It appears that Tom approved the patch, but Jan had some issues
with a compiler error that made the test fail on -m32 test runs.
He wrote up a hand-tweaked .S file to deal with it. Siddesh said
he would update tests. Then nothing.
Siddesh and Jan have both moved on since.
The patch originally required a large precursor patch to work.
I have whittled this down to/rewritten the bare minimum, and this
first patch is the result, changing the type of TYPE_LENGTH
to ULONGEST from unsigned int.
The majority of the changes involve changing printf format
strings to use %s and pulongest instead of %d.
gdb/ChangeLog:
* ada-lang.c (ada_template_to_fixed_record_type_1): Use
%s/pulongest for TYPE_LENGTH instead of %d in format
strings.
* ada-typerint.c (ada_print_type): Likewise.
* amd64-windows-tdep.c (amd64_windows_store_arg_in_reg): Likewise.
* compile/compile-c-support.c (generate_register_struct): Likewise.
* gdbtypes.c (recursive_dump_type): Likewise.
* gdbtypes.h (struct type) <length>: Change type to ULONGEST.
* m2-typeprint.c (m2_array): Use %s/pulongest for TYPE_LENGTH
instead of %d in format strings.
* riscv-tdep.c (riscv_type_alignment): Cast second argument
to std::min to ULONGEST.
* symmisc.c (print_symbol): Use %s/pulongest for TYPE_LENGTH
instead of %d in format strings.
* tracepoint.c (info_scope_command): Likewise.
* typeprint.c (print_offset_data::update)
(print_offset_data::finish): Likewise.
* xtensa-tdep.c (xtensa_store_return_value)
(xtensa_push_dummy_call): Likewise.
Max Filippov [Fri, 29 Mar 2019 00:03:57 +0000 (17:03 -0700)]
bfd: xtensa: fix shrink_dynamic_reloc_sections for export-dynamic
shrink_dynamic_reloc_sections must remove PLT entry that was created for
an undefined weak symbol in the presence of --export-dynamic option when
relaxation coalesces literals pointing to that symbol. This fixes the
following assertion:
ld: BFD (GNU Binutils) 2.31.1 internal error, aborting at
elf32-xtensa.c:3292 in elf_xtensa_finish_dynamic_sections
2019-03-29 Max Filippov <jcmvbkbc@gmail.com>
bfd/
* elf32-xtensa.c (shrink_dynamic_reloc_sections): Add
info->export_dynamic to the conditional.
ld/
* testsuite/ld-xtensa/relax-undef-weak-pie-export-dynamic.d: New
test definition.
* testsuite/ld-xtensa/xtensa.exp
(relax-undef-weak-pie-export-dynamic): Add new test.
broke many simulator targets. I fixed aarch64 in a previous commit
without realising how many other target were also broken.
This commit adds the missing includes (sim-assert.h and libiberty.h),
which seem to be needed by many simulator targets, in a central
location, this should fix most builds.
Fix gdb.multi/multi-arch-exec.exp blocking under high load/slow gdb
When running multi-arch-exec.exp under valgrind, the test succeeds
when the machine is not loaded, but blocks when the machine is highly
loaded (e.g. when running the testsuite with valgrind with -j X
where X is one more than the nr of available cores).
The problem is that the hello program dies too early due to the alarm (30).
So, increase the alarm timer.
Note that this does not make the test take longer (it takes about
3.5 seconds on my system). As I understand, the alarm is just there
to avoid hello staying there forever in case of another problem.
2019-03-28 Philippe Waroquiers <philippe.waroquiers@skynet.be>
Fix GDB being suspended SIGTTOU when running gdb.multi/multi-arch-exec.exp
When running under valgrind, multi-arch-exec.exp blocks forever.
Some (painful) investigation shows this is due to valgrind slowing
down GDB, and GDB has to output some messages at a different time,
when GDB does not have the terminal for output.
To reproduce the problem, you need to slow down GDB.
It can be reproduced by:
cd gdb/testsuite/outputs/gdb.multi/multi-arch-exec/
../../../../gdb -ex 'set debug lin-lwp 1' -ex 'break all_started' -ex 'run' ./2-multi-arch-exec
The above stops at a breakpoint. Do continue.
GDB is then suspended because of SIGTTOU.
The stacktrace that leads to the hanging GDB is:
(top-gdb) bt
at ../../binutils-gdb/gdb/exceptions.c:130
....
Alternatively, the same happens when doing
strace -o s.out ../../../../gdb -ex 'break all_started' -ex 'run' ./2-multi-arch-exec
And of course, valgrind is also sufficiently slowing down GDB to
reproduce this :).
Fix this by calling target_terminal::ours_for_output ();
at the beginning of follow_exec.
Note that all this terminal handling is not very clear to me:
* Some code takes the terminal, and then takes care to give it back to the inferior
if the terminal was belonging to the inferior.
(e.g. annotate_breakpoints_invalid).
* some code takes the terminal, but does not give it back
(e.g. update_inserted_breakpoint_locations).
* some code takes it, and unconditionally gives it back
(e.g. handle_jit_event)
* here and there, we also find
gdb::optional<target_terminal::scoped_restore_terminal_state> term_state;
before a (sometimes optional) call to ours_for_output.
And such calls to ours_for_output is sometimes protected by:
if (target_supports_terminal_ours ())
(e.g. exceptions.c: print_flush).
but most of the code calls it without checking if the target supports it.
* some code is outputting some errors, but only takes the terminal
after. E.g. infcmd.c: prepare_one_step
gdb/ChangeLog
2019-03-28 Philippe Waroquiers <philippe.waroquiers@skynet.be>
Sandra Loosemore [Thu, 28 Mar 2019 16:29:22 +0000 (09:29 -0700)]
Fix stepping past unwritable kernel helper on nios2-linux-gnu.
This patch fixes a problem on nios2-linux-gnu with stepping past the
kernel helper __kuser_cmpxchg, which was exposed by the testcase
gdb.threads/watchpoint-fork.exp. The kernel maps this function into
user space on an unwritable page. In this testcase, the cmpxchg
helper is invoked indirectly from the setbuf call in the test program.
