external/binutils.git
9 years ago[gold] The first patch to address cortex-a53 erratum-843419.
Han Shen [Thu, 16 Apr 2015 22:17:08 +0000 (15:17 -0700)]
[gold] The first patch to address cortex-a53 erratum-843419.

It implemented scanning the binary and reporting occurrences to users
when '--fix-cortex-a53' is turned on. With this, gold users will be
able to see if or not there are such erratum occurrences in the output
binary. Also included in the CL is reading/recording mapping symbols,
which is needed during scan.

gold/ChangeLog:
    * aarch64.cc (AArch64_insn_utilities): New utility class.
    (AArch64_relobj::Mapping_symbol_position): New struct.
    (AArch64_relobj::Mapping_symbol_info): New typedef.
    (AArch64_relobj::do_count_local_symbols): New function overriding
    parent's implementation.
    (AArch64_relobj::mapping_symbol_info_): New member
    (AArch64_relobj::scan_erratum_843419): New method.
    (Target_aarch64::scan_erratum_843419_span): New method.
    (Target_aarch64::is_erratum_843419_sequence): New method.
    * options.h (fix_cortex_a53): New option.

9 years agoFix {mi-tracepoint-changed, mi-tsv-changed}.exp with native-extended-gdbserver
Pedro Alves [Thu, 16 Apr 2015 11:23:26 +0000 (12:23 +0100)]
Fix {mi-tracepoint-changed, mi-tsv-changed}.exp with native-extended-gdbserver

Fixes:

 -FAIL: gdb.trace/mi-tracepoint-changed.exp: reconnect: break-info 1
 +PASS: gdb.trace/mi-tracepoint-changed.exp: reconnect: tracepoint created
 +PASS: gdb.trace/mi-tracepoint-changed.exp: reconnect: tracepoint on marker is installed
 +PASS: gdb.trace/mi-tracepoint-changed.exp: reconnect: break-info 1

 -FAIL: gdb.trace/mi-tsv-changed.exp: upload: tsv1 created
 -FAIL: gdb.trace/mi-tsv-changed.exp: upload: tsv2 created
 +PASS: gdb.trace/mi-tsv-changed.exp: upload: tsv1 created
 +PASS: gdb.trace/mi-tsv-changed.exp: upload: tsv2 created

These tests do something like this:

 #0 - start gdb/gdbserver normally
 #1 - setup some things in the debug session
 #2 - disconnect from gdbserver
 #3 - restart gdb
 #4 - reconnect to gdbserver

The problem is that the native-extended-gdbserver board always spawns
a new gdbserver instance in #3 (and has gdb connect to that).  So when
the test gets to #4, it connects to that new instance instead of the
old one:

 (gdb) spawn ../gdbserver/gdbserver --multi :2354
 Listening on port 2354
 target extended-remote localhost:2354
 Remote debugging using localhost:2354
 ...
 spawn ../gdbserver/gdbserver --multi :2355
 Listening on port 2355
 47-target-select extended-remote localhost:2355
 =tsv-created,name="trace_timestamp",initial="0"\n
 47^connected
 (gdb)
 ...
 47-target-select extended-remote localhost:2355
 47^connected
 (gdb)
 FAIL: gdb.trace/mi-tsv-changed.exp: upload: tsv1 created
 FAIL: gdb.trace/mi-tsv-changed.exp: upload: tsv2 created

testsuite/ChangeLog:
2015-04-16  Pedro Alves  <palves@redhat.com>

* boards/native-extended-gdbserver.exp (mi_gdb_start): Don't start
a new gdbserver if gdbserver_reconnect_p is set.

9 years agoFix gdbserver_reconnect_p handling
Pedro Alves [Thu, 16 Apr 2015 13:26:59 +0000 (14:26 +0100)]
Fix gdbserver_reconnect_p handling

Commit 6423214f (testsuite: Don't use expect_background to reap
gdbserver) broke a couple tests that set gdbserver_reconnect_p and
restart gdb before reconnecting, because a gdb_exit (e.g., through
clean_restart) exits gdbserver unconditionally.

Fixes, with --target_board=native-gdbserver:

 -FAIL: gdb.trace/mi-tracepoint-changed.exp: reconnect: break-info 1
 +PASS: gdb.trace/mi-tracepoint-changed.exp: reconnect: tracepoint created
 +PASS: gdb.trace/mi-tracepoint-changed.exp: reconnect: tracepoint on marker is installed
 +PASS: gdb.trace/mi-tracepoint-changed.exp: reconnect: break-info 1

 -FAIL: gdb.trace/mi-tsv-changed.exp: upload: tsv1 created
 -FAIL: gdb.trace/mi-tsv-changed.exp: upload: tsv2 created
 +PASS: gdb.trace/mi-tsv-changed.exp: upload: tsv1 created
 +PASS: gdb.trace/mi-tsv-changed.exp: upload: tsv2 created

gdb/testsuite/
2015-04-16  Pedro Alves  <palves@redhat.com>

* lib/gdbserver-support.exp (gdb_exit): If gdbserver_reconnect_p
is set, don't exit gdbserver.

9 years agoHonour software single step in fallback of displaced stepping
Yao Qi [Thu, 16 Apr 2015 12:48:10 +0000 (13:48 +0100)]
Honour software single step in fallback of displaced stepping

Hi,
When I run gdb.threads/non-stop-fair-events.exp on arm-linux target,
I see the following message in the debugging log,

  displaced: breakpoint is gone: Thread 22518, step(1)^M
  Sending packet: $vCont;s:p57f3.57f6#9d...
                  ^^^^^^^^^
GDB sends vCont;s by mistake, and GDBserver fails on assert.  GDB
doesn't consider software single step in infrun.c:displaced_step_fixup,

  /* Go back to what we were trying to do.  */
  step = currently_stepping (tp);

  if (debug_displaced)
    fprintf_unfiltered (gdb_stdlog,
"displaced: breakpoint is gone: %s, step(%d)\n",
target_pid_to_str (tp->ptid), step);

  target_resume (ptid, step, GDB_SIGNAL_0);

The patch is to let GDB consider software single step here.  It fixes
fails in gdb.threads/non-stop-fair-events.exp on arm.

gdb:

2015-04-16  Yao Qi  <yao.qi@linaro.org>

* infrun.c (maybe_software_singlestep): Declare.
(displaced_step_fixup): Call maybe_software_singlestep.

9 years agos390-vregs.exp: Avoid compile errors with older GCCs and on 31-bit targets
Andreas Arnez [Thu, 16 Apr 2015 11:03:48 +0000 (13:03 +0200)]
s390-vregs.exp: Avoid compile errors with older GCCs and on 31-bit targets

The test case s390-vregs.exp yields compile errors on 31-bit targets
as well as when using a GCC that defaults to an older "-march=".  This
patch fixes these issues.

gdb/testsuite/ChangeLog:

* gdb.arch/s390-vregs.S (change_vrs): Replace exrl by an
appropriate .insn, such that an older assembler can be used.
* gdb.arch/s390-vregs.exp: Add the compile flag -mzarch, to enable
the z/Architecture instruction set on 31-bit targets as well.

9 years agoGDB tests for Go language support: remove unnecessary first breakpoint
Andreas Arnez [Thu, 16 Apr 2015 11:03:47 +0000 (13:03 +0200)]
GDB tests for Go language support: remove unnecessary first breakpoint

On s390x targets some of the Go test cases fail because the first
breakpoint happens to be at the same spot as the breakpoint at
main.main.  When such a test case tries to continue to the first
breakpoint, the program runs until the end instead, and the test fails
like this:

FAIL: gdb.go/handcall.exp: Going to first breakpoint (the program exited)

This patch removes all the handling related to the first breakpoint in
those cases.  After applying the patch, the tests run successfully on
s390x.

gdb/testsuite/ChangeLog:

* gdb.go/handcall.exp: Remove all logic related to the first
breakpoint and rely on go_runto_main instead.
* gdb.go/strings.exp: Likewise.
* gdb.go/unsafe.exp: Likewise.
* gdb.go/hello.exp: Likewise.  Also rename the remaining
breakpoint marker to "breakpoint 1".
* gdb.go/handcall.go: Remove comment "set breakpoint 1 here".
* gdb.go/strings.go: Likewise.
* gdb.go/unsafe.go: Likewise.
* gdb.go/hello.go: Likewise.  Also remove the second occurrence of
"set breakpoint 2 here" and rename the remaining breakpoint marker
to "breakpoint 1".

9 years agosim: avr/mcore/moxie: fill out sim-cpu pc fetch/store helpers
Mike Frysinger [Thu, 16 Apr 2015 06:07:33 +0000 (02:07 -0400)]
sim: avr/mcore/moxie: fill out sim-cpu pc fetch/store helpers

This makes the common sim-cpu logic work.

9 years agoAutomatic date update in version.in
GDB Administrator [Thu, 16 Apr 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoRemove the unused PREFIX_UD_XXX
H.J. Lu [Wed, 15 Apr 2015 22:58:45 +0000 (15:58 -0700)]
Remove the unused PREFIX_UD_XXX

Remove the unused PREFIX_UD_XXX.  Invalid opcodes should be handled by
prefix_table.

* i386-dis.c (PREFIX_UD_SHIFT): Removed.
(PREFIX_UD_REPZ): Likewise.
(PREFIX_UD_REPNZ): Likewise.
(PREFIX_UD_DATA): Likewise.
(PREFIX_UD_ADDR): Likewise.
(PREFIX_UD_LOCK): Likewise.

9 years agoMake info fun|var|types interruptable for psyms.
Doug Evans [Wed, 15 Apr 2015 21:04:35 +0000 (14:04 -0700)]
Make info fun|var|types interruptable for psyms.

gdb/ChangeLog:

* psymtab.c (psym_expand_symtabs_matching): Add QUIT call.

9 years agoMake info fun|var|types interruptable.
Doug Evans [Wed, 15 Apr 2015 20:25:42 +0000 (13:25 -0700)]
Make info fun|var|types interruptable.

"info fun foo" can be a pain when it's not interruptable,
especially if you're not exactly sure of what you're looking for
and provide something that matches too much.

gdb/ChangeLog:

* dwarf2read.c (dw2_expand_symtabs_matching): Add some QUIT calls.

9 years agoCheck dp->prefix_requirement instead
H.J. Lu [Wed, 15 Apr 2015 18:28:16 +0000 (11:28 -0700)]
Check dp->prefix_requirement instead

This patch removes prefix_requirement and checks dp->prefix_requirement
instead.

* i386-dis.c (prefix_requirement): Removed.
(print_insn): Don't set prefix_requirement.  Check
dp->prefix_requirement instead of prefix_requirement.

9 years agoHandle invalid prefixes for rdrand and rdseed
H.J. Lu [Wed, 15 Apr 2015 16:53:13 +0000 (09:53 -0700)]
Handle invalid prefixes for rdrand and rdseed

This patch puts rdrand and rdseed in prefix_table so that invalid
prefixes for rdrand and rdseed are handled properly.

gas/testsuite/

PR binutils/17898
* gas/i386/prefix.s: Add rdrand/rdseed prefix tests.
* gas/i386/prefix.d: Updated.

opcodes/

PR binutils/17898
* i386-dis.c (PREFIX_0FC7_REG_6): Renamed to ...
(PREFIX_MOD_0_0FC7_REG_6): This.
(PREFIX_MOD_3_0FC7_REG_6): New.
(PREFIX_MOD_3_0FC7_REG_7): Likewise.
(prefix_table): Replace PREFIX_0FC7_REG_6 with
PREFIX_MOD_0_0FC7_REG_6.  Add PREFIX_MOD_3_0FC7_REG_6 and
PREFIX_MOD_3_0FC7_REG_7.
(mod_table): Replace PREFIX_0FC7_REG_6 with
PREFIX_MOD_0_0FC7_REG_6.  Use PREFIX_MOD_3_0FC7_REG_6 and
PREFIX_MOD_3_0FC7_REG_7.

