binutils-gdb.git
5 years agotestsuite: Add option to capture gdbserver debug
Alan Hayward [Fri, 12 Apr 2019 15:36:51 +0000 (16:36 +0100)]
testsuite: Add option to capture gdbserver debug

Add both board option and environment variable which enables gdbserver
debug and sends it to the file gdbserver.debug, located in the output
directory for the current test.  Document this.

Add support for the environment variable in the Makefile.

The testsuite can be run with gdbserver debug enabled in the following way:

make check GDBSERVER_DEBUG=all

Disable tspeed.exp when debugging to prevent the log file filling
many gigabytes then timing out.

gdb/testsuite/ChangeLog:

* Makefile.in: Pass through GDBSERVER_DEBUG.
        * README (Testsuite Parameters): Add GDBSERVER_DEBUG.
        (gdbserver,debug): Add board setting.
        * gdb.trace/tspeed.exp: Skip when debugging.
* lib/gdb.exp (gdbserver_debug_enabled): New procedure.
* lib/gdbserver-support.exp: Likewise

5 years agoLTO: Properly handle wrapper symbols in IR
H.J. Lu [Thu, 25 Apr 2019 14:53:46 +0000 (07:53 -0700)]
LTO: Properly handle wrapper symbols in IR

When a wrapper symbol, __wrap_FOO, is defined in IR, its resolution
should be LDPR_PREVAILING_DEF, not PREVAILING_DEF_IRONLY, since LTO
doesn't know that __wrap_FOO provides definition of FOO.  And resolution
of FOO should be LDPR_RESOLVED_IR since it is resolved by __wrap_FOO in
IR.

PR ld/24406
* ld.texi: Remove LTO warning from --wrap.
* plugin.c (get_symbols): Update resolution for wrapper and
wrapped symbols.
* testsuite/ld-plugin/lto.exp: Run ld/24406 tests.
* testsuite/ld-plugin/pr24406-1.c: New file.
* testsuite/ld-plugin/pr24406-2a.c: Likewise.
* testsuite/ld-plugin/pr24406-2b.c: Likewise.

5 years agoDetect invalid length field in debug frame FDE header.
Sandra Loosemore [Thu, 25 Apr 2019 14:27:02 +0000 (07:27 -0700)]
Detect invalid length field in debug frame FDE header.

GDB was failing to catch cases where a corrupt ELF or core file
contained an invalid length value in a Dwarf debug frame FDE header.
It was checking for buffer overflow but not cases where the length was
negative or caused pointer wrap-around.

In addition to the additional validity check, this patch cleans up the
multiple signed/unsigned conversions on the length field so that an
unsigned representation is used consistently throughout.

This patch fixes CVE-2017-9778 and PR gdb/21600.

2019-04-25  Sandra Loosemore  <sandra@codesourcery.com>
    Kang Li <kanglictf@gmail.com>

PR gdb/21600

* dwarf2-frame.c (read_initial_length): Be consistent about using
unsigned representation of length.
(decode_frame_entry_1): Likewise.  Check for wraparound of
end pointer as well as buffer overflow.

5 years ago[BFD, AArch64] Improve bti/pac plts.
Sudakshina Das [Thu, 25 Apr 2019 10:37:25 +0000 (11:37 +0100)]
[BFD, AArch64] Improve bti/pac plts.

This patch aims to improve the definitions of BTI and PAC based PLTs.
The following changes are made:
   * PLT0 does not need PAC instructions since the PLTGOT[2] (and PLTGOT[1])
     are readonly so they cannot be corrupted at runtime. Thus both PAC plt0
     and BTI+PAC plt0 are removed and we can use basic plt0 and BTI plt0
     instead, respectively.
   * We can remove the extra padding nops when we add the new bti instructions.
     BTI plt0 and BTI TLSDESC plt are updated.
   * For better performance PLTn could be padded to 24bytes. Both BTI pltn and
     PAC pltn are updated.

*** bfd/ChangeLog ***

2019-04-25  Sudakshina Das  <sudi.das@arm.com>

* elfnn-aarch64.c (PLT_BTI_ENTRY_SIZE): Remove.
(PLT_BTI_TLSDESC_ENTRY_SIZE): Remove.
(PLT_PAC_ENTRY_SIZE, PLT_BTI_PAC_ENTRY_SIZE): Remove.
(PLT_BTI_SMALL_ENTRY_SIZE, PLT_PAC_SMALL_ENTRY_SIZE): Update.
(elfNN_aarch64_small_plt0_pac_entry): Remove.
(elfNN_aarch64_small_plt0_bti_pac_entry): Remove.
(elfNN_aarch64_small_plt0_bti_entry): Update.
(elfNN_aarch64_small_plt_bti_entry): Update.
(elfNN_aarch64_small_plt_pac_entry): Update.
(elfNN_aarch64_tlsdesc_small_plt_bti_entry): Update.
(setup_plt_values): Setup new entries.
(elfNN_aarch64_finish_dynamic_sections): Remove size change.
(elfNN_aarch64_plt_sym_val): Likewise.

*** ld/ChangeLog ***

2019-04-25  Sudakshina Das  <sudi.das@arm.com>

* testsuite/ld-aarch64/bti-pac-plt-1.d: Update.
* testsuite/ld-aarch64/bti-pac-plt-2.d: Update.
* testsuite/ld-aarch64/bti-plt-1.d: Update.
* testsuite/ld-aarch64/bti-plt-3.d: Update.
* testsuite/ld-aarch64/bti-plt-5.d: Update.
* testsuite/ld-aarch64/pac-plt-1.d: Update.
* testsuite/ld-aarch64/pac-plt-2.d: Update.

5 years agoMIPS/include: opcode/mips.h: Update stale comment for CODE20 operand
Maciej W. Rozycki [Thu, 25 Apr 2019 00:28:49 +0000 (01:28 +0100)]
MIPS/include: opcode/mips.h: Update stale comment for CODE20 operand

Complement commit 1586d91e32ea ("/ 0 should send SIGFPE not SIGTRAP..."),
<https://sourceware.org/ml/binutils/2004-07/msg00260.html>, and update a
stale comment referring the 20-bit code field of the BREAK and SDBBP
instructions, by making it explicit that where permitted by choosing the
MIPS32 or a later ISA the whole field can now be set with a single
operand for the SDBBP instruction only.

include/
* opcode/mips.h: Update comment for MIPS32 CODE20 operand.

5 years agoAutomatic date update in version.in
GDB Administrator [Thu, 25 Apr 2019 00:00:15 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agoSpeed up locview resolution with relaxable frags
Alexandre Oliva [Sat, 13 Apr 2019 08:55:34 +0000 (05:55 -0300)]
Speed up locview resolution with relaxable frags

Targets such as xtensa incur a much higher overhead to resolve
location view numbers than e.g. x86, because the expressions used to
compute view numbers cannot be resolved soon enough.

Each view number is computed by incrementing the previous view, if
they are both at the same address, or by resetting it to zero
otherwise.  If PV is the previous view number, PL is its location, and
NL is the location of the next view, its number is computed by
evaluating NV = !(NL > PL) * (PV + 1).

set_or_check_view uses resolve_expression to decide whether portions
of this expression can be simplified to constants.  The (NL > PL)
subexpression is one that can often be resolved to a constant,
breaking chains of view number computations at instructions of nonzero
length, but not after alignment that might be unnecessary.

Alas, when nearly every frag ends with a relaxable instruction,
frag_offset_fixed_p will correctly fail to determine a known offset
between two unresolved addresses in neighboring frags, so the
unresolved symbolic operation will be constructed and used in the
computation of most view numbers.  This results in very deep
expressions.

As view numbers get referenced in location view lists, each operand in
the list goes through symbol_clone_if_forward_ref, which recurses on
every subexpression.  If each view number were to be referenced, this
would exhibit O(n^2) behavior, where n is the depth of the view number
expressions, i.e., the length of view number sequences without an
early resolution that cuts the expression short.

This patch enables address compares used by view numbering to be
resolved even when exact offsets are not known, using new logic to
determine when the location either remained the same or changed for
sure, even with the possibility of relaxation.  This enables most view
number expressions to be resolved with a small, reasonable depth.

PR gas/24444
* frags.c (frag_gtoffset_p): New.
* frags.h (frag_gtoffset_p): Declare it.
* expr.c (resolve_expression): Use it.

5 years agoFix Rust testing
Tom Tromey [Wed, 24 Apr 2019 19:43:27 +0000 (13:43 -0600)]
Fix Rust testing

This changes the gdb test suite to omit -fno-stack-protector when
compiling Rust code.  This makes Rust testing work again.

I think I saw this patch somewhere already, but I couldn't find it
again just now, so I'm checking this version in.

gdb/testsuite/ChangeLog
2019-04-24  Tom Tromey  <tromey@adacore.com>

* lib/gdb.exp (gdb_compile): Don't add -fno-stack-protector for
Rust.

5 years agoUse better test for usable compiler in ld testsuite.
Sandra Loosemore [Wed, 24 Apr 2019 19:14:56 +0000 (12:14 -0700)]
Use better test for usable compiler in ld testsuite.

The ld testsuite includes numerous tests that depend on being able to
compile and link programs with the C compiler.  Some of these tests
use [which $CC] to check for the presence of the compiler before
proceeding with the test, but run_ld_link_exec_tests and run_cc_link_tests
give ERRORs if compilation fails.  Also, even if $CC is defined and present,
it may not be usable due to missing libraries, etc.

This patch adds a new procedure check_compiler_available that attempts
to build an empty program and caches the result.  Uses of [which $CC]
are replaced with calls to this procedure, and run_ld_link_exec_tests
and run_cc_link_tests now also guard attempts to use $CC.

