platform/upstream/binutils.git
9 years agoAutomatic date update in version.in
GDB Administrator [Mon, 13 Jul 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoExtend --compress-debug-sections in gold
H.J. Lu [Sun, 12 Jul 2015 17:56:41 +0000 (10:56 -0700)]
Extend --compress-debug-sections in gold

This patch extends --compress-debug-sections= with [zlib-gnu|zlib-gabi]
in gold.

PR gold/18322
* compressed_output.cc (zlib_compress): Add argument for
compression header size.  Set header size to compression header
size if it isn't 0.  Don't write out the zlib header here.
(Output_compressed_section::set_final_data_size): Support
zlib-gnu and zlib-gabi compressions.  Pass compression header
size to zlib_compress and write out compression header.  Set
the SHF_COMPRESSED bit for zlib-gabi compression.  Otherwise
clear the SHF_COMPRESSED bit
* options.h (compress_debug_sections): Add zlib-gnu and
zlib-gabi.
* output.h (Output_section::set_flags): New.
* testsuite/Makefile.am (check_PROGRAMS): Add
flagstest_compress_debug_sections_none,
flagstest_compress_debug_sections_gnu and
flagstest_compress_debug_sections_gabi.
(check_DATA): Add flagstest_compress_debug_sections_none.stdout.
flagstest_compress_debug_sections.stdout,
flagstest_compress_debug_sections.cmp,
flagstest_compress_debug_sections.check,
flagstest_compress_debug_sections_gnu.stdout,
flagstest_compress_debug_sections_gnu.cmp,
flagstest_compress_debug_sections_gnu.check,
flagstest_compress_debug_sections_gabi.stdout,
flagstest_compress_debug_sections_gabi.cmp and
flagstest_compress_debug_sections_gabi.check.
(flagstest_compress_debug_sections_none): New.
(flagstest_compress_debug_sections_none.stdout): Likewise.
(flagstest_compress_debug_sections.stdout): Likewise.
(flagstest_compress_debug_sections.check): Likewise.
(flagstest_compress_debug_sections.cmp): Likewise.
(flagstest_compress_debug_sections_gnu): Likewise.
(flagstest_compress_debug_sections_gnu.stdout): Likewise.
(flagstest_compress_debug_sections_gnu.check): Likewise.
(flagstest_compress_debug_sections_gnu.cmp): Likewise.
(flagstest_compress_debug_sections_gabi): Likewise.
(flagstest_compress_debug_sections_gabi.stdout): Likewise.
(flagstest_compress_debug_sections_gnu.check): Likewise.
(flagstest_compress_debug_sections_gnu.cmp): Likewise.
* testsuite/Makefile.in: Regenerated.

9 years agoAdd SHF_COMPRESSED section decompression to gold
H.J. Lu [Sun, 12 Jul 2015 17:50:25 +0000 (10:50 -0700)]
Add SHF_COMPRESSED section decompression to gold

This patch adds SHF_COMPRESSED section decompression to gold.

PR gold/18321
* compressed_output.h (decompress_input_section): Add arguments
for ELF class, big endian and sh_flags.
* compressed_output.cc (decompress_input_section): Likewise.
Support the SHF_COMPRESSED section.
* dynobj.h (Dynobj): Add elfsize and is_big_endian member
functions.
* plugin.h (Pluginobj): Likewise.
* layout.cc (Layout::get_output_section_flags): Also clear the
SHF_COMPRESSED bit.
* object.h (Compressed_section_info): Add flag to store sh_flags.
(Object): Add pure virtual elfsize and is_big_endian member
functions.
* object.cc (need_decompressed_section): Don't skip the ".zdebug"
prefix here.
(build_compressed_section_map): Check SHF_COMPRESSED for
uncompressed size.  Store sh_flags in Compressed_section_info.
Pass size, big_endian and sh_flags to decompress_input_section.
Skip the ".debug"/".zdebug" prefix when passing section name to
need_decompressed_section.
(Sized_relobj_file<size, big_endian>::do_find_special_section):
Don't check ".zdebug_*" sections.
(Object::decompressed_section_contents): Pass ELF class, big
endian and sh_flags to decompress_input_section.
* reloc.cc (Sized_relobj_file<size, big_endian>::write_sections):
Likewise.
* testsuite/Makefile.am (check_DATA): Add
debug_msg_cdebug_gabi.err and gdb_index_test_2_gabi.stdout.
(MOSTLYCLEANFILES): Add debug_msg_cdebug_gabi.err and
gdb_index_test_2_gabi.stdout.
(debug_msg_cdebug_gabi.o): New.
(odr_violation1_cdebug_gabi.o): Likewise.
(odr_violation2_cdebug_gabi.o): Likewise.
(debug_msg_cdebug_gabi.err): Likewise.
(check_SCRIPTS): Add gdb_index_test_2_gabi.sh.
(gdb_index_test_cdebug_gabi.o): Likewise.
(gdb_index_test_2_gabi): Likewise.
(gdb_index_test_2_gabi.stdout): Likewise.
* testsuite/gdb_index_test_2_gabi.sh: New file.
* testsuite/Makefile.in: Regenerated.

9 years agoFix GDB breakage due to addition of new nios2 mach types in BFD.
Sandra Loosemore [Sun, 12 Jul 2015 14:13:12 +0000 (07:13 -0700)]
Fix GDB breakage due to addition of new nios2 mach types in BFD.

2015-07-12  Sandra Loosemore  <sandra@codesourcery.com>

gdb/
* nios2-linux-tdep.c (_initialize_nios2_linux_tdep): Register
all nios2 mach variants.

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

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

9 years agoProperly convert objects between different ELF classes
H.J. Lu [Fri, 10 Jul 2015 21:20:20 +0000 (14:20 -0700)]
Properly convert objects between different ELF classes

The output SHF_COMPRESSED section size is different from input if
ELF classes of input and output aren't the same.  We must adjust
the section sizes as well as the compression headers in
SHF_COMPRESSED sections when converting objects between different
ELF classes.

bfd/

 PR binutils/18656
 * bfd.c (bfd_convert_section_size): New function.
 (bfd_convert_section_contents): Likewise.
 * bfd-in2.h: Regenerated.

binutils/

2015-07-10  H.J. Lu  <hongjiu.lu@intel.com>

 PR binutils/18656
 * objcopy.c (setup_section): Call bfd_convert_section_size
 to get the output section size.
 (copy_section): Get the section size from the output section
 and call bfd_get_full_section_contents to convert section
 contents for output.

binutils/testsuite/

 PR binutils/18656
 * binutils-all/compress.exp (convert_test): New proc.
 Run conversion tests between x86-64 and x32.

9 years agorx-tdep.c: Use target_read_code instead of target_read_memory.
Kevin Buettner [Thu, 9 Jul 2015 23:05:31 +0000 (16:05 -0700)]
rx-tdep.c: Use target_read_code instead of target_read_memory.

This change causes the prologue scanner and the frame type scanner in
rx-tdep.c to use target_read_code() instead of target_read_memory().
This change allows these instruction scanners to operate much more
quickly due to the fact that target_read_code() can potentially read
from a cache maintained by GDB.

gdb/ChangeLog:

* rx-tdep.c (rx_get_opcode_byte): Use target_read_code instead
of target_read_memory.

9 years agogdbserver/event-loop.c: type mismatch
Pedro Alves [Fri, 10 Jul 2015 18:47:17 +0000 (19:47 +0100)]
gdbserver/event-loop.c: type mismatch

Building with C++ catches a buglet here:

../../../src/gdb/gdbserver/event-loop.c:205:19: warning: invalid conversion from ‘gdb_client_data {aka void*}’ to ‘void**’ [-fpermissive]
   event_ptr->data = data;
                   ^

This works in practice because gdb_client_data is a pointer already
(hence in C we get an implicit conversion), and nothing deferences the
pointer.  It's passed from client at event registration/creation time,
only to pass straight back to client callback.

Well, that and nothing in gdbserver uses the event data anyway.

gdb/gdbserver/ChangeLog:
2015-07-10  Pedro Alves  <palves@redhat.com>

* event-loop.c (struct callback_event) <data>: Change type to
gdb_client_data instance instead of gdb_client_data pointer.
(append_callback_event): Adjust.

9 years agoWarn unsupported compress type and corrupted compressed section
H.J. Lu [Fri, 10 Jul 2015 17:23:12 +0000 (10:23 -0700)]
Warn unsupported compress type and corrupted compressed section

* readelf.c (dump_section_as_strings): Warn unsupported compress
type and corrupted compressed section.
(dump_section_as_bytes): Likewise.
(load_specific_debug_section): Likewise.

9 years agoUpdated to accept .debug_* sections
H.J. Lu [Fri, 10 Jul 2015 15:41:10 +0000 (08:41 -0700)]
Updated to accept .debug_* sections

* gas/i386/dw2-compress-1.d: Updated to accept .debug_* sections.

9 years agoUpdated to accept .debug_* sections
H.J. Lu [Fri, 10 Jul 2015 15:14:59 +0000 (08:14 -0700)]
Updated to accept .debug_* sections

* binutils-all/dw2-3.W: Updated to accept .debug_* sections.

9 years agoFix testsuite regression by: Do not skip prologue for asm (.S) files
Jan Kratochvil [Fri, 10 Jul 2015 13:04:51 +0000 (15:04 +0200)]
Fix testsuite regression by: Do not skip prologue for asm (.S) files

I have somehow missed gdb.asm/asm-source.exp PASS->FAIL even on x86_64.