9 years agoReplace mandatory_prefix with prefix_requirement
H.J. Lu [Wed, 15 Apr 2015 16:24:45 +0000 (09:24 -0700)]
Replace mandatory_prefix with prefix_requirement

* i386-dis.c (PREFIX_MANDATORY_REPZ): Removed.
(PREFIX_MANDATORY_REPNZ): Likewise.
(PREFIX_MANDATORY_DATA): Likewise.
(PREFIX_MANDATORY_ADDR): Likewise.
(PREFIX_MANDATORY_LOCK): Likewise.
(PREFIX_MANDATORY): Likewise.
(PREFIX_UD_SHIFT): Set to 8
(PREFIX_UD_REPZ): Updated.
(PREFIX_UD_REPNZ): Likewise.
(PREFIX_UD_DATA): Likewise.
(PREFIX_UD_ADDR): Likewise.
(PREFIX_UD_LOCK): Likewise.
(PREFIX_IGNORED_SHIFT): New.
(PREFIX_IGNORED_REPZ): Likewise.
(PREFIX_IGNORED_REPNZ): Likewise.
(PREFIX_IGNORED_DATA): Likewise.
(PREFIX_IGNORED_ADDR): Likewise.
(PREFIX_IGNORED_LOCK): Likewise.
(PREFIX_OPCODE): Likewise.
(PREFIX_IGNORED): Likewise.
(Bad_Opcode): Replace PREFIX_MANDATORY with 0.
(dis386_twobyte): Replace PREFIX_MANDATORY with PREFIX_OPCODE.
(three_byte_table): Likewise.
(mod_table): Likewise.
(mandatory_prefix): Renamed to ...
(prefix_requirement): This.
(prefix_table): Replace PREFIX_MANDATORY with PREFIX_OPCODE.
Update PREFIX_90 entry.
(get_valid_dis386): Check prefix_requirement to see if a prefix
should be ignored.
(print_insn): Replace mandatory_prefix with prefix_requirement.

9 years ago[ARM] Disassembles SSAT and SSAT16 instructions incorrectly for Thumb-2
Renlin Li [Wed, 15 Apr 2015 16:44:03 +0000 (17:44 +0100)]
[ARM] Disassembles SSAT and SSAT16 instructions incorrectly for Thumb-2

2015-04-15  Renlin Li  <renlin.li@arm.com>
opcodes/:
    * arm-dis.c (thumb32_opcodes): Define 'D' format control code,
    use it for ssat and ssat16.
    (print_insn_thumb32): Add handle case for 'D' control code.

gas/testsuite/:
    * gas/arm/arch7em.d: Adjust required ssat and ssat16 immediate field.
    * gas/arm/thumb32.d: Likewise.

9 years agogdbserver: fix uClibc build whithout MMU.
Romain Naour [Tue, 14 Apr 2015 21:07:34 +0000 (23:07 +0200)]
gdbserver: fix uClibc build whithout MMU.

Since commit d86d4aafd4fa22fa4cccb83253fb187b03f97f48, the pid
must be retrieved from current_thread.

The change has not been made in the function linux_read_offsets().

Fixes:
http://autobuild.buildroot.net/results/9e4/9e4df085319e346803c26c65478accb27eb950ae/build-end.log

2015-04-14  Romain Naour <romain.naour@openwide.fr>  (tiny change)

* linux-low.c (linux_read_offsets): Remove get_thread_lwp.

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
9 years agoSome Python 3 fixes
Simon Marchi [Wed, 15 Apr 2015 15:54:33 +0000 (11:54 -0400)]
Some Python 3 fixes

Some missing parentheses and one itertools.imap (Py2) vs map (Py3) issue.

gdb/ChangeLog:

* python/lib/gdb/command/unwinders.py: Add parentheses.

gdb/testsuite/ChangeLog:

* gdb.python/py-framefilter.py (ErrorFilter.filter): Use map function
if itertools.imap is not present.
* gdb.python/py-objfile.exp: Add parentheses.
* gdb.python/py-type.exp: Same.
* gdb.python/py-unwind-maint.py: Same.

9 years ago[arm] Update displaced stepping debug message
Yao Qi [Wed, 15 Apr 2015 14:09:53 +0000 (15:09 +0100)]
[arm] Update displaced stepping debug message

When I "set debug displaced 1" to fix fail in
gdb.base/disp-step-syscall.exp, the debug message is wrong.  This
patch is to fix it.

gdb:

2015-04-15  Yao Qi  <yao.qi@linaro.org>

* arm-linux-tdep.c (arm_linux_copy_svc): Update debug message.

9 years agoFix code indentation
Yao Qi [Wed, 15 Apr 2015 13:55:04 +0000 (14:55 +0100)]
Fix code indentation

gdb:

2015-04-15  Yao Qi  <yao.qi@linaro.org>

* arm-linux-tdep.c (arm_linux_copy_svc): Fix indentation.

9 years ago[arm] Fix fails in gdb.base/disp-step-syscall.exp
Yao Qi [Wed, 15 Apr 2015 13:34:03 +0000 (14:34 +0100)]
[arm] Fix fails in gdb.base/disp-step-syscall.exp

Hi,
I see this fail on arm-linux target,

 FAIL: gdb.base/disp-step-syscall.exp: fork: single step over fork final pc

which is caused by the PC isn't expected after displaced stepping the
svc instruction.  The code is:

=> 0xb6ead9a4 <__libc_do_syscall+4>:    svc     0
   0xb6ead9a6 <__libc_do_syscall+6>:    pop     {r7, pc}
   0xb6ead9a8:  nop.w^M
   0xb6ead9ac:  nop.w

after single step svc instruction, pc should be 0xb6ead9a6, but the
actual value of pc is 0xb6ead9a8.  The problem is illustrated by
turning on debug message of displaced stepping,

stepi^M
displaced: stepping Thread 12031 now^M
displaced: saved 0x8574: 02 bc 6a 46 04 b4 01 b4 df f8 10 c0 4d f8 04 cd 03 48 04 4b ff f7 d2 ef ff f7 e8 ef 0d 87 00 00 ^M
displaced: process thumb insn df00 at b6ead9a4^M
displaced: copying svc insn df00^M
displaced: read r7 value 00000078^M
displaced: sigreturn/rt_sigreturn SVC call not in signal trampoline frame^M
displaced: writing insn df00 at 00008574^M
displaced: copy 0xb6ead9a4->0x8574: displaced: check mode of b6ead9a4 instead of 00008574^M
displaced: displaced pc to 0x8574^M
displaced: run 0x8574: 00 df 01 de ^M
displaced: restored Thread 12031 0x8574^M
displaced: PC is apparently 00008576 after SVC step (within scratch space)^M
displaced: writing pc b6ead9a8  <----- WRONG ADDRESS

GDB writes the wrong address back to pc because GDB thinks the
instruction size is 4, which isn't true for thumb instruction.
This patch is to replace 4 with dsc->insn_size.

gdb:

2015-04-15  Yao Qi  <yao.qi@linaro.org>

* arm-linux-tdep.c (arm_linux_cleanup_svc): Use
dsc->insn_size instead of 4.

9 years agoFix fails in gdb.dwarf2/dynarr-ptr.exp
Yao Qi [Wed, 15 Apr 2015 13:04:40 +0000 (14:04 +0100)]
Fix fails in gdb.dwarf2/dynarr-ptr.exp

I see many fails in gdb.dwarf2/dynarr-ptr.exp on arm-linux target,
started from this

print foo.three_ptr.all^M
Cannot access memory at address 0x107c8^M
(gdb) FAIL: gdb.dwarf2/dynarr-ptr.exp: print foo.three_ptr.all
print foo.three_ptr.all(1)^M
Cannot access memory at address 0x107c8

It turns out that ":$ptr_size" is used incorrectly.

             array_ptr_label: DW_TAG_pointer_type {
                 {DW_AT_byte_size :$ptr_size }
                                  ^^^^^^^^^^
                 {DW_AT_type :$array_label}
             }

Since the FORM isn't given, and it starts with the ":", it is regarded
as a label reference by dwarf assembler.  The generated asm file on
x86_64 is

        .uleb128        6               /* Abbrev (DW_TAG_pointer_type) */
        .4byte        8 - .Lcu1_begin   <----- WRONG
        .4byte        .Llabel2 - .Lcu1_begin

Looks .Lcu1_begin is 0 on x86_64 and that is why this test passes on
x86_64.  On arm, .Lcu1_begin is an address somewhere, and the value
of DW_AT_byte_size is a very large number, so memory read request
of such large length failed.

This patch is to remove ":" and set the form explicitly.  The generated
asm file on x86_64 becomes

        .uleb128        6               /* Abbrev (DW_TAG_pointer_type) */
        .byte        8
        .4byte        .Llabel2 - .Lcu1_begin

gdb/testsuite:

2015-04-15  Yao Qi  <yao.qi@linaro.org>

* gdb.dwarf2/dynarr-ptr.exp (assemble): Use $ptr_size instead
of ":$ptr_size" and set its form explicitly.

9 years agoMention --compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi]
H.J. Lu [Wed, 15 Apr 2015 12:25:58 +0000 (05:25 -0700)]
Mention --compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi]

binutils/

* NEWS: Mention
--compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi].

gas/

* NEWS: Mention
--compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi].

ld/

* NEWS: Mention
--compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi].

9 years agoIncrease timeout in watch-bitfields.exp for software watchpoint
Yao Qi [Wed, 15 Apr 2015 11:46:58 +0000 (12:46 +0100)]
Increase timeout in watch-bitfields.exp for software watchpoint

I see the following two timeout fails on pandaboard (arm-linux target),

 FAIL: gdb.base/watch-bitfields.exp: -location watch against bitfields: continue until exit (timeout)
 FAIL: gdb.base/watch-bitfields.exp: regular watch against bitfields: continue until exit (timeout)

In this test, more than one watchpoint is used, so the following
watchpoint requests fall back to software watchpoint, so that GDB
will single step all the way and it is very slow.

This patch is to copy the fix from

  [PATCH] GDB/testsuite: Correct gdb.base/watchpoint-solib.exp timeout tweak
  https://sourceware.org/ml/gdb-patches/2014-07/msg00716.html

I find the left-over of this patch review is to factor out code into
a procedure, so I do that in this patch.

Re-run tests watch-bitfields.exp, watchpoint-solib.exp, sigall-reverse.exp,
and until-precsave.exp on pandaboard, no regression.

gdb/testsuite:

2015-04-15  Pedro Alves  <palves@redhat.com>
    Yao Qi  <yao.qi@linaro.org>

* gdb.base/watch-bitfields.exp (test_watch_location): Increase
timeout by factor of 4.
(test_regular_watch): Likewise.
* gdb.base/watchpoint-solib.exp: Use with_timeout_factor.
* gdb.reverse/sigall-reverse.exp: Likewise.
* gdb.reverse/until-precsave.exp: Likewise.
* lib/gdb.exp (with_timeout_factor): New proc.
(gdb_expect): Move some code to ...
(get_largest_timeout): ... here.  New procedure.

9 years agosim: unify sim-cpu usage
Mike Frysinger [Wed, 15 Apr 2015 06:13:23 +0000 (02:13 -0400)]
sim: unify sim-cpu usage

Now that all the targets are utilizing CPU_PC_{FETCH,STORE}, and the
cpu state is multicore, and the STATE_CPU defines match, we can move
it all to the common code.

9 years agosim: cris/frv/h8300/iq2000/lm32/m32r/sh64: standardize cpu state
Mike Frysinger [Wed, 15 Apr 2015 05:22:34 +0000 (01:22 -0400)]
sim: cris/frv/h8300/iq2000/lm32/m32r/sh64: standardize cpu state

This sets up the sim_state structure and the cpu member to match what we
do in most other sims, and what the common code suggests.  This is a step
to unifying on the sim-cpu.o object.

9 years agoAdd --compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi]
H.J. Lu [Wed, 15 Apr 2015 05:01:25 +0000 (22:01 -0700)]
Add --compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi]

This patch adds --compress-debug-sections=[none|zlib|zlib-gnu|zlib-gabi]
to ld for ELF targets to support generating compressed DWARF debug
sections.  We always generate .zdebug_* section since section names have
been finalized and they can't be changed easily when compression is
being performed.

bfd/

* bfd-in.h (compressed_debug_section_type): New.
* compress.c (bfd_compress_section_contents): Add an argument
for linker write compression and always generate .zdebug_*
section when linking.
(bfd_init_section_compress_status): Pass FALSE to
bfd_compress_section_contents.
(bfd_compress_section): New function.
* elf.c (elf_fake_sections): For linking, set SEC_ELF_COMPRESS
on DWARF debug sections if COMPRESS_DEBUG is set and rename
section if COMPRESS_DEBUG_GABI_ZLIB isn't set.
(assign_file_positions_for_non_load_sections): Set sh_offset
to -1 if SEC_ELF_COMPRESS is set.
(assign_file_positions_except_relocs): Likwise.
(_bfd_elf_assign_file_positions_for_relocs): Renamed to ...
(_bfd_elf_assign_file_positions_for_non_load): This.  Change
return time to bfd_boolean.  Compress the section if
SEC_ELF_COMPRESS is set.
(_bfd_elf_write_object_contents): Updated.
(_bfd_elf_set_section_contents): Write section contents to
the buffer if SEC_ELF_COMPRESS is set.
* merge.c: Include "elf-bfd.h".
(sec_merge_emit): Add arguments for contents and offset.  Write
to contents with offset if contents isn't NULL.
(_bfd_write_merged_section): Write section contents to the
buffer if SEC_ELF_COMPRESS is set.  Pass contents and
output_offset to sec_merge_emit.
* elflink.c (bfd_elf_final_link): Allocate the buffer for
output section contents if SEC_ELF_COMPRESS is set.
* section.c (SEC_ELF_COMPRESS): New.
* bfd-in2.h: Regenerated.

gas/

* as.h (compressed_debug_section_type): Removed.

include/

* bfdlink.h (bfd_link_info): Add compress_debug.

ld/

* ld.texinfo: Document --compress-debug-sections=.
* ldmain.c (main): Set BFD_COMPRESS on output_bfd if
COMPRESS_DEBUG is set.  Set BFD_COMPRESS_GABI on output_bfd
for COMPRESS_DEBUG_GABI_ZLIB.
* lexsup.c (elf_static_list_options): Add
--compress-debug-sections=.
* emultempl/elf32.em (OPTION_COMPRESS_DEBUG): New.
(xtra_long): Add "compress-debug-sections".
(gld${EMULATION_NAME}_handle_option): Handle
OPTION_COMPRESS_DEBUG.

ld/testsuite/

* ld-elf/compress.exp (build_tests): Add tests for
--compress-debug-sections=.
(run_tests): Likewise.
Add additonal tests for --compress-debug-sections=.
* ld-elf/gabiend.rt: New file.
* ld-elf/gabinormal.rt: Likewise.
* ld-elf/gnubegin.rS: Likewise.
* ld-elf/gnunormal.rS: Likewise.
* ld-elf/zlibbegin.rS: Likewise.
* ld-elf/zlibnormal.rS: Likewise.

