be
skipped).  An internal error is always bad, so if it happens it should
noisily fail.

gdb/testsuite/ChangeLog:

* lib/gdb.exp (runto): Always emit fail on internal error.

Change-Id: I6e6faed4868ea821541a23042b2d01c30058b0d3

4 years agogdb: add debug_prefixed_vprintf
Simon Marchi [Mon, 24 Aug 2020 19:50:19 +0000 (15:50 -0400)]
gdb: add debug_prefixed_vprintf

To help ensure that all debug statements have the same format, introduce
the debug_prefixed_vprintf helper.  Implement linux_nat_debug_printf_1
and infrun_debug_printf_1 with it.

I would eventually like to style the module and function name with some
color, to help them stick out, but I don't really know how to do that
yet, it can always be done later.

gdb/ChangeLog:

* debug.h: New file.
* debug.c (debug_prefixed_vprintf): New function.
* infrun.c (infrun_debug_printf_1): Use debug_prefixed_vprintf.
* linux-nat.c (linux_nat_debug_printf_1): Likewise.

Change-Id: Iccc290a2dc6b5fffcbe1c2866ed8d804ad380764

4 years agogdb: add infrun_debug_printf macro
Simon Marchi [Mon, 24 Aug 2020 19:49:46 +0000 (15:49 -0400)]
gdb: add infrun_debug_printf macro

Introduce this macro to print debug statements in the infrun.c file,
same idea as what was done in 9327494e0eeb ("gdb: add
linux_nat_debug_printf macro").

Although in this case, there are places outside infrun.c that print
debug statements if debug_infrun is set.  So the macro has to be
declared in the header file, so that it can be used in these other
files.

Note one special case.  In stop_all_threads, I've used an explicit

    if (debug_infrun)
      infrun_debug_printf_1 ("stop_all_threads", "done");

for the message in the SCOPE_EXIT.  Otherwise, the message appears like
this:

  [infrun] operator(): done

Until we find a better solution for extracting a meaningful function
name for lambda functions, I think it's fine to handle these special
cases manually, they are quite rare.

Some tests need to be updated, because they rely on some infrun debug
statements.

gdb/ChangeLog:

* infrun.h (infrun_debug_printf_1): New function declaration.
(infrun_debug_printf): New macro.
* infrun.c (infrun_debug_printf_1): Use infrun_debug_printf
throughout.
(infrun_debug_printf): New function.
* breakpoint.c (should_be_inserted): Use infrun_debug_printf.
(handle_jit_event): Likewise.

gdb/testsuite/ChangeLog:

* gdb.base/gdb-sigterm.exp (do_test): Update expected regexp.
* gdb.threads/signal-while-stepping-over-bp-other-thread.exp:
Likewise.
* gdb.threads/stepi-random-signal.exp: Likewise.

Change-Id: I66433c8a9caa64c8525ab57c593022b9d1956d5c

4 years agoelf_hash_table_id access
Alan Modra [Mon, 24 Aug 2020 17:07:02 +0000 (02:37 +0930)]
elf_hash_table_id access

* elf-m10300.c (elf32_mn10300_hash_table): Test is_elf_hash_table
before accessing elf_hash_table_id.
* elf32-arc.c (elf_arc_hash_table): Likewise.
* elf32-arm.c (elf32_arm_hash_table): Likewise.
* elf32-avr.c (avr_link_hash_table): Likewise.
* elf32-bfin.c (bfinfdpic_hash_table): Likewise.
* elf32-cris.c (elf_cris_hash_table): Likewise.
* elf32-csky.c (csky_elf_hash_table): Likewise.
* elf32-frv.c (frvfdpic_hash_table): Likewise.
* elf32-hppa.c (hppa_link_hash_table): Likewise.
* elf32-lm32.c (lm32_elf_hash_table): Likewise.
* elf32-m32r.c (m32r_elf_hash_table): Likewise.
* elf32-m68hc1x.h (m68hc11_elf_hash_table): Likewise.
* elf32-m68k.c (elf_m68k_hash_table): Likewise.
* elf32-metag.c (metag_link_hash_table): Likewise.
* elf32-microblaze.c (elf32_mb_hash_table): Likewise.
* elf32-nds32.h (nds32_elf_hash_table): Likewise.
* elf32-or1k.c (or1k_elf_hash_table): Likewise.
* elf32-s390.c (elf_s390_hash_table): Likewise.
* elf32-sh.c (sh_elf_hash_table): Likewise.
* elf32-spu.c (spu_hash_table): Likewise.
* elf32-tilepro.c (tilepro_elf_hash_table): Likewise.
* elf32-xtensa.c (elf_xtensa_hash_table): Likewise.
* elf64-alpha.c (alpha_elf_hash_table): Likewise.
* elf64-hppa.c (hppa_link_hash_table): Likewise.
* elf64-ia64-vms.c (elf64_ia64_hash_table): Likewise.
* elf64-s390.c (elf_s390_hash_table): Likewise.
* elfnn-ia64.c (elfNN_ia64_hash_table): Likewise.
* elfnn-riscv.c (riscv_elf_hash_table): Likewise.
* elfxx-mips.c (mips_elf_hash_table): Likewise.
* elfxx-sparc.h (_bfd_sparc_elf_hash_table): Likewise.
* elfxx-tilegx.c (tilegx_elf_hash_table): Likewise.

4 years agoFix erroroneous use of '%zu' in elfcore_grok_win32pstatus
Jon Turney [Fri, 21 Aug 2020 15:30:00 +0000 (16:30 +0100)]
Fix erroroneous use of '%zu' in elfcore_grok_win32pstatus

As reported in [1], _bfd_error_handler() doesn't support '%zu'.

module_name_size is always 32-bits in the data structure we are
extracting it from, so use an unsigned int to store it instead.

[1] https://sourceware.org/pipermail/gdb-patches/2020-August/171391.html

bfd/ChangeLog:

2020-08-21  Jon Turney  <jon.turney@dronecode.org.uk>

* elf.c (elfcore_grok_win32pstatus): Change name_size to unsigned
int. Use '%u' format with  _bfd_error_handler to render it.

4 years agoPR26489, ASAN: ppc64_elf_size_stubs elf64-ppc.c:13389
Alan Modra [Mon, 24 Aug 2020 16:25:38 +0000 (01:55 +0930)]
PR26489, ASAN: ppc64_elf_size_stubs elf64-ppc.c:13389

Stub sections are inserted after sec_info is sized, so have higher ids.
Test flags that will exclude stub sections before looking at the
sec_info array.

PR 26489
* elf64-ppc.c (ppc64_elf_size_stubs): Test code_sec->has_toc_reloc
and code_sec->makes_toc_func_call before sec_info[code_sec->id].

4 years agoPR26492, ASAN: ppc64_elf_before_check_relocs elf64-ppc.c:4337
Alan Modra [Mon, 24 Aug 2020 16:21:39 +0000 (01:51 +0930)]
PR26492, ASAN: ppc64_elf_before_check_relocs elf64-ppc.c:4337

PR 26492
* elf64-ppc.c (ppc_hash_table): Test is_elf_hash_table before
accessing elf_hash_table_id.