Since this target lacks hardware breakpoint/watchpoint support, GDB
tries to single-step through the program by setting software
breakpoints, and was just giving an error when it reached the function
on the unwritable page.
The solution here is to always step over the call instead of stepping
into it; cmpxchg is supposed to be an atomic operation so this
behavior seems reasonable. The hook in nios2_get_next_pc is somewhat
generic, but at present cmpxchg is the only helper provided by the
Linux kernel that is invoked by an ordinary function call. (Signal
return trampolines also go through the unwritable page but not by a
function call.)
Fixing this issue also revealed that the testcase needs a much larger
timeout factor when software single-stepping is used. That has also
been fixed in this patch.
Alan Hayward [Thu, 28 Mar 2019 12:33:29 +0000 (12:33 +0000)]
Testsuite: set sysroot when using gdbserver
When testing using native-gdbserver and native-extended-gdbserver, the sysroot
is not set. This results in a warning from GDB and files are sent via the
remote protocol, which can be slow.
On Ubuntu 18.04 (unlike most distros) the debug versions of the standard
libraries are included by default in /usr/lib/debug/.
These file reads are causing a complete native-gdbserver run on the AArch64
buildbot slave to timeout after 2.5 hours. This is also causing the builds
to back up on the slave.
The solution is to ensure the sysroot is set to / for all local boards.
This drastically reduces the time of a test. For example, gdb.base/sigall.exp
drops from 23 seconds to 4 seconds.
A full native-gdbserver run on the AArch64 slave now takes 8 minutes.
Alan Hayward [Thu, 28 Mar 2019 11:23:25 +0000 (11:23 +0000)]
gdbserver: Ensure AT_HWCAP2 is defined
When using older compilers, AT_HWCAP2 may not be be defined.
It is defined in elf/common.h, however including this in
gdbserver/linux-low.c causes conflicts.
Manually add the define if it does not exist.
gdb/gdbserver/ChangeLog:
* linux-low.c (AT_HWCAP2): Add define if not already included.
Alan Modra [Thu, 28 Mar 2019 00:06:55 +0000 (10:36 +1030)]
PR24390, Don't decode mtfsb field as a cr field
"mtfsb0 4*cr7+lt" doesn't make all that much sense, but unfortunately
glibc uses just that instead of "mtfsb0 28" to clear the fpscr xe bit.
So for backwards compatibility accept cr field expressions when
assembling mtfsb operands, but disassemble to a plain number.
PR 24390
include/
* opcode/ppc.h (PPC_OPERAND_CR_REG): Comment.
opcodes/
* ppc-opc.c (BTF): Define.
(powerpc_opcodes): Use for mtfsb*.
* ppc-dis.c (print_insn_powerpc): Print fields with both
PPC_OPERAND_CR_REG and PPC_OPERAND_CR_BIT as a plain number.
gas/
* testsuite/gas/ppc/476.d: Update mtfsb*.
* testsuite/gas/ppc/a2.d: Likewise.
Stafford Horne [Wed, 27 Mar 2019 21:40:30 +0000 (06:40 +0900)]
sim/common: Fix warnings: "warning: implicit declaration of function..."
During building of several cgen simulator's I notices the below
warnings. Adding includes fixes these.
Including config.h allows stdio.h to properly configure itself to expose
asprintf().
The other warnings for abort, free, memset, strlen are trivial.
Warnings:
../../../binutils-gdb/sim/or1k/../common/sim-watch.c: In function ‘sim_watchpoint_install’:
../../../binutils-gdb/sim/or1k/../common/sim-watch.c:415:10: warning: implicit declaration of function ‘asprintf’; did you mean ‘vasprintf’? [-Wimplicit-function-declaration]
if (asprintf (&name, "watch-%s-%s",
^~~~~~~~
vasprintf
../../../binutils-gdb/sim/lm32/../common/hw-device.c: In function ‘hw_strdup’:
../../../binutils-gdb/sim/lm32/../common/hw-device.c:59:34: warning: implicit declaration of function ‘strlen’ [-Wimplicit-function-declaration]
char *dup = hw_zalloc (me, strlen (str) + 1);
^~~~~~
../../../binutils-gdb/sim/lm32/../common/hw-events.c: In function ‘hw_event_queue_schedule’:
../../../binutils-gdb/sim/lm32/../common/hw-events.c:92:3: warning: implicit declaration of function ‘memset’ [-Wimplicit-function-declaration]
memset (&dummy, 0, sizeof dummy);
^~~~~~
../../../binutils-gdb/sim/lm32/../common/hw-handles.c: In function ‘hw_handle_remove_ihandle’:
../../../binutils-gdb/sim/lm32/../common/hw-handles.c:211:4: warning: implicit declaration of function ‘free’ [-Wimplicit-function-declaration]
free (delete);
^~~~
../../../binutils-gdb/sim/lm32/../common/sim-fpu.c: In function ‘pack_fpu’:
../../../binutils-gdb/sim/lm32/../common/sim-fpu.c:292:7: warning: implicit declaration of function ‘abort’ [-Wimplicit-function-declaration]
abort ();
^~~~~
sim/common/ChangeLog:
* sim-options.c: Include "config.h".
Include <stdio.h>.
* sim-watch.c: Include "config.h".
Include <stdio.h>.
* hw-device.c: Include <string.h>.
* hw-events.c: Include <string.h>.
* hw-handles.c: Include <stdlib.h>.
* sim-fpu.c: Include <stdlib.h>.
/usr/bin/ld: libsim.a(mloop.o): in function `extract':
/home/shorne/work/openrisc/gdb-musl/sim/or1k/mloop.c:82: undefined reference to `sim_addr_range_hit_p'
/usr/bin/ld: /home/shorne/work/openrisc/gdb-musl/sim/or1k/mloop.c:83: undefined reference to `sim_addr_range_hit_p'
collect2: error: ld returned 1 exit status
make[3]: *** [Makefile:305: run] Error 1
sim/common/ChangeLog:
* Make-common.in (sim-arange_h): Remove sim-arange.c
* sim-arange.c: Remove SIM_ARANGE_C.