9 years agoxfail pr18223 test for tic6x
Alan Modra [Wed, 15 Apr 2015 02:29:10 +0000 (11:59 +0930)]
xfail pr18223 test for tic6x

Fails due to
warning: generating a shared library containing non-PIC/PID code

* ld-gc/pr18223.d: xfail tic6x.

9 years agoFix typo in last ChangeLog.
Hans-Peter Nilsson [Wed, 15 Apr 2015 02:11:18 +0000 (04:11 +0200)]
Fix typo in last ChangeLog.

9 years agoAdjust src-release.sh for sim using the gdb create-version.sh.
Hans-Peter Nilsson [Wed, 15 Apr 2015 02:08:21 +0000 (04:08 +0200)]
Adjust src-release.sh for sim using the gdb create-version.sh.

* sim-release.sh (tar_compress): If there's a fifth parameter,
use that in the getver call instead of $tool.
(sim_release): Pass gdb as fifth parameter to tar_compress.
(SIM_SUPPORT_DIRS): Add gdb/common/create-version.sh.

9 years agoFix typo in commit 6ff71e76
Alan Modra [Wed, 15 Apr 2015 01:54:30 +0000 (11:24 +0930)]
Fix typo in commit 6ff71e76

* elf32-rl78.c (rl78_elf_relocate_section): Typo fix.

9 years agoAutomatic date update in version.in
GDB Administrator [Wed, 15 Apr 2015 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoAdds support to the RL78 port for linker relaxation affecting .debug sections.
Nick Clifton [Tue, 14 Apr 2015 15:23:33 +0000 (16:23 +0100)]
Adds support to the RL78 port for linker relaxation affecting .debug sections.

gas * config/tc-rl78.h (TC_LINKRELAX_FIXUP): Define.
(TC_FORCE_RELOCATION_SUB_SAME): Define.
(DWARF2_USE_FIXED_ADVANCE_PC): Define.

* gas/lns/lns.exp: Add RL78 to list of targets using
DW_LNS_fixed_advance_pc.

bfd * elf32-rl78.c (RL78_OP_REL): New macro.
(rl78_elf_howto_table): Use it for complex relocs.
(get_symbol_value): Handle the cases when the info or status
arguments are NULL.
(get_romstart): Cache the status returned by get_symbol_value.
(get_ramstart): Likewise.
(RL78_STACK_PUSH): Generate an error message if the stack
overflows.
(RL78_STACK_POP): Likewise for underflows.
(rl78_compute_complex_reloc): New function.  Contains the basic
processing code for all RL78 complex relocs.
(rl78_special_reloc): New function.  Provides special reloc
handling for complex relocs.
(rl78_elf_relocate_section): Use rl78_compute_complex_reloc.
(rl78_offset_for_reloc): Likewise.

binutils* readelf.c (target_specific_reloc_handling): Add code to handle
RL78 complex relocs.

9 years agoHarden gdb.base/bp-permanent.exp
Luis Machado [Tue, 14 Apr 2015 12:22:53 +0000 (09:22 -0300)]
Harden gdb.base/bp-permanent.exp

Reinstate test message and replace hardcoded test command with a variable.

gdb/testsuite/ChangeLog:
2015-04-14  Luis Machado  <lgustavo@codesourcery.com>

* gdb.base/bp-permanent.exp (test): Reinstate correct test message.

9 years agoZero supplied stat buffers in functions that pretend to stat
Gary Benson [Tue, 14 Apr 2015 11:35:30 +0000 (12:35 +0100)]
Zero supplied stat buffers in functions that pretend to stat

GDB has five places where it pretends to stat for bfd_openr_iovec.
Four of these only set the incoming buffer's st_size, leaving the
other fields unchanged, which is to say very likely populated with
random values from the stack.  remote_bfd_iovec_stat was fixed in
0a93529c56714b1da3d7106d3e0300764f8bb81c; this commit fixes the
other four.

gdb/ChangeLog:

* jit.c (mem_bfd_iovec_stat): Zero supplied buffer.
* minidebug.c (lzma_stat): Likewise.
* solib-spu.c (spu_bfd_iovec_stat): Likewise.
* spu-linux-nat.c (spu_bfd_iovec_stat): Likewise.

9 years agoAdd -z noextern-protected-data to ld for ELF/x86
H.J. Lu [Tue, 14 Apr 2015 11:12:55 +0000 (04:12 -0700)]
Add -z noextern-protected-data to ld for ELF/x86

Address of protected data defined in the shared library may be external,
i.e., due to copy relocation.  By default, linker backend checks if
relocations against protected data symbols are valid for building shared
library and issues an error if relocation isn't allowed.  The new option
override linker backend default.  When -z noextern-protected-data is used,
updates on protected data symbols by another module won't be visibile
to the resulting shared library.  This option is specific to ELF/i386
and ELF/x86-64.

bfd/

PR ld/pr17709
* elflink.c (_bfd_elf_adjust_dynamic_copy): Check
info->extern_protected_data when warning copy relocs against
protected symbols.
(_bfd_elf_symbol_refs_local_p): Check info->extern_protected_data
when checking protected non-function symbols.

include/

PR ld/pr17709
* bfdlink.h (bfd_link_info): Add extern_protected_data.

ld/

PR ld/pr17709
* ld.texinfo: Document "-z noextern-protected-data".
* ldmain.c (main): Initialize link_info.extern_protected_data
to -1.
* lexsup.c (elf_shlib_list_options): Add
"-z [no]extern-protected-data".
* emulparams/elf32_x86_64.sh: Source extern_protected_data.sh.
* emulparams/elf_i386.sh: Likewise.
* emulparams/elf_i386_be.sh: Likewise.
* emulparams/elf_i386_chaos.sh: Likewise.
* emulparams/elf_i386_ldso.sh: Likewise.
* emulparams/elf_i386_vxworks.sh: Likewise.
* emulparams/elf_k1om.sh: Likewise.
* emulparams/elf_l1om.sh: Likewise.
* emulparams/elf_x86_64.sh: Source extern_protected_data.sh.
(PARSE_AND_LIST_OPTIONS): Renamed to ...
(PARSE_AND_LIST_OPTIONS_BNDPLT): This.
(PARSE_AND_LIST_ARGS_CASE_Z): Renamed to ...
(PARSE_AND_LIST_ARGS_CASE_Z_BNDPLT): This.
(PARSE_AND_LIST_OPTIONS): Append $PARSE_AND_LIST_OPTIONS_BNDPLT.
(PARSE_AND_LIST_ARGS_CASE_Z): Append
$PARSE_AND_LIST_ARGS_CASE_Z_BNDPLT.
* emulparams/extern_protected_data.sh: New file.

ld/testsuite/

PR ld/pr17709
* ld-i386/i386.exp: Run protected6b.
* ld-i386/protected6b.d: New file.
* ld-x86-64/protected6b.d: Likewise.
* ld-x86-64/x86-64.exp:  Run protected6b.

9 years agosim: ppc: fix up version script
Mike Frysinger [Tue, 14 Apr 2015 03:16:46 +0000 (23:16 -0400)]
sim: ppc: fix up version script

The common sim code has switched to using gdb directly; update the
ppc copy too.

9 years agoAutomatic date update in version.in
GDB Administrator [Tue, 14 Apr 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years ago * MAINTAINERS: Update my email address.
Stan Shebs [Fri, 10 Apr 2015 18:06:52 +0000 (11:06 -0700)]
    * MAINTAINERS: Update my email address.

diff --git a/gdb/MAINTAINERS b/gdb/MAINTAINERS
index a67a1a8..0fdd8e5 100644
--- a/gdb/MAINTAINERS
+++ b/gdb/MAINTAINERS
@@ -156,7 +156,7 @@ Doug Evans                  dje@google.com
 Daniel Jacobowitz              drow@false.org
 Mark Kettenis                  kettenis@gnu.org
 Yao Qi                         yao.qi@arm.com
-Stan Shebs                     stan@codesourcery.com
+Stan Shebs                     stanshebs@google.com
 Ulrich Weigand                 Ulrich.Weigand@de.ibm.com
 Elena Zannoni                  elena.zannoni@oracle.com
 Eli Zaretskii                  eliz@gnu.org
@@ -631,7 +631,7 @@ Keith Seitz                                 keiths@redhat.com
 Carlos Eduardo Seo                             cseo@linux.vnet.ibm.com
 Ozkan Sezer                                    sezeroz@gmail.com
 Marcus Shawcroft                               marcus.shawcroft@arm.com
-Stan Shebs                                     stan@codesourcery.com
+Stan Shebs                                     stanshebs@google.com
 Joel Sherrill                                  joel.sherrill@oarcorp.com
 Mark Shinwell                                  shinwell@codesourcery.com
 Craig Silverstein                              csilvers@google.com

9 years agoAdd support for the x86 XSAVE extended state on FreeBSD/x86.
John Baldwin [Sat, 21 Feb 2015 21:43:30 +0000 (16:43 -0500)]
Add support for the x86 XSAVE extended state on FreeBSD/x86.

Recognize NT_X86_XSTATE notes in FreeBSD process cores.  Recent
FreeBSD versions include a note containing the XSAVE state for each
thread in the process when XSAVE is in use.  The note stores a copy of
the current XSAVE mask in a reserved section of the machine-defined
XSAVE state at the same offset as Linux's NT_X86_XSTATE note.

For native processes, use the PT_GETXSTATE_INFO ptrace request to
determine if XSAVE is enabled, and if so the active XSAVE state mask
(that is, the value of %xcr0 for the target process) as well as the
size of XSAVE state area.  Use the PT_GETXSTATE and PT_SETXSTATE requests
to fetch and store the XSAVE state, respectively, in the BSD x86
native targets.

In addition, the FreeBSD amd64 and i386 native targets now include
"read_description" target methods to determine the correct x86 target
description for the current XSAVE mask.  On FreeBSD amd64 this also
properly returns an i386 target description for 32-bit binaries which
allows the 64-bit GDB to run 32-bit binaries.