2019-04-24  Sandra Loosemore  <sandra@codesourcery.com>

ld/
* testsuite/config/default.exp: Use [check_compiler_available]
instead of [which $CC].
* testsuite/ld-auto-import/auto-import.exp: Likewise.
* testsuite/ld-cygwin/exe-export.exp: Likewise.
* testsuite/ld-elf/audit.exp: Likewise.
* testsuite/ld-elf/compress.exp: Likewise.
* testsuite/ld-elf/dwarf.exp: Likewise.
* testsuite/ld-elf/elf.exp: Likewise.
* testsuite/ld-elf/indirect.exp: Likewise.
* testsuite/ld-elf/linux-x86.exp: Likewise.
* testsuite/ld-elf/shared.exp: Likewise.
* testsuite/ld-elf/tls.exp: Likewise.
* testsuite/ld-elf/wrap.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-gc/gc.exp: Likewise.
* testsuite/ld-i386/i386.exp: Likewise.
* testsuite/ld-i386/no-plt.exp: Likewise.
* testsuite/ld-i386/tls.exp: Likewise.
* testsuite/ld-ifunc/ifunc.exp: Likewise.
* testsuite/ld-mn10300/mn10300.exp: Likewise.
* testsuite/ld-pe/pe-compile.exp: Likewise.
* testsuite/ld-pe/pe-run.exp: Likewise.
* testsuite/ld-pe/pe-run2.exp: Likewise.
* testsuite/ld-pie/pie.exp: Likewise.
* testsuite/ld-plugin/lto.exp: Likewise.
* testsuite/ld-plugin/plugin.exp: Likewise.
* testsuite/ld-scripts/crossref.exp: Likewise.
* testsuite/ld-sh/sh.exp: Likewise.
* testsuite/ld-shared/shared.exp: Likewise.
* testsuite/ld-size/size.exp: Likewise.
* testsuite/ld-srec/srec.exp: Likewise.
* testsuite/ld-undefined/undefined.exp: Likewise.
* testsuite/ld-unique/unique.exp: Likewise.
* testsuite/ld-x86-64/mpx.exp: Likewise.
* testsuite/ld-x86-64/no-plt.exp: Likewise.
* testsuite/ld-x86-64/tls.exp: Likewise.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/lib/ld-lib.exp (run_ld_link_exec_tests): Call
check_compiler_available before trying to use the compiler.
(run_cc_link_tests): Likewise.
(check_compiler_available): New.  Use it instead of [which $CC].

5 years agoUse "pulongest" on aarch64-tdep.c:aarch64_gdbarch_init
Sergio Durigan Junior [Wed, 24 Apr 2019 18:53:59 +0000 (14:53 -0400)]
Use "pulongest" on aarch64-tdep.c:aarch64_gdbarch_init

While trying to build GDB on i686, I found the following error:

 In file included from ../../gdb/common/common-defs.h:105,
                  from ../../gdb/defs.h:28,
                  from ../../gdb/aarch64-tdep.c:21:
 ../../gdb/aarch64-tdep.c: In function 'gdbarch* aarch64_gdbarch_init(gdbarch_info, gdbarch_list*)':
 ../../gdb/aarch64-tdep.c:3176:43: error: format '%ld' expects argument of type 'long int', but argument 4 has type 'uint64_t' {aka 'long long unsigned int'} [-Werror=format=]
  3176 |     internal_error (__FILE__, __LINE__, _("VQ out of bounds: %ld (max %d)"),
       |                                           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 ../../gdb/common/gdb_locale.h:28:29: note: in definition of macro '_'
    28 | # define _(String) gettext (String)
       |                             ^~~~~~
 ../../gdb/aarch64-tdep.c:3176:64: note: format string is defined here
  3176 |     internal_error (__FILE__, __LINE__, _("VQ out of bounds: %ld (max %d)"),
       |                                                              ~~^
       |                                                                |
       |                                                                long int
       |                                                              %lld

This happens because aarch64-tdep.c:aarch64_gdbarch_init prints a
"uint64_t" variable using "%ld".  This patch fixes the build by using
"pulongest" instead.  As explained in a similar fix (commit
495143533ad95369811391c6e3c6dadd69d7dd67), this should be safe because
if aarch64-tdep.c is included in the build, then ULONGEST must be a
64-bit type.

gdb/ChangeLog:
2019-04-24  Sergio Durigan Junior  <sergiodj@redhat.com>

* aarch64-tdep.c (aarch64_gdbarch_init): Use "pulongest" to print
"vq".

5 years agoFix passing of struct with bitfields on x86-64
Tom Tromey [Tue, 16 Apr 2019 17:11:10 +0000 (11:11 -0600)]
Fix passing of struct with bitfields on x86-64

Commit 4aa866af ("Fix AMD64 return value ABI in expression
evaluation") introduced a regression when calling a function with a
structure that contains bitfields.

Because the caller of amd64_has_unaligned_fields handles bitfields
already, it seemed to me that the simplest fix was to ignore bitfields
here.

gdb/ChangeLog
2019-04-24  Tom Tromey  <tromey@adacore.com>

* amd64-tdep.c (amd64_has_unaligned_fields): Ignore bitfields.

gdb/testsuite/ChangeLog
2019-04-24  Tom Tromey  <tromey@adacore.com>

* gdb.arch/amd64-eval.exp: Test bitfield return.
* gdb.arch/amd64-eval.cc (struct Bitfields): New.
(class Foo) <return_bitfields>: New method.
(main): Call it.

5 years agoStop strip from merging notes when stripping debug or dwo information.
Nick Clifton [Wed, 24 Apr 2019 16:44:31 +0000 (17:44 +0100)]
Stop strip from merging notes when stripping debug or dwo information.

* objcopy.c (strip_main): Do not enable note merging by default if
just stripping debug or dwo information.
* doc/binutils.texi (strip): Update documentation.

5 years agoresolve_symbol_value vs. .loc view resolution
Alan Modra [Wed, 24 Apr 2019 01:56:51 +0000 (11:26 +0930)]
resolve_symbol_value vs. .loc view resolution

In most cases we don't want expression symbols, such as that created
for an expression like "symbol + (1f - .)", resolved down to a
constant.  Instead we'd like to leave the expression as "symbol +
constant" once the "1f - ." part has been resolved, and let the
backend decide whether "symbol" can be reduced further.

However, that doesn't work when trying to resolve .loc view symbols
early.  We get expression symbols left as an O_symbol expression
pointing at an absolute symbol, and marked as sy_flags.sy_resolved.
That wouldn't really be a problem, but when one of those expression
symbols is used in further .loc view expressions, its value is taken
as zero.

This patch fixes the symbol value mistake, and stops creation of
O_symbol expression symbols pointing to absolute symbols.  Either of
these fixes would cure the .loc view usage.

PR 24444
* symbols.c (resolve_symbol_value): When handling symbols
marked as sy_flags.resolved, return correct value for the
case of expression symbols left as an O_symbol expression.
Merge O_symbol code handling undefined and common symbols with
code handling special cases of expression symbols.  Use
seg_left to test for undefined and common symbols.  Don't
leave an O_symbol expression when X_add_symbol resolves to
the absolute_section.  Init final_val later.
* testsuite/gas/mmix/basep-7.d: Adjust expected output.

5 years agoS12Z: Opcodes: Handle bit map operations with non-canonical operands.
John Darrington [Wed, 24 Apr 2019 07:41:23 +0000 (09:41 +0200)]
S12Z: Opcodes: Handle bit map operations with non-canonical operands.

opcodes/
* s12z-opc.c (bm_decode): Handle the RESERVERD0 case.

gas/
* testsuite/gas/s12z/bit-manip-invalid.d: Extend the test.
* testsuite/gas/s12z/bit-manip-invalid.s: Extend the test.

5 years agoS12Z: s12z-opc.h: Add extern "C" bracketing
John Darrington [Mon, 15 Apr 2019 07:25:23 +0000 (09:25 +0200)]
S12Z: s12z-opc.h: Add extern "C" bracketing

opcodes/
* s12z-opc.h: Add extern "C" bracketing to help
  users who wish to use this interface in c++ code.

5 years agoAutomatic date update in version.in
GDB Administrator [Wed, 24 Apr 2019 00:00:16 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agogdb/s12z: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:38:42 +0000 (20:38 +0000)]
gdb/s12z: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_unwind_pc, and
gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* s12z-tdep.c (s12z_unwind_pc): Delete.
(s12z_unwind_sp): Delete.
(s12z_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/rl78: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:36:34 +0000 (20:36 +0000)]
gdb/rl78: Use default gdbarch methods where possible

Make use of the default gdbarch method gdbarch_unwind_sp where
possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* rl78-tdep.c (rl78_unwind_sp): Delete.
(rl78_gdbarch_init): Don't register deleted function with gdbarch.

5 years agogdb/xstormy16: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:32:40 +0000 (20:32 +0000)]
gdb/xstormy16: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* xstormy16-tdep.c (xstormy16_unwind_sp): Delete.
(xstormy16_unwind_pc): Delete.
(xstormy16_dummy_id): Delete.
(xstormy16_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/vax: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:30:32 +0000 (20:30 +0000)]
gdb/vax: Use default gdbarch methods where possible

Make use of the default gdbarch method gdbarch_unwind_pc where
possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* vax-tdep.c (vax_unwind_pc): Delete.
(vax_gdbarch_init): Don't register deleted function with gdbarch.

5 years agogdb/v850: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:28:24 +0000 (20:28 +0000)]
gdb/v850: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* v850-tdep.c (v850_unwind_sp): Delete.
(v850_unwind_pc): Delete.
(v850_dummy_id): Delete.
(v850_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/tilegx: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:25:59 +0000 (20:25 +0000)]
gdb/tilegx: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* tilegx-tdep.c (tilegx_unwind_sp): Delete.
(tilegx_unwind_pc): Delete.
(tilegx_unwind_dummy_id): Delete.
(tilegx_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/tic6x: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:23:44 +0000 (20:23 +0000)]
gdb/tic6x: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id, and
gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* tic6x-tdep.c (tic6x_unwind_sp): Delete.
(tic6x_dummy_id): Delete.
(tic6x_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/sparc: Use default_unwind_pc
Andrew Burgess [Thu, 10 Jan 2019 20:20:25 +0000 (20:20 +0000)]
gdb/sparc: Use default_unwind_pc

Make use of the default gdbarch method gdbarch_unwind_pc where
possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* sparc-tdep.c (sparc_unwind_pc): Delete.
(sparc32_gdbarch_init): Don't register deleted function with
gdbarch.