Add ifdef for _SIM_ARANGE_C_.
Include "sim-arange.h".
Remove include for unused "sim-assert.h".
Remove DEFINE_INLINE_P. Remove DEFINE_NON_INLINE_P.
(sim_addr_range_add): Declare as INLINE_SIM_ARANGE.
(sim_addr_range_delete): Declare as INLINE_SIM_ARANGE.
(sim_addr_range_hit_p): Change from SIM_ARANGE_INLINE to
INLINE_SIM_ARANGE.
* sim-arange.h (sim_addr_range_add): Declare as
INLINE_SIM_ARANGE.
(sim_addr_range_delete): Declare as INLINE_SIM_ARANGE.
(sim_addr_range_hit_p) Declare as INLINE_SIM_ARANGE.
Remove definition of SIM_ARANGE_INLINE.
Remove [HAVE_INLINE].
Wrap include "sim-arange.c" in H_REVEALS_MODULE_P.
* sim-base.h: Include "sim-arange.h"
* sim-basics.h: Remove include of "sim-arange.h"
* sim-inline.c: Include "sim-arange.c"
* sim-inline.h: Define INLINE_SIM_ARANGE.
Define SIM_ARANGE_INLINE. Define EXTERN_SIM_ARANGE_P.
Define STATIC_INLINE_SIM_ARANGE. Define STATIC_SIM_ARANGE.
Fix buffer overflow regression due to minsym malloc-ed instead of obstack-ed.
Valgrind detects the following error in a bunch of tests,
e.g. in gdb.base/foll-fork.exp.
==15155== VALGRIND_GDB_ERROR_BEGIN
==15155== Invalid read of size 8
==15155== at 0x55BE04: minimal_symbol_upper_bound(bound_minimal_symbol) (minsyms.c:1504)
==15155== by 0x3B2E9C: find_pc_partial_function(unsigned long, char const**, unsigned long*, unsigned long*, block const**) (blockframe.c:340)
==15155== by 0x3B3135: find_function_entry_range_from_pc(unsigned long, char const**, unsigned long*, unsigned long*) (blockframe.c:385)
==15155== by 0x4F5597: fill_in_stop_func(gdbarch*, execution_control_state*) [clone .part.16] (infrun.c:4124)
==15155== by 0x4FBE01: fill_in_stop_func (infrun.c:7636)
==15155== by 0x4FBE01: process_event_stop_test(execution_control_state*) (infrun.c:6279)
...
==15155== Address 0x715bec8 is 0 bytes after a block of size 2,952 alloc'd
==15155== at 0x4C2E2B3: realloc (vg_replace_malloc.c:836)
==15155== by 0x405F2C: xrealloc (common-utils.c:62)
==15155== by 0x55BA4E: xresizevec<minimal_symbol> (poison.h:170)
==15155== by 0x55BA4E: minimal_symbol_reader::install() (minsyms.c:1399)
==15155== by 0x4981C7: elf_read_minimal_symbols (elfread.c:1165)
...
This seems to be a regression created by:
commit 042d75e42c5572f333e0e06dabd3c5c4afab486c
Author: Tom Tromey <tom@tromey.com>
AuthorDate: Sat Mar 2 12:29:48 2019 -0700
Commit: Tom Tromey <tom@tromey.com>
CommitDate: Fri Mar 15 16:02:10 2019 -0600
Allocate minimal symbols with malloc
Before this commit, the array of 'struct minimal_symbol'
contained a last element that was a "null symbol". The comment in
minimal_symbol_reader::install was:
/* We also terminate the minimal symbol table with a "null symbol",
which is *not* included in the size of the table. This makes it
easier to find the end of the table when we are handed a pointer
to some symbol in the middle of it. Zero out the fields in the
"null symbol" allocated at the end of the array. Note that the
symbol count does *not* include this null symbol, which is why it
is indexed by mcount and not mcount-1. */
However, minimal_symbol_upper_bound was still based on the assumption
that the array of minsym is terminated by a minsym with a null symbol:
it is looping with:
for (i = 1; MSYMBOL_LINKAGE_NAME (msymbol + i) != NULL; i++)
Replace this NULL comparison by a logic that calculates how
many msymbol are following the msymbols from which we are starting from.
(Re-)tested on debian/amd64, natively and under valgrind.
gdb/ChangeLog
2019-03-24 Philippe Waroquiers <philippe.waroquiers@skynet.be>
Tom Tromey <tromey@adacore.com>
Joel Brobecker [Tue, 26 Mar 2019 22:30:21 +0000 (17:30 -0500)]
gdb-gdb.py.in: Fix error when printing range type
I noticed that trying to print the contents of a struct main_type
would fail when the type was a TYPE_CODE_RANGE:
(gdb) p *type.main_type
$1 = Python Exception <class 'gdb.error'> There is no member named low_undefined.:
And indeed, Python is right, fields "low_undefined" has been removed
from struct range_bounds back in ... 2014! It was done when we introduced
dynamic bounds handling. This patch fixes gdb-gdb.py.in according to
the new structure.
gdb/ChangeLog:
* gdb-gdb.py.in (StructMainTypePrettyPrinter.bound_img): New method.
(StructMainTypePrettyPrinter.bounds_img): Use new "bound_img"
method to compute the bounds of range types. Also print "[evaluated]"
if the bounds' values come from a dynamic evaluation.
Andrew Burgess [Thu, 21 Mar 2019 16:29:14 +0000 (16:29 +0000)]
gdb: Make python display_hint None handling defined behaviour
The documentation say that the display_hint method must return a
string to serve as a display hint, and then goes on to list some
acceptable strings.
However, if we don't supply the display_hint method then we get a
default display style behaviour and there's currently no way (in the
python api) to force this default behaviour.
The guile api allows #f to be used in order to force the default
display style behaviour, and this is documented.
Currently, using None in the python api also forces the default
display behaviour.