Note that the ptrace changes are in the BSD native targets, not the
FreeBSD-specific native targets since that is where the other ptrace
register accesses occur.  Of the other BSDs, NetBSD and DragonFly use
XSAVE in the kernel but do not currently export the extended state via
ptrace(2).  OpenBSD does not currently support XSAVE.

bfd/ChangeLog:

* elf.c (elfcore_grok_note): Recognize NT_X86_XSTATE on
FreeBSD.
(elfcore_write_xstatereg): Use correct note name on FreeBSD.

gdb/ChangeLog:

* amd64-tdep.c (amd64_target_description): New function.
* amd64-tdep.h: Export amd64_target_description and tdesc_amd64.
* amd64bsd-nat.c [PT_GETXSTATE_INFO]: New variable amd64bsd_xsave_len.
(amd64bsd_fetch_inferior_registers) [PT_GETXSTATE_INFO]: Handle
x86 extended save area.
(amd64bsd_store_inferior_registers) [PT_GETXSTATE_INFO]: Likewise.
* amd64bsd-nat.h: Export amd64bsd_xsave_len.
* amd64fbsd-nat.c (amd64fbsd_read_description): New function.
(_initialize_amd64fbsd_nat): Set "to_read_description" to
"amd64fbsd_read_description".
* amd64fbsd-tdep.c (amd64fbsd_core_read_description): New function.
(amd64fbsd_supply_xstateregset): New function.
(amd64fbsd_collect_xstateregset): New function.
Add "amd64fbsd_xstateregset".
(amd64fbsd_iterate_over_regset_sections): New function.
(amd64fbsd_init_abi): Set "xsave_xcr0_offset" to
"I386_FBSD_XSAVE_XCR0_OFFSET".
Add "iterate_over_regset_sections" gdbarch method.
Add "core_read_description" gdbarch method.
* i386-tdep.c (i386_target_description): New function.
* i386-tdep.h: Export i386_target_description and tdesc_i386.
* i386bsd-nat.c [PT_GETXSTATE_INFO]: New variable i386bsd_xsave_len.
(i386bsd_fetch_inferior_registers) [PT_GETXSTATE_INFO]: Handle
x86 extended save area.
(i386bsd_store_inferior_registers) [PT_GETXSTATE_INFO]: Likewise.
* i386bsd-nat.h: Export i386bsd_xsave_len.
* i386fbsd-nat.c (i386fbsd_read_description): New function.
(_initialize_i386fbsd_nat): Set "to_read_description" to
"i386fbsd_read_description".
* i386fbsd-tdep.c (i386fbsd_core_read_xcr0): New function.
(i386fbsd_core_read_description): New function.
(i386fbsd_supply_xstateregset): New function.
(i386fbsd_collect_xstateregset): New function.
Add "i386fbsd_xstateregset".
(i386fbsd_iterate_over_regset_sections): New function.
(i386fbsd4_init_abi): Set "xsave_xcr0_offset" to
"I386_FBSD_XSAVE_XCR0_OFFSET".
Add "iterate_over_regset_sections" gdbarch method.
Add "core_read_description" gdbarch method.
* i386fbsd-tdep.h: New file.

9 years agoFix reading of .debug_str_offsets{,.dwo} twice.
Doug Evans [Mon, 13 Apr 2015 19:50:17 +0000 (12:50 -0700)]
Fix reading of .debug_str_offsets{,.dwo} twice.

PR binutils/18218
* readelf.c (printable_section_name): Constify sec argument.
(apply_relocations): Ditto.  New arg "size".  All callers updated.
(load_specific_debug_section): Constify sec argument.
Remove side-effect of modifying sec->sh_size.

9 years agoHarden gdb.base/bp-permanent.exp
Luis Machado [Mon, 13 Apr 2015 17:45:56 +0000 (14:45 -0300)]
Harden gdb.base/bp-permanent.exp

This testcase does not work as expected in QEMU (aarch64 QEMU in my case). It
fails when trying to manually write the breakpoint instruction to a certain
PC address.

(gdb) p /x addr_bp[0] = buffer[0]^M
Cannot access memory at address 0x400834^M
(gdb) PASS: gdb.base/bp-permanent.exp: always_inserted=off, sw_watchpoint=0: setup: p /x addr_bp[0] = buffer[0]
p /x addr_bp[1] = buffer[1]^M
Cannot access memory at address 0x400835^M
(gdb) PASS: gdb.base/bp-permanent.exp: always_inserted=off, sw_watchpoint=0: setup: p /x addr_bp[1] = buffer[1]
p /x addr_bp[2] = buffer[2]^M
Cannot access memory at address 0x400836^M
(gdb) PASS: gdb.base/bp-permanent.exp: always_inserted=off, sw_watchpoint=0: setup: p /x addr_bp[2] = buffer[2]
p /x addr_bp[3] = buffer[3]^M
Cannot access memory at address 0x400837^M
(gdb) PASS: gdb.base/bp-permanent.exp: always_inserted=off, sw_watchpoint=0: setup: p /x addr_bp[3] = buffer[3]

The following patch prevents a number of failures by detecting this and bailing out in case the target has such a restriction. Writing to .text from within the program isn't any better. It just leads to a SIGSEGV.

Before the patch:

                === gdb Summary ===

After the patch:

                === gdb Summary ===

gdb/testsuite/ChangeLog:
2015-04-13  Luis Machado  <lgustavo@codesourcery.com>

* gdb.base/bp-permanent.exp (test): Handle the case of being unable
to write to the .text section.

9 years agoHarden gdb.base/coredump-filter.exp
Luis Machado [Mon, 13 Apr 2015 17:42:48 +0000 (14:42 -0300)]
Harden gdb.base/coredump-filter.exp

This testcase seems to assume the target is running Linux, so bare metal,
simulators and other debugging stubs running different OS' will have a
hard time executing some of the commands the testcase issues.

Even restricting the testcase to Linux systems (which the patch below does),
there are still problems with, say, QEMU not providing PID information when
"info inferior" is issued. As a consequence, the subsequent tests will either
fail or will not make much sense.

The attached patch checks if PID information is available. If not, it just
bails out and avoids running into a number of failures.

gdb/testsuite/ChangeLog:
2015-04-13  Luis Machado  <lgustavo@codesourcery.com>

* gdb.base/coredump-filter.exp: Restrict test to Linux systems only.
Handle the case of targets that do not provide PID information.

9 years agoCatch exception in lib/gdbserver-support.exp:gdb_exit
Yao Qi [Mon, 13 Apr 2015 11:36:56 +0000 (12:36 +0100)]
Catch exception in lib/gdbserver-support.exp:gdb_exit

I see the error when I run gdb-sigterm.exp with native-gdbserver
on x86_64-linux.

infrun: prepare_to_wait^M
Cannot execute this command while the target is running.^M
Use the "interrupt" command to stop the target^M
and then try again.^M
gdb.base/gdb-sigterm.exp: expect eof #0: got eof
gdb.base/gdb-sigterm.exp: expect eof #0: stepped 12 times
ERROR OCCURED: : spawn id exp8 not open
    while executing
"expect {
-i exp8 -timeout 10
            -re "$gdb_prompt $" {
                exp_continue
            }
            -i "$server_spawn_id" eof {
                wait -i $expect_out(spawn_id)
                unse..."
    ("uplevel" body line 1)
    invoked from within

In gdb-sigterm.exp, SIGTERM is sent to GDB and it exits.  However,
Dejagnu or tcl doesn't know this.

This patch is to catch the exception, but error messages are still
shown in the console and gdb.log.  In order to avoid this, we also
replace gdb_expect with expect.

gdb/testsuite:

2015-04-13  Yao Qi  <yao.qi@linaro.org>

* lib/gdbserver-support.exp (gdb_exit): Catch exception
and use expect instead of gdb_expect.

9 years agoDon't check object claimed by plugin
H.J. Lu [Mon, 13 Apr 2015 11:58:16 +0000 (04:58 -0700)]
Don't check object claimed by plugin

When ELF linker backend searchs the symbol table of an archive element,
it should skip the object which has been claimed by plugin.

PR ld/18250
* elflink.c (elf_link_is_defined_archive_symbol): Return FALSE
if the object has been claimed by plugin.

9 years agoRename variable "addr" to "coredump_var_addr" in gdb.base/coredump-filter.exp
Sergio Durigan Junior [Mon, 13 Apr 2015 06:40:08 +0000 (02:40 -0400)]
Rename variable "addr" to "coredump_var_addr" in gdb.base/coredump-filter.exp

This commit renames the global array variable "addr" to an unique name
"coredump_var_addr" in the test gdb.base/coredump-filter.exp.  This is
needed because global arrays can have name conflicts between tests.
For example, this specific test was conflicting with dmsym.exp,
causing errors like:

  ERROR: tcl error sourcing ../../../../../binutils-gdb/gdb/testsuite/gdb.base/dmsym.exp.
  ERROR: can't set "addr": variable is array
      while executing
  "set addr "0x\[0-9a-zA-Z\]+""
      (file "../../../../../binutils-gdb/gdb/testsuite/gdb.base/dmsym.exp" line 45)
      invoked from within
  "source ../../../../../binutils-gdb/gdb/testsuite/gdb.base/dmsym.exp"
      ("uplevel" body line 1)
      invoked from within
  "uplevel #0 source ../../../../../binutils-gdb/gdb/testsuite/gdb.base/dmsym.exp"
      invoked from within
  "catch "uplevel #0 source $test_file_name""

This problem was reported by Yao Qi at:

  <https://sourceware.org/ml/gdb-patches/2015-04/msg00373.html>
  Message-Id: <1428666671-12926-1-git-send-email-qiyaoltc@gmail.com>

gdb/testsuite/ChangeLog:
2015-04-13  Sergio Durigan Junior  <sergiodj@redhat.com>

* gdb.base/coredump-filter.exp: Rename variable "addr" to
"coredump_var_addr" to avoid naming conflict with other testcases.

9 years agosim: fix the PKGVERSION define
Mike Frysinger [Mon, 13 Apr 2015 06:38:57 +0000 (02:38 -0400)]
sim: fix the PKGVERSION define

This should be SIM, not GDB.

9 years agosim: options: add --version support
Mike Frysinger [Mon, 13 Apr 2015 06:33:26 +0000 (02:33 -0400)]
sim: options: add --version support

The old run frontend had a --version option, but the new common
sim-options file does not.  Restore support for that so we can
get version info out of `run` when using the new frontend.

9 years agosim: switch to gdb version script
Mike Frysinger [Fri, 10 Apr 2015 23:40:54 +0000 (19:40 -0400)]
sim: switch to gdb version script

Since the local create-version.sh already points directly into the gdb
source tree, we might as well use the gdb script directly too.

9 years agosim: mn10300: convert to sim-cpu
Mike Frysinger [Mon, 13 Apr 2015 06:13:48 +0000 (02:13 -0400)]
sim: mn10300: convert to sim-cpu

Make cpu allocation fully dynamic so we can leverage the common
sim-cpu and its APIs.

9 years agosim: v850: convert to sim-cpu
Mike Frysinger [Mon, 13 Apr 2015 06:11:24 +0000 (02:11 -0400)]
sim: v850: convert to sim-cpu

Make cpu allocation fully dynamic so we can leverage the common
sim-cpu and its APIs.

9 years agosim: mips: convert to sim-cpu
Mike Frysinger [Mon, 13 Apr 2015 06:09:55 +0000 (02:09 -0400)]
sim: mips: convert to sim-cpu

Make cpu allocation fully dynamic so we can leverage the common
sim-cpu and its APIs.

9 years agosim: m68hc11: convert to sim-cpu
Mike Frysinger [Mon, 13 Apr 2015 06:07:23 +0000 (02:07 -0400)]
sim: m68hc11: convert to sim-cpu

Make cpu allocation fully dynamic so we can leverage the common
sim-cpu and its APIs.

9 years agosim: mips: fix prototype warnings
Mike Frysinger [Mon, 13 Apr 2015 05:56:13 +0000 (01:56 -0400)]
sim: mips: fix prototype warnings

Convert a bunch of old style prototypes and tweak various casts
to match the function signatures.

9 years agosim: ft32: fix ft32_pc_get logic
Mike Frysinger [Mon, 13 Apr 2015 05:54:33 +0000 (01:54 -0400)]
sim: ft32: fix ft32_pc_get logic

9 years agoAutomatic date update in version.in
GDB Administrator [Mon, 13 Apr 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoDo unset_currtarget_info ldscript for all simulator testsuites.
Hans-Peter Nilsson [Sun, 12 Apr 2015 23:54:43 +0000 (01:54 +0200)]
Do unset_currtarget_info ldscript for all simulator testsuites.

sim/testsuite:
* sim-defs.exp (sim_init): Unset target ldscript here.

sim/testsuite/sim/mips:
* basic.exp: Don't unset target ldscript here.

9 years agosim: arm/mips: fix sim_read/sim_write linkage errors
Mike Frysinger [Sun, 12 Apr 2015 09:47:39 +0000 (05:47 -0400)]
sim: arm/mips: fix sim_read/sim_write linkage errors

With sim-hrw.o being built & linked in the common list, some people are
getting linking errors now for these targets.  Move the main objects that
provide these functions before the common list to avoid that.

9 years agosim: ft32: delete sim_read/sim_write funcs
Mike Frysinger [Sun, 12 Apr 2015 09:13:23 +0000 (05:13 -0400)]
sim: ft32: delete sim_read/sim_write funcs

The common sim-hrw.o provides both of these, so simply use them.