5 years agogdb/sh: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:17:00 +0000 (20:17 +0000)]
gdb/sh: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* sh-tdep.c (sh_unwind_sp): Delete.
(sh_unwind_pc): Delete.
(sh_dummy_id): Delete.
(sh_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/score: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:15:03 +0000 (20:15 +0000)]
gdb/score: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* score-tdep.c (score_unwind_sp): Delete.
(score_unwind_pc): Delete.
(score_dummy_id): Delete.
(score_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/rx: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:12:28 +0000 (20:12 +0000)]
gdb/rx: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* rx-tdep.c (rx_unwind_pc): Delete.
(rx_unwind_sp): Delete.
(rx_dummy_id): Delete.
(rx_gdbarch_init): Don't register deleted functions with
gdbarch.  Update comment.

5 years agogdb/rs6000: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 20:08:18 +0000 (20:08 +0000)]
gdb/rs6000: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* rs6000-tdep.c (rs6000_unwind_pc): Delete.
(rs6000_dummy_id): Delete.
(rs6000_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/or1k: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:58:18 +0000 (18:58 +0000)]
gdb/or1k: Use default gdbarch methods where possible

Make use of the default gdbarch method gdbarch_dummy_id where
possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

This commit leaves or1k_unwind_sp and or1k_unwind_pc in place.  These
functions do match the default methods except that they add additional
debugging code.  In order to preserve the debug I have left these
functions unchanged.

gdb/ChangeLog:

* or1k-tdep.c (or1k_dummy_id): Delete.
(or1k_gdbarch_init): Don't register deleted function with gdbarch.

5 years agogdb/nios2: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:55:23 +0000 (18:55 +0000)]
gdb/nios2: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id, and
gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* nios2-tdep.c (nios2_dummy_id): Delete.
(nios2_unwind_sp): Delete.
(nios2_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/nds32: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:53:31 +0000 (18:53 +0000)]
gdb/nds32: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* nds32-tdep.c (nds32_dummy_id): Delete.
(nds32_unwind_pc): Delete.
(nds32_unwind_sp): Delete.
(nds32_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/msp430: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:51:11 +0000 (18:51 +0000)]
gdb/msp430: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* msp430-tdep.c (msp430_unwind_pc): Delete.
(msp430_unwind_sp): Delete.
(msp430_dummy_id): Delete.
(msp430_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/moxie: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:48:53 +0000 (18:48 +0000)]
gdb/moxie: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* moxie-tdep.c (moxie_unwind_sp): Delete.
(moxie_unwind_pc): Delete.
(moxie_dummy_id): Delete.
(moxie_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/mn10300: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:46:27 +0000 (18:46 +0000)]
gdb/mn10300: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* mn10300-tdep.c (mn10300_dummy_id): Delete.
(mn10300_unwind_pc): Delete.
(mn10300_unwind_sp): Delete.
(mn10300_push_dummy_call): Use gdbarch_unwind_sp not
mn10300_unwind_sp.
(mn10300_frame_unwind_init): Don't register deleted functions with
gdbarch.

5 years agogdb/mep: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:37:03 +0000 (18:37 +0000)]
gdb/mep: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* mep-tdep.c (mep_unwind_pc): Delete.
(mep_unwind_sp): Delete.
(mep_dummy_id): Delete.
(mep_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/m68hc11: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:33:50 +0000 (18:33 +0000)]
gdb/m68hc11: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_unwind_pc, and
gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* m68hc11-tdep.c (m68hc11_unwind_pc): Delete.
(m68hc11_unwind_sp): Delete.
(m68hc11_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/m32r: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:29:47 +0000 (18:29 +0000)]
gdb/m32r: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* m32r-tdep.c (m32r_unwind_sp): Delete.
(m32r_unwind_pc): Delete.
(m32r_dummy_id): Delete.
(m32r_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/m32c: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:27:12 +0000 (18:27 +0000)]
gdb/m32c: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* m32c-tdep.c (m32c_unwind_pc): Delete.
(m32c_unwind_sp): Delete.
(m32c_dummy_id): Delete.
(m32c_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/lm32: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:21:40 +0000 (18:21 +0000)]
gdb/lm32: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* gdb/lm32-tdep.c (lm32_unwind_sp): Delete.
(lm32_unwind_pc): Delete.
(lm32_dummy_id): Delete.
(lm32_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/iq2000: Use default gdbarch methods where possible
Andrew Burgess [Thu, 10 Jan 2019 18:18:39 +0000 (18:18 +0000)]
gdb/iq2000: Use default gdbarch methods where possible

Make use of the default gdbarch methods for gdbarch_dummy_id,
gdbarch_unwind_pc, and gdbarch_unwind_sp where possible.

I have not tested this change but, by inspecting the code, I believe
the default methods are equivalent to the code being deleted.

gdb/ChangeLog:

* gdb/iq2000-tdep.c (iq2000_unwind_sp): Delete.
(iq2000_unwind_pc): Delete.
(iq2000_dummy_id): Delete.
(iq2000_gdbarch_init): Don't register deleted functions with
gdbarch.

5 years agogdb/nds32: Use type_align instead of nds32_type_align
Andrew Burgess [Fri, 12 Apr 2019 14:07:51 +0000 (15:07 +0100)]
gdb/nds32: Use type_align instead of nds32_type_align

The general type_align method should be a suitable alternative to
nds32_type_align, so switch to use that.

The only change this will introduce is related to static fields in a
struct or union, the existing code doesn't take account of static
fields when computing the alignment for structs of unions, though this
is probably a bug - which would probably be exposed by the test case
gdb.cp/many-args.exp, though I don't have any way to test this target
right now.

gdb/ChangeLog:

* nds32-tdep.c (nds32_type_align): Delete.
(nds32_push_dummy_call): Use type_align instead.

5 years agogdb/arm: Use type_align instead of arm_type_align
Andrew Burgess [Fri, 12 Apr 2019 13:25:32 +0000 (14:25 +0100)]
gdb/arm: Use type_align instead of arm_type_align

Replaces use of arm_type_align with common type_align function.

Doing this fixes a bug in arm_type_align where static fields are
considered as part of the alignment calculation of a struct, which
results in arguments passed on the stack being misaligned, this bug
was causing a failure in gdb.cp/many-args.exp.

Part of the old arm_type_align is retained and used as the gdbarch
type align callback in order to correctly align vectors.

gdb/ChangeLog:

* arm-tdep.c (arm_type_align): Only handle vector override case.
(arm_push_dummy_call): Use type_align.
(arm_gdbarch_init): Register arm_type_align gdbarch function.

5 years agogdb/aarch64: Use type_align instead of aarch64_type_align
Andrew Burgess [Sun, 7 Apr 2019 21:48:00 +0000 (22:48 +0100)]
gdb/aarch64: Use type_align instead of aarch64_type_align

Replaces use of aarch64_type_align with common type_align function.

Doing this fixes a bug in aarch64_type_align where static fields are
considered as part of the alignment calculation of a struct, which
results in arguments passed on the stack being misaligned.  This bug
is exposed in the new test gdb.cp/many-args.exp.

Part of the old aarch64_type_align is retained and used as the gdbarch
type align callback in order to correctly align vectors.

gdb/ChangeLog:

* aarch64-tdep.c (aarch64_type_align): Only handle vector override
case.
(pass_on_stack): Use type_align.
(aarch64_gdbarch_init): Register aarch64_type_align gdbarch
function.

gdb/testsuite/ChangeLog:

* gdb.cp/many-args.cc: New file.
* gdb.cp/many-args.exp: New file.

5 years agoRemove unused overload of line_header::file_name_at
Tom Tromey [Tue, 23 Apr 2019 16:48:40 +0000 (10:48 -0600)]
Remove unused overload of line_header::file_name_at

I noticed that one of the overloads of line_header::file_name_at is
unused.  This patch removes it.

gdb/ChangeLog
2019-04-23  Tom Tromey  <tromey@adacore.com>

* dwarf2read.c (line_header::file_name_at): Remove unused
overload.

5 years ago[gdb/testsuite] Fix gdb.btrace/reconnect.exp with native-gdbserver
Tom de Vries [Tue, 23 Apr 2019 13:49:52 +0000 (15:49 +0200)]
[gdb/testsuite] Fix gdb.btrace/reconnect.exp with native-gdbserver

When running gdb.btrace/reconnect.exp with native-gdbserver, we run into:
...
FAIL: gdb.btrace/reconnect.exp: first: stepi 19
...
due to the fact that we're trying to match:
...
stepi 19^M
0x00007ffff7dd8b57 in _dl_start () from /lib64/ld-linux-x86-64.so.2^M
...
using pattern:
...
  gdb_test "stepi 19" "0x.* in .* from target.*"
...

Fix this by changing the pattern to:
...
  gdb_test "stepi 19" "0x.* in .* from .*"
...

Tested on x86_64-linux with native and native-gdbserver.

gdb/testsuite/ChangeLog:

2019-04-23  Tom de Vries  <tdevries@suse.de>

PR gdb/24433
* gdb.btrace/reconnect.exp: Fix stepi 19 pattern.

5 years ago[gdb/contrib] Remove superfluous .alt file after dwz invocation in cc-with-tweaks.sh
Tom de Vries [Tue, 23 Apr 2019 13:35:21 +0000 (15:35 +0200)]
[gdb/contrib] Remove superfluous .alt file after dwz invocation in cc-with-tweaks.sh

The -m option of cc-with-tweaks.sh sets want_multi to true, invoking dwz like
this:
...
elif [ "$want_multi" = true ]; then
    cp $output_file ${output_file}.alt
    $DWZ -m ${output_file}.dwz "$output_file" ${output_file}.alt \
        > /dev/null 2>&1
fi
...

The problem that is being solved here, is that we want to test dwz in
multifile mode, which requires more than one input file, while we only have
(at the scope of cc-with-tweaks.sh) one executable.  We handle this by copying
the executable and offering this as a second input (and using a copy has the
additional benefit that it maximally enables dwz transformation).

However, after the dwz invocation, the copy is no longer used, and the
presence of the file actually causes a test regression:
...
FAIL: gdb.base/jit-so.exp: test jit-reader-load filename completion
...

Fix this by removing the superflous copy after dwz invocation.

Tested on x86_64-linux.

gdb/ChangeLog:

2019-04-23  Tom de Vries  <tdevries@suse.de>

PR gdb/24438
* contrib/cc-with-tweaks.sh: Remove superfluous .alt file after dwz
invocation.

5 years agoTestsuite: Remove pie from trace tests
Alan Hayward [Tue, 23 Apr 2019 13:12:25 +0000 (14:12 +0100)]
Testsuite: Remove pie from trace tests

Ubuntu/Debian defaults PIE to enabled.  This causes the trace tests
to fall over due to variables being returned as "unavailable".  The
tests were never designed to work with pie.