This commit extends the documentation to make returning None from the
display_hint method an official mechanism by which the user can get
the default display style.
I've extended one of the existing tests to cover this case.
gdb/doc/ChangeLog:
* python.texi (Pretty Printing API): Document use of None for the
display_hint.
gdb/testsuite/ChangeLog:
* gdb.python/py-prettyprint.c (struct container) <is_map_p>: New
field.
(make_container): Initialise new field.
* gdb.python/py-prettyprint.exp: Add new tests.
* gdb.python/py-prettyprint.py (class ContainerPrinter)
<display_hint>: New method.
Andrew Burgess [Sat, 23 Mar 2019 22:49:09 +0000 (22:49 +0000)]
gdb/testsuite: Make test names unique in gdb.python/py-prettyprint.exp
This makes the test names unique in gdb.python/py-prettyprint.exp, it
also switches to use gdb_breakpoint and gdb_continue_to_breakpoint
more so that we avoid test names with the source line number in - this
is bad if the test source ever changes as the test names will then
change.
One final change is to switch from using gdb_py_test_silent_cmd to use
gdb_test_no_output, the former should be used for running python
commands and can catch any thrown exception. However, in this case
the command being run is not a python command, its just a normal GDB
CLI command that produces no output, so lets use the appropriate
wrapper function.
gdb/testsuite/ChangeLog:
* gdb.python/py-prettyprint.exp: Use gdb_breakpoint and
gdb_continue_to_breakpoint more throughout this test.
(run_lang_tests) Supply unique test names, and use
gdb_test_no_output.
Andrew Burgess [Thu, 21 Mar 2019 12:25:47 +0000 (12:25 +0000)]
gdb: Avoid trailing whitespace when pretty printing
While writing a new test for 'set print pretty on' I spotted that GDB
will sometimes add a trailing whitespace character when pretty
printing. This commit removes the trailing whitespace and updates the
expected results in one tests where this was an issue.
I've added an extra test for 'set print pretty on' as it doesn't seem
to have much testing.
gdb/ChangeLog:
* cp-valprint.c (cp_print_value_fields): Don't print trailing
whitespace when pretty printing is on.
gdb/testsuite/ChangeLog:
* gdb.base/finish-pretty.exp: Update expected results.
* gdb.base/pretty-print.c: New file.
* gdb.base/pretty-print.exp: New file.
Tamar Christina [Tue, 26 Mar 2019 16:46:42 +0000 (16:46 +0000)]
AArch64/Arm: Update testcases fixing endiannes and linux targets
This fixes the testcases that are failing due to my recent patch.
It turns out that the start address across baremetal and linux builds
isn't entirely predictable without a linker script. Since the address
themselves are not the important thing I am ignoring them now.
Secondly I was encoding data using .word using non 0 values, however
because .word is subjected to endiannes these non-zero values under
big-endian happen to fall into the encoding space of instructions which
changes the disassembly. Using 0 fixes this problem and the purpose of
the test still holds, though objdump will dump ... for data only sections,
which is ok as the data/insn mixed sections will test the patch.
The ARM Attributes sections is not important and is ignored.
marxin [Tue, 26 Mar 2019 09:40:04 +0000 (10:40 +0100)]
Provide string description of definition, visibility and resolution in LTO plug-in.
ld/ChangeLog:
2019-02-26 Martin Liska <mliska@suse.cz>
* plugin.c (get_symbols): Add lto_kind_str, lto_resolution_str,
lto_visibility_str and use then to inform about plugin-symbols.
* testsuite/ld-plugin/plugin-12.d: Adjust expected pattern.
Tamar Christina [Mon, 25 Mar 2019 12:16:17 +0000 (12:16 +0000)]
Arm: Fix Arm disassembler mapping symbol search.
Similar to the AArch64 patches the Arm disassembler has the same issues with
out of order sections but also a few short comings.
For one thing there are multiple code blocks to determine mapping symbols, and
they all work slightly different, and neither fully correct. The first thing
this patch does is centralise the mapping symbols search into one function
mapping_symbol_for_insn. This function is then updated to perform a search in
a similar way as AArch64.
Their used to be a value has_mapping_symbols which was used to determine the
default disassembly for objects that have no mapping symbols. The problem with
the approach was that it was determining this value in the same loop that needed
it, which is why this field could take on the states -1, 0, 1 where -1 means
"don't know". However this means that until you actually find a mapping symbol
or reach the end of the disassembly glob, you don't know if you did the right
action or not, and if you didn't you can't correct it anymore.
This is why the two jump-reloc-veneers-* testcases end up disassembling some
insn as data when they shouldn't.
Out of order here refers to an object file where sections are not listed in a
monotonic increasing VMA order.
The ELF ABI for Arm [1] specifies the following for mapping symbols:
1) A text section must always have a corresponding mapping symbol at it's
start.
2) Data sections do not require any mapping symbols.
3) The range of a mapping symbol extends from the address it starts on up to
the next mapping symbol (exclusive) or section end (inclusive).
However there is no defined order between a symbol and it's corresponding
mapping symbol in the symbol table. This means that while in general we look
up for a corresponding mapping symbol, we have to make at least one check of
the symbol below the address being disassembled.
When disassembling different PCs within the same section, the search for mapping
symbol can be cached somewhat. We know that the mapping symbol corresponding to
the current PC is either the previous one used, or one at the same address as
the current PC.
However this optimization and mapping symbol search must stop as soon as we
reach the end or start of the section. Furthermore if we're only disassembling
a part of a section, the search is a allowed to search further than the current
chunk, but is not allowed to search past it (The mapping symbol if there, must
be at the same address, so in practice we usually stop at PC+4).
lastly, since only data sections don't require a mapping symbol the default
mapping type should be DATA and not INSN as previously defined, however if the
binary has had all its symbols stripped than this isn't very useful. To fix
this we determine the default based on the section flags. This will allow the
disassembler to be more useful on stripped binaries. If there is no section
than we assume you to be disassembling INSN.