9 years agoRevert the last 2 commits on pagesize_m1
H.J. Lu [Sun, 12 Apr 2015 00:05:04 +0000 (17:05 -0700)]
Revert the last 2 commits on pagesize_m1

9 years agoAutomatic date update in version.in
GDB Administrator [Sun, 12 Apr 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoDon't call getpagesize more than once
H.J. Lu [Sat, 11 Apr 2015 23:49:09 +0000 (16:49 -0700)]
Don't call getpagesize more than once

9 years agoInitialize pagesize_m1 in bfd_cache_init
H.J. Lu [Sat, 11 Apr 2015 22:12:08 +0000 (15:12 -0700)]
Initialize pagesize_m1 in bfd_cache_init

There is no need to check and initialize pagesize_m1 in cache_bmmap.

* cache.c (cache_bmmap): Move pagesize_m1 ... to
(pagesize_m1): Here.
(bfd_cache_init): Initialize pagesize_m1.

9 years agoRemove an extra ';'
H.J. Lu [Sat, 11 Apr 2015 19:06:33 +0000 (12:06 -0700)]
Remove an extra ';'

* plugin.c (plugin_load_plugins): Removed an extra ';'.

9 years agoRemove --xdb
Jan Kratochvil [Sat, 11 Apr 2015 17:49:03 +0000 (19:49 +0200)]
Remove --xdb

Pedro Alves:

The commands that enables aren't even documented in the manual.
Judging from that, I assume that only wdb users would ever really
be using the --xdb switch.

I think it's time to drop "support" for the --xdb switch too.  I
looked through the commands that that exposes, the only that looked
potentially interesting was "go", but then it's just an alias
for "tbreak+jump", which can easily be done with "define go...end".
I'd rather free up the "go" name for something potentially
more interesting (either run control, or maybe even unrelated,
e.g., for golang).

gdb/ChangeLog
2015-04-11  Jan Kratochvil  <jan.kratochvil@redhat.com>

* NEWS (Changes since GDB 7.9): Add removed -xdb.
* breakpoint.c (command_line_is_silent): Remove xdb_commands
conditional.
(_initialize_breakpoint): Remove xdb_commands for bc, ab, sb, db, ba
and lb.
* cli/cli-cmds.c (_initialize_cli_cmds): Remove xdb_commands for v and
va.
* cli/cli-decode.c (find_command_name_length): Remove xdb_commands
conditional.
* defs.h (xdb_commands): Remove declaration.
* f-valprint.c (_initialize_f_valprint): Remove xdb_commands for lc.
* guile/scm-cmd.c (command_classes): Remove xdb from comment.
* infcmd.c (run_no_args_command, go_command): Remove.
(_initialize_infcmd): Remove xdb_commands for S, go, g, R and lr.
* infrun.c (xdb_handle_command): Remove.
(_initialize_infrun): Remove xdb_commands for lz and z.
* main.c (xdb_commands): Remove variable.
(captured_main): Remove "xdb" from long_options.
(print_gdb_help): Remove --xdb from help.
* python/py-cmd.c (gdbpy_initialize_commands): Remove xdb from comment.
* source.c (_initialize_source): Remove xdb_commands for D, ld, / and ?.
* stack.c (backtrace_full_command, args_plus_locals_info)
(current_frame_command): Remove.
(_initialize_stack): Remove xdb_commands for t, T and l.
* symtab.c (_initialize_symtab): Remove xdb_commands for lf and lg.
* thread.c (_initialize_thread): Remove xdb_commands condition.
* tui/tui-layout.c (tui_toggle_layout_command)
(tui_toggle_split_layout_command, tui_handle_xdb_layout): Remove.
(_initialize_tui_layout): Remove xdb_commands for td and ts.
* tui/tui-regs.c (tui_scroll_regs_forward_command)
(tui_scroll_regs_backward_command): Remove.
(_initialize_tui_regs): Remove xdb_commands for fr, gr, sr, +r and -r.
* tui/tui-win.c (tui_xdb_set_win_height_command): Remove.
(_initialize_tui_win): Remove xdb_commands for U and w.
* utils.c (pagination_on_command, pagination_off_command): Remove.
(initialize_utils): Remove xdb_commands for am and sm.

gdb/doc/ChangeLog
2015-04-11  Jan Kratochvil  <jan.kratochvil@redhat.com>

* gdb.texinfo (Mode Options): Remove -xdb.

9 years agoReplace SYMBOLIC_BIND with SYMBOL_REFERENCES_LOCAL
H.J. Lu [Sat, 11 Apr 2015 14:34:49 +0000 (07:34 -0700)]
Replace SYMBOLIC_BIND with SYMBOL_REFERENCES_LOCAL

When checking R_386_GOTOFF/R_X86_64_GOTOFF64 for building shared library,
we should check SYMBOL_REFERENCES_LOCAL instead of SYMBOLIC_BIND to cover
more cases.

bfd/

* elf32-i386.c (elf_i386_relocate_section): Replace SYMBOLIC_BIND
with SYMBOL_REFERENCES_LOCAL when checking R_386_GOTOFF against
protected data symbol when building shared library.
* elf64-x86-64.c (elf_x86_64_relocate_section): Check
R_X86_64_GOTOFF64 against undefined symbol and replace
SYMBOLIC_BIND with SYMBOL_REFERENCES_LOCAL when checking
R_X86_64_GOTOFF64 against protected data symbol when building
shared library.

ld/testsuite/

* ld-i386/i386.exp: Run protected6a.
* ld-i386/protected6.d: Renamed to ...
* ld-i386/protected6a.d: This.
* ld-x86-64/hidden4.d: New file.
* ld-x86-64/hidden4.s: Likewise.
* ld-x86-64/hidden5.d: Likewise.
* ld-x86-64/hidden5.s: Likewise.
* ld-x86-64/protected6.d: Renamed to ...
* ld-x86-64/protected6a.d: This.
* ld-x86-64/x86-64.exp: Run hidden4, hidden5, protected6a,
protected7a and protected7b.

9 years agoAutomatic date update in version.in
GDB Administrator [Sat, 11 Apr 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoCheck GOTOFF reloc against protected data on x86
H.J. Lu [Fri, 10 Apr 2015 21:02:23 +0000 (14:02 -0700)]
Check GOTOFF reloc against protected data on x86

R_386_GOTOFF/R_X86_64_GOTOFF64 relocation shouldn't be used against
protected data symbol on x86 since with copy relocation, address of
protected data defined in the shared library may be external.

This patch will break building shared libraries with protected data
symbols using GCCs older than GCC 5 without the bug fix for

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65248

GCC backport request should be made in the GCC bug report above.

bfd/

PR ld/pr17709
* elf32-i386.c (elf_i386_relocate_section): Also check R_386_GOTOFF
against protected data symbol when building shared library.
* elf64-x86-64.c (elf_x86_64_relocate_section): Also check
R_X86_64_GOTOFF64 against protected data symbol when building
shared library.

ld/testsuite/

PR ld/pr17709
* ld-i386/protected6.d: New file.
* ld-i386/protected6.s: Likewise.
* ld-x86-64/protected6.d: Likewise.
* ld-x86-64/protected6.s: Likewise.
* ld-x86-64/protected7.d: Likewise.
* ld-x86-64/protected7.s: Likewise.
* ld-x86-64/protected7a.d: Likewise.
* ld-x86-64/protected7b.d: Likewise.

9 years agoCleanup signal-while-stepping-over-bp-other-thread.exp
Pedro Alves [Mon, 9 Mar 2015 19:02:30 +0000 (19:02 +0000)]
Cleanup signal-while-stepping-over-bp-other-thread.exp

gdb/testsuite/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* gdb.threads/signal-while-stepping-over-bp-other-thread.exp: Use
gdb_test_sequence and gdb_assert.

9 years agostep-over-trips-on-watchpoint.exp: Don't put addresses in test messages
Pedro Alves [Fri, 10 Apr 2015 18:23:24 +0000 (19:23 +0100)]
step-over-trips-on-watchpoint.exp: Don't put addresses in test messages

Diffing test results, I noticed:

 -PASS: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: with thread-specific bp: next: b *0x0000000000400811 thread 1
 +PASS: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: with thread-specific bp: next: b *0x00000000004007d1 thread 1

gdb/testsuite/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* gdb.threads/step-over-trips-on-watchpoint.exp (do_test): Use
test messages that don't include the breakpoint address.

9 years agoAdd extern_protected_data and set it for x86
H.J. Lu [Thu, 5 Mar 2015 14:34:39 +0000 (06:34 -0800)]
Add extern_protected_data and set it for x86

Re-apply: commit ca3fe95e469b9daec153caa2c90665f5daaec2b5

With copy relocation, address of protected data defined in the shared
library may be external.  This patch adds extern_protected_data and
changes _bfd_elf_symbol_refs_local_p to return false for protected data
if extern_protected_data is true.

This patch will break building shared libraries with protected data
symbols using GCCs older than GCC 5 without the bug fix for

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65248

GCC backport request should be made in the GCC bug report above.

To get correct run-time behavior on Linux, glibc 2.22 or above are
required, which have the bug fix for

https://sourceware.org/bugzilla/show_bug.cgi?id=17711

Backports for glibc 2.21, 2.20, 2.19 and 2.18 are on hjl/pr17711/2.21,
hjl/pr17711/2.20, hjl/pr17711/2.19 and hjl/pr17711/2.18 branches,
respectively, at

https://sourceware.org/git/?p=glibc.git;a=summary

bfd/

PR ld/pr15228
PR ld/pr17709
* elf-bfd.h (elf_backend_data): Add extern_protected_data.
* elf32-i386.c (elf_backend_extern_protected_data): New.
Defined to 1.
* elf64-x86-64.c (elf_backend_extern_protected_data): Likewise.
* elflink.c (_bfd_elf_adjust_dynamic_copy): Don't error on
copy relocs against protected symbols if extern_protected_data
is true.
(_bfd_elf_symbol_refs_local_p): Don't return true on protected
non-function symbols if extern_protected_data is true.
* elfxx-target.h (elf_backend_extern_protected_data): New.
Default to 0.
(elfNN_bed): Initialize extern_protected_data with
elf_backend_extern_protected_data.

ld/testsuite/

PR ld/pr15228
PR ld/pr17709
* ld-i386/i386.exp (i386tests): Add a test for PR ld/17709.
* ld-i386/pr17709-nacl.rd: New file.
* ld-i386/pr17709.rd: Likewise.
* ld-i386/pr17709a.s: Likewise.
* ld-i386/pr17709b.s: Likewise.
* ld-i386/protected3.d: Updated.
* ld-i386/protected3.s: Likewise.
* ld-x86-64/pr17709-nacl.rd: New file.
* ld-x86-64/pr17709.rd: Likewise.
* ld-x86-64/pr17709a.s: Likewise.
* ld-x86-64/pr17709b.s: Likewise.
* ld-x86-64/protected3.d: Updated.
* ld-x86-64/protected3.s: Likewise.
* ld-x86-64/x86-64.exp (x86_64tests): Add a test for PR ld/17709.

9 years ago[arm] watchpoint-reuse-slot.exp: skip setting HW points on some address
Yao Qi [Fri, 10 Apr 2015 15:23:13 +0000 (16:23 +0100)]
[arm] watchpoint-reuse-slot.exp: skip setting HW points on some address

Hi,
ARM linux kernel has some requirements on the address/length setting
for HW breakpoints/watchpoints, but watchpoint-reuse-slot.exp doesn't
consider them and sets HW points on various addresses.  Many fails
are causes as a result:

stepi^M
Warning:^M
Could not insert hardware watchpoint 20.^M
Could not insert hardware breakpoints:^M
You may have requested too many hardware breakpoints/watchpoints.^M
^M
(gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted off: watch x watch: : width 2, iter 2: base + 1: stepi advanced

watch *(buf.byte + 2 + 1)@2^M
Hardware watchpoint 388: *(buf.byte + 2 + 1)@2^M
Warning:^M
Could not insert hardware watchpoint 388.^M
Could not insert hardware breakpoints:^M
You may have requested too many hardware breakpoints/watchpoints.^M
^M
(gdb) FAIL: gdb.base/watchpoint-reuse-slot.exp: always-inserted on: watch x watch: : width 2, iter 2: base + 1: watch *(buf.byte + 2 + 1)@2

This patch is to reflect kernel requirements in watchpoint-reuse-slot.exp
in order to skip some tests.

gdb/testsuite:

2015-04-10  Yao Qi  <yao.qi@linaro.org>

* gdb.base/watchpoint-reuse-slot.exp (valid_addr_p): Return
false for some offset and width combinations which aren't
supported by linux kernel.