It has no longer valid assumption that "break" breaks after the prologue even
in assembler.  So I have changed this assumption of the testfile.

gdb/testsuite/ChangeLog
2015-07-10  Jan Kratochvil  <jan.kratochvil@redhat.com>

* gdb.asm/asm-source.exp (f at main): Stop at gdbasm_enter.
(n at main): New.
* gdb.asm/asmsrc1.s: Add comment "mark: main enter".

9 years ago[GDBserver][AArch64] Cleanup comments for each linux_target_ops method
Pierre Langlois [Thu, 2 Jul 2015 11:11:47 +0000 (12:11 +0100)]
[GDBserver][AArch64] Cleanup comments for each linux_target_ops method

This patch cleans up the comments for each linux_target_ops methods.  We
should mention which method each function implements but there is no
need to duplicate information already mentionned in the base target_ops
or linux_target_ops definitions.

gdb/gdbserver/ChangeLog:

* linux-aarch64-low.c: Add comments for each linux_target_ops
method.  Remove comments already covered in target_ops and
linux_target_ops definitions.
(the_low_target): Add comments for each unimplemented method.

9 years agogdb/tui: Use cleanups to free string copies.
Andrew Burgess [Mon, 6 Jul 2015 17:16:24 +0000 (18:16 +0100)]
gdb/tui: Use cleanups to free string copies.

In parse_scrolling_args it is possible for a string copy to leak if an
error occurs.  Switching to using a cleanup fixes this leak.

In tui_set_win_height the string can't be leaked, but switching to using
a cleanup guards against the possibility that a leak could be introduced
in the future (by adding an error somewhere in the call stack).

gdb/ChangeLog:

* tui/tui-win.c (tui_set_win_height): Use a cleanup to free the
string copy.
(parse_scrolling_args): Likewise.

9 years agogdb/tui: Define tui window names once.
Andrew Burgess [Mon, 6 Jul 2015 16:50:28 +0000 (17:50 +0100)]
gdb/tui: Define tui window names once.

Don't duplicate the window names inside the completion function.
Instead make use of the existing defines, and the tui_win_name function
to obtain the window names.

gdb/ChangeLog:

* tui/tui-win.c (focus_completer): Don't duplicate the tui window
names in this function.

9 years agogdb: Convert TUI windows names to lower case.
Andrew Burgess [Mon, 6 Jul 2015 10:09:05 +0000 (11:09 +0100)]
gdb: Convert TUI windows names to lower case.

This commit converts the window names for the TUI windows to lower case.
The windows names are already lower case in the documentation, and are
shown as lower case in some of the command completion results.

Given that all the commands that take a window name currently accept
upper or lower case, this commit just changes the window names to lower
case in the remaining places they are displayed by gdb.

gdb/ChangeLog:

* tui/tui-data.h (SRC_NAME): Convert to lower case.
(CMD_NAME): Likewise.
(DATA_NAME): Likewise.
(DISASSEM_NAME): Likewise.
* tui/tui-win.c (tui_set_focus): Window names are now lower case.
(tui_set_win_height): Likewise.
(parse_scrolling_args): Likewise.

9 years agoMake ALIGN(x) behave as ALIGN(.,x)
Alan Modra [Wed, 8 Jul 2015 12:48:34 +0000 (22:18 +0930)]
Make ALIGN(x) behave as ALIGN(.,x)

Inside output sections, ALIGN(.,x) uses a section-relative value for
dot.  The unary ALIGN always used the absolute value of dot.

* ldexp.c (align_dot_val): New function.
(fold_unary <ALIGN_K, NEXT>): Use it.

9 years agoAdd missing changelog entries
Alan Modra [Fri, 10 Jul 2015 10:36:25 +0000 (20:06 +0930)]
Add missing changelog entries

9 years agognu_vector.exp: Skip infcall tests on x86/x86_64
Andreas Arnez [Fri, 10 Jul 2015 08:36:02 +0000 (10:36 +0200)]
gnu_vector.exp: Skip infcall tests on x86/x86_64

Since the new KFAILs/KPASSs for the infcall tests on x86 and x86_64
targets generated unnecessary noise, this change skips them with
UNSUPPORTED instead.

gdb/testsuite/ChangeLog:

* gdb.base/gnu_vector.exp: On x86 and x86_64 targets, skip the
infcall tests instead of setting up for KFAIL.

9 years agorecord: set stop_pc in "record goto" command
Markus Metzger [Mon, 6 Jul 2015 14:36:45 +0000 (16:36 +0200)]
record: set stop_pc in "record goto" command

When navigating in the recorded execution trace via "record goto", we do not
set stop_pc.  This may trigger an internal error in infrun.c when stepping
from that location.  Set it.

(gdb) rec full
(gdb) c
Continuing.

Breakpoint 1, foo (void) at foo.c:42
42             x = y
(gdb) rn
foo (void)
    at foo.c:41
41             y = x
(gdb) rec go end
Go forward to insn number 98724
    at foo.c:42
42             x = y
(gdb) n
infrun.c:2382: internal-error: resume: Assertion `sig != GDB_SIGNAL_0' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)

This happens because there's a breakpoint at PC when the "next"
is issued, so that breapoint should be immediately stepped over.
That should have been detected/done by proceed, here:

  if (addr == (CORE_ADDR) -1)
    {
      if (pc == stop_pc
  && breakpoint_here_p (aspace, pc) == ordinary_breakpoint_here
  && execution_direction != EXEC_REVERSE)
/* There is a breakpoint at the address we will resume at,
   step one instruction before inserting breakpoints so that
   we do not stop right away (and report a second hit at this
   breakpoint).

   Note, we don't do this in reverse, because we won't
   actually be executing the breakpoint insn anyway.
   We'll be (un-)executing the previous instruction.  */
tp->stepping_over_breakpoint = 1;

But since stop_pc was stale, the pc == stop_pc check failed, and left the
breakpont at PC inserted.

gdb/
* record-btrace.c (record_btrace_goto_begin, record_btrace_goto_end)
record_btrace_goto): Move call to print_stack_frame ...
(record_btrace_set_replay): ... here.  Set stop_pc.
* record-full.c (record_full_goto_entry): Set stop_pc.

testsuite/
* gdb.btrace/record_goto-step.exp: New.

9 years agoUse "gold_info" instead of "gold_warning" for erratum fix.
Han Shen [Thu, 9 Jul 2015 22:55:55 +0000 (15:55 -0700)]
Use "gold_info" instead of "gold_warning" for erratum fix.

gold/ChangeLog:

* aarch64.cc (Target_aarch64::scan_erratum_843419_span): Use 'gold_info'.
(Target_aarch64::scan_erratum_835769_span): USe 'gold_info'.

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

9 years agoDrop 'missing mapping symbols' warning for arm and aarch64.
Han Shen [Thu, 9 Jul 2015 22:38:12 +0000 (15:38 -0700)]
Drop 'missing mapping symbols' warning for arm and aarch64.

The psABI (AAELF) says this about mapping symbols:

    4.6.5.1 Section-relative mapping symbols

    Mapping symbols defined in a section define a sequence of
    half-open address intervals that cover the address range of the
    section. Each interval starts at the address defined by the
    mapping symbol, and continues up to, but not including, the
    address defined by the next (in address order) mapping symbol or
    the end of the section. A section must have a mapping symbol
    defined at the beginning of the section; however, if the section
    contains only data then the mapping symbol may be omitted.

That makes it pretty clear that, in the absence of a mapping symbol,
the entire section can be treated as data.

gold/ChangeLog:
* aarch64.cc (AArch64_relobj::scan_errata): Drop missing symbol warning.
* arm.cc (Arm_relobj::scan_section_for_cortex_a8_erratum): Drop missing
symbol warning.

9 years agoAdd NEWS entry for tracepoints support on aarch64-linux
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
Add NEWS entry for tracepoints support on aarch64-linux

gdb/ChangeLog:

* NEWS: Mention support for tracepoints on aarch64-linux.

9 years ago[GDBServer][AArch64] Enable support for tracepoints
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
[GDBServer][AArch64] Enable support for tracepoints

gdb/gdbserver/ChangeLog:

* linux-aarch64-low.c (aarch64_supports_tracepoints): New
function.  Return 1.
(the_low_target): Install it.

9 years ago[testsuite][AArch64] Port gdb.trace
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
[testsuite][AArch64] Port gdb.trace

This patch adds support for AArch64 to the gdb.trace testsuite.

Note that it does not add support for testing fast tracepoint as it
isn't supported.  Therefore the test cases with inline assembly are not
ported in this patch, as we do not know what the conditions for
inserting a fast tracepoint on AArch64 would be.

gdb/testsuite/ChangeLog:

* gdb.trace/backtrace.exp: Set registers for aarch64 target.
* gdb.trace/collection.exp: Likewise.
* gdb.trace/mi-trace-frame-collected.exp: Likewise.
* gdb.trace/mi-trace-unavailable.exp: Likewise.
* gdb.trace/report.exp: Likewise.
* gdb.trace/trace-break.exp: Likewise.
* gdb.trace/unavailable.exp: Likewise.
* gdb.trace/while-dyn.exp: Likewise.

9 years ago[AArch64] Implement gdbarch_gen_return_address gdbarch method
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
[AArch64] Implement gdbarch_gen_return_address gdbarch method

This patch implements the 'collect $_ret' command to collect the return
address of a function in a tracepoint.  It marks the LR register for
collection.

gdb/ChangeLog:

* aarch64-tdep.c: Add ax.h and ax-gdb.h includes.
(aarch64_gen_return_address): New function.
(aarch64_gdbarch_init): Hook it.