* testsuite/binutils-all/arm/in-order-all.d: New test.
* testsuite/binutils-all/arm/in-order.d: New test.
* testsuite/binutils-all/arm/objdump.exp: Support .d tests.
* testsuite/binutils-all/arm/out-of-order-all.d: New test.
* testsuite/binutils-all/arm/out-of-order.T: New test.
* testsuite/binutils-all/arm/out-of-order.d: New test.
* testsuite/binutils-all/arm/out-of-order.s: New test.
Tamar Christina [Mon, 25 Mar 2019 12:14:37 +0000 (12:14 +0000)]
AArch64: Have -D override mapping symbol as documented.
The documentation for -D says that on Arm platforms -D should disassemble
data as instructions.
"If the target is an ARM architecture this switch also has the effect of
forcing the disassembler to decode pieces of data found in code sections
as if they were instructions. "
This makes it do as it says on the tincan so it's more consistent with
aarch32. The usecase here is for baremetal developers who have created
their instructions using .word directives instead if .insn.
Though for Linux users I do find this behavior somewhat non-optimal.
Perhaps there should be a new flag that just disassembles the values
following the actual mapping symbol?
binutils/ChangeLog:
* testsuite/binutils-all/aarch64/in-order-all.d: New test.
* testsuite/binutils-all/aarch64/out-of-order-all.d: New test.
* testsuite/binutils-all/aarch64/out-of-order.d:
Tamar Christina [Mon, 25 Mar 2019 12:12:03 +0000 (12:12 +0000)]
AArch64: Fix AArch64 disassembler mapping symbol search
My previous patch for AArch64 was not enough to catch all the cases where
disassembling an out-of-order section could go wrong. It had missed the case
DATA sections could be incorrectly disassembled as TEXT.
Out of order here refers to an object file where sections are not listed in a
monotonic increasing VMA order.
The ELF ABI for AArch64 [1] specifies the following for mapping symbols:
1) A text section must always have a corresponding mapping symbol at it's
start.
2) Data sections do not require any mapping symbols.
3) The range of a mapping symbol extends from the address it starts on up to
the next mapping symbol (exclusive) or section end (inclusive).
However there is no defined order between a symbol and it's corresponding
mapping symbol in the symbol table. This means that while in general we look
up for a corresponding mapping symbol, we have to make at least one check of
the symbol below the address being disassembled.
When disassembling different PCs within the same section, the search for mapping
symbol can be cached somewhat. We know that the mapping symbol corresponding to
the current PC is either the previous one used, or one at the same address as
the current PC.
However this optimization and mapping symbol search must stop as soon as we
reach the end or start of the section. Furthermore if we're only disassembling
a part of a section, the search is a allowed to search further than the current
chunk, but is not allowed to search past it (The mapping symbol if there, must
be at the same address, so in practice we usually stop at PC+4).
lastly, since only data sections don't require a mapping symbol the default
mapping type should be DATA and not INSN as previously defined, however if the
binary has had all its symbols stripped than this isn't very useful. To fix this
we determine the default based on the section flags. This will allow the
disassembler to be more useful on stripped binaries. If there is no section than
we assume you to be disassembling INSN.
Tamar Christina [Mon, 25 Mar 2019 12:08:53 +0000 (12:08 +0000)]
AArch64: Fix disassembler bug with out-of-order sections
The AArch64 disassembler has an optimization that it uses to reduce the amount
it has to search for mapping symbols during disassembly. This optimization
assumes that sections are listed in the section header in monotonic increasing
VMAs. However this is not a requirement for the ELF specification.
Because of this when such "out of order" sections occur the disassembler would
pick the wrong mapping symbol to disassemble the section with.
This fixes it by explicitly passing along the stop offset for the current
disassembly glob and when this changes compared to the previous one we've seen
the optimization won't be performed. In effect this restarts the search from
a well defined starting point. Usually the symbol's address.
The existing stop_vma can't be used for this as it is allowed to be unset and
setting this unconditionally would change the semantics of this field.
binutils/ChangeLog:
* objdump.c (disassemble_bytes): Pass stop_offset.
* testsuite/binutils-all/aarch64/out-of-order.T: New test.
* testsuite/binutils-all/aarch64/out-of-order.d: New test.
* testsuite/binutils-all/aarch64/out-of-order.s: New test.
Pedro Alves [Mon, 25 Mar 2019 13:26:23 +0000 (13:26 +0000)]
Fix testsuite hangs when gdb_test_multiple body errors out
This commit fixes a regression in the testsuite itself, triggered by
errors being raised from within gdb_test_multiple, originally reported
by Pedro Franco de Carvalho's at
<https://sourceware.org/ml/gdb-patches/2019-03/msg00160.html>. Parts
of the commit message are based on his report.
This started happening due to a commit that was introduced recently,
and it can cause the testsuite to hang.
That commit introduces a new "eof" block in gdb_test_multiple. That
is not incorrect itself, but dejagnu's remote_expect is picking that
block as the "default" action when an error is raised from within the
commands inside a call to gdb_test_multiple:
# remote_expect works basically the same as standard expect, but it
# also takes care of getting the file descriptor from the specified
# host and also calling the timeout/eof/default section if there is an
# error on the expect call.
#
proc remote_expect { board timeout args } {
I find that "feature" surprising, and I don't really know why it
exists, but this means that the eof section that remote_expect picks
as the error block can be executed even when there was no actual eof
and the GDB process is still running, so the wait introduced in the
commit that tries to get the exit status of GDB hangs forever, while
GDB itself waits for input.
This only happens when there are internal testsuite errors (not
testcase failures). This can be reproduced easily with a testcase
such as:
I think that working around this in GDB is useful so that the
testsuite doesn't hang in these cases.
Adding an empty "default" block at the end of the expect body in
gdb_test_multiple doesn't work, because dejagnu gives preference to
"eof" blocks:
if { $x eq "eof" } {
set save_next 1
} elseif { $x eq "default" || $x eq "timeout" } {
if { $error_sect eq "" } {
set save_next 1
}
}
And we do have "eof" blocks. So we need to make sure that the last
"eof" block is safe to use as the default error block. It's also
pedantically incorrect to print
"ERROR: Process no longer exists"
which is what we'd get if the last eof block we have was selected
(more below on this).