9 years agoPPC64: Fix step-over-trips-on-watchpoint.exp with displaced stepping on
Pedro Alves [Fri, 10 Apr 2015 12:08:32 +0000 (13:08 +0100)]
PPC64: Fix step-over-trips-on-watchpoint.exp with displaced stepping on

PPC64 currently fails this test like:

 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: no thread-specific bp: step: step
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: no thread-specific bp: next: next
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: no thread-specific bp: continue: continue (the program exited)
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: with thread-specific bp: step: step
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: with thread-specific bp: next: next
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: displaced=on: with thread-specific bp: continue: continue (the program exited)

The problem is that PPC is a non-continuable watchpoints architecture
and the displaced stepping code isn't coping with that correctly.  On
such targets/architectures, a watchpoint traps _before_ the
instruction executes/completes.  On a watchpoint trap, the PC points
at the instruction that triggers the watchpoint (side effects haven't
happened yet).  In order to move past the watchpoint, GDB needs to
remove the watchpoint, single-step, and reinsert the watchpoint, just
like when stepping past a breakpoint.

The trouble is that if GDB is stepping over a breakpoint with
displaced stepping, and the instruction under the breakpoint triggers
a watchpoint, we get the watchpoint SIGTRAP, expecting a finished
(hard or software) step trap.  Even though the thread's PC hasn't
advanced yet (must remove watchpoint for that), since we get a
SIGTRAP, displaced_step_fixup thinks the single-step finished
successfuly anyway, and calls gdbarch_displaced_step_fixup, which then
adjusts the thread's registers incorrectly.

The fix is to cancel the displaced step if we trip on a watchpoint.
handle_inferior_event then processes the watchpoint event, and starts
a new step-over, here:

...
      /* At this point, we are stopped at an instruction which has
         attempted to write to a piece of memory under control of
         a watchpoint.  The instruction hasn't actually executed
         yet.  If we were to evaluate the watchpoint expression
         now, we would get the old value, and therefore no change
         would seem to have occurred.
...
      ecs->event_thread->stepping_over_watchpoint = 1;
      keep_going (ecs);
      return;
...

but this time, since we have a watchpoint to step over, watchpoints
are removed from the target, so the step-over succeeds.

The keep_going/resume changes are necessary because if we're stepping
over a watchpoint, we need to remove it from the target - displaced
stepping doesn't help, the copy of the instruction in the scratch pad
reads/writes to the same addresses, thus triggers the watchpoint
too...  So without those changes we keep triggering the watchpoint
forever, never making progress.  With non-stop that means we'll need
to pause all threads momentarily, which we can't today.  We could
avoid that by removing the watchpoint _only_ from the thread that is
moving past the watchpoint, but GDB is not prepared for that today
either.  For remote targets, that would need new packets, so good to
be able to step over it in-line as fallback anyway.

gdb/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* infrun.c (displaced_step_fixup): Switch to the event ptid
earlier.  If the thread stopped for a watchpoint and the
target/arch has non-continuable watchpoints, cancel the displaced
step.
(resume): Don't start a displaced step if in-line step-over info
is valid.

9 years agoTest step-over-{lands-on-breakpoint|trips-on-watchpoint}.exp with displaced stepping
Pedro Alves [Fri, 10 Apr 2015 14:22:38 +0000 (15:22 +0100)]
Test step-over-{lands-on-breakpoint|trips-on-watchpoint}.exp with displaced stepping

These tests exercise the infrun.c:proceed code that needs to know to
start new step overs (along with switch_back_to_stepped_thread, etc.).
That code is tricky to get right in the multitude of possible
combinations (at least):

 (native | remote)
  X (all-stop | all-stop-but-target-always-in-non-stop)
  X (displaced-stepping | in-line step-over).

The first two above are properties of the target, but the different
step-over-breakpoint methods should work with any target that supports
them.  This patch makes sure we always test both methods on all
targets.

Tested on x86-64 Fedora 20.

gdb/testsuite/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* gdb.threads/step-over-lands-on-breakpoint.exp (do_test): New
procedure, factored out from ...
(top level): ... here.  Add "set displaced-stepping" testing axis.
* gdb.threads/step-over-trips-on-watchpoint.exp (do_test): New
parameter "displaced".  Use it.
(top level): Use foreach and add "set displaced-stepping" testing
axis.

9 years agoMake gdb.threads/step-over-trips-on-watchpoint.exp effective on !x86
Pedro Alves [Fri, 10 Apr 2015 12:11:32 +0000 (13:11 +0100)]
Make gdb.threads/step-over-trips-on-watchpoint.exp effective on !x86

This test is currently failing like this on (at least) PPC64 and s390x:

 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: no thread-specific bp: step: step
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: no thread-specific bp: next: next
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: with thread-specific bp: step: step
 FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: with thread-specific bp: next: next

gdb.log:

 (gdb) PASS: gdb.threads/step-over-trips-on-watchpoint.exp: no thread-specific bp: step: set scheduler-locking off
 step
 wait_threads () at ../../../src/gdb/testsuite/gdb.threads/step-over-trips-on-watchpoint.c:49
 49        return 1; /* in wait_threads */
 (gdb) FAIL: gdb.threads/step-over-trips-on-watchpoint.exp: no thread-specific bp: step: step

The problem is that the test assumes that both the "watch_me = 1;" and
the "other = 1;" lines compile to a single instruction each, which
happens to be true on x86, but no necessarily true everywhere else.
The result is that the test doesn't really test what it wants to test.

Fix it by looking for the instruction that triggers the watchpoint.

gdb/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* gdb.threads/step-over-trips-on-watchpoint.c (child_function):
Remove comment.
* gdb.threads/step-over-trips-on-watchpoint.exp (do_test): Find
both the address of the instruction that triggers the watchpoint
and the address of the instruction immediately after, and use
those addresses for the test.  Fix comment.

9 years agoAdd a testcase for PR ld/18223
H.J. Lu [Fri, 10 Apr 2015 11:15:53 +0000 (04:15 -0700)]
Add a testcase for PR ld/18223

PR ld/18223
* ld-gc/gc.exp: Run pr18223.
* ld-gc/pr18223.d: New file.
* ld-gc/pr18223.s: Likewise.

9 years agoUse bfd_alloc for compressed section contents
H.J. Lu [Fri, 10 Apr 2015 10:54:41 +0000 (03:54 -0700)]
Use bfd_alloc for compressed section contents

Compressed section contents should be bfd_alloced to avoid memory leak.
This patch replaces bfd_malloc and free with bfd_alloc and bfd_release
on compressed buffer in bfd_compress_section_contents.  There is still
a very small memory leak when compressed section isn't smaller.

* compress.c (bfd_compress_section_contents): Replace bfd_malloc
and free with bfd_alloc and bfd_release on compressed buffer.
Release buffer if compressed section isn't smaller.

9 years agoFix gdb.base/sigstep.exp with displaced stepping on software single-step targets
Pedro Alves [Fri, 10 Apr 2015 09:55:09 +0000 (10:55 +0100)]
Fix gdb.base/sigstep.exp with displaced stepping on software single-step targets

TL;DR:

When stepping over a breakpoint with displaced stepping, the core must
be notified of all signals, otherwise the displaced step fixup code
confuses a breakpoint trap in the signal handler for the expected trap
indicating the displaced instruction was single-stepped
normally/successfully.

Detailed version:

Running sigstep.exp with displaced stepping on, against my x86
software single-step branch, I got:

 FAIL: gdb.base/sigstep.exp: step on breakpoint, to handler: performing step
 FAIL: gdb.base/sigstep.exp: next on breakpoint, to handler: performing next
 FAIL: gdb.base/sigstep.exp: continue on breakpoint, to handler: performing continue

Turning on debug logs, we see:

 (gdb) step
 infrun: clear_proceed_status_thread (process 32147)
 infrun: proceed (addr=0xffffffffffffffff, signal=GDB_SIGNAL_DEFAULT)
 infrun: resume (step=1, signal=GDB_SIGNAL_0), trap_expected=1, current thread [process 32147] at 0x400842
 displaced: stepping process 32147 now
 displaced: saved 0x400622: 49 89 d1 5e 48 89 e2 48 83 e4 f0 50 54 49 c7 c0
 displaced: %rip-relative addressing used.
 displaced: using temp reg 2, old value 0x3615eafd37, new value 0x40084c
 displaced: copy 0x400842->0x400622: c7 81 1c 08 20 00 00 00 00 00
 displaced: displaced pc to 0x400622
 displaced: run 0x400622: c7 81 1c 08
 LLR: Preparing to resume process 32147, 0, inferior_ptid process 32147
 LLR: PTRACE_CONT process 32147, 0 (resume event thread)
 linux_nat_wait: [process -1], [TARGET_WNOHANG]
 LLW: enter
 LNW: waitpid(-1, ...) returned 32147, No child processes
 LLW: waitpid 32147 received Alarm clock (stopped)
 LLW: PTRACE_CONT process 32147, Alarm clock (preempt 'handle')
 LNW: waitpid(-1, ...) returned 0, No child processes
 LLW: exit (ignore)
 sigchld
 infrun: target_wait (-1.0.0, status) =
 infrun:   -1.0.0 [process -1],
 infrun:   status->kind = ignore
 infrun: TARGET_WAITKIND_IGNORE
 infrun: prepare_to_wait
 linux_nat_wait: [process -1], [TARGET_WNOHANG]
 LLW: enter
 LNW: waitpid(-1, ...) returned 32147, No child processes
 LLW: waitpid 32147 received Trace/breakpoint trap (stopped)
 CSBB: process 32147 stopped by software breakpoint
 LNW: waitpid(-1, ...) returned 0, No child processes
 LLW: trap ptid is process 32147.
 LLW: exit
 infrun: target_wait (-1.0.0, status) =
 infrun:   32147.32147.0 [process 32147],
 infrun:   status->kind = stopped, signal = GDB_SIGNAL_TRAP
 infrun: TARGET_WAITKIND_STOPPED
 displaced: restored process 32147 0x400622
 displaced: fixup (0x400842, 0x400622), insn = 0xc7 0x81 ...
 displaced: restoring reg 2 to 0x3615eafd37
 displaced: relocated %rip from 0x400717 to 0x400937
 infrun: stop_pc = 0x400937
 infrun: delayed software breakpoint trap, ignoring
 infrun: no line number info
 infrun: stop_waiting
 0x0000000000400937 in __dso_handle ()
 1: x/i $pc
 => 0x400937:    and    %ah,0xa0d64(%rip)        # 0x4a16a1
 (gdb) FAIL: gdb.base/sigstep.exp: displaced=on: step on breakpoint, to handler: performing step

What should have happened is that the breakpoint hit in the signal
handler should have been presented to the user.  But note that
"preempt 'handle'" -- what happened instead is that
displaced_step_fixup confused the breakpoint in the signal handler for
the expected SIGTRAP indicating the displaced instruction was
single-stepped normally/successfully.

This should be affecting all software single-step targets in the same
way.

The fix is to make sure the core sees all signals when displaced
stepping, just like we already must see all signals when doing an
stepping over a breakpoint in-line.  We now get:

 infrun: target_wait (-1.0.0, status) =
 infrun:   570.570.0 [process 570],
 infrun:   status->kind = stopped, signal = GDB_SIGNAL_ALRM
 infrun: TARGET_WAITKIND_STOPPED
 displaced: restored process 570 0x400622
 infrun: stop_pc = 0x400842
 infrun: random signal (GDB_SIGNAL_ALRM)
 infrun: signal arrived while stepping over breakpoint
 infrun: inserting step-resume breakpoint at 0x400842
 infrun: resume (step=0, signal=GDB_SIGNAL_ALRM), trap_expected=0, current thread [process 570] at 0x400842
 LLR: Preparing to resume process 570, Alarm clock, inferior_ptid process 570
 LLR: PTRACE_CONT process 570, Alarm clock (resume event thread)
 infrun: prepare_to_wait
 linux_nat_wait: [process -1], [TARGET_WNOHANG]
 LLW: enter
 LNW: waitpid(-1, ...) returned 0, No child processes
 LLW: exit (ignore)
 infrun: target_wait (-1.0.0, status) =
 infrun:   -1.0.0 [process -1],
 infrun:   status->kind = ignore
 sigchld
 infrun: TARGET_WAITKIND_IGNORE
 infrun: prepare_to_wait
 linux_nat_wait: [process -1], [TARGET_WNOHANG]
 LLW: enter
 LNW: waitpid(-1, ...) returned 570, No child processes
 LLW: waitpid 570 received Trace/breakpoint trap (stopped)
 CSBB: process 570 stopped by software breakpoint
 LNW: waitpid(-1, ...) returned 0, No child processes
 LLW: trap ptid is process 570.
 LLW: exit
 infrun: target_wait (-1.0.0, status) =
 infrun:   570.570.0 [process 570],
 infrun:   status->kind = stopped, signal = GDB_SIGNAL_TRAP
 infrun: TARGET_WAITKIND_STOPPED
 infrun: stop_pc = 0x400717
 infrun: BPSTAT_WHAT_STOP_NOISY
 infrun: stop_waiting

 Breakpoint 3, handler (sig=14) at /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.base/sigstep.c:35
 35        done = 1;

Hardware single-step targets already behave this way, because the
Linux backends (both native and gdbserver) always report signals to
the core if the thread was single-stepping.