4 years agoPR26483, ASAN: ppc_elf_link_params elf32-ppc.c:2314
Alan Modra [Mon, 24 Aug 2020 16:18:08 +0000 (01:48 +0930)]
PR26483, ASAN: ppc_elf_link_params elf32-ppc.c:2314

PR 26483
* elf32-ppc.c (ppc_elf_hash_table): Test is_elf_hash_table before
accessing elf_hash_table_id.

4 years agobfd: Handle DW_FORM_data16 for .debug_line tables containing MD5
Mark Wielaard [Mon, 24 Aug 2020 14:52:53 +0000 (15:52 +0100)]
bfd: Handle DW_FORM_data16 for .debug_line tables containing MD5

* dwarf2.c (read_attribute_value): Handle DW_FORM_data16.
(read_formatted_entries): Likewise. And skip zero entry.

4 years agoCSKY: Add new arch CK860.
Cooper Qu [Mon, 24 Aug 2020 12:13:47 +0000 (20:13 +0800)]
CSKY: Add new arch CK860.

bfd/
        * bfd-in2.h (bfd_mach_ck860): New.
        * cpu-csky.c (arch_info_struct): Add item for CK860.

gas/
        * config/tc-csky.c (csky_archs): Add item for CK860,
        change ck810 and ck807's arch_flag.
        (csky_cpus): Add item for CK860.
        (md_begin): Enable DSP for CK810 and CK807 by default.
        (md_apply_fix): Fix CKCORE_TLS_IE32 relocation failure.
        * gas/testsuite/gas/csky/cskyv2_all.d: Change 'sync 0'
        to 'sync'.
        * gas/testsuite/gas/csky/cskyv2_all.s: Likewise.
        * gas/testsuite/gas/csky/cskyv2_ck860.d: New.
        * gas/testsuite/gas/csky/cskyv2_ck860.s: New.
        * gas/testsuite/gas/csky/enhance_dsp.d: Change plsli.u16
        to plsli.16.
        * gas/testsuite/gas/csky/enhance_dsp.s: Likewise.

include/
        * opcode/csky.h (CSKYV2_ISA_10E60): New.
        (CSKY_ARCH_860): New.

opcode/
        * csky-dis.c (csky_find_inst_info): Skip CK860's instructions
        in other CPUs to speed up disassembling.
        * csky-opc.h (csky_v2_opcodes): Add CK860's instructions,
        Change plsli.u16 to plsli.16, change sync's operand format.

Change-Id: I80ec1a9c0cc600d668082a9b91ae6d45b33ec0fc

4 years agosparc testsuite fallout
Alan Modra [Mon, 24 Aug 2020 08:00:30 +0000 (17:30 +0930)]
sparc testsuite fallout

4 years agotic54x fixes
Alan Modra [Sun, 23 Aug 2020 11:35:34 +0000 (21:05 +0930)]
tic54x fixes

A number of the tic54x tests were failing, and I thought it worth
investigating since the target makes use of a lot of hash tables, and
we've just changed gas hash tables.  It turns out none of the gas
failures were due to hashing.

* config/tc-tic54x.c (stag_add_field_symbols): Don't free "name"
in case where it isn't copied.
* config/tc-tic54x.h (LOCAL_LABELS_FB): Undef.
* testsuite/gas/tic54x/field.d: Dump section contents and symbols
rather than disassembling.
* testsuite/gas/tic54x/set.d: Adjust for newer disassembly.

4 years agoPowerPC TPREL_HA/LO optimisation
Alan Modra [Mon, 24 Aug 2020 07:02:57 +0000 (16:32 +0930)]
PowerPC TPREL_HA/LO optimisation

ppc64 ld optimises sequences like the following
addis 3,13,wot@tprel@ha
lwz 3,wot@tprel@l(3)
to
nop
lwz 3,wot@tprel(13)
when "wot" is located near enough to the thread pointer.
However, the ABI doesn't require that R_PPC64_TPREL16_HA always be on
an addis rt,13,imm instruction, and while ld checked for that on the
high-part instruction it didn't disable the optimisation on the
low-part instruction.  This patch fixes that problem, disabling the
tprel optimisation globally if high-part instructions don't pass
sanity checks.  The optimisation is also enabled for ppc32, where
before ld.bfd had the code in the wrong place and ld.gold had it in a
block only enabled for ppc64.

bfd/
* elf32-ppc.c (ppc_elf_check_relocs): Set has_tls_reloc for
high part tprel16 relocs.
(ppc_elf_tls_optimize): Sanity check high part tprel16 relocs.
Clear do_tls_opt on odd instructions.
(ppc_elf_relocate_section): Move TPREL16_HA/LO optimisation later.
Don't sanity check them here.
* elf64-ppc.c (ppc64_elf_check_relocs): Set has_tls_reloc for
high part tprel16 relocs.
(ppc64_elf_tls_optimize): Sanity check high part tprel16 relocs.
Clear do_tls_opt on odd instructions.
(ppc64_elf_relocate_section): Don't sanity check TPREL16_HA.
ld/
* testsuite/ld-powerpc/tls32.d: Update for TPREL_HA/LO optimisation.
* testsuite/ld-powerpc/tlsexe32.d: Likewise.
* testsuite/ld-powerpc/tlsldopt32.d: Likewise.
* testsuite/ld-powerpc/tlsmark32.d: Likewise.
* testsuite/ld-powerpc/tlsopt4_32.d: Likewise.
* testsuite/ld-powerpc/tprel.s,
* testsuite/ld-powerpc/tprel.d,
* testsuite/ld-powerpc/tprel32.d: New tests.
* testsuite/ld-powerpc/tprelbad.s,
* testsuite/ld-powerpc/tprelbad.d: New test.
* testsuite/ld-powerpc/powerpc.exp: Run them.
gold/
* powerpc.cc (Target_powerpc): Add tprel_opt_ and accessors.
(Target_powerpc::Scan::local): Sanity check tprel high relocs.
(Target_powerpc::Scan::global): Likewise.
(Target_powerpc::Relocate::relocate): Control tprel optimisation
with tprel_opt_ and enable for 32-bit.

4 years agoRemove "memory exhausted" messages
Alan Modra [Mon, 24 Aug 2020 03:51:31 +0000 (13:21 +0930)]
Remove "memory exhausted" messages

Since we use xcalloc to set up hash table memory, htab_create won't
ever return a failure.

* config/tc-aarch64.c (md_begin): Don't bother checking for
out of memory failure from str_htab_create.
* config/tc-arc.c (arc_insert_opcode, md_begin): Likewise.
(arc_extcorereg, arc_stralloc): Likewise.
* config/tc-arm.c (md_begin): Likewise.
* config/tc-cr16.c (initialise_reg_hash_table, md_begin): Likewise.
* config/tc-cris.c (md_begin): Likewise.
* config/tc-crx.c (md_begin): Likewise.
* config/tc-pdp11.c (md_begin): Likewise.
* config/tc-score.c (s3_build_reg_hsh, s3_begin): Likewise.
* config/tc-score7.c (s7_build_reg_hsh, s7_begin): Likewise.