So this commit solves this by appending an "eof" with an empty
spawn_id list, so that it won't ever match.
Now, why is the first "eof" block selected today as the error block,
instead of the last one?
The reason is that remote_expect, while parsing the body to select the
default block to execute after an error, is affected by the comments
in the body (since they are also parsed).
If this comment in gdb_test_multiple
# patterns below apply to any spawn id specified.
is changed to
# The patterns below apply to any spawn id specified.
then the second eof block is selected and there is no hang.
Any comment at that same place with an even number of tokens also
works.
This is IMO a coincidence caused by how comments work in TCL.
Comments should only appear in places where a command can appear. And
here, remote_expect is parsing a list of options, not commands, so
it's not unreasonable to not parse comments, similarly to how this:
set a_list {
an_element
# another_element
}
results in a list with three elements, not one element.
The fact that comments with an even number of tokens work is just a
coincidence of how remote_expect's little state machine is
implemented.
I thought we could solve this by stripping out comment lines in
gdb_expect, but I didn't find an easy way to do that. Particularly, a
couple naive approaches I tried run into complications. For example,
we have gdb_test calls with regular expressions that include sequences
like "\r\n#", and by the time we get to gdb_expect, the \r\n have
already been expanded to a real newline, so just splitting the whole
body at newline boundaries, looking for lines that start with #
results in incorrectly stripping out half of the gdb_text regexp. I
think it's better (at least in this commit), to move the comments out
of the list, because it's much simpler and risk free.
gdb/testsuite/ChangeLog:
2019-03-25 Pedro Alves <palves@redhat.com>
* lib/gdb.exp (gdb_test_multiple): Split appends to $code and
move comments outside list. Append '-i "" eof' section.
Tom Tromey [Fri, 22 Mar 2019 22:53:12 +0000 (16:53 -0600)]
Remove null_block_symbol
This removes null_block_symbol. It seemed simpler to me to change
initializations and returns to use value initialization rather than
null_block_symbol. This also fixes up a few spots where
initialization was done piecemeal.
gdb/ChangeLog
2019-03-24 Tom Tromey <tom@tromey.com>
Tom Tromey [Fri, 22 Mar 2019 22:06:36 +0000 (16:06 -0600)]
Don't include symtab.h from expression.h
expression.h includes symtab.h, but apparently only for the
declaration of struct block. This patch changes it to foward-declare
the structure, and remove the include.
gdb/ChangeLog
2019-03-24 Tom Tromey <tom@tromey.com>
* expression.h: Don't include symtab.h.
(struct block): Forward declare.
Tom Tromey [Fri, 22 Mar 2019 21:54:10 +0000 (15:54 -0600)]
More block constification
I noticed that there are still many places referring to non-const
blocks. This constifies all the remaining ones that I found that
could be constified.
In a few spots, this search found unused variables or fields. I
removed these. I've also removed some unnecessary casts to
"struct block *".
gdb/ChangeLog
2019-03-24 Tom Tromey <tom@tromey.com>
* c-exp.y (typebase): Remove casts.
* gdbtypes.c (lookup_unsigned_typename, )
(lookup_signed_typename): Remove cast.
* eval.c (parse_to_comma_and_eval): Remove cast.
* parse.c (write_dollar_variable): Remove cast.
* block.h (struct block) <superblock>: Now const.
* symfile-debug.c (debug_qf_map_matching_symbols): Update.
* psymtab.c (psym_map_matching_symbols): Make "block" const.
(map_block): Make "block" const.
* symfile.h (struct quick_symbol_functions)
<map_matching_symbols>: Constify block argument to "callback".
* symtab.c (basic_lookup_transparent_type_quick): Make "block"
const.
(find_pc_sect_compunit_symtab): Make "b" const.
(find_symbol_at_address): Likewise.
(search_symbols): Likewise.
* dwarf2read.c (dw2_lookup_symbol): Make "block" const.
(dw2_debug_names_lookup_symbol): Likewise.
(dw2_map_matching_symbols): Update.
* p-valprint.c (pascal_val_print): Remove "block".
* ada-lang.c (ada_add_global_exceptions): Make "b" const.
(aux_add_nonlocal_symbols): Make "block" const.
(resolve_subexp): Remove cast.
* linespec.c (iterate_over_all_matching_symtabs): Make "block"
const.
(iterate_over_file_blocks): Likewise.
* f-exp.y (%union) <bval>: Remove.
* coffread.c (patch_opaque_types): Make "b" const.
* spu-tdep.c (spu_catch_start): Make "block" const.
* c-valprint.c (print_unpacked_pointer): Remove "block".
* symmisc.c (dump_symtab_1): Make "b" const.
(block_depth): Make "block" const.
* d-exp.y (%union) <bval>: Remove.
* cp-support.h (cp_lookup_rtti_type): Update.
* cp-support.c (cp_lookup_rtti_type): Make "block" const.
* psymtab.c (psym_lookup_symbol): Make "block" const.
(maintenance_check_psymtabs): Make "b" const.
* python/py-framefilter.c (extract_sym): Make "sym_block" const.
(enumerate_locals, enumerate_args): Update.
* python/py-symtab.c (stpy_global_block): Make "block" const.
(stpy_static_block): Likewise.
* inline-frame.c (block_starting_point_at): Make "new_block"
const.
* block.c (find_block_in_blockvector): Make return type const.
(blockvector_for_pc_sect): Make "b" const.
(find_block_in_blockvector): Make "b" const.
(re-)fix the regcache leaks when detaching from an executable.
Commit 799efbe8e01ab8292c01f46ac59a6fb2349d4535 was supposed to fix
the below leak. However, for this fix to work, it is critical to
save the ptid before detach.
This commit (pushed as OBVIOUS, as the change was already reviewed/approved)
saves the ptid before the detach, as in the original reviewed patch
(see https://sourceware.org/ml/gdb-patches/2019-02/msg00263.html).