Simply ensure the nopie flag is always used for the failing tests.

This removes 100+ failures when running native-gdbserver on Ubuntu 18.04.

gdb/testsuite/ChangeLog:

* gdb.trace/backtrace.exp: Use nopie flag.
* gdb.trace/circ.exp: Likewise.
* gdb.trace/collection.exp: Likewise.
* gdb.trace/ftrace.exp: Likewise.
* gdb.trace/mi-trace-unavailable.exp: Likewise.
* gdb.trace/mi-traceframe-changed.exp: Likewise.
* gdb.trace/qtro.exp: Likewise.
* gdb.trace/read-memory.exp: Likewise.
* gdb.trace/report.exp: Likewise.
* gdb.trace/tfile.exp: Likewise.
* gdb.trace/tfind.exp: Likewise.
* gdb.trace/unavailable.exp: Likewise.

5 years agoUpdate binutils release making documenation to mention changing the symbolic document...
Nick Clifton [Tue, 23 Apr 2019 10:19:58 +0000 (11:19 +0100)]
Update binutils release making documenation to mention changing the symbolic documentation link.

* README-how-to-make-a-release: Add note to update the symbolic
link from "docs" to "docs-2.x" on the sourceware website.

5 years agoFix automatic makefile dependencies for generated ld/e*.c
Alan Modra [Tue, 23 Apr 2019 07:12:27 +0000 (16:42 +0930)]
Fix automatic makefile dependencies for generated ld/e*.c

Commit c40e31a121 broke --enable-dependency-tracking=no.

* Makefile.am (GENDEPDIR): New var, used..
(GENSCRIPTS): ..here.
* Makefile.in: Regenerate.
* genscripts.sh: Test for $DEPDIR set before every use.

5 years agoFix M5100 flags test with interAptiv-MR2
Matthew Fortune [Fri, 19 Apr 2019 21:03:18 +0000 (21:03 +0000)]
Fix M5100 flags test with interAptiv-MR2

ld/
* testsuite/ld-mips-elf/mips-elf-flags.exp: Fix expected ASEs
for M5100.

5 years agoSupport for DW_OP_addrx and DW_FORM_addrx tags
Ali Tamur [Sat, 30 Mar 2019 02:29:24 +0000 (19:29 -0700)]
Support for DW_OP_addrx and DW_FORM_addrx tags

DW_OP_addrx is the new name of DW_OP_GNU_addr_index, and DW_FORM_addrx
is the name of DW_FORM_addr_index in the Dwarf 5 standard. This is a small
step towards supporting Dwarf 5 in gdb.

Note: I could not find any tests specifically for *_GNU_addr_index, and
I did not add any new tests, please advise.

5 years agoAutomatic date update in version.in
GDB Administrator [Tue, 23 Apr 2019 00:00:17 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years ago[FYI] Add myself to gdb/MAINTAINERS
Ali Tamur [Tue, 16 Apr 2019 22:35:31 +0000 (15:35 -0700)]
[FYI] Add myself to gdb/MAINTAINERS

5 years agoRISC-V: Enable 32-bit linux gdb core file support.
Jim Wilson [Mon, 22 Apr 2019 21:17:55 +0000 (14:17 -0700)]
RISC-V: Enable 32-bit linux gdb core file support.

bfd/
* elfnn-riscv.c (PRSTATUS_SIZE) [ARCH_SIZE==32]: Change from 0 to 204.

5 years agosolib-svr4: Pass down svr4_info as much as possible
Simon Marchi [Mon, 22 Apr 2019 18:02:36 +0000 (14:02 -0400)]
solib-svr4: Pass down svr4_info as much as possible

While reviewing

  https://sourceware.org/ml/gdb-patches/2019-04/msg00141.html

I noticed that we relied heavily on global state through the
get_svr4_info function, which uses current_program_space.  I thought we
could improve this (make things more explicit and easier to follow) by

- Making get_svr4_info accept a program_space parameter, making it
  return the SVR4 info for that program space.
- Passing down the svr4_info object from callers as much as possible.

This means looking up the svr4_info for the appropriate program space at
the entry points of the solib-svr4.c file and passing it down.  For now,
these entry points (most of them are "methods" of svr4_so_ops) rely on
current_program_space, but we can later try to change the target_so_ops
interface to pass down the program space.

gdb/ChangeLog:

* solib-svr4.c (get_svr4_info): Add pspace parameter.
(svr4_keep_data_in_core): Pass current_program_space to get_svr4_info.
(open_symbol_file_object): Likewise.
(svr4_default_sos): Add info parameter.
(svr4_read_so_list): Likewise.
(svr4_current_sos_direct): Adjust functions calls to pass down
info.
(svr4_current_sos_1): Add info parameter.
(svr4_current_sos): Call get_svr4_info, pass info down to
svr4_current_sos_1.
(svr4_fetch_objfile_link_map): Pass objfile->pspace to
get_svr4_info.
(svr4_in_dynsym_resolve_code): Pass current_program_space to
get_svr4_info.
(probes_table_htab_remove_objfile_probes): Pass objfile->pspace
to get_svr4_info.
(probes_table_remove_objfile_probes): Likewise.
(register_solib_event_probe): Add info parameter.
(solist_update_incremental): Pass info parameter down to
svr4_read_so_list.
(disable_probes_interface): Add info parameter.
(svr4_handle_solib_event): Pass current_program_space to
get_svr4_info.  Adjust disable_probes_interface cleanup.
(svr4_create_probe_breakpoints): Add info parameter, pass it
down to register_solib_event_probe.
(svr4_create_solib_event_breakpoints): Add info parameter,
pass it down to svr4_create_probe_breakpoints.
(enable_break): Pass info down to
svr4_create_solib_event_breakpoints.
(svr4_solib_create_inferior_hook): Pass current_program_space to
get_svr4_info.
(svr4_clear_solib): Likewise.

5 years agoFix "nosharedlibrary + continue + shared lib event" crash
Pedro Alves [Mon, 22 Apr 2019 13:20:59 +0000 (14:20 +0100)]
Fix "nosharedlibrary + continue + shared lib event" crash

On systems that use the probes-based solib interface, GDB misbehaves
if you run the "nosharelibrary" command, continue execution, and then
the program hits the shared library event breakpoint.  On my system it
aborts like this:

 (gdb) nosharedlibrary
 (gdb) c
 Continuing.
 pure virtual method called
 terminate called without an active exception
 Aborted (core dumped)

Though it's really undefined behavior territory, caused by deferencing
a dangling solib event probe pointer.

I've observed this by running "nosharedlibrary" when stopped at the
entry point, but it should happen at any other point, if the program
does a dlopen/dlclose after.

The fix is to discard an objfile's probes from the svr4 probes table
when an objfile is about to be released.

New test included, works with both native and gdbserver testing.

Valgrind log:

 (gdb) starti
 (gdb) nosharedlibrary
 (gdb) c
 Continuing.
 ==24895== Invalid read of size 8
 ==24895==    at 0x89E5FB: solib_event_probe_action(probe_and_action*) (solib-svr4.c:1735)
 ==24895==    by 0x89E95A: svr4_handle_solib_event() (solib-svr4.c:1872)
 ==24895==    by 0x8A7198: handle_solib_event() (solib.c:1274)
 ==24895==    by 0x4E3407: bpstat_stop_status(address_space const*, unsigned long, thread_info*, target_waitstatus const*, bpstats*) (breakpoint.c:5407)
 ==24895==    by 0x721F41: handle_signal_stop(execution_control_state*) (infrun.c:5685)
 ==24895==    by 0x720B11: handle_inferior_event(execution_control_state*) (infrun.c:5129)
 ==24895==    by 0x71DD93: fetch_inferior_event(void*) (infrun.c:3748)
 ==24895==    by 0x7059C3: inferior_event_handler(inferior_event_type, void*) (inf-loop.c:43)
 ==24895==    by 0x874DF0: remote_async_serial_handler(serial*, void*) (remote.c:14039)
 ==24895==    by 0x894101: run_async_handler_and_reschedule(serial*) (ser-base.c:137)
 ==24895==    by 0x8941E6: fd_event(int, void*) (ser-base.c:188)
 ==24895==    by 0x67AFEF: handle_file_event(file_handler*, int) (event-loop.c:732)
 ==24895==  Address 0x18b63860 is 0 bytes inside a block of size 136 free'd
 ==24895==    at 0x4C2E616: operator delete(void*, unsigned long) (vg_replace_malloc.c:585)
 ==24895==    by 0x8C6A12: stap_probe::~stap_probe() (stap-probe.c:124)
 ==24895==    by 0x66F7DB: probe_key_free(bfd*, void*) (elfread.c:1382)
 ==24895==    by 0x69B705: bfdregistry_callback_adaptor(void (*)(registry_container*, void*), registry_container*, void*) (gdb_bfd.c:131)
 ==24895==    by 0x855A57: registry_clear_data(registry_data_registry*, void (*)(void (*)(registry_container*, void*), registry_container*, void*), registry_container*, registry_fields*) (registry.c:79)
 ==24895==    by 0x855B01: registry_container_free_data(registry_data_registry*, void (*)(void (*)(registry_container*, void*), registry_container*, void*), registry_container*, registry_fields*) (registry.c:92)
 ==24895==    by 0x69B783: bfd_free_data(bfd*) (gdb_bfd.c:131)
 ==24895==    by 0x69C4BA: gdb_bfd_unref(bfd*) (gdb_bfd.c:609)
 ==24895==    by 0x7CC33F: objfile::~objfile() (objfiles.c:651)
 ==24895==    by 0x7CD559: objfile_purge_solibs() (objfiles.c:1021)
 ==24895==    by 0x8A7132: no_shared_libraries(char const*, int) (solib.c:1252)
 ==24895==    by 0x548E3D: do_const_cfunc(cmd_list_element*, char const*, int) (cli-decode.c:106)
 ==24895==  Block was alloc'd at
 ==24895==    at 0x4C2D42A: operator new(unsigned long) (vg_replace_malloc.c:334)
 ==24895==    by 0x8C527C: handle_stap_probe(objfile*, sdt_note*, std::vector<probe*, std::allocator<probe*> >*, unsigned long) (stap-probe.c:1561)
 ==24895==    by 0x8C5535: stap_static_probe_ops::get_probes(std::vector<probe*, std::allocator<probe*> >*, objfile*) const (stap-probe.c:1656)
 ==24895==    by 0x66F71B: elf_get_probes(objfile*) (elfread.c:1365)
 ==24895==    by 0x7EDD85: find_probes_in_objfile(objfile*, char const*, char const*) (probe.c:227)
 ==24895==    by 0x4DF382: create_longjmp_master_breakpoint() (breakpoint.c:3275)
 ==24895==    by 0x4F6562: breakpoint_re_set() (breakpoint.c:13828)
 ==24895==    by 0x8A66AA: solib_add(char const*, int, int) (solib.c:1010)
 ==24895==    by 0x89F7C6: enable_break(svr4_info*, int) (solib-svr4.c:2360)
 ==24895==    by 0x8A104C: svr4_solib_create_inferior_hook(int) (solib-svr4.c:2992)
 ==24895==    by 0x8A70B9: solib_create_inferior_hook(int) (solib.c:1215)
 ==24895==    by 0x70C073: post_create_inferior(target_ops*, int) (infcmd.c:467)
 ==24895==
 pure virtual method called
 terminate called without an active exception
 ==24895==
 ==24895== Process terminating with default action of signal 6 (SIGABRT): dumping core
 ==24895==    at 0x7CF3750: raise (raise.c:51)
 ==24895==    by 0x7CF4D30: abort (abort.c:79)
 ==24895==    by 0xB008F4: __gnu_cxx::__verbose_terminate_handler() (in build/gdb/gdb)
 ==24895==    by 0xAFF845: __cxxabiv1::__terminate(void (*)()) (in build/gdb/gdb)
 ==24895==    by 0xAFF890: std::terminate() (in build/gdb/gdb)
 ==24895==    by 0xAFF95E: __cxa_pure_virtual (in build/gdb/gdb)
 ==24895==    by 0x89E610: solib_event_probe_action(probe_and_action*) (solib-svr4.c:1735)
 ==24895==    by 0x89E95A: svr4_handle_solib_event() (solib-svr4.c:1872)
 ==24895==    by 0x8A7198: handle_solib_event() (solib.c:1274)
 ==24895==    by 0x4E3407: bpstat_stop_status(address_space const*, unsigned long, thread_info*, target_waitstatus const*, bpstats*) (breakpoint.c:5407)
 ==24895==    by 0x721F41: handle_signal_stop(execution_control_state*) (infrun.c:5685)
 ==24895==    by 0x720B11: handle_inferior_event(execution_control_state*) (infrun.c:5129)
 ==24895==