4 years agogcc-4 -Og false positive "may be used uninitialised"
Alan Modra [Mon, 24 Aug 2020 03:28:14 +0000 (12:58 +0930)]
gcc-4 -Og false positive "may be used uninitialised"

binutils/
* readelf.c (dump_section_as_strings) Avoid false positive
"may be used uninitialised".
gas/
* config/tc-arm.c (move_or_literal_pool): Avoid false positive
"may be used uninitialised".
(opcode_lookup): Likewise.

4 years agoPR26526, 5014c2d22b2 breaks compiling the Linux kernel for ARM
Alan Modra [Mon, 24 Aug 2020 03:36:46 +0000 (13:06 +0930)]
PR26526, 5014c2d22b2 breaks compiling the Linux kernel for ARM

PR 26526
* symbols.c (local_symbol_convert): Clear out xtra.

4 years agoCSKY: Add ck803r2 series cpu.
Cooper Qu [Fri, 21 Aug 2020 10:08:10 +0000 (18:08 +0800)]
CSKY: Add ck803r2 series cpu.

gas/
        * config/tc-csky.c (CSKY_ISA_803R2): New.
        (csky_archs): Add ck803r2 series.
        (md_begin): Fix warning about -medsp.
        (csky_get_freg_val): Support lowercase of fpu register name.
        * testsuite/gas/csky/cskyv2_ck803r2.s: New file.
        * testsuite/gas/csky/cskyv2_ck803r2.d: New file.

include/
        * csky.h (CSKYV2_ISA_3E3R2): New.

opcodes/
        * csky-opc.h (csky_v2_opcodes): Add instruction bnezad.

4 years agoAutomatic date update in version.in
GDB Administrator [Mon, 24 Aug 2020 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agoDisable default ELF plugins on 32-bit hppa*-*-hpux*.
John David Anglin [Sun, 23 Aug 2020 16:44:22 +0000 (16:44 +0000)]
Disable default ELF plugins on 32-bit hppa*-*-hpux*.

PR binutils/26357
* configure.ac: Disable plugins by default on 32-bit hppa*-*-hpux*.
* configure: Regenerate.

4 years agoPR26513, 629310abec breaks assembling PowerPC Linux kernels
Alan Modra [Sat, 22 Aug 2020 08:29:57 +0000 (17:59 +0930)]
PR26513, 629310abec breaks assembling PowerPC Linux kernels

Inserting with replacement is wrong for some gas hash table uses.
This patch implements an htab_insert that conditionally replaces, and
similarly for str_hash_insert.  str_hash_insert with replace=0 is
roughly equivalent to the older hash_insert, and str_hash_insert with
replace=1 to the older hash_jam, but return values are different.  I
found it useful to know whether the slot was occupied prior to
inserting/replacing.  I've also reinstated the fatal errors on messing
up opcode tables with duplicates.

PR 26513
* hash.h (htab_insert): Update prototype and comment.
(struct string_tuple): Make "value" a const void*.
(string_tuple_alloc): Likewise.
(str_hash_find, str_hash_find_n): Cast returned value.
(str_hash_insert): Add "replace" parameter, and return slot pointer.
Free alloc'd element when not inserted.
* hash.c (htab_insert): Likewise.  Return slot when element exists,
otherwise return NULL.
* read.c (pop_insert): Insert into hash table without first searching.
* config/tc-avr.c (md_begin): Likewise.
* config/tc-msp430.c (md_begin): Likewise.
* config/tc-nds32.c (nds32_init_nds32_pseudo_opcodes): Likewise.
* config/tc-v850.c (md_begin): Likewise.
* macro.c (do_formals, define_macro, macro_expand_body): Likewise.
(delete_macro): Delete from hash table.
* config/tc-tic54x.c (subsym_create_or_replace): Correct logic.

* symbols.c (local_symbol_make, symbol_table_insert): Allow
replacement of hash table entries.
* config/obj-coff-seh.c (seh_hash_insert): Likewise.
* config/obj-coff.c (tag_insert): Likewise.
* config/tc-iq2000.c (iq2000_add_macro): Likewise.
* config/tc-m68k.c (md_begin): Likewise for aliases.
* config/tc-tic4x.c (tic4x_asg): Likewise.
* config/tc-tic6x.c (md_begin): Likewise.

* dw2gencfi.c (dwcfi_hash_find_or_make): Disallow replacement of
hash table entries.
* ecoff.c (add_string, get_tag): Likewise.
* macro.c (expand_irp): Likewise.
* config/obj-elf.c (build_additional_section_info): Likewise.
* config/tc-aarch64.c (insert_reg_alias): Likewise.
(checked_hash_insert): Likewise.
* config/tc-alpha.c (get_alpha_reloc_tag, md_begin): Likewise.
* config/tc-arc.c (arc_insert_opcode, declare_register): Likewise.
(declare_addrtype, md_begin, arc_extcorereg): Likewise.
* config/tc-arm.c (insert_reg_alias): Likewise.
(arm_tc_equal_in_insn, md_begin): Likewise.
* config/tc-cr16.c (initialise_reg_hash_table, md_begin): Likewise.
* config/tc-cris.c (md_begin): Likewise.
* config/tc-crx.c (md_begin): Likewise.
* config/tc-csky.c (md_begin): Likewise.
* config/tc-d10v.c (md_begin): Likewise.
* config/tc-dlx.c (md_begin): Likewise.
* config/tc-ft32.c (md_begin): Likewise.
* config/tc-h8300.c (md_begin): Likewise.
* config/tc-hppa.c (md_begin): Likewise.
* config/tc-i386.c (md_begin): Likewise.
* config/tc-ia64.c (dot_rot, dot_entry, declare_register): Likewise.
(md_begin, dot_alias): Likewise.
* config/tc-m68hc11.c (md_begin): Likewise.
* config/tc-m68k.c (md_begin): Likewise.
* config/tc-mcore.c (md_begin): Likewise.
* config/tc-microblaze.c (md_begin): Likewise.
* config/tc-mips.c (md_begin): Likewise.
* config/tc-mmix.c (md_begin): Likewise.
* config/tc-mn10200.c (md_begin): Likewise.
* config/tc-mn10300.c (md_begin): Likewise.
* config/tc-moxie.c (md_begin): Likewise.
* config/tc-nds32.c (nds32_relax_hint, md_begin): Likewise.
* config/tc-nios2.c (md_begin): Likewise.
* config/tc-ns32k.c (md_begin): Likewise.
* config/tc-pdp11.c (md_begin): Likewise.
* config/tc-pj.c (fake_opcode, md_begin): Likewise.
* config/tc-ppc.c (ppc_setup_opcodes): Likewise.
* config/tc-pru.c (md_begin): Likewise.
* config/tc-riscv.c (init_ext_version_hash): Likewise.
(init_opcode_names_hash, hash_reg_name, init_opcode_hash): Likewise.
(riscv_init_csr_hash): Likewise.
* config/tc-s390.c (s390_setup_opcodes, md_begin): Likewise.
* config/tc-score.c (s3_insert_reg): Likewise.
(s3_build_score_ops_hsh, s3_build_dependency_insn_hsh): Likewise.
* config/tc-score7.c (s7_build_score_ops_hsh): Likewise.
(s7_build_dependency_insn_hsh, s7_insert_reg): Likewise.
* config/tc-sh.c (md_begin): Likewise.
* config/tc-sparc.c (md_begin): Likewise.
* config/tc-spu.c (md_begin): Likewise.
* config/tc-tic30.c (md_begin): Likewise.
* config/tc-tic4x.c (tic4x_inst_insert): Likewise.
* config/tc-tic54x.c (stag_add_field_symbols, md_begin): Likewise.
(tic54x_endstruct, tic54x_var, tic54x_macro_info): Likewise.
(subsym_substitute): Likewise.
* config/tc-tilegx.c (md_begin): Likewise.
* config/tc-tilepro.c (md_begin): Likewise.
* config/tc-vax.c (vip_begin): Likewise.
* config/tc-wasm32.c (md_begin): Likewise.
* config/tc-xgate.c (md_begin): Likewise.
* config/tc-z8k.c (md_begin): Likewise.
* testsuite/gas/ppc/dcbt.d,
* testsuite/gas/ppc/dcbt.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