Re-tested on debian/amd64, natively and under valgrind.
==7426== 1,123 (72 direct, 1,051 indirect) bytes in 1 blocks are definitely lost in loss record 2,872 of 3,020
==7426== at 0x4C2C4CC: operator new(unsigned long) (vg_replace_malloc.c:344)
==7426== by 0x5BD1E1: get_thread_arch_aspace_regcache(ptid_t, gdbarch*, address_space*) (regcache.c:330)
==7426== by 0x5BD39A: get_thread_regcache (regcache.c:366)
==7426== by 0x5BD39A: get_current_regcache() (regcache.c:372)
==7426== by 0x4B1EB4: get_current_frame() (frame.c:1588)
...
Tom Tromey [Sat, 23 Mar 2019 16:11:51 +0000 (10:11 -0600)]
Have parser reset the innermost block tracker
I ran across a comment in symfile.c today:
/* Clear globals which might have pointed into a removed objfile.
FIXME: It's not clear which of these are supposed to persist
between expressions and which ought to be reset each time. */
It seems to me that this can be clarified: the parser entry points
ought to reset the innermost block tracker (and the expression context
block), and these should not be considered valid for code to use at
arbitrary times -- only immediately after an expression has been
parsed.
This patch implements this idea. This could be further improved by
removing the parser globals and changing the parser functions to
return this information, but I have not done this.
Tested by the buildbot.
gdb/ChangeLog
2019-03-23 Tom Tromey <tom@tromey.com>
Alan Hayward [Fri, 22 Mar 2019 10:41:15 +0000 (10:41 +0000)]
AArch64: Prologue scan unwinder support for signed return addresses
Pauth address signing is enabled at binary compile time. When enabled the
return addresses for functions may be mangled. This patch adds functionality
to restore the original address for use in the prologue scan unwinder.
In the prologue analyzer, check for PACIASP/PACIBSP (enable address mangling)
and AUTIASP/AUTIBSP (disable address mangling).
When unwinding the PC from the prologue, unmask the register if required.
Add a test case to the prologue tests.
gdb/ChangeLog:
* aarch64-tdep.c (aarch64_analyze_prologue): Check for pauth
instructions.
(aarch64_analyze_prologue_test): Add PACIASP test.
(aarch64_prologue_prev_register): Unmask PC value.
Alan Hayward [Fri, 22 Mar 2019 10:39:11 +0000 (10:39 +0000)]
AArch64: DWARF unwinder support for signed return addresses
Pauth address signing is enabled at binary compile time. When enabled the
return addresses for functions may be mangled. This patch adds functionality
to restore the original address for use in the DWARF unwinder.
DW_CFA_AARCH64_negate_ra_state in a binary indicates the toggling of address
signing between enabled and disabled. Ensure the state is stored in the DWARF
register ra_state.
Ensure the pauth DWARF registers are initialised.
gdb/ChangeLog:
* aarch64-tdep.c (aarch64_frame_unmask_address): New function.
(aarch64_dwarf2_prev_register): Unmask PC value.
(aarch64_dwarf2_frame_init_reg): Init pauth registers.
(aarch64_execute_dwarf_cfa_vendor_op): Check for
DW_CFA_AARCH64_negate_ra_state.
(aarch64_gdbarch_init): Add aarch64_execute_dwarf_cfa_vendor_op.
Alan Hayward [Fri, 22 Mar 2019 10:37:46 +0000 (10:37 +0000)]
AArch64: Add pauth DWARF registers
Map the pauth registers to DWARF.
Add a new pseudo register ra_state and also map this to DWARF. This register
is hidden from the user - prevent it from being read or written to. It will
be used for the unmangling of addresses.
gdb/ChangeLog:
* aarch64-tdep.c (aarch64_dwarf_reg_to_regnum): Check for pauth
registers.
(aarch64_pseudo_register_name): Likewise.
(aarch64_pseudo_register_type): Likewise.
(aarch64_pseudo_register_reggroup_p): Likewise.
(aarch64_gdbarch_init): Add pauth registers.
* aarch64-tdep.h (AARCH64_DWARF_PAUTH_RA_STATE): New define.
(AARCH64_DWARF_PAUTH_DMASK): Likewise.
(AARCH64_DWARF_PAUTH_CMASK): Likewise.
(struct gdbarch_tdep): Add regnum for ra_state.
Alan Hayward [Fri, 22 Mar 2019 10:34:09 +0000 (10:34 +0000)]
AArch64: gdbserver: read pauth registers
Add the pauth registers to the regset lists.
Add a new regset type OPTIONAL_REGS which allows for the regset read to fail.
Once the read fails, it will not be checked again. This allows targets with
optional features to keep a single static regset_info structure.
gdb/ChangeLog:
* arch/aarch64.h (AARCH64_PAUTH_REGS_SIZE): New define.
gdb/gdbserver/ChangeLog:
* linux-aarch64-low.c (aarch64_store_pauthregset): New function.
* linux-low.c (regsets_store_inferior_registers): Allow optional reads
to fail.
* linux-low.h (enum regset_type): Add OPTIONAL_REGS.
Alan Hayward [Wed, 13 Feb 2019 12:28:38 +0000 (12:28 +0000)]
Testsuite: Ensure pie is disabled on some tests
Recent versions of Ubuntu and Debian default GCC to enable pie.
In dump.exp, pie will causes addresses to be out of range for IHEX.
In break-interp.exp, pie is explicitly set for some tests and assumed
to be disabled for the remainder.
Ensure pie is disabled for these tests when required.
In addition, add a pie option to gdb_compile to match the nopie option
and simplify use.
gdb/testsuite/ChangeLog:
* README: Add pie options.
* gdb.base/break-interp.exp: Ensure pie is disabled.
* gdb.base/dump.exp: Likewise.
* lib/gdb.exp (gdb_compile): Add pie option.