Note, this little bit in the patch is just a cleanup that I noticed:

 -  lookup.prob = prob;
    lookup.address = address;

That line isn't necessary because hashing/comparison only looks at the
address.

gdb/ChangeLog:
2019-04-22  Pedro Alves  <palves@redhat.com>

* solib-svr4.c (svr4_free_objfile_observer): New.
(probe_and_action::objfile): New field.
(probes_table_htab_remove_objfile_probes)
(probes_table_remove_objfile_probes): New functions.
(register_solib_event_probe): Add 'objfile' parameter.  Store it
in the new probe_and_action.  Don't store the probe in 'lookup'.
(svr4_create_probe_breakpoints): Pass objfile to
register_solib_event_probe.
(_initialize_svr4_solib): Register a free_objfile observer.

gdb/testsuite/ChangeLog:
2019-04-22  Pedro Alves  <palves@redhat.com>

* gdb.base/solib-probes-nosharedlibrary.c,
gdb.base/solib-probes-nosharedlibrary.exp: New files.

5 years agoImprove reverse debugging docs, mention built-in support and supports archs
Pedro Alves [Mon, 22 Apr 2019 11:42:21 +0000 (12:42 +0100)]
Improve reverse debugging docs, mention built-in support and supports archs

gdb/doc/ChangeLog:
2019-04-22  Pedro Alves  <palves@redhat.com>

* gdb.texinfo (Reverse Execution): Mention and xref process record
and replay.  Mention remote and system emulators.
(Process Record and Replay): List supported architectures.
Mention that "record btrace" is only supported on Intel
processors.

5 years agoAutomatic date update in version.in
GDB Administrator [Mon, 22 Apr 2019 00:00:46 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agoAutomatic date update in version.in
GDB Administrator [Sun, 21 Apr 2019 00:01:05 +0000 (00:01 +0000)]
Automatic date update in version.in

5 years agoFix GDB crash when registers cannot be modified.
Philippe Waroquiers [Sat, 13 Apr 2019 08:22:41 +0000 (10:22 +0200)]
Fix GDB crash when registers cannot be modified.

This crash was detected when using GDB with the valgrind gdbserver.
To reproduce:

valgrind sleep 10000

In another window:
gdb
target remote | vgdb
p printf("make sleep print something\n")
=>
terminate called after throwing an instance of 'gdb_exception_error'
Aborted

The problem is that the valgrind gdbserver does not allow to change
registers when the inferior is blocked in a system call.
GDB then raises an exception.  The exception causes the destructor
of
 typedef std::unique_ptr<infcall_suspend_state, infcall_suspend_state_deleter>
    infcall_suspend_state_up;
to be called.  This destructor itself tries to restore the value of
the registers, and fails similarly.  We must catch the exception in
the destructor to avoid crashing GDB.
If the destructor encounters a problem, no warning is produced if
there is an uncaught exception, as in this case, the user will already
be informed of a problem via this exception.

With this change, no crash anymore, and all the valgrind 3.15 tests
pass succesfully.

gdb/ChangeLog
2019-04-19  Philippe Waroquiers  <philippe.waroquiers@skynet.be>

* inferior.h (struct infcall_suspend_state_deleter):
Catch exception in destructor to avoid crash.

5 years agoAutomatic date update in version.in
GDB Administrator [Sat, 20 Apr 2019 00:00:21 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agoRemove common/queue.h
Tom Tromey [Sun, 7 Apr 2019 22:32:54 +0000 (16:32 -0600)]
Remove common/queue.h

gdb no longer needs common/queue.h, so this removes it.

gdb/ChangeLog
2019-04-19  Tom Tromey  <tom@tromey.com>

* common/queue.h: Remove.

5 years agoRemove an include of common/queue.h
Tom Tromey [Sun, 7 Apr 2019 22:32:19 +0000 (16:32 -0600)]
Remove an include of common/queue.h

event-loop.c does not need to include common/queue.h, so this removes
it.

gdb/ChangeLog
2019-04-19  Tom Tromey  <tom@tromey.com>

* event-loop.c: Don't include "common/queue.h".

5 years agoUse std::list for remote_notif_state::notif_queue
Tom Tromey [Sun, 7 Apr 2019 22:31:01 +0000 (16:31 -0600)]
Use std::list for remote_notif_state::notif_queue

This changes remote_notif_state::notif_queue to be a std::list and
updates all the uses.

gdb/ChangeLog
2019-04-19  Tom Tromey  <tom@tromey.com>

* remote.c (remote_target): Use delete.
* remote-notif.h: Include <list>, not "common/queue.h".
(notif_client_p): Remove typedef.
(remote_notif_state): Add constructor, destructor, initializer.
<notif_queue>: Now a std::list.
(remote_notif_state_xfree): Don't declare.
* remote-notif.c (remote_notif_process, handle_notification)
(remote_notif_state_allocate): Update.
(~remote_notif_state): Rename from remote_notif_state_xfree.

5 years agoUse std::list for event notifications in gdbserver
Tom Tromey [Sun, 7 Apr 2019 22:17:40 +0000 (16:17 -0600)]
Use std::list for event notifications in gdbserver

This changes gdbserver to use std::list rather than common/queue.h for
event notifications.

gdb/gdbserver/ChangeLog
2019-04-19  Tom Tromey  <tom@tromey.com>

* server.c (struct vstop_notif): Derive from notif_event.
<base>: Remove.
(queue_stop_reply): Update.
(remove_all_on_match_ptid): Change type.  Rewrite.
(discard_queued_stop_replies): Rewrite.
(in_queued_stop_replies_ptid): Change type.
(in_queued_stop_replies): Rewrite.
(notif_stop): Update.
(queue_stop_reply_callback): Update.
(captured_main): Don't call initialize_notif.
(push_stop_notification): Update.
* notif.c (notif_write_event, handle_notif_ack)
(notif_event_enque, notif_push): Update.
(notif_event_xfree, initialize_notif): Remove.
* notif.h (struct notif_event): Include <list>, not
"common/queue.h".
(struct notif_server) <queue>: Now a std::list.
(notif_event_p): Remove typedef.
(initialize_notif): Don't declare.
(struct notif_event): Add virtual destructor.

5 years agoMake objfile::static_links an htab_up
Tom Tromey [Sun, 7 Apr 2019 21:39:37 +0000 (15:39 -0600)]
Make objfile::static_links an htab_up

This changes objfile::static_links to be an htab_up, so that ~objfile
no longer has to explicitly destroy it.

Tested by the buildbot.

gdb/ChangeLog
2019-04-19  Tom Tromey  <tom@tromey.com>

* symfile.c (reread_symbols): Update.
* objfiles.c (objfile_register_static_link)
(objfile_lookup_static_link): Update
(~objfile) Don't delete static_links.
* objfiles.h (struct objfile) <static_links>: Now an htab_up.

5 years agoMake copy_name return std::string
Tom Tromey [Sun, 7 Apr 2019 21:29:58 +0000 (15:29 -0600)]
Make copy_name return std::string

This changes copy_name to return a std::string, updating all the
callers.  In some cases, an extra copy was removed.  This also
required a little bit of constification.

Tested by the buildbot.

gdb/ChangeLog
2019-04-19  Tom Tromey  <tom@tromey.com>

* type-stack.h (struct type_stack) <insert>: Constify string.
* type-stack.c (type_stack::insert): Constify string.
* gdbtypes.h (lookup_template_type): Update.
(address_space_name_to_int): Update.
* gdbtypes.c (address_space_name_to_int): Make space_identifier
const.
(lookup_template_type): Make name const.
* c-exp.y: Update rules.
(lex_one_token, classify_name, classify_inner_name)
(c_print_token): Update.
* p-exp.y: Update rules.
(yylex): Update.
* f-exp.y: Update rules.
(yylex): Update.
* d-exp.y: Update rules.
(lex_one_token, classify_name, classify_inner_name): Update.
* parse.c (write_dollar_variable, copy_name): Return std::string.
* parser-defs.h (copy_name): Change return type.
* m2-exp.y: Update rules.
(yylex): Update.
* go-exp.y (lex_one_token): Update.
Update rules.
(classify_unsafe_function, classify_packaged_name)
(classify_name, yylex): Update.