* ecoff.c (add_string): Report fatal error on duplicates.
* config/tc-alpha.c (md_begin): Likewise.
* config/tc-arc.c (arc_insert_opcode, declare_register): Likewise.
(declare_addrtype, md_begin, arc_extcorereg): Likewise.
* config/tc-cr16.c (initialise_reg_hash_table, md_begin): Likewise.
* config/tc-cris.c (md_begin): Likewise.
* config/tc-crx.c (md_begin): Likewise.
* config/tc-dlx.c (md_begin): Likewise.
* config/tc-hppa.c (md_begin): Likewise.
* config/tc-i386.c (md_begin): Likewise.
* config/tc-ia64.c (dot_rot, dot_entry, declare_register): Likewise.
(md_begin): Likewise.
* config/tc-m68k.c (md_begin): Likewise.
* config/tc-mips.c (md_begin): Likewise.
* config/tc-nios2.c (md_begin): Likewise.
* config/tc-ns32k.c (md_begin): Likewise.
* config/tc-ppc.c (ppc_setup_opcodes): Likewise.
* config/tc-pru.c (md_begin): Likewise.
* config/tc-riscv.c (init_ext_version_hash): Likewise.
(init_opcode_names_hash, hash_reg_name, init_opcode_hash): Likewise.
* config/tc-s390.c (s390_setup_opcodes, md_begin): Likewise.
* config/tc-sparc.c (md_begin): Likewise.
* config/tc-tic30.c (md_begin): Likewise.
* config/tc-tic4x.c (tic4x_inst_insert): Likewise.
* config/tc-tilegx.c (md_begin): Likewise.
* config/tc-tilepro.c (md_begin): Likewise.
* config/tc-vax.c (vip_begin): Likewise.

* config/tc-alpha.c,
* config/tc-arm.c,
* config/tc-avr.c,
* config/tc-cr16.c,
* config/tc-csky.c,
* config/tc-i386.c,
* config/tc-m68hc11.c,
* config/tc-m68k.c,
* config/tc-microblaze.c,
* config/tc-ns32k.c,
* config/tc-pj.c,
* config/tc-ppc.c,
* config/tc-score.c,
* config/tc-score7.c,
* config/tc-tic4x.c,
* config/tc-tic54x.c,
* config/tc-tilegx.c,
* config/tc-tilepro.c,
* config/tc-xgate.c: Formatting.

4 years agoada-lex.l: Ignore register diagnostic also for g++ defaulting to ISO C++17
Mark Wielaard [Fri, 21 Aug 2020 22:20:55 +0000 (00:20 +0200)]
ada-lex.l: Ignore register diagnostic also for g++ defaulting to ISO C++17

Building with a really old flex and a really new g++ is probably not
recommended, but it should not cause compile errors.

gdb/ChangeLog:

* ada-lex.l: Extend register warnings diagnostics comment for g++.

include/ChangeLog:

* diagnostics.h (DIAGNOSTIC_IGNORE_DEPRECATED_REGISTER): Also define
for GCC version 7.0 or higher.

4 years agoAutomatic date update in version.in
GDB Administrator [Sun, 23 Aug 2020 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agoelf: Keep only one '@' for undefined versioned symbols
H.J. Lu [Sat, 22 Aug 2020 15:31:53 +0000 (08:31 -0700)]
elf: Keep only one '@' for undefined versioned symbols

The symbol string table in the .symtab section is optional and cosmetic.
Keep only one '@' for undefined versioned symbols, which are defined in
shared objects, in the symbol string table.  Update "nm -D" to display
only one '@' for undefined versioned symbols.

bfd/

PR ld/26382
* elflink.c (elf_link_output_symstrtab): Keep only one '@' for
versioned symbols, which are defined in shared objects, in
symbol string table.

binutils/

PR ld/26382
* nm.c (print_symname): Display only one '@' for undefined
versioned symbols.
* doc/binutils.texi: Update nm version information.

ld/

PR ld/26382
* testsuite/ld-elf/pr26302.nd: Updated.
* testsuite/ld-elf/pr26302.rd: New file.
* testsuite/ld-elf/shared.exp: Add a test for readelf -sW.

4 years agogdb: assert that we don't try to get a frame's id while it is computed
Simon Marchi [Sat, 22 Aug 2020 15:14:46 +0000 (11:14 -0400)]
gdb: assert that we don't try to get a frame's id while it is computed

I'm dealing these days with a class of bugs that involve trying to get a
certain frame's id while we are in the process of computing it.  In other
words, compute_frame_id being called for a frame, eventually calling
get_frame_id for that same frame.  I don't think this is ever supposed to
happen, as that creates a cyclic dependency.

Usually, these problems cause some failure down the line.  I'm proposing with
this patch to catch them as early as possible, as soon as the situation
described above happens.  I think that helps because the failed assertion will
be closer to the root of the problem.

To do so, the patch changes the frame_info::this_id::p flag from a boolean (is
the frame id computed or not) to a tri-state:

- the frame id is not computed
- the frame id is being computed
- the frame id is computed

Then, we can properly assert that get_frame_id doesn't get called for a frame
whose id is being computed.

To illustrate how that can help, let's imagine we apply the following change to
frame_unwind_got_optimized:

    --- a/gdb/frame-unwind.c
    +++ b/gdb/frame-unwind.c
    @@ -260,8 +260,7 @@ frame_unwind_got_optimized (struct frame_info *frame, int regnum)
       mark_value_bytes_optimized_out (val, 0, TYPE_LENGTH (type));
       VALUE_LVAL (val) = lval_register;
       VALUE_REGNUM (val) = regnum;
    -  VALUE_NEXT_FRAME_ID (val)
    -    = get_frame_id (get_next_frame_sentinel_okay (frame));
    +  VALUE_NEXT_FRAME_ID (val) = get_frame_id (frame);
       return val;
     }

... and run the following command, which leads to a failed assertion (you need
to run the corresponding test to generate the binary first):

    $ ./gdb -q -nx testsuite/outputs/gdb.dwarf2/dw2-undefined-ret-addr/dw2-undefined-ret-addr -ex "b stop_frame" -ex r

Without this patch applied, we catch the issue indirectly, when the top-level
get_frame_id tries to stash the frame:

    /home/smarchi/src/binutils-gdb/gdb/frame.c:593: internal-error: frame_id get_frame_id(frame_info*): Assertion `stashed' failed.