9 years ago[AArch64] Teach stub unwinder to terminate gracefully
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
[AArch64] Teach stub unwinder to terminate gracefully

The stub unwinder is used on AArch64 if the target's memory is not
readable at the current PC.  For example, the user could try to call at
an invalid address such as 0x0, as covered in the gdb.base/signull.exp
test case.  Many GDB ports use a similar unwinder to handle this case
too.

If we purposely kill the inferior before examining the trace then we get
the following issue:

~~~
...
(gdb) trace f
Tracepoint 3 at 0x7fb7fc28c0
(gdb) tstart
(gdb) continue
...
(gdb) tstop
(gdb) tsave /tmp/trace
(gdb) kill
...
(gdb) target tfile /tmp/trace
...
(gdb) tfind
Register 31 is not available.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Found trace frame 0, tracepoint 3
#-1 0x0000007fb7fc28c0 in f () ...
^^^
~~~

This patch teaches the stub unwinder to report to the core frame code
with UNWIND_UNAVAILABLE when either the stack pointer of the return
address are unavailable to read from the target.

gdb/ChangeLog:

* aarch64-tdep.c (aarch64_make_stub_cache): Set available_p and
swallow NOT_AVAILABLE_ERROR.
(aarch64_stub_this_id): Call frame_id_build_unavailable_stack if
available_p is not set.
(aarch64_stub_frame_unwind_stop_reason): New function.
(aarch64_stub_unwind): Install it.

9 years ago[AArch64] Teach prologue unwinder to terminate gracefully
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
[AArch64] Teach prologue unwinder to terminate gracefully

Without debugging information, we have the following issue when
examining a trace buffer:

~~~
...
(gdb) trace f
Tracepoint 3 at 0x7fb7fc28c0
(gdb) tstart
(gdb) continue
...
(gdb) tstop
(gdb) tfind start
Register 31 is not available.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Found trace frame 0, tracepoint 3
#-1 0x0000007fb7fc28c0 in f () ...
^^^
~~~

The reason for this is that the target's stack pointer is unavailable
when examining the trace buffer.  What we are seeing is due to the
'tfind' command creating a sentinel frame and unwinding it.  If an
exception is thrown, we are left with the sentinel frame being displayed
at level #-1.  The exception is thrown when the prologue unwinder tries
to read the stack pointer to construct an ID for the frame.

This patch fixes this by making the prologue unwinder catch
NOT_AVAILABLE_ERROR exceptions when either registers or memory is
unreadable and report back to the frame core code with
UNWIND_UNAVAILABLE.

gdb/ChangeLog:

* aarch64-tdep.c (aarch64_prologue_cache) <available_p>: New
field.
(aarch64_make_prologue_cache_1): New function, factored out from
aarch64_make_prologue_cache.  Do not allocate cache.  Set
available_p.
(aarch64_make_prologue_cache): Reimplement wrapping
aarch64_make_prologue_cache_1, and swallowing
NOT_AVAILABLE_ERROR.
(aarch64_prologue_frame_unwind_stop_reason): New function.
Return UNWIND_UNAVAILABLE if available_p is not set.
(aarch64_prologue_unwind): Install it.
(aarch64_prologue_this_id): Move prev_pc and prev_sp limit
checks into aarch64_prologue_frame_unwind_stop_reason.  Call
frame_id_build_unavailable_stack if available_p is not set.

9 years ago[AArch64] Only access inferior registers when creating a frame cache
Pierre Langlois [Thu, 9 Jul 2015 15:35:11 +0000 (16:35 +0100)]
[AArch64] Only access inferior registers when creating a frame cache

This patch moves the address of the start of a function (func) and the
address from which it was called (prev_pc) into aarch64_prologue_cache.
The idea is to keep accesses to the inferior's registers into
aarch64_make_prologue_cache and aarch64_make_stub_cache.

gdb/ChangeLog:

* aarch64-tdep.c (aarch64_prologue_cache) <func, prev_pc>: New
fields.
(aarch64_scan_prologue): Set prev_pc.
(aarch64_make_prologue_cache): Set func.
(aarch64_make_stub_cache): Set prev_pc.
(aarch64_prologue_this_id): Remove local variables id, pc and
func.  Read prev_pc and func from cache.
(aarch64_stub_this_id): Read prev_pc from cache.

9 years ago[AArch64] Refactor aarch64_make_stub_cache
Pierre Langlois [Thu, 9 Jul 2015 15:35:10 +0000 (16:35 +0100)]
[AArch64] Refactor aarch64_make_stub_cache

We would previously have to make sure the frame cache was not already
created before calling aarch64_make_stub_cache.  This patch makes this
function check it so the caller does not need to do so.

gdb/ChangeLog:

* aarch64-tdep.c (aarch64_make_stub_cache): Update comment.  New
argument this_cache.  Remove unused local variables reg and
unwound_fp.  Return early if this_cache is already set.  Set
this_cache.
(aarch64_stub_this_id): Update call to aarch64_make_stub_cache.

9 years ago[AArch64] Refactor aarch64_make_prologue_cache
Pierre Langlois [Thu, 9 Jul 2015 15:35:10 +0000 (16:35 +0100)]
[AArch64] Refactor aarch64_make_prologue_cache

We would previously have to make sure the frame cache was not already
created before calling aarch64_make_prologue_cache.  This patch makes
this function check it so that the caller does not need to do so.

gdb/ChangeLog:

* aarch64-tdep.c (aarch64_make_prologue_cache): Update comment.
New argument this_cache.  Return early if this_cache is already
set.  Set this_cache.
(aarch64_prologue_this_id): Update call to
aarch64_make_prologue_cache.
(aarch64_prologue_prev_register): Likewise.
(aarch64_normal_frame_base): Likewise.

9 years ago2015-07-09 Catherine Moore <clm@codesourcery.com>
Catherine Moore [Thu, 9 Jul 2015 15:24:54 +0000 (08:24 -0700)]
2015-07-09  Catherine Moore  <clm@codesourcery.com>

        include/
        * elf/mips/mips.h (Val_GNU_MIPS_ABI_FP_NAN2008): New.

        gas/
        * config/tc-mips.c (check_fpabi):  Handle
        VAL_GNU_MIPS_ABI_FP_NAN2008.

        binutils/
        * readelf.c (print_mips_fp_abi_value): Handle
        Val_GNU_MIPS_ABI_FP_NAN2008.

        ld/testsuite/
        * ld-mips-elf/attr-gnu-4-08.d: Update expected output.
        * ld-mips-elf/attr-gnu-4-09.d: New.
        * ld-mips-elf/attr-gnu-4-19.d: New.
        * ld-mips-elf/attr-gnu-4-29.d: New.
        * ld-mips-elf/attr-gnu-4-39.d: New.
        * ld-mips-elf/attr-gnu-4-49.d: New.
        * ld-mips-elf/attr-gnu-4-59.d: New.
        * ld-mips-elf/attr-gnu-4-69.d: New.
        * ld-mips-elf/attr-gnu-4-79.d: New.
        * ld-mips-elf/attr-gnu-4-89.d: New.
        * ld-mips-elf/attr-gnu-4-9.s: New.
        * ld-mips-elf/mips-elf.exp: Run new tests.

9 years agoFactor out memberptr printing code from c_val_print
Simon Marchi [Thu, 9 Jul 2015 15:19:25 +0000 (11:19 -0400)]
Factor out memberptr printing code from c_val_print

gdb/ChangeLog:

* c-valprint.c (c_val_print): Factor out memberptr printing code
from c_val_print to ...
(c_val_print_memberptr): ... this new function.

9 years agoFactor out int printing code from c_val_print
Simon Marchi [Thu, 9 Jul 2015 15:18:59 +0000 (11:18 -0400)]
Factor out int printing code from c_val_print

gdb/ChangeLog:

* c-valprint.c (c_val_print): Factor out int printing code to ...
(c_val_print_int): ... this new function.

9 years agoFactor out struct and union printing code from c_val_print
Simon Marchi [Thu, 9 Jul 2015 15:18:39 +0000 (11:18 -0400)]
Factor out struct and union printing code from c_val_print

gdb/ChangeLog:

* c-valprint.c (c_val_print): Factor out struct and union
printing code to ...
(c_val_print_struct): ... this new function ...
(c_val_print_union): ... and this new function.

9 years agoFactor out pointer printing code from c_val_print
Simon Marchi [Thu, 9 Jul 2015 15:18:12 +0000 (11:18 -0400)]
Factor out pointer printing code from c_val_print

gdb/ChangeLog:

* c-valprint.c (c_val_print): Factor out pointer printing code
to ...
(c_val_print_ptr): ... this new function.

9 years agoFactor out array printing code from c_val_print
Simon Marchi [Thu, 9 Jul 2015 15:17:20 +0000 (11:17 -0400)]
Factor out array printing code from c_val_print

gdb/ChangeLog:

* c-valprint.c (c_valprint): Factor our array printing code from
c_val_print to ...
(c_val_print_array): ... this new function.

9 years agoFactor out print_unpacked_pointer from c_val_print
Simon Marchi [Thu, 9 Jul 2015 15:16:22 +0000 (11:16 -0400)]
Factor out print_unpacked_pointer from c_val_print

Turn this code into a function, instead of a goto.

gdb/ChangeLog:

* c-valprint.c (c_val_print): Factor out pointer printing code
to ...
(print_unpacked_pointer): ... this new function.