5 years agogdb/configure.ac: add --enable-source-highlight
Sergei Trofimovich [Sun, 17 Mar 2019 22:48:02 +0000 (22:48 +0000)]
gdb/configure.ac: add --enable-source-highlight

Allow disabling source-highlight dependency autodetection even
it exists in the system. More details on problem of automatic
dependencies:
https://wiki.gentoo.org/wiki/Project:Quality_Assurance/Automagic_dependencies

Noticed by Jeroen Roovers in https://bugs.gentoo.org/680238

* configure.ac: add --enable-source-highlight switch.
* configure: Regenerate.
* top.c (print_gdb_version): plumb --enable-source-highlight
status to "show configuration".

gdb/ChangeLog
2019-04-19  Sergei Trofimovich <siarheit@google.com>

* configure.ac: add --enable-source-highlight switch.
* configure: Regenerate.
* top.c (print_gdb_version): plumb --enable-source-highlight
status to "show configuration".

5 years agoPrint non-Ada unions without crashing
Tom Tromey [Tue, 9 Apr 2019 13:46:39 +0000 (07:46 -0600)]
Print non-Ada unions without crashing

ada-lang.c is a bit too eager trying to decode unions in the Ada style
-- looking for discriminants and such.  This causes crashes when
printing a non-Ada union in Ada mode, something that can easily happen
when printing a value from history or certain registers on AArch64.

This patch fixes the bug by changing ada-lang.c to only apply special
Ada treatment to types coming from an Ada CU.  This in turn required a
couple of surprising changes.

First, some of the Ada code was already using HAVE_GNAT_AUX_INFO to
decide whether a type had already been fixed -- such types had
INIT_CPLUS_SPECIFIC called on them.  This patch changes these spots to
use the "none" identifier instead.

This then required changing value_rtti_type to avoid changing the
language-specific object attached to an Ada type, which seems like a
good change regardless.

Tested on x86-64 Fedora 29.

gdb/ChangeLog
2019-04-19  Tom Tromey  <tromey@adacore.com>

* ada-lang.c (ada_is_variant_part, ada_to_fixed_type_1):
Check ADA_TYPE_P.
(empty_record, ada_template_to_fixed_record_type_1)
(template_to_static_fixed_type)
(to_record_with_fixed_variant_part): Use INIT_NONE_SPECIFIC.
* cp-abi.c (value_rtti_type): Check HAVE_CPLUS_STRUCT.
* gdbtypes.h (INIT_NONE_SPECIFIC, ADA_TYPE_P): New
macros.

gdb/testsuite/ChangeLog
2019-04-19  Tom Tromey  <tromey@adacore.com>

* gdb.ada/ptype_union.c: New file.
* gdb.ada/ptype_union.exp: New file.

5 years agoFix "list" when control characters are seen
Tom Tromey [Tue, 9 Apr 2019 18:52:46 +0000 (12:52 -0600)]
Fix "list" when control characters are seen

PR symtab/24423 points out that control characters in a source file
cause a hang in the "list" command, a regression introduced by the
styling changes.

This patch, from the PR, fixes the bug.  I've included a minimal
change to the "list" test that exercises this code.

I recall that this bug was discussed on gdb-patches, and I thought
there was a patch there as well, but I was unable to find it.

gdb/ChangeLog
2019-04-19  Ilya Yu. Malakhov  <malakhov@mcst.ru>

PR symtab/24423:
* source.c (print_source_lines_base): Advance "iter" when a
control character is seen.

gdb/testsuite/ChangeLog
2019-04-19  Tom Tromey  <tromey@adacore.com>

PR symtab/24423:
* gdb.base/list0.h (foo): Add a control-l character.

5 years agoFix GDB crash when registers cannot be modified.
Philippe Waroquiers [Sat, 13 Apr 2019 08:22:41 +0000 (10:22 +0200)]
Fix GDB crash when registers cannot be modified.

This crash was detected when using GDB with the valgrind gdbserver.
To reproduce:

valgrind sleep 10000

In another window:
gdb
target remote | vgdb
p printf("make sleep print something\n")
=>
terminate called after throwing an instance of 'gdb_exception_error'
Aborted

The problem is that the valgrind gdbserver does not allow to change
registers when the inferior is blocked in a system call.
GDB then raises an exception.  The exception causes the destructor
of
 typedef std::unique_ptr<infcall_suspend_state, infcall_suspend_state_deleter>
    infcall_suspend_state_up;
to be called.  This destructor itself tries to restore the value of
the registers, and fails similarly.  We must catch the exception in
the destructor to avoid crashing GDB.
If the destructor encounters a problem, no warning is produced if
there is an uncaught exception, as in this case, the user will already
be informed of a problem via this exception.

With this change, no crash anymore, and all the valgrind 3.15 tests
pass succesfully.

gdb/ChangeLog
2019-04-19  Philippe Waroquiers  <philippe.waroquiers@skynet.be>

* inferior.h (struct infcall_suspend_state_deleter):
Catch exception in destructor to avoid crash.

5 years agoOBVIOUS move add_comm_alias "!" <=> "shell" near the add_com "shell"
Philippe Waroquiers [Fri, 19 Apr 2019 11:45:00 +0000 (13:45 +0200)]
OBVIOUS move add_comm_alias "!" <=> "shell" near the add_com "shell"

gdb/ChangeLog

2019-04-19  Philippe Waroquiers  <philippe.waroquiers@skynet.be>

* cli/cli-cmds.c (_initialize_cli_cmds): Move "shell" "!" alias
close to the add_com "shell".

5 years agoRX Assembler: Ensure that the internal limit on the number of relaxation iterations...
Nick Clifton [Fri, 19 Apr 2019 09:39:47 +0000 (10:39 +0100)]
RX Assembler: Ensure that the internal limit on the number of relaxation iterations is not larger that the external limit.

PR 24464
* config/tc-rx.h (md_relax_frag): Pass the max_iterations variable
to the relaxation function.
* config/tc-rx.c (rx_relax_frag): Add new parameter - the maximum
number of iterations.  Make sure that our internal iteration limit
does not exceed this external iteration limit.

5 years agos12z and h8300 no-print-map-discarded fails
Alan Modra [Fri, 19 Apr 2019 03:03:46 +0000 (12:33 +0930)]
s12z and h8300 no-print-map-discarded fails

This tidies the remaining --no-print-map-discarded fails.  h8300-elf
warns on a section without flags, and s12z doesn't support
--gc-sections.

bfd/
* elf32-s12z.c (elf_backend_can_gc_sections): Don't define
ld/
* testsuite/ld-gc/skip-map-discarded.s: Add section attributes.
* testsuite/lib/ld-lib.exp (check_gc_sections_available): Add
s12z to list of targets not supporting --gc-sections.

5 years agoAutomatic date update in version.in
GDB Administrator [Fri, 19 Apr 2019 00:00:12 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years ago[gdb/testsuite] Fix gdb.base/break-probes.exp with native-gdbserver
Tom de Vries [Thu, 18 Apr 2019 21:37:33 +0000 (23:37 +0200)]
[gdb/testsuite] Fix gdb.base/break-probes.exp with native-gdbserver

When running break-probes.exp with native-gdbserver, we run into:
...
FAIL: gdb.base/break-probes.exp: run til our library loads (the program exited)
FAIL: gdb.base/break-probes.exp: call (int) foo(23)
...
due to the fact that we're trying to match:
...
Inferior loaded /data/gdb_versions/devel/build/gdb/testsuite/outputs/gdb.base\
  /break-probes/break-probes-solib.so
...
using pattern:
...
Inferior loaded $sysroot$binfile_lib
...
which expands into:
...
Inferior loaded //data/gdb_versions/devel/build/gdb/testsuite/outputs/gdb.base\
  /break-probes/break-probes-solib.so
...

Fix by setting sysroot to "" in local-board.exp.

Tested on x86_64-linux with native-gdbserver.

gdb/testsuite/ChangeLog:

2019-04-18  Tom de Vries  <tdevries@suse.de>

PR gdb/24433
* boards/local-board.exp: Set sysroot to "".

5 years agoMake process_stratum_target::stratum "final"
Tom Tromey [Thu, 18 Apr 2019 15:42:46 +0000 (09:42 -0600)]
Make process_stratum_target::stratum "final"

It seemed to me that process_stratum_target::stratum ought to be
"final".

Tested by rebuilding, let me know what you think.

gdb/ChangeLog
2019-04-18  Tom Tromey  <tromey@adacore.com>

* process-stratum-target.h (class process_stratum_target)
<stratum>: Add "final".

5 years agoImprove warning message for $0 constraint on MIPSR6 branches
Matthew Fortune [Wed, 17 Apr 2019 04:07:19 +0000 (04:07 +0000)]
Improve warning message for $0 constraint on MIPSR6 branches

gas/
* config/tc-mips.c (match_non_zero_reg_operand): Update
warning message.
* testsuite/gas/mips/r6-branch-constraints.l: Likewise.

5 years ago[gdb] Handle vfork in thread with follow-fork-mode child
Tom de Vries [Thu, 18 Apr 2019 16:05:43 +0000 (17:05 +0100)]
[gdb] Handle vfork in thread with follow-fork-mode child

When debugging any of the testcases added by this commit, which do a
vfork in a thread with "set follow-fork-mode child" + "set
detach-on-fork on", we run into this assertion:

...
src/gdb/nat/x86-linux-dregs.c:146: internal-error: \
  void x86_linux_update_debug_registers(lwp_info*): \
  Assertion `lwp_is_stopped (lwp)' failed.
...