    ...
    #9  0x0000000001af1c3a in internal_error (file=0x1cea060 "/home/smarchi/src/binutils-gdb/gdb/frame.c", line=593, fmt=0x1ce9f80 "%s: Assertion `%s' failed.") at /home/smarchi/src/binutils-gdb/gdbsupport/errors.cc:55
    #10 0x0000000000e9b413 in get_frame_id (fi=0x6210005105e0) at /home/smarchi/src/binutils-gdb/gdb/frame.c:593
    #11 0x0000000000e99e35 in scoped_restore_selected_frame::scoped_restore_selected_frame (this=0x7fff1d8b9760) at /home/smarchi/src/binutils-gdb/gdb/frame.c:308
    #12 0x000000000149a261 in print_frame_args (fp_opts=..., func=0x6210000dd7d0, frame=0x6210005105e0, num=-1, stream=0x60300008a580) at /home/smarchi/src/binutils-gdb/gdb/stack.c:750
    #13 0x000000000149d938 in print_frame (fp_opts=..., frame=0x6210005105e0, print_level=0, print_what=SRC_AND_LOC, print_args=1, sal=...) at /home/smarchi/src/binutils-gdb/gdb/stack.c:1394
    #14 0x000000000149c0c8 in print_frame_info (fp_opts=..., frame=0x6210005105e0, print_level=0, print_what=SRC_AND_LOC, print_args=1, set_current_sal=1) at /home/smarchi/src/binutils-gdb/gdb/stack.c:1119
    #15 0x0000000001498100 in print_stack_frame (frame=0x6210005105e0, print_level=0, print_what=SRC_AND_LOC, set_current_sal=1) at /home/smarchi/src/binutils-gdb/gdb/stack.c:366
    #16 0x00000000010234b7 in print_stop_location (ws=0x7fff1d8ba1f0) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:8366
    #17 0x000000000102362d in print_stop_event (uiout=0x607000018660, displays=true) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:8382
    ...

It freaks out because the frame is already in the stash: it was added by an
inner call to get_frame_id, called indirectly by compute_frame_id.  Debugging
this failure is difficult because we have to backtrack to where this happened.

With the patch applied, we catch the issue earlier, here:

    /home/smarchi/src/binutils-gdb/gdb/frame.c:601: internal-error: frame_id get_frame_id(frame_info*): Assertion `fi->this_id.p != frame_id_status::COMPUTING' failed

    ...
    #9  0x0000000001af22bc in internal_error (file=0x1cea6e0 "/home/smarchi/src/binutils-gdb/gdb/frame.c", line=601, fmt=0x1cea600 "%s: Assertion `%s' failed.") at /home/smarchi/src/binutils-gdb/gdbsupport/errors.cc:55
    #10 0x0000000000e9b7e3 in get_frame_id (fi=0x62100050dde0) at /home/smarchi/src/binutils-gdb/gdb/frame.c:601
    #11 0x0000000000e989b3 in frame_unwind_got_optimized (frame=0x62100050dde0, regnum=16) at /home/smarchi/src/binutils-gdb/gdb/frame-unwind.c:264
    #12 0x0000000000cbe386 in dwarf2_frame_prev_register (this_frame=0x62100050dde0, this_cache=0x62100050ddf8, regnum=16) at /home/smarchi/src/binutils-gdb/gdb/dwarf2/frame.c:1267
    #13 0x0000000000e9f569 in frame_unwind_register_value (next_frame=0x62100050dde0, regnum=16) at /home/smarchi/src/binutils-gdb/gdb/frame.c:1266
    #14 0x0000000000e9eaab in frame_register_unwind (next_frame=0x62100050dde0, regnum=16, optimizedp=0x7ffca814ade0, unavailablep=0x7ffca814adf0, lvalp=0x7ffca814ae10, addrp=0x7ffca814ae20, realnump=0x7ffca814ae00, bufferp=0x7ffca814aec0 "") at /home/smarchi/src/binutils-gdb/gdb/frame.c:1169
    #15 0x0000000000e9f233 in frame_unwind_register (next_frame=0x62100050dde0, regnum=16, buf=0x7ffca814aec0 "") at /home/smarchi/src/binutils-gdb/gdb/frame.c:1225
    #16 0x0000000000f84262 in i386_unwind_pc (gdbarch=0x6210000eed10, next_frame=0x62100050dde0) at /home/smarchi/src/binutils-gdb/gdb/i386-tdep.c:1969
    #17 0x0000000000ec95dd in gdbarch_unwind_pc (gdbarch=0x6210000eed10, next_frame=0x62100050dde0) at /home/smarchi/src/binutils-gdb/gdb/gdbarch.c:3062
    #18 0x0000000000cb5e9d in dwarf2_tailcall_sniffer_first (this_frame=0x62100050dde0, tailcall_cachep=0x62100050dee0, entry_cfa_sp_offsetp=0x7ffca814b160) at /home/smarchi/src/binutils-gdb/gdb/dwarf2/frame-tailcall.c:387
    #19 0x0000000000cbdd38 in dwarf2_frame_cache (this_frame=0x62100050dde0, this_cache=0x62100050ddf8) at /home/smarchi/src/binutils-gdb/gdb/dwarf2/frame.c:1198
    #20 0x0000000000cbe026 in dwarf2_frame_this_id (this_frame=0x62100050dde0, this_cache=0x62100050ddf8, this_id=0x62100050de40) at /home/smarchi/src/binutils-gdb/gdb/dwarf2/frame.c:1226
    #21 0x0000000000e9b447 in compute_frame_id (fi=0x62100050dde0) at /home/smarchi/src/binutils-gdb/gdb/frame.c:580
    #22 0x0000000000e9b89e in get_frame_id (fi=0x62100050dde0) at /home/smarchi/src/binutils-gdb/gdb/frame.c:613
    #23 0x0000000000e99e35 in scoped_restore_selected_frame::scoped_restore_selected_frame (this=0x7ffca814b610) at /home/smarchi/src/binutils-gdb/gdb/frame.c:315
    #24 0x000000000149a8e3 in print_frame_args (fp_opts=..., func=0x6210000dd7d0, frame=0x62100050dde0, num=-1, stream=0x60300008a520) at /home/smarchi/src/binutils-gdb/gdb/stack.c:750
    #25 0x000000000149dfba in print_frame (fp_opts=..., frame=0x62100050dde0, print_level=0, print_what=SRC_AND_LOC, print_args=1, sal=...) at /home/smarchi/src/binutils-gdb/gdb/stack.c:1394
    #26 0x000000000149c74a in print_frame_info (fp_opts=..., frame=0x62100050dde0, print_level=0, print_what=SRC_AND_LOC, print_args=1, set_current_sal=1) at /home/smarchi/src/binutils-gdb/gdb/stack.c:1119
    #27 0x0000000001498782 in print_stack_frame (frame=0x62100050dde0, print_level=0, print_what=SRC_AND_LOC, set_current_sal=1) at /home/smarchi/src/binutils-gdb/gdb/stack.c:366
    #28 0x0000000001023b39 in print_stop_location (ws=0x7ffca814c0a0) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:8366
    #29 0x0000000001023caf in print_stop_event (uiout=0x607000018660, displays=true) at /home/smarchi/src/binutils-gdb/gdb/infrun.c:8382
    ...