9 years agoRemove unneeded variable assignment
Simon Marchi [Thu, 9 Jul 2015 15:15:40 +0000 (11:15 -0400)]
Remove unneeded variable assignment

The assignment to i in the TYPE_CODE_PTR section is not useful.
Removing it allows to move i in a narrower scope, which will help
things somewhere in the next patches.

gdb/ChangeLog:

* c-valprint.c (c_val_print): Remove an assignment to i and move
its declaration.

9 years ago2015-07-09 Catherine Moore <clm@codesourcery.com>
Catherine Moore [Thu, 9 Jul 2015 15:07:47 +0000 (08:07 -0700)]
2015-07-09  Catherine Moore  <clm@codesourcery.com>

        * elflink.c (bfd_elf_size_dynamic_sections): Call to
        _bfd_elf_maybe_strip_eh_frame_hdr should be before
        bed->elf_backend_size_dynamic_sections.

9 years agoPass ptid to aarch64_linux_get_debug_reg_capacity
Yao Qi [Thu, 9 Jul 2015 11:55:45 +0000 (12:55 +0100)]
Pass ptid to aarch64_linux_get_debug_reg_capacity

This patch is to pass ptid to aarch64_linux_get_debug_reg_capacity,
and stop using global variable inferior_ptid.

gdb:

2015-07-09  Yao Qi  <yao.qi@linaro.org>

* aarch64-linux-nat.c (aarch64_linux_get_debug_reg_capacity): Add
argument ptid.  Update comments.  Caller update.

9 years ago[gdbserver, aarch64] Set usrregs to NULL in regs_info
Yao Qi [Thu, 9 Jul 2015 11:47:06 +0000 (12:47 +0100)]
[gdbserver, aarch64] Set usrregs to NULL in regs_info

Hi,
I happen to read the comments in regs_info below,

struct regs_info
{
  ...
  /* Info used when accessing registers with PTRACE_PEEKUSER /
     PTRACE_POKEUSER.  This can be NULL if all registers are
     transferred with regsets  .*/
  struct usrregs_info *usrregs;

that usrregs can be NULL if all registers are transferred with
regsets, which is exactly what aarch64-linux does.  This patch
is to set usrregs to NULL in regs_info and remove
aarch64_usrregs_info and aarch64_regmap.

gdb/gdbserver:

2015-07-09  Yao Qi  <yao.qi@linaro.org>

* linux-aarch64-low.c (aarch64_regmap): Remove.
(aarch64_usrregs_info): Remove.
(regs_info): Set field usrregs to NULL.

9 years agobtrace, pt: support new packets
Markus Metzger [Thu, 2 Jul 2015 13:16:09 +0000 (15:16 +0200)]
btrace, pt: support new packets

Add support for dumping new Intel(R) Processor Trace packets in the
"maint btrace packet-history" command.

gdb/
* btrace.c (pt_print_packet): Print stop, vmcs, tma, mtc, cyc, and
mnt packets.

9 years agobtrace: fix build fail with 32-bit BFD
Markus Metzger [Tue, 7 Jul 2015 11:54:34 +0000 (13:54 +0200)]
btrace: fix build fail with 32-bit BFD

When compiling GDB with 32-bit BFD, the build fails with:

In file included from btrace.h:33:0,
                 from btrace.c:23:
/usr/include/intel-pt.h:1643:51: note: expected 'int (*)(uint8_t *, size_t,
 const struct pt_asid *, uint64_t, void *)' but argument is of type 'int
 (*)(gdb_byte *, size_t, const struct pt_asid *, CORE_ADDR, void *)' extern
 pt_export int pt_image_set_callback(struct pt_image *image, ^

gdb/
* btrace.c (btrace_pt_readmem_callback): Change type of PC argument.

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

9 years agoAdd missing spaces in previous patch
Simon Marchi [Wed, 8 Jul 2015 19:48:02 +0000 (15:48 -0400)]
Add missing spaces in previous patch

gdb/ChangeLog:

* progspace.c (delete_program_space): Add missing spaces.

9 years agoDelete program spaces directly when removing inferiors
Simon Marchi [Wed, 8 Jul 2015 19:41:01 +0000 (15:41 -0400)]
Delete program spaces directly when removing inferiors

When deleting an inferior, delete the associated program space as well
if it becomes unused. This replaces the "pruning" approach, with which
you could forget to call prune_program_spaces (as seen, with the
-remove-inferior command, see [1]).

This allows to remove the prune_program_spaces function. At the same
time, I was able to clean up the delete_inferior* family:

 - delete_inferior is unused
 - delete_inferior_silent is only used in monitor_close, but is replaced
   with discard_all_inferiors [2], so it becomes unused
 - All remaining calls to delete_inferior_1 are with silent=1, so the
   parameter is removed
 - delete_inferior_1 is renamed to delete_inferior

I renamed pspace_empty_p to program_space_empty_p. I prefer if the
"exported" functions have a more explicit and standard name.

Tested on Ubuntu 14.10.

[1] https://sourceware.org/ml/gdb-patches/2014-09/msg00717.html
[2] See https://sourceware.org/ml/gdb-patches/2015-07/msg00228.html and
    follow-ups for details.

gdb/Changelog:

* inferior.c (delete_inferior_1): Rename to ...
(delete_inferior): ..., remove 'silent' parameter, delete
program space when unused and remove call to prune_program_spaces.
Remove the old, unused, delete_inferior.
(delete_inferior_silent): Remove.
(prune_inferiors): Change call from delete_inferior_1 to
delete_inferior and remove 'silent' parameter. Remove call to
prune_program_spaces.
(remove_inferior_command): Idem.
* inferior.h (delete_inferior_1): Rename to...
(delete_inferior): ..., remove 'silent' parameter and remove the
original delete_inferior.
(delete_inferior_silent): Remove.
* mi/mi-main.c (mi_cmd_remove_inferior): Change call from
delete_inferior_1 to delete_inferior and remove 'silent'
parameter.
* progspace.c (prune_program_spaces): Remove.
(pspace_empty_p): Rename to...
(program_space_empty_p): ... and make non-static.
(delete_program_space): New.
* progspace.h (prune_program_spaces): Remove declaration.
(program_space_empty_p): New declaration.
(delete_program_space): New declaration.
* monitor.c (monitor_close): Replace call to
delete_thread_silent and delete_inferior_silent with
discard_all_inferiors.

9 years agoClarify case requirements for gas pseudo-ops
Richard Sandiford [Wed, 8 Jul 2015 19:11:03 +0000 (20:11 +0100)]
Clarify case requirements for gas pseudo-ops

gas/
2015-07-08  Ciro Santilli  <ciro.santilli@gmail.com>

* doc/as.texinfo: Clarify case requirements for pseudo ops.

9 years agotui: replace deprecated_register_changed_hook with observer
Patrick Palka [Thu, 2 Jul 2015 16:27:50 +0000 (12:27 -0400)]
tui: replace deprecated_register_changed_hook with observer

This is a straightforward replacement of the TUI's use of the
aforementioned hook with the register_changed observer.  Since this was
the only user of the hook, this patch also removes the hook.

gdb/ChangeLog:

* defs.h (deprecated_register_changed_hook): Remove prototype.
* interps.c (clear_iterpreter_hooks): Remove reference to
deprecated_register_changed_hook.
* top.c (deprecated_register_changed_hook): Remove prototype.
* valops.c (value_assign): Remove reference to
deprecated_register_changed_hook.
* tui/tui-hooks.c (tui_register_changed): Add parameter "frame".
Add comment documenting the function.
(tui_register_changed_observer): Define.
(tui_install_hooks): Remove reference to
deprecated_register_changed_hook.  Set
tui_register_changed_observer.
(tui_remove_hooks): Remove reference to
deprecated_register_changed_hook.  Unset
tui_register_changed_observer.

9 years agoDefine DIFF_EXPR_OK for avr target to allow PC relative difference relocation.
Denis Chertykov [Wed, 8 Jul 2015 18:35:19 +0000 (21:35 +0300)]
Define DIFF_EXPR_OK for avr target to allow PC relative difference relocation.

When generating relocation (tc_gen_reloc) 32 bit relocation fixup
is changed to new 32 bit PC relative relocation if the fixup has pc-relative
flag set.

bfd/ChangeLog

2015-07-06  Pitchumani Sivanupandi <pitchumani.s@atmel.com>

    * elf32-avr.c: Add 32 bit PC relative relocation for AVR target.

gas/ChangeLog

2015-07-06  Pitchumani Sivanupandi <pitchumani.s@atmel.com>

    * config/tc-avr.c (tc_gen_reloc): Change 32 bit relocation to
    32 bit PC relative and update offset if the fixup is pc-relative.
    * config/tc-avr.h (DIFF_EXPR_OK): Define to enable PC relative diff
    relocs.

gas/testsuite/ChangeLog

2015-07-06  Pitchumani Sivanupandi <pitchumani.s@atmel.com>

    * gas/avr/pc-relative-reloc.d: New test for 32 bit pc relative reloc.
    * gas/avr/per-function-debugline.s: New test source.

include/ChangeLog

2015-07-06  Pitchumani Sivanupandi <pitchumani.s@atmel.com>

    * elf/avr.h: Add new 32 bit PC relative relocation.

ld/testsuite/ChangeLog

2015-07-06  Pitchumani Sivanupandi <pitchumani.s@atmel.com>

    * ld-avr/gc-section-debugline.d: New test.
    * ld-avr/per-function-debugline.s: Source for new test.