The assert is caused by the following: the vfork-child exit or exec
event is handled by handle_vfork_child_exec_or_exit, which calls
target_detach to detach from the vfork parent.  During target_detach
we call linux_nat_target::detach, which:

#1 - stops all the threads
#2 - waits for all the threads to be stopped
#3 - detaches all the threads

However, during the second step we run into this code in
stop_wait_callback:

...
  /* If this is a vfork parent, bail out, it is not going to report
     any SIGSTOP until the vfork is done with.  */
  if (inf->vfork_child != NULL)
    return 0;
...

and we don't wait for the threads to be stopped, which results in this
assert in x86_linux_update_debug_registers triggering during the third
step:

...
  gdb_assert (lwp_is_stopped (lwp));
...

The fix is to reset the vfork parent's vfork_child field before
calling target_detach in handle_vfork_child_exec_or_exit.  There's
already similar code for the other paths handled by
handle_vfork_child_exec_or_exit, so this commit refactors the code a
bit so that all paths share the same code.

The new tests cover both a vfork child exiting, and a vfork child
execing, since both cases would trigger the assertion.

The new testcases also exercise following the vfork children with "set
detach-on-fork off", since it doesn't seem to be tested anywhere.

Tested on x86_64-linux, using native and native-gdbserver.

gdb/ChangeLog:
2019-04-18  Tom de Vries  <tdevries@suse.de>
    Pedro Alves  <palves@redhat.com>

PR gdb/24454
* infrun.c (handle_vfork_child_exec_or_exit): Reset vfork parent's
vfork_child field before calling target_detach.

gdb/testsuite/ChangeLog:
2019-04-18  Tom de Vries  <tdevries@suse.de>
    Pedro Alves  <palves@redhat.com>

PR gdb/24454
* gdb.threads/vfork-follow-child-exec.c: New file.
* gdb.threads/vfork-follow-child-exec.exp: New file.
* gdb.threads/vfork-follow-child-exit.c: New file.
* gdb.threads/vfork-follow-child-exit.exp: New file.

5 years agoMSP430 Assembler: Define symbols for functions to run through.
Jozef Lawrynowicz [Thu, 18 Apr 2019 12:15:09 +0000 (13:15 +0100)]
MSP430 Assembler: Define symbols for functions to run through.

gas * config/tc-msp430.c (msp430_make_init_symbols): Define
__crt0_run_{preinit,init,fini}_array symbols if
.{preinit,init,fini}_array sections exist.
* testsuite/gas/msp430/fini-array.d: New test.
* testsuite/gas/msp430/init-array.d: New test.
* testsuite/gas/msp430/preinit-array.d: New test.
* testsuite/gas/msp430/fini-array.s: New test source.
* testsuite/gas/msp430/init-array.s: New test source.
* testsuite/gas/msp430/preinit-array.s: New test source.
* testsuite/gas/msp430/msp430.exp: Add new tests to driver.

5 years agoAutomatic date update in version.in
GDB Administrator [Thu, 18 Apr 2019 00:00:22 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agox86: Suggest -fPIE when not building shared object
H.J. Lu [Wed, 17 Apr 2019 17:22:19 +0000 (10:22 -0700)]
x86: Suggest -fPIE when not building shared object

When PIC is needed, linker should suggest -fPIE, instead of -fPIC, when
not building shared object.

bfd/

* elf64-x86-64.c (elf_x86_64_need_pic): Suggest -fPIE when not
building shared object.
* elfxx-x86.c (_bfd_x86_elf_size_dynamic_sections): Likewise.

ld/

* testsuite/ld-i386/i386.exp: Run pr18801a and pr18801b instead
of pr18801.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
* testsuite/ld-i386/pr18801.d: Removed.
* testsuite/ld-x86-64/pr18801.d: Likewise.
* testsuite/ld-i386/pr18801a.d: New file.
* testsuite/ld-i386/pr18801b.d: Likewise.
* testsuite/ld-x86-64/pr18801a.d: Likewise.
* testsuite/ld-x86-64/pr18801b.d: Likewise.
* testsuite/ld-x86-64/pie2.d: Suggest -fPIE instead of -fPIC.
* testsuite/ld-x86-64/pie2.d: Likewise.
* testsuite/ld-x86-64/pr19719.d: Likewise.
* testsuite/ld-x86-64/pr19807-2a.d: Likewise.
* testsuite/ld-x86-64/pr19969.d: Likewise.
* testsuite/ld-x86-64/pr21997-1a.err: Likewise.
* testsuite/ld-x86-64/pr21997-1b.err: Likewise.
* testsuite/ld-x86-64/pr22001-1a.err: Likewise.
* testsuite/ld-x86-64/pr22001-1b.err: Likewise.
* testsuite/ld-x86-64/pr22791-1.err: Likewise.

5 years agox86: Also check x86 linker_def for non-shared definition
H.J. Lu [Wed, 17 Apr 2019 16:08:46 +0000 (09:08 -0700)]
x86: Also check x86 linker_def for non-shared definition

Since elf_x86_linker_defined sets linker_def in elf_x86_link_hash_entry
for linker defined symbols, SYMBOL_DEFINED_NON_SHARED_P should also check
linker_def in elf_x86_link_hash_entry.

bfd/

PR ld/24458
* elfxx-x86.h (SYMBOL_DEFINED_NON_SHARED_P): Also check x86
linker_def.

ld/

PR ld/24458
* testsuite/ld-x86-64/x86-64.exp: Run PR ld/24458 tests.
* testsuite/ld-x86-64/pr24458.s: New file.
* testsuite/ld-x86-64/pr24458a-x32.d: Likewise.
* testsuite/ld-x86-64/pr24458a.d: Likewise.
* testsuite/ld-x86-64/pr24458b-x32.d: Likewise.
* testsuite/ld-x86-64/pr24458b.d: Likewise.
* testsuite/ld-x86-64/pr24458c-x32.d: Likewise.
* testsuite/ld-x86-64/pr24458c.d: Likewise.

5 years agoMSP430 Linker: Define __crt0_init_bss/__crt0_movedata symbols when .lower or .either...
Jozef Lawrynowicz [Wed, 17 Apr 2019 14:05:08 +0000 (15:05 +0100)]
MSP430 Linker: Define __crt0_init_bss/__crt0_movedata symbols when .lower or .either prefixed sections are present.

ld * config/tc-msp430.c (msp430_make_init_symbols): Define __crt0_init_bss
symbol when .lower.bss or .either.bss sections exist.
Define __crt0_movedata when .lower.data or .either.data sections exist.
* testsuite/gas/msp430/either-data-bss-sym.d: New test.
* testsuite/gas/msp430/low-data-bss-sym.d: New test.
* testsuite/gas/msp430/either-data-bss-sym.s: New test source.
* testsuite/gas/msp430/low-data-bss-sym.s: New test source.
* testsuite/gas/msp430/msp430.exp: Run new tests.
Enable large code model when running -mdata-region={upper,either}
tests.

5 years agoMSP430 Assembler: Leave placement of .lower and .upper sections to generic linker...
Jozef Lawrynowicz [Wed, 17 Apr 2019 14:03:27 +0000 (15:03 +0100)]
MSP430 Assembler: Leave placement of .lower and .upper sections to generic linker code.

* config/tc-msp430.c (msp430_make_init_symbols): Define __crt0_init_bss
symbol when .lower.bss or .either.bss sections exist.
Define __crt0_movedata when .lower.data or .either.data sections exist.
* testsuite/gas/msp430/either-data-bss-sym.d: New test.
* testsuite/gas/msp430/low-data-bss-sym.d: New test.
* testsuite/gas/msp430/either-data-bss-sym.s: New test source.
* testsuite/gas/msp430/low-data-bss-sym.s: New test source.
* testsuite/gas/msp430/msp430.exp: Run new tests.
Enable large code model when running -mdata-region={upper,either}
tests.

5 years agoMSP420 assembler: Add -m{u,U} options to enable/disable NOP warnings for unknown...
Jozef Lawrynowicz [Wed, 17 Apr 2019 14:01:28 +0000 (15:01 +0100)]
MSP420 assembler:  Add -m{u,U} options to enable/disable NOP warnings for unknown interrupt state changes

gas * config/tc-msp430.c (options): New OPTION_UNKNOWN_INTR_NOPS,
OPTION_NO_UNKNOWN_INTR_NOPS and do_unknown_interrupt_nops.
(md_parse_option): Handle OPTION_UNKNOWN_INTR_NOPS and
OPTION_NO_UNKNOWN_INTR_NOPS by setting do_unknown_interrupt_nops
accordingly.
(md_show_usage): Likewise.
(md_shortopts): Add "mu" for OPTION_UNKNOWN_INTR_NOPS and
"mU" for OPTION_NO_UNKNOWN_INTR_NOPS.
(md_longopts): Likewise.
(warn_eint_nop): Update comment.
(warn_unsure_interrupt): Don't warn if prev_insn_is_nop or
prev_insn_is_dint or we are assembling for 430 ISA.
(msp430_operands): Only call warn_unsure_interrupt if
do_unknown_interrupt_nops == TRUE.
* testsuite/gas/msp430/nop-unknown-intr.s: New test source file.
* testsuite/gas/msp430/nop-unknown-intr-430.d: New test.
* testsuite/gas/msp430/nop-unknown-intr-430x.d: New test.
* testsuite/gas/msp430/nop-unknown-intr-430x-ignore.d: New test.
* testsuite/gas/msp430/nop-unknown-intr-430.l: Warning output for new
test.
* testsuite/gas/msp430/nop-unknown-intr-430x.l: Likewise.
* testsuite/gas/msp430/msp430.exp: Add new tests to driver.

5 years agoAvoid crash in dwarf2_init_complex_target_type
Tom Tromey [Tue, 16 Apr 2019 18:12:09 +0000 (12:12 -0600)]
Avoid crash in dwarf2_init_complex_target_type

After commit 35add35 ("gdb: Fix failure in gdb.base/complex-parts.exp
for x86-32"), dwarf2_init_complex_target_type can crash if "tt" is
nullptr.  This patch avoids the problem by checking for this case.

No test case because I don't know a good way to write one; it was
found by an internal AdaCore test case that apparently uses a 16 bit
floating point type.

gdb/ChangeLog:
* dwarf2read.c (dwarf2_init_complex_target_type): Check "tt"
against nullptr before use.

gdb/ChangeLog
2019-04-17  Tom Tromey  <tromey@adacore.com>

* dwarf2read.c (dwarf2_init_complex_target_type): Check "tt"
against nullptr before use.