Now, we can clearly see that get_frame_id for frame `fi=0x62100050dde0` gets
called while compute_frame_id is active for that frame.  The backtrace is more
helpful to identify the root of the problem.

gdb/ChangeLog:

* frame.c (enum class frame_id_status): New.
(struct frame_info) <this_id::p>: Change type to frame_id_status.
(fprintf_frame): Update.
(compute_frame_id): Set frame id status to "computing" on entry.
Set it back to "not_computed" on failure and to "computed" on
success.
(get_frame_id): Assert the frame id is not being computed.
(create_sentinel_frame): Use frame_id_status::COMPUTED.
(create_new_frame): Likewise.
(frame_cleanup_after_sniffer): Update assert.

Change-Id: I5f1a25fafe045f756bd75f358892720b30ed20c9

4 years agoAutomatic date update in version.in
GDB Administrator [Sat, 22 Aug 2020 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agosim/bfin: include config/pkg.m4 in configure.ac
Simon Marchi [Fri, 21 Aug 2020 15:55:08 +0000 (11:55 -0400)]
sim/bfin: include config/pkg.m4 in configure.ac

When trying to re-generate configure in sim/bfin, I get:

    $ autoreconf -vf
    autoreconf: Entering directory `.'
    autoreconf: configure.ac: not using Gettext
    autoreconf: running: aclocal --force
    autoreconf: configure.ac: tracing
    autoreconf: configure.ac: not running libtoolize: --install not given
    autoreconf: running: /opt/autostuff/bin/autoconf --force
    configure.ac:57: error: possibly undefined macro: AC_CHECK_LIB
          If this token and others are legitimate, please use m4_pattern_allow.
          See the Autoconf documentation.
    autoreconf: /opt/autostuff/bin/autoconf failed with exit status: 1

This happens since commit f693213d126a ("Run `autoreconf -vf` throughout").

The problem (not clear from the error message) is that the
PKG_CHECK_MODULES macro used in configure.ac is undefined.  In the past, I
suppose that it relied on the person running autoconf having pkg.m4 (from
pkg-config) in their include path.  That's not my case.

Since we recently added a local version of PKG_CHECK_MODULES to our tree,
we can just make sim/bfin/configure.ac use it.  This patch makes
configure.ac include config/pkg.m4, and re-generates configure.  With this,
the configure script appears to be generated correctly, I am able to
configure and build the bfin simulator.

Note: using sinclude to include the required m4 files makes no sense to
me.  These files contain macros we need, if they are not defined then
the resulting file is unusable.  And sinclude fails silently if the file
is not found.  So, better use include/m4_include.

sim/bfin/ChangeLog:

* configure.ac: Include config/pkg.m4.

Change-Id: I7d8012e5ed510cd7746b94e918f0feb1c701cd83

4 years agoFix problems with the AArch64 linker exposed by testing it with sanitization enabled.
Nick Clifton [Fri, 21 Aug 2020 14:32:20 +0000 (15:32 +0100)]
Fix problems with the AArch64 linker exposed by testing it with sanitization enabled.

bfd * elfnn-aarch64.c (_bfd_aarch64_erratum_835769_scan): Only sort
the data map if there are entries in it.
(_bfd_aarch64_erratum_843419_scan): Likewise.

opcodes * aarch64-dis.c (get_sym_code_type): Return FALSE for non-ELF
symbols.

4 years agoRecognize DW_LANG_C11 in readelf testsuite.
Mark Wielaard [Fri, 21 Aug 2020 11:26:07 +0000 (13:26 +0200)]
Recognize DW_LANG_C11 in readelf testsuite.

readelf_wi_test checks the DW_AT_LANGUAGE is recognized by matching
against ANSI C. But for the new DWARF5 DW_LANG_C11 readelf simply
prints "C11", not prefixed with ANSI (DW_LANG_C is "non-ANSI C",
DW_LANG_C89 is "ANSI C" and DW_LANG_C99 is "ANSI-C99"). Adjust the
testcase instead of readelf because C11 seems a fine name.

binutils/ChangeLog:

* testsuite/binutils-all/readelf.exp (readelf_wi_test): Also
recognize DW_LANG_C11.

4 years agoRedo gas local symbol support
Alan Modra [Thu, 20 Aug 2020 06:06:46 +0000 (15:36 +0930)]
Redo gas local symbol support

gas handles local symbols specially in order to save memory, but the
implementation using two separate hash tables is inefficient,
particularly the scheme of duplicating a struct local_symbol when it
needs to be converted to a full struct symbol.  Also, updating symbol
pointers with LOCAL_SYMBOL_CHECK is horrible and has led to some hard
to find bugs.

This changes the implementation to use a single hash table and avoids
another copy of the symbol name in symbol_entry_t.  When converting
local symbols the struct local_symbol memory is reused.  Not only
does that save memory, but there is no need to twiddle symbol pointers
with LOCAL_SYMBOL_CHECK.

Assembling gcc-10 -g -Og gold/powerpc.cc output shows the following:

old:
symbol table hash statistics:
1371192 searches
1290398 collisions
143585 elements
262139 table size
mini local symbol table hash statistics:
2966204 searches
2707489 collisions
523533 elements
1048573 table size
523533 mini local symbols created, 140453 converted

new:
symbol table hash statistics:
2828883 searches
2453138 collisions
526665 elements
1048573 table size
523533 mini local symbols created, 140453 converted

* symbols.c (struct local_symbol): Add "hash" entry.  Reorder fields.
Delete union.  Adjust code throughout file.
(struct symbol): Add "hash", "name" and "x" entries.  Reorder fields.
Split off some to..
(struct xsymbol): ..this.  New struct.  Adjust code throughout file
accessing these fields.
(struct symbol_entry): Delete.
(union symbol_entry): New.
(hash_symbol_entry): Adjust for symbol_entry_t change.
(symbol_entry_find): Likewise.
(eq_symbol_entry): Compare hash values too.
(symbol_entry_alloc): Delete.
(local_symbol_converted_p, local_symbol_mark_converted): Delete.
(local_symbol_get_real_symbol, local_symbol_set_real_symbol): Delete.
(local_hash): Delete.
(abs_symbol_x, dot_symbol_x): New static var.
(symbol_init): New function.
(symbol_create): Rewrite.
(LOCAL_SYMBOL_CHECK): Delete.  Replace uses throughout with simple
test of flags.local_symbol.
(local_symbol_make): Adjust for struct local_symbol changes.
(local_symbol_convert): Rewrite.  Adjust all callers.
(symbol_table_insert): Simplify.
(symbol_clone): Comment on local sym cloning.  Handle split symbol
struct.
(get_real_sym): Delete.  Remove all uses.
(symbol_find_exact_noref): Simplify.
(resolve_local_symbol): Don't resolve non-locals.
(S_SET_SEGMENT): Don't special case reg_section.
(S_SET_NAME): Set both name and bsym->name.
(symbol_mark_resolved, symbol_resolved_p): Simplify.
(symbol_symbolS): Update comment.
(symbol_begin): Don't create local_hash.  Adjust abs_symbol setup.
(dot_symbol_init): Adjust dot_symbol setup.
(symbol_print_statistics): Delete local_hash stats.