9 years agogdb/doc: Update 'frame' command documentation.
Andrew Burgess [Wed, 8 Jul 2015 14:02:32 +0000 (15:02 +0100)]
gdb/doc: Update 'frame' command documentation.

The documentation for the 'frame' command has gotten a little out of
date, it still mentions architecturally specific details that are no
longer relevant.

This commit removes the old details that no longer apply, and tries to
expand the existing text a little to make the usage clearer for some
cases.

gdb/doc/ChangeLog:

* gdb.texinfo (Selection): Update documentation for 'frame'
command.

9 years agoRevert "Delete program spaces directly when removing inferiors"
Simon Marchi [Wed, 8 Jul 2015 15:40:35 +0000 (11:40 -0400)]
Revert "Delete program spaces directly when removing inferiors"

Reverted, since it causes a build failure.  It turns out that
delete_inferior_silent wasn't actually unused.

This reverts commit 0560c645c02eba2828a053039dcfdf676cdd1d00.

9 years agoDelete program spaces directly when removing inferiors
Simon Marchi [Mon, 29 Sep 2014 20:33:10 +0000 (16:33 -0400)]
Delete program spaces directly when removing inferiors

When deleting an inferior, delete the associated program space as well
if it becomes unused. This replaces the "pruning" approach, with which
you could forget to call prune_program_spaces (as seen, with the
-remove-inferior command, see [1]).

This allows to remove the prune_program_spaces function. At the same
time, I was able to clean up the delete_inferior* family.
delete_inferior_silent and delete_inferior were unused, which allowed
renaming delete_inferior_1 to delete_inferior. Also, since all calls to
it were with silent=1, I removed that parameter completely.

I renamed pspace_empty_p to program_space_empty_p. I prefer if the
"exported" functions have a more explicit and standard name.

Tested on Ubuntu 14.10.

This obsoletes my previous patch "Add call to prune_program_spaces in
mi_cmd_remove_inferior" [1].

[1] https://sourceware.org/ml/gdb-patches/2014-09/msg00717.html

gdb/Changelog:

* inferior.c (delete_inferior_1): Rename to ...
(delete_inferior): ..., remove 'silent' parameter, delete
program space when unused and remove call to prune_program_spaces.
Remove the old, unused, delete_inferior.
(delete_inferior_silent): Remove.
(prune_inferiors): Change call from delete_inferior_1 to
delete_inferior and remove 'silent' parameter. Remove call to
prune_program_spaces.
(remove_inferior_command): Idem.
* inferior.h (delete_inferior_1): Rename to...
(delete_inferior): ..., remove 'silent' parameter and remove the
original delete_inferior.
(delete_inferior_silent): Remove.
* mi/mi-main.c (mi_cmd_remove_inferior): Change call from
delete_inferior_1 to delete_inferior and remove 'silent'
parameter.
* progspace.c (prune_program_spaces): Remove.
(pspace_empty_p): Rename to...
(program_space_empty_p): ... and make non-static.
(delete_program_space): New.
* progspace.h (prune_program_spaces): Remove declaration.
(program_space_empty_p): New declaration.
(delete_program_space): New declaration.

9 years agocompile: Warn for old GCC on cv-qualified self-reference
Jan Kratochvil [Wed, 8 Jul 2015 12:42:19 +0000 (14:42 +0200)]
compile: Warn for old GCC on cv-qualified self-reference

GDB could:

compile code struct_object.selffield = &struct_object
./compile/compile-c-types.c:83: internal-error: insert_type: Assertion `add == NULL || add->gcc_type == gcc_type' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) FAIL: gdb.compile/compile.exp: compile code struct_object.selffield = &struct_object (GDB internal
error)

The bug was not in GDB but in the GCC part interfacing with GDB.

Alexandre Oliva has fixed it the right way:
https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;a=commitdiff;h=072dfdba0ea62abb65514cb3a90cdf3868efe286
git://gcc.gnu.org/git/gcc.git
aoliva/libcp1

Attaching this GDB testsuite update + info to user s/he should upgrade GCC.
After Alex upstreams the fix I can update the message to contain the specific
GCC release.

gdb/ChangeLog
2015-07-08  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR compile/18484
* compile/compile-c-types.c (insert_type): Change gdb_assert to error.

gdb/testsuite/ChangeLog
2015-07-08  Jan Kratochvil  <jan.kratochvil@redhat.com>

PR compile/18484
* gdb.compile/compile.c (struct struct_type): Add volatile to
selffield's type.
* gdb.compile/compile.exp
(compile code struct_object.selffield = &struct_object): Skip further
struct_object tests if this one xfails.

9 years agoPR18617 - Incorrect expression bytecode generated for narrowing conversions
Robert O'Callahan [Wed, 8 Jul 2015 09:53:36 +0000 (10:53 +0100)]
PR18617 - Incorrect expression bytecode generated for narrowing conversions

The existing code preserves 'from' bits, which is incorrect.  E.g.

 (gdb) maint agent-eval (char)255L
 Scope: 0x4008d6
 Reg mask: 00
   0  const16 255
   3  ext 64
   5  end

'ext 64' should be 'ext 8'; this bytecode evaluates to 255 instead of
the correct result of -1.  The fix is simple.  I ran the entire test
suite on x86-64 and there were no new test failures.

gdb/ChangeLog:
2015-07-08  Robert O'Callahan  <robert@ocallahan.org>

PR exp/18617
* ax-gdb.c (gen_conversion): Extend to 'to' bits, not 'from'.

gdb/testsuite/ChangeLog:
2015-07-08  Robert O'Callahan  <robert@ocallahan.org>

PR exp/18617
* gdb.trace/ax.exp: Add test.

9 years agoadd myself to write-after-approval list
Markus Metzger [Mon, 6 Jul 2015 13:40:45 +0000 (15:40 +0200)]
add myself to write-after-approval list

gdb/
* MAINTAINERS (Write After Approval): Add Markus T. Metzger.

9 years agoari: fix strerror() use
Markus Metzger [Mon, 6 Jul 2015 13:21:37 +0000 (15:21 +0200)]
ari: fix strerror() use

Do not use strerror(), instead use safe_strerror().

gdb/
* nat/linux-btrace.c (kernel_supports_bts, kernel_supports_pt):
Use safe_strerror() instead of strerror().

9 years agoAutomatic date update in version.in
GDB Administrator [Wed, 8 Jul 2015 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoMake gold aarch64 accept long form of mapping symbols.
Han Shen [Mon, 6 Jul 2015 22:17:25 +0000 (15:17 -0700)]
Make gold aarch64 accept long form of mapping symbols.

2015-07-07  Han Shen  <shenhan@google.com>

gold/ChangeLog:
2015-07-06 Han Shen  <shenhan@google.com>
* aarch64.cc (AArch64_relobj::do_count_local_symbols): Make legal
of mapping symbols.

9 years agoNew proc is_aarch64_target
Yao Qi [Tue, 7 Jul 2015 15:58:20 +0000 (16:58 +0100)]
New proc is_aarch64_target

Some tests expect the the target is aarch64, but checking target
triplet is not accurate, because target triplet can be aarch64 but
the program is in arm (or aarch32) state.

This patch addes a new proc is_aarch64_target which returns true
if the target is on aarch64 state.

gdb/testsuite:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* gdb.arch/aarch64-atomic-inst.exp: Check is_aarch64_target
instead of istarget "aarch64*-*-*".
* gdb.arch/aarch64-fp.exp: Likewise.
* gdb.base/float.exp: Likewise.
* gdb.reverse/aarch64.exp: Likewise.
* lib/gdb.exp (is_aarch64_target): New proc.

9 years agoNew proc is_aarch32_target
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
New proc is_aarch32_target

GDB tests running on arm target should be also run on aarch32
(32-bit mode on aarch64).  There should be no difference.  It is not
precise to check target triplet to decide which tests should be run,
because if I compiler all the test binary in 32-bit (arm program),
but target triplet is still aarch64, so that these arm specific tests
are skipped.

This patch is to add a new proc is_aarch32_target which return true
if target triplet is arm or the test binary is compiled for arm.

gdb/testsuite:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* lib/gdb.exp (is_aarch32_target): New proc.
* gdb.arch/arm-bl-branch-dest.exp: Check is_aarch32_target
instead of "istarget "arm*-*-*"".
* gdb.arch/arm-disp-step.exp: Likewise.
* gdb.arch/thumb-bx-pc.exp: Likewise.
* gdb.arch/thumb-prologue.exp: Likewise.
* gdb.arch/thumb-singlestep.exp: Likewise.
* gdb.base/disp-step-syscall.exp: Likewise.
* gdb.base/float.exp: Likewise.

9 years agoEnable multi-arch test in catch-syscall.exp on aarch64
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
Enable multi-arch test in catch-syscall.exp on aarch64

This patch is to enable test_catch_syscall_multi_arch on aarch64.

gdb/testsuite:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* gdb.base/catch-syscall.exp (test_catch_syscall_multi_arch):
Set arch1, arch2, syscall1_name, syscall2_name and syscall_number.

9 years agoAdjust gdb.multi tests for aarch64
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
Adjust gdb.multi tests for aarch64

Multi-arch related tests under gdb.multi are to compile programs with
the same compiler but different compiler options (-m64 vs -m32).  However,
different compilers are needed to compile both aarch64 program and
arm (aarch32) program.  This patch is to adjust these test cases to
compile programs in different modes with different compiler.