As mentioned in the new comment in do_target_resume, we can't fix this
by instead making the displaced_step_fixup phase skip fixing up the PC
if the single step stopped somewhere we didn't expect.  Here's what
the backtrace would look like if we did that:

 Breakpoint 3, handler (sig=14) at /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.base/sigstep.c:35
 35        done = 1;
 1: x/i $pc
 => 0x400717 <handler+7>:        movl   $0x1,0x200943(%rip)        # 0x601064 <done>
 (gdb) bt
 #0  handler (sig=14) at /home/pedro/gdb/mygit/src/gdb/testsuite/gdb.base/sigstep.c:35
 #1  <signal handler called>
 #2  0x0000000000400622 in _start ()
 (gdb) FAIL: gdb.base/sigstep.exp: displaced=on: step on breakpoint, to handler: backtrace

gdb/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* infrun.c (displaced_step_in_progress): New function.
(do_target_resume): Advise target to report all signals if
displaced stepping.

gdb/testsuite/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

* gdb.base/sigstep.exp (breakpoint_to_handler)
(breakpoint_to_handler_entry): New parameter 'displaced'.  Use it.
Test "backtrace" in handler.
(breakpoint_over_handler): New parameter 'displaced'.  Use it.
(top level): Add new "displaced" test axis to
breakpoint_to_handler, breakpoint_to_handler_entry and
breakpoint_over_handler.

9 years agoDowngrade linker error on protected symbols in .dynbss to a warning
Alan Modra [Fri, 10 Apr 2015 00:39:34 +0000 (10:09 +0930)]
Downgrade linker error on protected symbols in .dynbss to a warning

PR ld/18222
* elflink.c (_bfd_elf_adjust_dynamic_copy): Don't report an error
on adding a protected visibility variable to .dynbss.

9 years agoRevert "Add extern_protected_data and set it for x86"
Alan Modra [Thu, 9 Apr 2015 12:06:24 +0000 (21:36 +0930)]
Revert "Add extern_protected_data and set it for x86"

This reverts commit ca3fe95e469b9daec153caa2c90665f5daaec2b5.

9 years agogdb/18216: displaced step+deliver signal, a thread needs step-over, crash
Pedro Alves [Fri, 10 Apr 2015 09:36:23 +0000 (10:36 +0100)]
gdb/18216: displaced step+deliver signal, a thread needs step-over, crash

The problem is that with hardware step targets and displaced stepping,
"signal FOO" when stopped at a breakpoint steps the breakpoint
instruction at the same time it delivers a signal.  This results in
tp->stepped_breakpoint set, but no step-resume breakpoint set.  When
the next stop event arrives, GDB crashes.  Irrespective of whether we
should do something more/different to step past the breakpoint in this
scenario (e.g., PR 18225), it's just wrong to assume there'll be a
step-resume breakpoint set (and was not the original intention).

gdb/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

PR gdb/18216
* infrun.c (process_event_stop_test): Don't assume a step-resume
is set if tp->stepped_breakpoint is true.

gdb/testsuite/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

PR gdb/18216
* gdb.threads/multiple-step-overs.exp: Remove expected eof.

9 years ago[arm] Fix displaced stepping for thumb alu reg instruction
Yao Qi [Fri, 10 Apr 2015 09:33:01 +0000 (10:33 +0100)]
[arm] Fix displaced stepping for thumb alu reg instruction

Recent patch series "V2 All-stop on top of non-stop" causes a SIGSEGV
in the test case,

> -PASS: gdb.base/info-shared.exp: continue to breakpoint: library function #4
> +FAIL: gdb.base/info-shared.exp: continue to breakpoint: library function #4
>
> continue^M
> Continuing.^M
> ^M
> Program received signal SIGSEGV, Segmentation fault.^M
> 0x40021564 in ?? () gdb/testsuite/gdb.base/info-shared-solib1.so^M
> (gdb) FAIL: gdb.base/info-shared.exp: continue to breakpoint: library function #4

and an ARM displaced stepping bug is exposed.  It can be reproduced by
the modified gdb.arch/arm-disp-step.exp as below,

continue^M
Continuing.^M
^M
Program received signal SIGSEGV, Segmentation fault.^M
0xa713cfcc in ?? ()^M
(gdb) FAIL: gdb.arch/arm-disp-step.exp: continue to breakpoint: continue to test_add_rn_pc_end

This patch is to fix it.

gdb:

2015-04-10  Yao Qi  <yao.qi@linaro.org>

* arm-tdep.c (install_alu_reg): Update comment.
(thumb_copy_alu_reg): Remove local variable rn.  Update
debugging message.  Use r2 instead of r1 in the modified
instruction.

gdb/testsuite:

2015-04-10  Yao Qi  <yao.qi@linaro.org>

* gdb.arch/arm-disp-step.S (main): Call test_add_rn_pc.
(test_add_rn_pc): New function.
* gdb.arch/arm-disp-step.exp (test_add_rn_pc): New proc.
(top level): Invoke test_add_rn_pc.

9 years agoPR13858 - Can't do displaced stepping with no symbols
Pedro Alves [Fri, 10 Apr 2015 09:07:02 +0000 (10:07 +0100)]
PR13858 - Can't do displaced stepping with no symbols

Running break-interp.exp with the target always in non-stop mode trips
on PR13858, as enabling non-stop also enables displaced stepping.

The problem is that when GDB doesn't know where the entry point is, it
doesn't know where to put the displaced stepping scratch pad.  The
test added by this commit exercises this.  Without the fix, we get:

 (gdb) PASS: gdb.base/step-over-no-symbols.exp: displaced=on: break *$pc
 set displaced-stepping on
 (gdb) PASS: gdb.base/step-over-no-symbols.exp: displaced=on: set displaced-stepping on
 stepi
 0x00000000004005be in ?? ()
 Entry point address is not known.
 (gdb) PASS: gdb.base/step-over-no-symbols.exp: displaced=on: stepi
 p /x $pc
 $2 = 0x4005be
 (gdb) PASS: gdb.base/step-over-no-symbols.exp: displaced=on: get after PC
 FAIL: gdb.base/step-over-no-symbols.exp: displaced=on: advanced

The fix switches all GNU/Linux ports to get the entry point from
AT_ENTRY in the target auxiliary vector instead of from symbols.  This
is currently only done by PPC when Cell debugging is enabled, but I
think all archs should be able to do the same.  Note that
ppc_linux_displaced_step_location cached the result, I'm guessing to
avoid constantly re-fetching the auxv out of remote targets, but
that's no longer necessary nowadays, as the auxv blob is itself cached
in the inferior object.  The ppc_linux_entry_point_addr global is
obviously bad for multi-process too nowadays.

Tested on x86-64 (-m64/-m32), PPC64 (-m64/-m32) and S/390 GNU/Linux.
Yao tested the new test on ARM as well.

gdb/ChangeLog:
2015-04-10  Pedro Alves  <palves@redhat.com>

PR gdb/13858
* amd64-linux-tdep.c (amd64_linux_init_abi_common): Install
linux_displaced_step_location as gdbarch_displaced_step_location
hook.
* arm-linux-tdep.c (arm_linux_init_abi): Likewise.
* i386-linux-tdep.c (i386_linux_init_abi): Likewise.
* linux-tdep.c (linux_displaced_step_location): New function,
based on ppc_linux_displaced_step_location.
* linux-tdep.h (linux_displaced_step_location): New declaration.
* ppc-linux-tdep.c (ppc_linux_entry_point_addr): Delete.
(ppc_linux_inferior_created, ppc_linux_displaced_step_location):
Delete.
(ppc_linux_init_abi): Install linux_displaced_step_location as
gdbarch_displaced_step_location hook, even without Cell/B.E..
(_initialize_ppc_linux_tdep): Don't install
ppc_linux_inferior_created as inferior_created observer.
* s390-linux-tdep.c (s390_gdbarch_init): Install
linux_displaced_step_location as gdbarch_displaced_step_location
hook.

gdb/testsuite/
2015-04-10  Pedro Alves  <palves@redhat.com>

PR gdb/13858
* gdb.base/step-over-no-symbols.exp: New file.

9 years agoCorrect setting of elf_list_options
Alan Modra [Fri, 10 Apr 2015 08:51:32 +0000 (18:21 +0930)]
Correct setting of elf_list_options

* configure.ac: Set elf_list_options etc. in proper case.
* configure: Regenerate.

9 years agodoc patch: compile: missing bits
Jan Kratochvil [Fri, 10 Apr 2015 08:34:51 +0000 (10:34 +0200)]
doc patch: compile: missing bits

gdb/doc/ChangeLog
2015-04-10  Jan Kratochvil  <jan.kratochvil@redhat.com>
    Eli Zaretskii  <eliz@gnu.org>

* gdb.texinfo (Compiling and Injecting Code): Describe set debug
compile, show debug compile.  New subsection Compilation options for
the compile command.  New subsection Compiler search for the compile
command.

9 years agoAdd documentation about the interation of the ARM assembler's -EB option and the...
Nick Clifton [Fri, 10 Apr 2015 07:26:07 +0000 (08:26 +0100)]
Add documentation about the interation of the ARM assembler's -EB option and the linker's --be8 option.

PR binutils/18198
* ld.texinfo (--be8): Add a note about the interaction of this
option with the assembler's -EB option.

* doc/c-arm.texi (ARM Options): Add a note about the interaction of
the -EB option with the linker's --be8 option.

9 years agoMark _init and _fini for --gc-sections
Alan Modra [Thu, 9 Apr 2015 09:08:03 +0000 (18:38 +0930)]
Mark _init and _fini for --gc-sections

gold marks _init and _fini via symbol, ld marks them via section
(default scripts set .init and .fini section KEEP).  This makes it
possible for people to write their own _init and not bother to put the
function into the right section.

PR ld/18223
* ldlang.c (lang_process): Add _init and _fini to gc_sym_list.

9 years agoAutomatic date update in version.in
GDB Administrator [Fri, 10 Apr 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoAllow gold to resolve defined TLS symbols in a PIE link.
Cary Coutant [Thu, 9 Apr 2015 22:47:13 +0000 (15:47 -0700)]
Allow gold to resolve defined TLS symbols in a PIE link.

This patch fixes Symbol::final_value_is_known so that a defined TLS symbol
in a PIE link is treated as having a known final value, thus allowing
GD->LE TLS optimization, eliminating an unnecessary GOT entry and dynamic
relocation.

gold/
* symtab.cc (Symbol::final_value_is_known): Check for TLS symbol
in a PIE link.
* testsuite/Makefile.am (tls_pie_test.sh): New test.
* testsuite/Makefile.in: Regenerate.
* testsuite/tls_pie_test.sh: New.

9 years agoUpdate comments in compress.c
H.J. Lu [Thu, 9 Apr 2015 19:48:49 +0000 (12:48 -0700)]
Update comments in compress.c

* compress.c (bfd_compress_section_contents): Update comments.
(bfd_init_section_decompress_status): Likewise.
(bfd_init_section_compress_status): Likewise.

9 years agodoc/c-rx.texi: Fix markup typos in last change.
Hans-Peter Nilsson [Thu, 9 Apr 2015 19:09:02 +0000 (21:09 +0200)]
doc/c-rx.texi: Fix markup typos in last change.

9 years agoImprove ODR checking in gold.
Cary Coutant [Thu, 9 Apr 2015 18:52:21 +0000 (11:52 -0700)]
Improve ODR checking in gold.

gold/
* debug.h (DEBUG_LOCATION): New.
(DEBUG_ALL): Include DEBUG_LOCATION.
(debug_string_to_enum): Add DEBUG_LOCATION.
* dwarf_reader.cc (Sized_dwarf_line_info::read_lines): Fix debug
output to print correct context.
(Sized_dwarf_line_info::do_addr2line): Add debug output. Return
up to 4 more locations at the beginning of the function.
* symtab.cc (Symbol_table::detect_odr_violations): Get canonical
result before sorting list of line numbers.
* testsuite/debug_msg.sh: Allow range of line numbers for
canonical results on optimized code.

9 years agoProperly check --compress-debug-sections=XXX
H.J. Lu [Thu, 9 Apr 2015 17:56:28 +0000 (10:56 -0700)]
Properly check --compress-debug-sections=XXX

We can't check if input is ELF in copy_file since some targets may only
set xvec after bfd_check_format_matches is called.  This patch moves
this check to copy_object.

bfd/

* elfxx-target.h (TARGET_BIG_SYM): Add BFD_COMPRESS_GABI to
object_flags.
(TARGET_LITTLE_SYM): Likewise.

binutils/

* objcopy.c (do_debug_sections): Use bit patterns.
(copy_object): Return FALSE for compress_zlib, compress_gnu_zlib
and compress_gabi_zlib on non-ELF input.
(copy_file): Don't check non-ELF input here.