4 years agogas symbol struct field renaming
Alan Modra [Fri, 21 Aug 2020 03:31:37 +0000 (13:01 +0930)]
gas symbol struct field renaming

Get rid of sy_ prefix, and some unused fields.

* symbols.c (struct symbol_flags): Rename sy_volatile to volatil,
and remove sy_ from other field names.  Update throughout.
(struct symbol): Remove sy_ from field names.  Delete unused
TARGET_SYMBOL_FIELDS.  Update throughout file.  Move after..
(struct local_symbol): ..here.  Remove lsy_ from field names.
Delete unused TC_LOCAL_SYMFIELD_TYPE.  Update throughout file.
(local_symbol_resolved_p, local_symbol_mark_resolved): Delete.
Expand uses throughout file.
(local_symbol_get_frag, local_symbol_set_frag): Likewise.
(symbol_new): Move symbol_table_frozen test to..
(symbol_append): ..here, and..
(symbol_insert): ..here.
(resolve_symbol_value, symbol_relc_make_expr): White space fixes.
(HANDLE_XADD_OPT1, HANDLE_XADD_OPT2): Likewise.
* config/obj-coff.h (RESOLVE_SYMBOL_REDEFINITION): Update.

4 years agoRearrange symbol_create parameters
Alan Modra [Thu, 20 Aug 2020 23:47:53 +0000 (09:17 +0930)]
Rearrange symbol_create parameters

These functions take an offset within frag, frag within section, and
section parameter.  So it makes sense to order the parameters as
section, frag, offset.

* symbols.h (symbol_new, symbol_create, local_symbol_make),
(symbol_temp_new): Arrange params as section, frag, offset.
* symbols.c: Adjust to suit.
* as.c: Likewise.
* cgen.c: Likewise.
* dwarf2dbg.c: Likewise.
* ecoff.c: Likewise.
* expr.c: Likewise.
* itbl-ops.c: Likewise.
* read.c: Likewise.
* stabs.c: Likewise.
* subsegs.c: Likewise.
* config/obj-coff.c: Likewise.
* config/obj-elf.c: Likewise.
* config/obj-macho.c: Likewise.
* config/tc-aarch64.c: Likewise.
* config/tc-alpha.c: Likewise.
* config/tc-arc.c: Likewise.
* config/tc-arm.c: Likewise.
* config/tc-avr.c: Likewise.
* config/tc-cr16.c: Likewise.
* config/tc-cris.c: Likewise.
* config/tc-csky.c: Likewise.
* config/tc-dlx.c: Likewise.
* config/tc-hppa.c: Likewise.
* config/tc-i386.c: Likewise.
* config/tc-ia64.c: Likewise.
* config/tc-m32r.c: Likewise.
* config/tc-m68k.c: Likewise.
* config/tc-mips.c: Likewise.
* config/tc-mmix.c: Likewise.
* config/tc-mn10200.c: Likewise.
* config/tc-mn10300.c: Likewise.
* config/tc-nds32.c: Likewise.
* config/tc-nios2.c: Likewise.
* config/tc-ppc.c: Likewise.
* config/tc-riscv.c: Likewise.
* config/tc-s390.c: Likewise.
* config/tc-sh.c: Likewise.
* config/tc-tic4x.c: Likewise.
* config/tc-tic54x.c: Likewise.
* config/tc-xtensa.c: Likewise.

4 years agoCSKY: Support two operands form for bloop.
Cooper Qu [Fri, 21 Aug 2020 08:17:36 +0000 (16:17 +0800)]
CSKY: Support two operands form for bloop.

gas/
        * config/tc-csky.c (csky_insn_info): Add member last_isize.
        (md_assemble): Assign value to csky_insn.last_isize.
        * testsuite/gas/csky/enhance_dsp.d: Test bloop's two operands form.
        * testsuite/gas/csky/enhance_dsp.s: Likewise.

opcodes/
        * csky-opc.h (csky_v2_opcodes): Add two operands form for bloop.

4 years agofix objcopy of PE images with .buildid section
Jan Beulich [Fri, 21 Aug 2020 08:28:35 +0000 (10:28 +0200)]
fix objcopy of PE images with .buildid section

Xen Project embeds a build ID in its hypervisor binary (including its
EFI variant), living in a standalone section. This usually gets placed
right after .rodata, and due to the rounding done on the (file) size of
.rodata the two sections appear to overlap (as far as e.g.
find_section_by_vma() is concerned). With the first byte "found" in
.rodata, nothing guarantees that the entire debug dir fits in that
section, leading to apparently random failure of objcopy on such an
image.

Possible alternatives to the solution chosen:
- make find_section_by_vma() honor virt_size,
- correct the recording of sizes elsewhere (ibfd has size == virt_size,
  while obfd doesn't),
- fix the linker to avoid producing apparently overlapping sections.

While touching the condition around and the contents of the disgnostic,
pull it up ahead of the bfd_malloc_and_get_section() call: There's no
point first obtaining the section contents, in order to then fail.

4 years agoAutomatic date update in version.in
GDB Administrator [Fri, 21 Aug 2020 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

4 years agogdb: handle the `ptid.is_pid ()` case in registers_changed_ptid
Simon Marchi [Thu, 20 Aug 2020 14:11:09 +0000 (10:11 -0400)]
gdb: handle the `ptid.is_pid ()` case in registers_changed_ptid

As reported by Tom here [1], commit 888bdb2b7445 ("gdb: change regcache
list to be a map") overlooked an important case, causing a regression.
When registers_changed_ptid is called with a pid-like ptid, it used to
clear all the regcaches for that pid.  That commit accidentally removed
that behavior.  We need to handle the `ptid.is_pid ()` case in
registers_changed_ptid.

The most trivial way of fixing it would be to iterate on all ptids of a
target and delete the entries where the ptid match the pid.  But the
point of that commit was to avoid having to iterate on ptids to
invalidate regcaches, so that would feel like a step backwards.

The only logical solution I see is to add yet another map level, so that
we now have:

  target -> (pid -> (ptid -> regcaches))

This patch implements that and adds a test for the case of calling
registers_changed_ptid with a pid-like ptid.

[1] https://sourceware.org/pipermail/gdb-patches/2020-August/171222.html

gdb/ChangeLog:

* regcache.c (pid_ptid_regcache_map): New type.
(target_ptid_regcache_map): Remove.
(target_pid_ptid_regcache_map): New type.
(regcaches): Change type to target_pid_ptid_regcache_map.
(get_thread_arch_aspace_regcache): Update.
(regcache_thread_ptid_changed): Update, handle pid-like ptid
case.
(regcaches_size): Update.
(regcache_count): Update.
(registers_changed_ptid_target_pid_test): New.
(_initialize_regcache): Register new test.

Change-Id: I4c46e26d8225c177dbac9488b549eff4c68fa0d8

4 years agogdb: split regcaches management selftest
Simon Marchi [Thu, 20 Aug 2020 14:10:59 +0000 (10:10 -0400)]
gdb: split regcaches management selftest

The selftest `regcaches` selftest is a bit too broad for my taste,
testing the behavior of get_thread_arch_aspace_regcache and various
cases of registers_changed_ptid.  Since I'll want to test even more
scenarios of registers_changed_ptid, passing different sets of
parameters, it will be difficult to do in a single test case.  It is
difficult to change something at some point in the test case while make
sure it doesn't compromise what comes after, that we still test the
scenarios that we intended to test.  So, split the test case in multiple
smaller ones.