When we use gcc for arm-linux target, its file name can be different,
arm-linux-gnueabihf-gcc, arm-linux-gnueabi-gcc, or arm-none-linux-gnueabi-gcc,
so I add a variable ARM_CC_FOR_TARGET, so that user can set the name
of gcc for arm-linux target on aarch64, like:

 $ make check RUNTESTFLAGS='ARM_CC_FOR_TARGET=arm-linux-gnueabihf-gcc multi-arch.exp'

gdb/testsuite:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* gdb.multi/multi-arch-exec.exp: Set march1 and march2 to "" if target
is aarch64.  If target is aarch64, set compiler=${ARM_CC_FOR_TARGET}
if it exists.
* gdb.multi/multi-arch.exp: Likewise.

9 years agoSet architecture to arm in arm-*.xml files
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
Set architecture to arm in arm-*.xml files

This patch is to add the following line to various arm target description
xml files,

<architecture>arm</architecture>

in order to fix problems I've seen on aarch64 multi-arch debugging,

 detach^M
 Detaching from program: build-gdb/gdb/testsuite/gdb.base/attach, process 17145^M
 (gdb) PASS: gdb.base/attach.exp: attach1 detach file^M
 No executable file now.^M
 Architecture of file not recognized.^M
 (gdb) FAIL: gdb.base/attach.exp: attach1, purging symbols after detach

Without this patch, struct target_desc *tdesc_* are not initialised
properly, that is, fields arch and osabi in 'struct target_desc' are
not set properly.  This doesn't cause any problems on single arch
debugging, because arch-utils.c:gdbarch_info_fill will guess correctly.
However, in multi-arch debugging, gdbarch_info_fill gets the aarch64
arch, but the target description is for arm (because the current
inferior is 32-bit arm).

It is a surprise to me we didn't set architecture to "arm" before in *.xml
files, and I didn't find out why didn't do so.  AFAICS,
gdb/features/arm-with-iwmmxt.xml was added firstly (in patch
https://sourceware.org/ml/gdb-patches/2007-01/msg00593.html)
which had <architecture>iwmmxt</architecture>, however, afterwards,
architecture isn't set anymore in features/arm-*.xml files (in patches
https://sourceware.org/ml/gdb-patches/2009-07/msg00689.html and
https://sourceware.org/ml/gdb-patches/2010-08/msg00225.html).

gdb:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* features/arm-with-m-fpa-layout.xml: Set architecture to arm.
* features/arm-with-m-fpa-layout.c: Regenerated.
* features/arm-with-m-vfp-d16.xml: Likewise.
* features/arm-with-m-vfp-d16.c: Regenerated.
* features/arm-with-m.xml: Likewise.
* features/arm-with-m.c: Regenerated.
* features/arm-with-neon.xml: Likewise.
* features/arm-with-neon.c: Regenerated.
* features/arm-with-vfpv2.xml: Likewise.
* features/arm-with-vfpv2.c: Regenerated.
* features/arm-with-vfpv3.xml: Likewise.
* features/arm-with-vfpv3.c: Regenerated.

9 years agoNative debug arm program by aarch64 GDB
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
Native debug arm program by aarch64 GDB

This patch is to let aarch64 GDB debug 32-bit arm program natively.  In
each function for fetching and storing registers, GDB will check
gdbarch_bfd_arch_info (gdbarch)->bits_per_word, if it is 32, call
the corresponding aarch32 functions in aarch32-linux-nat.c, otherwise
fall back to aarch64 code to fetch and store registers.

aarch64_linux_read_description has to return the right target description,
but we don't have gdbarch available there, so GDB fetches auxv and gets
AT_PHENT, in order to determine whether the target is 32-bit or 64-bit.
I learned this trick from solib-svr4.c.

gdb:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* aarch32-linux-nat.h (VFP_REGS_SIZE): New macro, moved from
arm-linux-nat.c.
* aarch64-linux-nat.c: Include aarch32-linux-nat.h and
elf/external.h.
(fetch_gregs_from_thread): Call aarch32_gp_regcache_supply
if target is 32-bit.
(store_gregs_to_thread): Call aarch32_gp_regcache_collect
if target is 32-bit.
(fetch_fpregs_from_thread): Call aarch32_vfp_regcache_supply
if target is 32-bit.
(store_fpregs_to_thread): Call aarch32_vfp_regcache_collect
if target is 32-bit.
(tdesc_arm_with_vfpv3, tdesc_arm_with_neon): Declare.
(aarch64_linux_read_description): Return the right target
description.
* arm-linux-nat.c (VFP_REGS_SIZE): Moved to aarch32-linux-nat.h.
* config/aarch64/linux.mh (NATDEPFILES): Add aarch32-linux-nat.o.
* configure.tgt (aarch64*-*-linux*): Add arm-tdep.o and
arm-linux-tdep.o

9 years agoNew aarch32-linux-nat.c
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
New aarch32-linux-nat.c

This patch is to move all the code about transferring
regcache <-> byte buffer for arm (aarch32) to a separate file
aarch32-linux-nat.c.  Then, in the following patch, aarch64 code
can use it to do multi-arch debugging.  This is a refactory patch.

gdb:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* aarch32-linux-nat.c: New file.
* aarch32-linux-nat.h: New file.
* arm-linux-nat.c: Include aarch32-linux-nat.h.
(fetch_regs): Move code to aarch32-linux-nat.c.  Call
aarch32_gp_regcache_supply.
(store_regs): Move code to aarch32-linux-nat.c.  Call
aarch32_gp_regcache_collect.
(fetch_vfp_regs): Move code to aarch32-linux-nat.c.  Call
aarch32_vfp_regcache_supply.
(store_vfp_regs): Move code to aarch32-linux-nat.c.  Call
aarch32_vfp_regcache_collect.
* config/arm/linux.mh (NATDEPFILES): Add aarch32-linux-nat.o.

9 years agoRemove {fetch,store}_fpregister and {fetch,store}_register
Yao Qi [Tue, 7 Jul 2015 15:58:19 +0000 (16:58 +0100)]
Remove {fetch,store}_fpregister and {fetch,store}_register

This patch is to remove fetch_fpregister, fech_register,
store_fpregister and store_register, and use fetch_fpregs,
fetch_regs, store_regs, and store_fpregs instead.

gdb:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* arm-linux-nat.c (store_fpregister): Remove.
(store_register): Likewise.
(fetch_fpregister): Likewise.
(fetch_register): Likewise.
(arm_linux_store_inferior_registers): Call store_regs and
store_fpregs instead.
(arm_linux_fetch_inferior_registers): Call fetch_fpregs and
fetch_regs instead.

9 years agoFix test for Python 3
Simon Marchi [Tue, 7 Jul 2015 15:39:06 +0000 (11:39 -0400)]
Fix test for Python 3

gdb/testsuite/ChangeLog:

* gdb.python/py-lazy-string.exp: Add missing parentheses to
print.

9 years ago[arm] Fix regression by Do not skip prologue for asm (.S) files
Yao Qi [Tue, 7 Jul 2015 09:37:28 +0000 (10:37 +0100)]
[arm] Fix regression by Do not skip prologue for asm (.S) files

Patch "Do not skip prologue for asm (.S) files" [1] changes GDB's
behaviour on which test gdb.arch/thumb-singlestep.exp depends, so
it causes the fail below:

 (gdb) si^M
 37              blx     foo^M
 (gdb) FAIL: gdb.arch/thumb-singlestep.exp: step into foo

the test assumes the program will stop at the instruction after "push"
but it doesn't.  The fix to this fail is to do one more single step.

[1] https://sourceware.org/ml/gdb-patches/2015-06/msg00561.html

gdb/testsuite:

2015-07-07  Yao Qi  <yao.qi@linaro.org>

* gdb.arch/thumb-singlestep.exp: Do one more single step.

9 years agoAutomatic date update in version.in
GDB Administrator [Tue, 7 Jul 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoAlign text section ends so that stub tables appears at expected addresses.
Doug Kwan [Mon, 6 Jul 2015 22:02:40 +0000 (15:02 -0700)]
Align text section ends so that stub tables appears at expected addresses.

9 years agogdb: Rewrite documentation for layout and focus commands.
Andrew Burgess [Mon, 6 Jul 2015 10:33:31 +0000 (11:33 +0100)]
gdb: Rewrite documentation for layout and focus commands.

Changes the documentation for the layout and focus commands.

Instead of documenting each layout (or focus) sub-command as a separate
command, document a single layout (and focus) command which takes a
parameter, then list the possible parameters in a table nested under
each command.

The documentation for the layout command has been extended little to
make it clearer which windows are shown in each layout.

gdb/ChangeLog:

* doc/gdb.texinfo (TUI): Restructure documentation on TUI layout
and focus commands.

9 years agoUpdate NEWS post GDB 7.10 branch creation.
Joel Brobecker [Mon, 6 Jul 2015 20:02:47 +0000 (13:02 -0700)]
Update NEWS post GDB 7.10 branch creation.

gdb/ChangeLog:

* NEWS: Create a new section for the next release branch.
Rename the section of the current branch, now that it has
been cut.

9 years agoBump version to 7.10.50.DATE-cvs.
Joel Brobecker [Mon, 6 Jul 2015 19:58:29 +0000 (12:58 -0700)]
Bump version to 7.10.50.DATE-cvs.

Now that the GDB 7.10 branch has been created, we can
bump the version number.

gdb/ChangeLog:

GDB 7.10 branch created (66c4b3e8a628a207bc6aafef6af0c4128195f56e):
* version.in: Bump version to 7.10.50.DATE-cvs.