9 years agoxtensa: update bfd/Changelog
Max Filippov [Thu, 9 Apr 2015 16:27:53 +0000 (19:27 +0300)]
xtensa: update bfd/Changelog

9 years agoxtensa: replace action list with splay tree
Max Filippov [Sun, 5 Apr 2015 14:04:22 +0000 (17:04 +0300)]
xtensa: replace action list with splay tree

text_action_add uses linear list search to order text actions list by
action VMA. The list is used at the first relaxation pass, when it's not
fixed yet.
Replace the list with splay tree from libiberty.

Original profile:

% time    self  children    called     name
-----------------------------------------
          0.00    0.00      14/158225      compute_text_actions
          3.62    0.00   25211/158225      remove_dead_literal
          8.42    0.00   58645/158225      coalesce_shared_literal
         10.68    0.00   74355/158225      text_action_add_proposed
  38.8   22.73    0.00  158225         text_action_add
          0.00    0.00  144527/293246      bfd_zmalloc
-----------------------------------------

Same data, after optimization:

% time    self  children    called     name
-----------------------------------------
          0.00    0.00      14/158225      compute_text_actions
          0.00    0.00   25211/158225      remove_dead_literal
          0.00    0.01   58645/158225      coalesce_shared_literal
          0.00    0.01   74355/158225      text_action_add_proposed
   0.1    0.00    0.02  158225         text_action_add
          0.01    0.00  144527/144527      splay_tree_insert
          0.00    0.00  144527/195130      splay_tree_lookup
          0.00    0.00  144527/293246      bfd_zmalloc
-----------------------------------------

2015-04-03  Max Filippov  <jcmvbkbc@gmail.com>
bfd/
* elf32-xtensa.c (splay-tree.h): include header.
(text_action_struct): drop next pointer.
(text_action_list_struct): drop head pointer, add count and
tree fields.
(find_fill_action): instead of linear search in text_action_list
search in the tree.
(text_action_compare, action_first, action_next): new functions.
(text_action_add, text_action_add_literal): instead of linear
search and insertion insert new node into the tree.
(removed_by_actions): pass additional parameter: action_list,
use it to traverse the tree.
(offset_with_removed_text): pass additional action_list parameter
to removed_by_actions.
(map_action_fn_context): new typedef.
(map_action_fn_context_struct): new structure.
(map_action_fn): new function.
(map_removal_by_action): use splay_tree_foreach to build map.
(find_insn_action): replace linear search in text_action_list
with series of splay_tree_lookups.
(print_action, print_action_list_fn): new functions.
(print_action_list): use splay_tree_foreach.
(init_xtensa_relax_info): drop action_list.head initialization.
Initialize the tree.
(compute_text_actions): use non-zero action_list_count instead of
non-NULL action list.
(xlate_map_context): new typedef.
(xlate_map_context_struct): new structure.
(xlate_map_fn): new function.
(build_xlate_map): use splay_tree_foreach to build map.
(action_remove_bytes_fn): new function.
(relax_section): use zero action_list_count instead of NULL
action list. Use splay_tree_foreach to count final section size.
Drop unused variable 'removed'.

9 years agoxtensa: optimize find_removed_literal
Max Filippov [Sat, 4 Apr 2015 11:49:42 +0000 (14:49 +0300)]
xtensa: optimize find_removed_literal

find_removed_literal uses linear search to find removed literal by its
VMA. The list of literals is fixed at that point, build an ordered index
array and use binary search instead.

Original profile:

% time    self  children    called     name
-----------------------------------------
         56.72    0.00  297578/669392      translate_reloc
         70.86    0.00  371814/669392      relax_section
  67.9  127.58    0.00  669392         find_removed_literal
-----------------------------------------

Same data, after optimization:

% time    self  children    called     name
-----------------------------------------
          0.00    0.00  297578/669392      translate_reloc
          0.00    0.00  371814/669392      relax_section
   0.0    0.00    0.00  669392         find_removed_literal
          0.00    0.00   23838/23838       map_removed_literal
-----------------------------------------

2015-04-03  Max Filippov  <jcmvbkbc@gmail.com>
bfd/
* elf32-xtensa.c (removed_literal_map_entry): new typedef.
(removed_literal_map_entry_struct): new structure.
(removed_literal_list_struct): add new fields: n_map and map.
(map_removed_literal, removed_literal_compare): new functions.
(find_removed_literal): build index array for literals ordered
by VMA, use binary search to find removed literal.

9 years agoxtensa: optimize removed_by_actions
Max Filippov [Sat, 28 Mar 2015 05:46:28 +0000 (08:46 +0300)]
xtensa: optimize removed_by_actions

The function removed_by_actions iterates through text actions to
calculate an offset applied by text actions to a given VMA. Although it
has a parameter p_start_action that allows for incremental offset
calculation, in many places it's used with p_start_action explicitly set
to the first action. After the first relaxation pass when the list of
text actions is finalized, an array of offsets sorted by VMA may be used
to speed up this function.

Original profile:

% time    self  children    called     name
-----------------------------------------
          0.35    0.00   33872/4808961     relax_section_symbols
          3.32    0.00  326022/4808961     relax_property_section
         12.83    0.00 1259379/4808961     offset_with_removed_text
         32.50    0.00 3189688/4808961     translate_reloc
  71.5   49.00    0.00 4808961         removed_by_actions
-----------------------------------------

Same data, after optimization:

% time    self  children    called     name
-----------------------------------------
          0.00    0.00   33872/4808537     relax_section_symbols
          0.01    0.00  326022/4808537     relax_property_section
          0.05    0.00 1258955/4808537     offset_with_removed_text_map
          0.13    0.00 3189688/4808537     translate_reloc
   1.0    0.20    0.00 4808537         removed_by_actions_map
          0.00    0.00     120/120         map_removal_by_action
-----------------------------------------

2015-04-01  Max Filippov  <jcmvbkbc@gmail.com>
bfd/
* elf32-xtensa.c (removal_by_action_entry_struct,
removal_by_action_map_struct): new structures.
(removal_by_action_entry, removal_by_action_map): new typedefs.
(text_action_list_struct): add new field: map.
(map_removal_by_action, removed_by_actions_map,
offset_with_removed_text_map): new functions.
(relax_section): replace offset_with_removed_text with
offset_with_removed_text_map.
(translate_reloc, relax_property_section, relax_section_symbols):
replace removed_by_actions with removed_by_actions_map.

9 years agoxtensa: optimize check_section_ebb_pcrels_fit
Max Filippov [Fri, 27 Mar 2015 04:13:55 +0000 (07:13 +0300)]
xtensa: optimize check_section_ebb_pcrels_fit

The original check_section_ebb_pcrels_fit algorithm checks that text
actions proposed for current EBB are OK for every relocation in a
section. There's no need to check every relocation, because text actions
for EBB can only change size of that EBB, thus only affecting
relocations that in any way cross that EBB. In addition EBBs are
iterated in ascending order of their VMA, making it easier to track
relevant relocations.

Introduce a structure that can track relocations that cross the range of
VMAs of EBB and use it to only check relocations relevant to current EBB
in check_section_ebb_pcrels_fit.
It takes O(N log N) operations to build it and O(N) operations to move
current EBB VMA window through its entire range, where N is the number
of relocations in a section. The resulting complexity of
compute_text_actions is thus reduced from O(N^2) to O(N log N + N * M),
where M is the average number of relocations crossing each EBB.

Original profile:

% time    self  children    called     name
-----------------------------------------
         44.26   71.53    6429/6429        compute_text_actions
  50.2   44.26   71.53    6429         check_section_ebb_pcrels_fit
          1.16   20.12 347506666/347576152     pcrel_reloc_fits
          2.95   16.52 347506666/348104944     get_relocation_opnd
          2.01    9.74 347575100/361252208     r_reloc_init
          0.55    7.53 347575100/363381467     r_reloc_get_section
          5.76    0.02 695013332/695013332     xlate_offset_with_removed_text
          0.68    3.89 347575100/363483827     bfd_octets_per_byte
          0.32    0.00 347506666/349910253     is_alt_relocation
          0.18    0.11    6391/6391        build_xlate_map
          0.00    0.00    6429/19417168     get_xtensa_relax_info
          0.00    0.00    6391/6391        free_xlate_map
-----------------------------------------

Same data, after optimization:

% time    self  children    called     name
-----------------------------------------
          2.56    3.08    6429/6429        compute_text_actions
   8.2    2.56    3.08    6429         check_section_ebb_pcrels_fit
          0.08    0.91 17721075/17790561     pcrel_reloc_fits
          0.17    0.47 17721075/31685977     r_reloc_init
          0.43    0.00 35442150/35442150     xlate_offset_with_removed_text
          0.02    0.37 17721075/33815236     r_reloc_get_section
          0.22    0.11    6391/6391        build_xlate_map
          0.05    0.22 17721075/33917596     bfd_octets_per_byte
          0.03    0.00 17721075/20405299     is_alt_relocation
          0.01    0.00    6429/6429        reloc_range_list_update_range
          0.00    0.00    6429/19417168     get_xtensa_relax_info
          0.00    0.00    6391/6391        free_xlate_map
-----------------------------------------

2015-04-01  Max Filippov  <jcmvbkbc@gmail.com>
bfd/
* elf32-xtensa.c (reloc_range_list, reloc_range_list_entry,
reloc_range): new typedef.
(reloc_range_list_struct, reloc_range_list_entry_struct,
reloc_range_struct): new structures.
(reloc_range_compare, build_reloc_ranges,
reloc_range_list_append, reloc_range_list_remove,
reloc_range_list_update_range, free_reloc_range_list): new
functions.
(compute_text_actions): precompute relocation opcodes before the
loop. Add relevant_relocs variable, initialize it before the
loop, pass it to the check_section_ebb_pcrels_fit.
(check_section_ebb_pcrels_fit): add new parameter:
relevant_relocs. Update address range in the relevant_relocs if
it's non-NULL and iterate only over relevant relocations.

9 years agoRemove BNDPLT and add PARSE_AND_LIST_ARGS_CASE_Z
H.J. Lu [Thu, 9 Apr 2015 14:46:21 +0000 (07:46 -0700)]
Remove BNDPLT and add PARSE_AND_LIST_ARGS_CASE_Z

* emulparams/elf_x86_64.sh (BNDPLT): Removed.
(PARSE_AND_LIST_OPTIONS): New.
(PARSE_AND_LIST_ARGS_CASE_Z): Likewise.
* emultempl/elf32.em (gld${EMULATION_NAME}_handle_option): Remove
BNDPLT check.  Handle PARSE_AND_LIST_ARGS_CASE_Z.
(gld${EMULATION_NAME}_list_options): Remove BNDPLT check.

9 years agoRename common-remote-fileio.[ch] as fileio.[ch]
Gary Benson [Thu, 9 Apr 2015 13:22:56 +0000 (14:22 +0100)]
Rename common-remote-fileio.[ch] as fileio.[ch]

This commit renames common-remote-fileio.[ch] as fileio.[ch]
and renames all functions in these files.

gdb/ChangeLog:

* common/common-remote-fileio.h: Rename to...
* common/fileio.h: ...this.  Update all references.
(remote_fileio_to_fio_error): Rename to...
(host_to_fileio_error): ...this.
(remote_fileio_to_be): Rename to...
(host_to_bigendian): ...this.  Update all callers.
(remote_fileio_to_fio_uint): Rename to...
(host_to_fileio_uint): ...this.  Update all callers.
(remote_fileio_to_fio_time): Rename to...
(host_to_fileio_time): ...this.  Update all callers.
(remote_fileio_to_fio_stat): Rename to...
(host_to_fileio_stat): ...this.
Update all references.
* common/common-remote-fileio.c: Rename to...
* common/fileio.c: ...this.  Update all references.
(remote_fileio_to_fio_error): Rename to...
(host_to_fileio_error): ...this.  Update all callers.
(remote_fileio_mode_to_target): Rename to...
(fileio_mode_pack): ...this.  Update all callers.
(remote_fileio_to_fio_mode): Rename to...
(host_to_fileio_mode): ...this.  Update all callers.
(remote_fileio_to_fio_ulong): Rename to...
(host_to_fileio_ulong): ...this.  Update all callers.
(remote_fileio_to_fio_stat): Rename to...
(host_to_fileio_stat): ...this.  Update all callers.

9 years agoUpdate French translation of ld strings.
Nick Clifton [Thu, 9 Apr 2015 13:56:29 +0000 (14:56 +0100)]
Update French translation of ld strings.

* po/fr.po: Updated French translation.