- Split the test case in multiple, where each test case starts from
  scratch and tests one specific scenario.

- Introduce the populate_regcaches_for_test function, which is used by
  the various test cases to start with a regcache container populated
  with a few regcaches for two different targets.

- populate_regcaches_for_test returns a regcache_test_data object, which
  contains the test targets that were used to create the regcaches.  It
  also takes care to call registers_changed at the beginning and end of
  the test to ensure the test isn't influenced by existing regcaches,
  and cleans up after itself.

- Move the regcache_count lambda function out of
  regcache_thread_ptid_changed, so it can be used in
  other tests.

- For get_thread_arch_aspace_regcache, test that getting a regcache that
  already exists does not increase the count of existing regcaches.

- For registers_changed_ptid, test the three cases we handle today:
  (nullptr, minus_one_ptid), (target, minus_one_ptid) and (target,
  ptid).  The (target, minus_one_ptid) case was not tested prior to this
  patch.

gdb/ChangeLog:

* regcache.c (regcache_count): New.
(struct regcache_test_data): New.
(regcache_test_data_up): New.
(populate_regcaches_for_test): New.
(regcaches_test): Remove.
(get_thread_arch_aspace_regcache_test): New.
(registers_changed_ptid_all_test): New.
(registers_changed_ptid_target_test): New.
(registers_changed_ptid_target_ptid_test): New.
(regcache_thread_ptid_changed): Remove regcache_count lambda.
(_initialize_regcache): Register new tests.

Change-Id: Id4280879fb40ff3aeae49b01b95359e1359c3d4b

4 years agogdb: refactor test_get_thread_arch_aspace_regcache
Simon Marchi [Thu, 20 Aug 2020 14:10:47 +0000 (10:10 -0400)]
gdb: refactor test_get_thread_arch_aspace_regcache

Do these misc changes to test_get_thread_arch_aspace_regcache:

- Rename to get_thread_arch_aspace_regcache_and_check.  The following
  patch introduces a selftest for get_thread_arch_aspace_regcache, named
  get_thread_arch_aspace_regcache_test.  To avoid confusion between the
  two functions, rename this one to
  get_thread_arch_aspace_regcache_and_check, I think it describes better
  what it does.

- Remove gdbarch parameter.  We always pass the same gdbarch (the
  current inferior's gdbarch), so having a parameter is not useful.  It
  would be interesting to actually test with multiple gdbarches, to
  verify that the regcache container can hold multiple regcaches (with
  different architectures) for a same (target, ptid).  But it's not the
  case as of this patch.

- Verify that the regcache's arch is correctly set.

- Remove the aspace parameter.  We always pass NULL here, so it's not
  useful to have it as a parameter.  Also, instead of passing a NULL
  aspace to get_thread_arch_aspace_regcache and verifying that we get a
  NULL aspace back, pass the current inferior's aspace (just like we use
  the current inferior's gdbarch).

gdb/ChangeLog:

* regcache.c (test_get_thread_arch_aspace_regcache): Rename to...
(get_thread_arch_aspace_regcache_and_check): ... this.  Remove
gdbarch and aspace parameter.  Use current inferior's aspace.
Validate regcache's arch value.
(regcaches_test): Update.

Change-Id: I8b4c2303b4f91f062269043d1f7abe1650232010

4 years agogdb: clear regcaches at the start of regcaches selftest
Simon Marchi [Thu, 20 Aug 2020 14:07:21 +0000 (10:07 -0400)]
gdb: clear regcaches at the start of regcaches selftest

It currently does not work to run the `regcaches` selftest while
debugging something.  This is because we expect that there exists no
regcache at the start of the test.  If we are debugging something, there
might exist some regcaches.

Fix it by making the test clear regcaches at the start.

While at it, make the test clean up after it self and clear the
regcaches at the end too.

gdb/ChangeLog:

* regcache.c (regcaches_test): Call registers_changed.

Change-Id: I9d4f83ecb0ff9721a71e2c5cbd19e6e6d4e6c30c

4 years agoEnsure that compressed sections that have an ELF compression header structure at...
Nick Clifton [Thu, 20 Aug 2020 14:03:21 +0000 (15:03 +0100)]
Ensure that compressed sections that have an ELF compression header structure at the start are correctly aligned.

PR 26428
bfd * bfd.c (bfd_update_compression_header): Also set the sh_addralign
field in the ELF header of the compressed sections.

ld * testsuite/ld-elf/zlibbegin.rS: Update expected output.
* testsuite/ld-elf/zlibnormal.rS: Likewise.

4 years agogdb: fix typo "breapoint" -> "breakpoint"
Tankut Baris Aktemur [Thu, 20 Aug 2020 12:23:34 +0000 (14:23 +0200)]
gdb: fix typo "breapoint" -> "breakpoint"

gdb/ChangeLog:
2020-08-20  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* infrun.c (process_event_stop_test): Fix typo "breapoint".

gdb/testsuite/ChangeLog:
2020-08-20  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* gdb.base/print-file-var.exp: Fix typo "breapoint".
* gdb.trace/strace.exp: Ditto.

4 years agoApply a workaround to mitigate a quadratic performance hit in the linker when writing...
Nick Clifton [Thu, 20 Aug 2020 09:19:47 +0000 (10:19 +0100)]
Apply a workaround to mitigate a quadratic performance hit in the linker when writing out secondary reloc sections.

PR 26406
* elf-bfd.h (struct bfd_elf_section_data): Add
has_secondary_relocs field.
* elf.c (_bfd_elf_copy_special_section_fields): Set the
has_secondary_relocs field for sections which have associated
secondary relocs.
* elfcode.h (elf_write_relocs): Only call write_secondary_relocs
on sections which have associated secondary relocs.

4 years agoFix for incorrect breakpoint set in case of flang compiled binary
Alok Kumar Sharma [Thu, 20 Aug 2020 05:05:27 +0000 (10:35 +0530)]
Fix for incorrect breakpoint set in case of flang compiled binary

  Currently, GDB is not able to set a breakpoint at subprogram post
prologue for flang generated binaries. This is due to clang having
two line notes one before and another after the prologue.
  Now the end of prologue is determined using symbol table, which was
the way for clang generated binaries already. Since clang and flang
both share same back-end it is true for flang as well.

gdb/ChangeLog

      * amd64-tdep.c (amd64_skip_prologue): Using symbol table
      to find the end of prologue for flang compiled binaries.
      * arm-tdep.c (arm_skip_prologue): Likewise.
      * i386-tdep.c (i386_skip_prologue): Likewise.
      * producer.c (producer_is_llvm): New function.
      (producer_parsing_tests): Added new tests for clang/flang.
      * producer.h (producer_is_llvm): New declaration.

gdb/testsuite/ChangeLog

      * gdb.fortran/vla-type.exp: Skip commands not required for
      the Flang compiled binaries after prologue fix.