5 years agogdbserver: Ensure all debug output uses debug functions
Alan Hayward [Tue, 16 Apr 2019 09:37:47 +0000 (10:37 +0100)]
gdbserver: Ensure all debug output uses debug functions

All debug output needs to go via debug functions to ensure it writes to the
correct output stream.

gdb/ChangeLog:

* nat/linux-waitpid.c (linux_debug): Call debug_vprintf.

gdb/gdbserver/ChangeLog:

* ax.c (ax_vdebug): Call debug_printf.
* debug.c (debug_write): New function.
* debug.h (debug_write): New declaration.
* linux-low.c (sigchld_handler): Call debug_write.

5 years agogdbserver: Add debug-file option
Alan Hayward [Fri, 12 Apr 2019 14:49:11 +0000 (15:49 +0100)]
gdbserver: Add debug-file option

Add command line option to send all debug output to a given file.
Always default back to stderr.

Add matching monitor command. Add documentation.

gdb/doc/ChangeLog:

* gdb.texinfo
(Other Command-Line Arguments for gdbserver): Add debug-file
option.
(Monitor Commands for gdbserver): Likewise.
(gdbserver man): Likewise.

gdb/gdbserver/ChangeLog:

* debug.c (debug_set_output): New function.
(debug_vprintf): Send output to debug_file.
(debug_flush): Likewise.
* debug.h (debug_set_output): New declaration.
* server.c (handle_monitor_command): Add debug-file option.
(captured_main): Likewise.

5 years agogdbserver: Move remote_debug to a single place
Alan Hayward [Fri, 12 Apr 2019 13:21:55 +0000 (14:21 +0100)]
gdbserver: Move remote_debug to a single place

A comment in debug.h (written in 2014) states: "We declare debug format
variables here, and debug_threads but no other debug content variables
(e.g., not remote_debug) because while this file is not currently used by
IPA it may be some day, and IPA may have its own set of debug content
variables".

This has resulted in remote_debug being declared in many .c/.h files
throughout gdbserver.

It would be much simplier to define it one place.  The most logical place to
define it is in debug.h, surrounded by #define guards.  If IPA is changed,
then at that point the variable can be moved elsewhere.

gdb/gdbserver/ChangeLog:

* debug.c (remote_debug): Add definition.
* debug.h (remote_debug): Add declaration.
* hostio.c (remote_debug): Remove declaration.
* remote-utils.c (struct ui_file): Likewise.
(remote_debug): Likewise.
* remote-utils.h (remote_debug): Likewise,
* server.c (remote_debug): Remove definition.

5 years agoAutomatic date update in version.in
GDB Administrator [Wed, 17 Apr 2019 00:00:31 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agogdb/riscv: Allow breakpoints to be created at invalid addresses
Andrew Burgess [Tue, 16 Apr 2019 23:31:43 +0000 (00:31 +0100)]
gdb/riscv: Allow breakpoints to be created at invalid addresses

Some testsuite cases (gdb.cp/nsalias.exp for example) construct dwarf2
debug info for fake functions to test that this debug info is handled
correctly.

We currently get an error trying to read from an invalid address while
creating breakpoints for these fake functions.

Other targets allow creating breakpoints on invalid addresses, and
only error when GDB actually tries to insert the breakpoints.

In order to make RISC-V behave in the same way as other targets, this
commit makes the failure to read memory during breakpoint creation
non-fatal, we then expect to see a failure when GDB tries to insert
the breakpoint, just like other targets.

Tested with a riscv64-linux native testsuite run.

gdb/ChangeLog:

* riscv-tdep.c (riscv_breakpoint_kind_from_pc): Hanndle case where
code read might fail, assume 4-byte breakpoint in that case.

5 years agoxfail gas weakref1 test for nds32
Alan Modra [Tue, 16 Apr 2019 12:04:22 +0000 (21:34 +0930)]
xfail gas weakref1 test for nds32

Oops, I removed the wrong xfail from gas.exp in last commit, fix it
here.

* testsuite/gas/all/weakref1.d: xfail nds32.

5 years agons32k testsuite tidy
Alan Modra [Tue, 16 Apr 2019 08:38:11 +0000 (18:08 +0930)]
ns32k testsuite tidy

Some of these tests were excluded for ns32k-netbsd, exclude for all
ns32k instead.

binutils/
* testsuite/binutils-all/copy-2.d: Don't run for ns32k-*-*.
* testsuite/binutils-all/copy-3.d: Likewise.
gas/
* testsuite/gas/all/gas.exp: Remove ns32k xfails.
* testsuite/gas/all/weakref1u.d: Don't run for ns32k-*-*.
ld/
* testsuite/ld-scripts/pr20302.d: Don't run for ns32k-*-*.
* testsuite/ld-scripts/section-match-1.d: Likewise.
* testsuite/ld-undefined/require-defined.exp: Likewise.

5 years agoMove fixup fx_bit_fixP and fx_im_disp fields to TC_FIX_TYPE
Alan Modra [Tue, 16 Apr 2019 06:17:02 +0000 (15:47 +0930)]
Move fixup fx_bit_fixP and fx_im_disp fields to TC_FIX_TYPE

These are only used by dlx and ns32k.

* write.h: Don't include bit_fix.h.
(struct fix): Rearrange some fields.  Delete fx_im_disp and
fx_bit_fixP.  Use bitfields for fx_size and fx_pcrel_adjust.
* write.c (fix_new_internal): Don't init fx_im_disp and fx_bit_fixP.
(fixup_segment): Don't exclude overflow checks on fx_bit_fixP.
(print_fixup): Don't print im_disp.
* config/tc-cris.c (md_apply_fix): Remove tests of fx_bit_fixP
and fx_im_disp.
* config/tc-dlx.c (md_apply_fix): Remove wrong debug code.  Set
fx_no_overflow when fx_bit_fixP.
* config/tc-dlx.h: Include bit_fix.h.
(TC_FIX_TYPE, tc_fix_data, TC_INIT_FIX_DATA): Define.
* config/tc-ns32k.c (fix_new_ns32k, fix_new_ns32k_exp): Set
fx_no_overflow when bit_fixP.
* config/tc-ns32k.h (TC_FIX_TYPE): Add fx_bit_fixP and fx_im_disp.
(fix_im_disp, fix_bit_fixP): Adjust to suit.
(TC_INIT_FIX_DATA, TC_FIX_DATA_PRINT): Likewise.

5 years agoMake fixup fx_where unsigned
Alan Modra [Tue, 16 Apr 2019 01:40:44 +0000 (11:10 +0930)]
Make fixup fx_where unsigned

Another field that only stores unsigned values.

* write.h (struct fix <fx_where>): Make unsigned.
(fix_new, fix_at_start, fix_new_exp): Adjust prototypes.
* write.c (fix_new, fix_new_exp, fix_at_start): Make "where" and
"size" parameters unsigned long.
(fix_new_internal): Likewise.  Adjust error format string to suit.
* config/tc-mips.c (md_convert_frag): Remove cast of fx_where.
* config/tc-sparc.c (md_apply_fix): Likewise.
* config/tc-score.c (s3_convert_frag): Adjust for unsigned fx_where.
* config/tc-score7.c (s7_convert_frag): Likewise.

5 years agoMake frag fr_fix unsigned
Alan Modra [Mon, 15 Apr 2019 12:21:44 +0000 (21:51 +0930)]
Make frag fr_fix unsigned

The field only stores unsigned values, so let's make it unsigned to
stop people worrying about the possibility of negative values.

* frags.h (struct frag <fr_fix>): Use unsigned type.
* frags.c (frag_new): Assert that current size exceeds
old_frags_var_max_size.
* ehopt.c (get_cie_info): Adjust for unsigned fr_fix.
* listing.c (calc_hex): Likewise.
* write.c (cvt_frag_to_fill, write_relocs): Likewise.
* config/tc-arc.c (md_convert_frag): Likewise.
* config/tc-avr.c (avr_patch_gccisr_frag): Likewise.
* config/tc-mips.c (md_convert_frag): Likewise.
* config/tc-rl78.c (md_convert_frag): Likewise.
* config/tc-rx.c (md_convert_frag): Likewise.
* config/tc-sparc.c (md_apply_fix): Likewise.
* config/tc-xtensa.c (next_instrs_are_b_retw): Likewise.
(unrelaxed_frag_min_insn_count, unrelaxed_frag_has_b_j): Likewise.

5 years agoAutomatic date update in version.in
GDB Administrator [Tue, 16 Apr 2019 00:00:22 +0000 (00:00 +0000)]
Automatic date update in version.in

5 years agoFix AMD64 return value ABI in expression evaluation
Leszek Swirski [Mon, 15 Apr 2019 15:56:43 +0000 (11:56 -0400)]
Fix AMD64 return value ABI in expression evaluation

The AMD64 System V ABI specifies that when a function has a return type
classified as MEMORY, the caller provides space for the value and passes
the address to this space as the first argument to the function (before
even the "this" pointer). The classification of MEMORY is applied to
struct that are sufficiently large, or ones with unaligned fields.

The expression evaluator uses call_function_by_hand to call functions,
and the hand-built frame has to push arguments in a way that matches the
ABI of the called function. call_function_by_hand supports ABI-based
struct returns, based on the value of gdbarch_return_value, however on
AMD64 the implementation of the classifier incorrectly assumed that all
non-POD types (implemented as "all types with a base class") should be
classified as MEMORY and use the struct return.

This ABI mismatch resulted in issues when calling a function that returns
a class of size <16 bytes which has a base class, including issues such
as the "this" pointer being incorrect (as it was passed as the second
argument rather than the first).

This is now fixed by checking for field alignment rather than POD-ness,
and a testsuite is added to test expression evaluation for AMD64.

gdb/ChangeLog:

* amd64-tdep.c (amd64_classify_aggregate): Use cp_pass_by_reference
rather than a hand-rolled POD check when checking for forced MEMORY
classification.

gdb/testsuite/ChangeLog:

* gdb.arch/amd64-eval.cc: New file.
* gdb.arch/amd64-eval.exp: New file.

5 years agoSkip print-map-discarded test for non-ELF targets
Faraz Shahbazker [Mon, 15 Apr 2019 15:48:42 +0000 (08:48 -0700)]
Skip print-map-discarded test for non-ELF targets

ld/
* testsuite/ld-gc/gc.exp: Skip print-map-discarded test
for non-ELF targets.