9 years agoFix problems with finishing a dummy function call on simulators.
Luis Machado [Mon, 6 Jul 2015 19:09:21 +0000 (16:09 -0300)]
Fix problems with finishing a dummy function call on simulators.

This fixes regressions introduced with the original change to not
consider permanent breakpoints always inserted:

  6ae8866180bf90e9ec76c2dd34c07fd826d11a83 is the first bad commit
  commit 6ae8866180bf90e9ec76c2dd34c07fd826d11a83
  Author: Luis Machado <lgustavo@codesourcery.com>
  Date:   Wed Jun 17 16:50:57 2015 -0300

      Fix problems with finishing a dummy function call on simulators.

Some checks were mistakenly left out of the original patch, which
caused the following failures:

-PASS: gdb.base/shlib-call.exp: print mainshr1(1)
-PASS: gdb.base/shlib-call.exp: step into mainshr1
+FAIL: gdb.base/shlib-call.exp: print mainshr1(1)
+FAIL: gdb.base/shlib-call.exp: step into mainshr1

-PASS: gdb.cp/chained-calls.exp: q(p())
+FAIL: gdb.cp/chained-calls.exp: q(p())

-PASS: gdb.cp/chained-calls.exp: q(p() + r())
+FAIL: gdb.cp/chained-calls.exp: q(p() + r())

-PASS: gdb.cp/chained-calls.exp: g(f(g(f() + f())) + f())
+FAIL: gdb.cp/chained-calls.exp: g(f(g(f() + f())) + f())

-PASS: gdb.cp/chained-calls.exp: *c
-PASS: gdb.cp/chained-calls.exp: *c + *c
-PASS: gdb.cp/chained-calls.exp: q(*c + *c)
+FAIL: gdb.cp/chained-calls.exp: *c
+FAIL: gdb.cp/chained-calls.exp: *c + *c
+FAIL: gdb.cp/chained-calls.exp: q(*c + *c)

-PASS: gdb.cp/classes.exp: calling method for small class
+FAIL: gdb.cp/classes.exp: calling method for small class

The above is likely caused by GDB not removing the permanent
breakpoints from the target, leading to the inferior executing
the breakpoint instruction and tripping on a SIGSEGV.

gdb/ChangeLog:
2015-07-06  Luis Machado  <lgustavo@codesourcery.com>

* breakpoint.c (remove_breakpoint_1): Don't handle permanent
breakpoints in a special way.
(remove_breakpoint): Likewise.
(mark_breakpoints_out): Likewise.

9 years agoRemove the merge conflict introduced by
H.J. Lu [Mon, 6 Jul 2015 19:00:45 +0000 (12:00 -0700)]
Remove the merge conflict introduced by

commit 2f0c68f23bb3132cd5ac466ca8775c0d9e4960cd
Author: Catherine Moore <clm@codesourcery.com>
Date:   Thu May 28 14:50:36 2015 -0700

    Compact EH Support

9 years agogdb/doc: Fix incorrect use of @xref.
Andrew Burgess [Mon, 6 Jul 2015 10:37:24 +0000 (11:37 +0100)]
gdb/doc: Fix incorrect use of @xref.

All uses of @xref must be followed by either '.' or ','.  In commit
a4ea0946c an incorrect use of @xref was introduced.  This commit
adds a comma after the use of @xref.

gdb/ChangeLog:

* doc/gdb.texinfo (TUI): Add comma after @xref.

9 years agogdb/tui: Don't cast between window types.
Andrew Burgess [Mon, 6 Jul 2015 15:56:42 +0000 (16:56 +0100)]
gdb/tui: Don't cast between window types.

Instead of casting between structure types to get the 'tui_gen_win_info'
info from a 'tui_win_info' access the generic member variable.  This is
inline with what is done throughout the rest of the tui code.

gdb/ChangeLog:

* tui/tui-win.c (tui_set_focus): Use structure member 'generic'
instead of casting the structure type.

9 years agosearch_struct_field: remove OFFSET parameter
Simon Marchi [Mon, 6 Jul 2015 17:10:56 +0000 (13:10 -0400)]
search_struct_field: remove OFFSET parameter

I was trying to understand what the OFFSET parameter was for, and
realized it was set to 0 in every call to search_struct_field.  I
assume that it was used at some point, but some subsequent changes
made it useless.

gdb/ChangeLog:

* valops.c (search_struct_field): Remove OFFSET parameter.
(value_cast_structs): Adjust calls to search_struct_field.
(value_struct_elt): Same.
(find_overload_match): Same.

9 years agoCleanup value_fetch_lazy's comment and return value
Simon Marchi [Mon, 6 Jul 2015 17:04:11 +0000 (13:04 -0400)]
Cleanup value_fetch_lazy's comment and return value

The comment for value_fetch_lazy seems outdated. It says that it's only
called from the value_contents and value_contents_all (macros!), which
is not true.  Also, the return value seems useless now, despite what the
comment says.

gdb/ChangeLog:

* value.c (value_fetch_lazy): Update comment, change return
value to void.
* value.h (value_fetch_lazy): Change return value to void.

9 years agogdb: tui_win_name: Make parameter and result const.
Andrew Burgess [Mon, 6 Jul 2015 14:32:11 +0000 (15:32 +0100)]
gdb: tui_win_name: Make parameter and result const.

This commit makes the parameter and the result for 'tui_win_name'
constant.  There's one place in the code that is then updated as a
result of this change.

gdb/ChangeLog:

* tui/tui-data.c (tui_partial_win_by_name): Window name is const.
(tui_win_name): Make parameter and result const.
* tui/tui-data.h (tui_win_name): Make parameter and result const.

9 years agoDon't throw an error in "show mpx bound" implementation
Patrick Palka [Thu, 2 Jul 2015 15:55:01 +0000 (11:55 -0400)]
Don't throw an error in "show mpx bound" implementation

"show" functions should not throw an exception in part because it causes
the output of the commands "info set" and "show" to get truncated.

This fixes the following fails:

    FAIL: gdb.base/default.exp: info set
    FAIL: gdb.base/default.exp: show

gdb/ChangeLog:

* i386-tdep.c (i386_mpx_info_bounds): Don't call error, instead
use printf_unfiltered.
(set_mpx_cmd): Add missing trailing space to command string
literal.
(_initialize_i386_tdep): Give the "mpx" prefix command its
correct name.

9 years agoAutomatic date update in version.in
GDB Administrator [Mon, 6 Jul 2015 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoFix an opd->append index in elf64-ppc.c
Richard Sandiford [Sun, 5 Jul 2015 18:56:59 +0000 (19:56 +0100)]
Fix an opd->append index in elf64-ppc.c

bfd/
* elf64-ppc.c (toc_adjusting_stub_needed): Use the symbol value
plus addend rather than the original st_value when looking up
entries in opd->adjust.

ld/testsuite/
* ld-powerpc/tocopt6-inc.s, ld-powerpc/tocopt6a.s,
ld-powerpc/tocopt6b.s, ld-powerpc/tocopt6c.s,
ld-powerpc/tocopt6.d: New test.
* ld-powerpc/powerpc.exp (ppc64elftests): Add it.

9 years agoAutomatic date update in version.in
GDB Administrator [Sun, 5 Jul 2015 00:00:07 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoAutomatic date update in version.in
GDB Administrator [Sat, 4 Jul 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agoAdd experimental support for --gc-sections with COFF and PE based targets.
Kai Tietz [Fri, 3 Jul 2015 14:50:29 +0000 (15:50 +0100)]
Add experimental support for --gc-sections with COFF and PE based targets.

PR ld/11539
bfd * coffcode.h (coff_bfd_gc_sections): Define default
to bfd_coff_gc_sections function.
* cofflink.c (init_reloc_cookie): Copy and adjust coff
related code about gc-sections from elflink.c to here.
(fini_reloc_cookie): Likewise.
(init_reloc_cookie_rels): Likewise.
(fini_reloc_cookie_rels): Likewise.
(init_reloc_cookie_for_section): Likewise.
(fini_reloc_cookie_for_section): Likewise.
(_bfd_coff_gc_mark_hook): Likewise.
(_bfd_coff_gc_mark_rsec): Likewise.
(_bfd_coff_gc_mark_reloc): Likewise.
(_bfd_coff_gc_mark): Likewise.
(_bfd_coff_gc_mark_extra_sections): Likewise.
(coff_gc_sweep_symbol_info): Likewise.
(coff_gc_sweep_symbol): Likewise.
(gc_sweep_hook_fn): Likewise.
(coff_gc_sweep): Likewise.
(bfd_coff_gc_sections): Likewise.
(_bfd_coff_gc_keep): Likewise.
* libcoff.h (coff_reloc_cookie): New struct.
(bfd_coff_gc_sections): New prototype.
(coff_gc_mark_hook_fn): New type.

ld * scripttempl/pep.sc: Mark .idata*, .CRT*, .tls*,
.rsrc*, .init, .ctor*, .dtor*, .fini, .jcr,
.eh_frame, .pdata. .xdata, and .gcc_except_table sections
as KEEP.
* scripttempl/pe.sc: Likewise.

9 years agoCorrect ld Makefile dependendcy for or1k
Alan Modra [Fri, 3 Jul 2015 01:28:22 +0000 (10:58 +0930)]
Correct ld Makefile dependendcy for or1k

* Makefile.am (eelf32or1k.c, eelf32or1k_linux.c): Depend on ELF_DEPS.
* Makefile.in: Regenerate.

9 years agoRemove ppc860, ppc750cl, ppc7450 insns from common ppc.
Alan Modra [Fri, 3 Jul 2015 01:26:26 +0000 (10:56 +0930)]
Remove ppc860, ppc750cl, ppc7450 insns from common ppc.

Back in the day support for these processors was added, we probably
didn't want to waste PPC_OPCODE bits on minor variations.  I've had a
complaint that disassembly of mfspr/mtspr was wrong for power8.  This
patch fixes that problem.

Note that since -m860/-m850/-m821 are new gas options enabling the
mpc8xx specific mfspr/mtspr variants it is possible that this change
will break some mpc8xx assembly code.  ie. you might need to modify
makefiles to pass -m860 to gas.

include/opcode/
* ppc.h (PPC_OPCODE_750, PPC_OPCODE_7450, PPC_OPCODE_860): Define.
opcodes/
* ppc-opc.c (PPC750, PPC7450, PPC860): Define using PPC_OPCODE_*.
* ppc-dis.c (ppc_opts): Add 821, 850 and 860 entries.  Add
PPC_OPCODE_7450 to 7450 entry.  Add PPC_OPCODE_750 to 750cl entry.
gas/
* config/tc-ppc.c (md_show_usage): Add -m821, -m850, -m860.
* doc/c-ppc.texi (PowerPC-Opts): Likewise.
gas/testsuite/
* gas/ppc/titan.d: Correct mfmcsrr0 disassembly.

9 years agoAdd support for backtracing through Renesas RX exception frames.
Kevin Buettner [Thu, 2 Jul 2015 23:46:31 +0000 (16:46 -0700)]
Add support for backtracing through Renesas RX exception frames.

This change adds support for backtracing through Renesas RX exception
frames.

Determination about the type of frame is made by scanning the
remainder of the function for a return instruction and then looking at
which, if any, return instruction is found.  A normal RTS instruction
indicates that the frame is a normal frame.  An RTFI instruction
indicates that it's a fast interrupt, and an RTE instruction indicates
that the frame is a (normal) exception frame.  If no return instruction
is found within the scanned region - which can happen when the end of
the function cannot be found - it is assumed to be a normal frame.

I was able to test that normal prologue scanning still works by
disabling the dwarf2 sniffer.  I've tested this code for normal
interrupts.  The fast interrupt case has not been tested.

gdb/ChangeLog:

* rx-tdep.c (RX_USP_REGNUM, RX_BPC_REGNUM): New constants.
(enum rx_frame_type): New.
(struct rx_prologue): Add new field `frame_type'.
(rx_analyze_prologue): Add `frame_type' parameter. Cache this
parameter in the prologue struct.  Add code for recording
locations of PC and PSW for fast interrupt and exception frames.
(rx_skip_prologue): Adjust call to rx_analyze_prologue.
(rx_analyze_frame_prologue): Add `frame_type' parameter.
(rx_frame_type): New function.
(rx_frame_base): Fetch frame type and pass it to rx_analyze_prologue.
(rx_frame_this_id): Rename parameter `this_prologue_cache' to
`this_cache'.
(rx_frame_prev_register): Rename parameter `this_prologue_cache' to
`this_cache'.  Add cases for RX_FRAME_TYPE_EXCEPTION and
RX_FRAME_TYPE_FAST_INTERRUPT.
(normal_frame_p, exception_frame_p, rx_frame_sniffer_common)
(rx_frame_sniffer, rx_exception_sniffer): New functions.
(rx_frame_unwind): Use rx_frame_sniffer instead of
default_frame_sniffer.
(rx_frame_unwind): New unwinder.
(rx_gdbarch_init): Register new unwinder.

9 years agoAutomatic date update in version.in
GDB Administrator [Fri, 3 Jul 2015 00:00:08 +0000 (00:00 +0000)]
Automatic date update in version.in

9 years agorx: Create and use flags types for psw, bpsw, and fpsw registers.
Kevin Buettner [Thu, 2 Jul 2015 22:02:56 +0000 (15:02 -0700)]
rx: Create and use flags types for psw, bpsw, and fpsw registers.

This change adds two flags types for the (Renesas RX) psw, bpsw, and
fpsw registers.  As a result, symbolic flags are displayed for these
registers in the output of GDB's "info registers" command as well as
in output from other commands, such as "print".

gdb/ChangeLog:

* rx-tdep.c (RX_BPSW_REGNUM, RX_FPSW_REGNUM): New constants.
(struct gdbarch_tdep): Add fields rx_psw_type and rx_fpsw_type.
(rx_register_type): Add cases for RX_PSW_REGNUM, RX_BPSW_REGNUM,
and RX_FPSW_REGNUM.
(rx_gdbarch_init): Initialize PSW, BPSW, and FPSW flags types.

9 years agoFix GCC false warning
Jan Kratochvil [Thu, 2 Jul 2015 20:39:57 +0000 (22:39 +0200)]
Fix GCC false warning

At least on
gcc-4.4.7-11.el6.i686
./configure --enable-64-bit-bfd --enable-targets=all
GDB does not build due to:
cc1: warnings being treated as errors
s390-linux-tdep.c: In function ‘s390_handle_arg’:
s390-linux-tdep.c:2575: error: ‘val’ may be used uninitialized in this function

gdb/ChangeLog
2015-07-02  Jan Kratochvil  <jan.kratochvil@redhat.com>

Fix GCC false warning.
* s390-linux-tdep.c (s390_handle_arg): Initialize VAL.

9 years agoFix typo in aarch64_linux_insert_hw_breakpoint
Yao Qi [Thu, 2 Jul 2015 13:03:54 +0000 (14:03 +0100)]
Fix typo in aarch64_linux_insert_hw_breakpoint

It should be "insert_hw_breakpoint" rather than "insert_hw_watchpoint".

gdb:

2015-07-02  Yao Qi  <yao.qi@linaro.org>

* aarch64-linux-nat.c (aarch64_linux_insert_hw_breakpoint): Fix
typo in the debugging message.

9 years agoFix snafu with latest addition to the ARM sim.
Nick Clifton [Thu, 2 Jul 2015 15:19:09 +0000 (16:19 +0100)]
Fix snafu with latest addition to the ARM sim.

* Makefile.in (SIM_EXTRA_CFLAGS): Revert previous delta.
(SIM_EXTRA_LIBS): Add -lm.

9 years agobtrace: maintenance commands
Markus Metzger [Mon, 3 Feb 2014 13:35:28 +0000 (14:35 +0100)]
btrace: maintenance commands

Add maintenance commands that help debugging the btrace record target.
The following new commands are added:

maint info btrace
  Print information about branch tracing internals.

maint btrace packet-history
  Print the raw branch tracing data.

maint btrace clear-packet-history
  Discard the stored raw branch tracing data.

maint btrace clear
  Discard all branch tracing data.  It will be fetched and processed
  anew by the next "record" command.

maint set|show btrace pt skip-pad
  Set and show whether PAD packets are skipped when computing the
  packet history.

gdb/
* btrace.c: Include gdbcmd.h, cli/cli-utils.h, and ctype.h.
(maint_btrace_cmdlist, maint_btrace_set_cmdlist)
(maint_btrace_show_cmdlist, maint_btrace_pt_set_cmdlist)
(maint_btrace_pt_show_cmdlist, maint_btrace_pt_skip_pad)
(btrace_maint_clear): New.
(btrace_fetch, btrace_clear): Call btrace_maint_clear.
(pt_print_packet, btrace_maint_decode_pt)
(btrace_maint_update_pt_packets, btrace_maint_update_packets)
(btrace_maint_print_packets, get_uint, get_context_size, no_chunk)
(maint_btrace_packet_history_cmd)
(maint_btrace_clear_packet_history_cmd, maint_btrace_clear_cmd)
(maint_btrace_cmd, maint_btrace_set_cmd, maint_btrace_show_cmd)
(maint_btrace_pt_set_cmd, maint_btrace_pt_show_cmd)
(maint_info_btrace_cmd, _initialize_btrace): New.
* btrace.h (btrace_pt_packet, btrace_pt_packet_s)
(btrace_maint_packet_history, btrace_maint_info): New.
(btrace_thread_info) <maint>: New.
* NEWS: Announce it.

doc/
* gdb.texinfo (Maintenance Commands): Document "maint btrace"
commands.

9 years agobtrace: store raw btrace data
Markus Metzger [Mon, 3 Feb 2014 10:40:50 +0000 (11:40 +0100)]
btrace: store raw btrace data

Store the raw branch trace data that has been read from the target.

This data can be used for maintenance commands as well as for generating
a core file for the "record save" command.

gdb/
* btrace.c (btrace_fetch): Append the new trace data.
(btrace_clear): Clear the stored trace data.
* btrace.h (btrace_thread_info) <data>: New.
* common/btrace-common.h (btrace_data_clear)
(btrace_data_append): New.
* common/btrace-common.c (btrace_data_clear)
(btrace_data_append): New.

9 years agobtrace, linux: use data_size and data_offset
Markus Metzger [Mon, 17 Nov 2014 10:18:05 +0000 (11:18 +0100)]
btrace, linux: use data_size and data_offset

In struct perf_event_mmap_page there are new fields data_size and data_offset
that give the location of the perf_event data buffer relative to the mmap
page.  Use them if they are present.

gdb/
* nat/linux-btrace.c (linux_enable_bts): Check for
PERF_ATTR_SIZE_VER5.
Check for data_offset and data_size fields.  Use them.