Jim Wilson [Thu, 21 Mar 2019 22:08:48 +0000 (15:08 -0700)]
RISC-V: Fix linker crash in section symbol check.
sym is only set for local symbols. h is only set for global symbols. Gas
won't let me create a global section symbol, but bfd appears to have some
support for that, and I can't rule out that other assemblers might do this.
So we need to support both, and verify sym and h are non-NULL before using.
bfd/
PR 24365
* elfnn-riscv.c (riscv_elf_relocate_section): For STT_SECTION check,
verify sym non-NULL before using. Add identical check using h.
Sudakshina Das [Thu, 21 Mar 2019 16:20:21 +0000 (16:20 +0000)]
[BFD, AArch64, x86] Improve warning for --force-bti
The AArch64 linker option to turn on BTI (--force-bti) warns in case there are
input objects which have a missing GNU NOTE section for BTI. This patch is trying
to improve the warnings that come out.
In order to do so, I propose adding a new argument to elf_merge_gnu_properties
and the backend function merge_gnu_properties. This new argument makes sure
that we now pass both the objects along with the properties to which they
belong to. The x86 backend function has also been updated to match this
change.
*** bfd/ChangeLog ***
2019-03-21 Sudakshina Das <sudi.das@arm.com>
* elf-bfd.h (struct elf_backend_data): Add argument to
merge_gnu_properties.
* elf-properties.c (elf_merge_gnu_properties): Add argument to
itself and while calling bed->merge_gnu_properties.
(elf_merge_gnu_property_list): Update the calls for
elf_merge_gnu_properties.
* elfnn-aarch64.c (elfNN_aarch64_merge_gnu_properties): Update handling
of --force-bti warning and add argument.
* elfxx-aarch64.c (_bfd_aarch64_elf_link_setup_gnu_properties): Add
warning.
* elfxx-x86.c (_bfd_x86_elf_merge_gnu_properties): Add argument.
* elfxx-x86.h (_bfd_x86_elf_merge_gnu_properties): Likewise in
declaration.
*** ld/ChangeLog ***
2019-03-21 Sudakshina Das <sudi.das@arm.com>
* testsuite/ld-aarch64/aarch64-elf.exp: Add new test.
* testsuite/ld-aarch64/bti-plt-1.s: Add .ifdef for PAC note section.
* testsuite/ld-aarch64/bti-plt-6.d: Update warning.
* testsuite/ld-aarch64/bti-plt-7.d: Likewise.
* testsuite/ld-aarch64/bti-warn.d: New test.
Alan Modra [Thu, 21 Mar 2019 04:28:35 +0000 (14:58 +1030)]
Add SORT_NONE to .init and .fini in scripts
The special case for .init and .fini in update_wild_statements is
ineffective for .init or .fini wildcards inside other output sections.
The special case needs to be on the wildcard, not the output section.
This patch is belt and braces, both fixing update_wild_statements and
the scripts.
Alan Modra [Wed, 20 Mar 2019 22:09:18 +0000 (08:39 +1030)]
lm32-linux ld testsuite fails
A number of the fails are due to ld supporting the creation of shared
libraries but not allowing linking against them without using an
option like -Bdynamic.
FAIL: Symbol export class test (final shared object)
FAIL: PROVIDE_HIDDEN test 4
FAIL: PROVIDE_HIDDEN test 6
FAIL: PROVIDE_HIDDEN test 10
FAIL: PROVIDE_HIDDEN test 12
FAIL: Build pr22471b.so
FAIL: Build pr22649-2b.so
FAIL: Build pr22649-2d.so
FAIL: PR ld/20828 dynamic symbols with section GC (plain)
FAIL: PR ld/20828 dynamic symbols with section GC (version script)
FAIL: PR ld/20828 dynamic symbols with section GC (versioned)
FAIL: PR ld/21233 dynamic symbols with section GC (--undefined)
FAIL: PR ld/21233 dynamic symbols with section GC (--require-defined)
FAIL: PR ld/21233 dynamic symbols with section GC (EXTERN)
FAIL: Build pr22150
FAIL: PR ld/14170
FAIL: Link using broken linker script
FAIL: pr17068 link --as-needed lib in group
FAIL: ld-gc/pr20022
Alan Modra [Wed, 20 Mar 2019 08:05:16 +0000 (18:35 +1030)]
Remove strip_underscore from struct emulation
This field is unused. I would have liked to also remove
leading_underscore and fake_label_name but in an i386-elf/coff/aout
multi-obj setup those fields are used to select a different
fake_label_name for aout.
Alan Modra [Tue, 19 Mar 2019 22:50:55 +0000 (09:20 +1030)]
Teach a few targets to resolve BFD_RELOC_8
and tidy "forward" test. I've removed some checks in d30v
md_apply_fix that have no business being there. Any symbol problems
will be caught later in tc_gen_reloc, and overflow checking is done in
gas/write.c.
* config/tc-d10v.c (md_apply_fix): Apply BFD_RELOC_8.
* config/tc-pdp11.c (md_apply_fix): Likewise.
* config/tc-d30v.c (md_apply_fix): Don't emit errors for BFD_RELOC_8,
BFD_RELOC_16, and BFD_RELOC_64.
* testsuite/gas/all/gas.exp: Move target exclusions for forward
test, but not cr16, to..
* testsuite/gas/all/forward.d: ..here, with explanation. Remove
d10v, d30v, and pdp11 xfails.
Alan Modra [Tue, 19 Mar 2019 05:11:58 +0000 (15:41 +1030)]
Fix some dlx fails
Generic linker ELF targets using CREATE_OBJECT_SYMBOLS in their
scripts run into a problem. The file symbols are created by
_bfd_generic_link_output_symbols in each object file, in the section
corresponding to the CREATE_OBJECT_SYMBOLS section, typically .text.
If it so happens that the output .text section is stripped due to
being empty, then elf.c:assign_section_numbers won't assign an ELF
section number and swap_out_syms will report "unable to find
equivalent output section" for the object symbols. Fix this by
always keeping an output section with CREATE_OBJECT_SYMBOLS.
* ldlang.c (lang_size_sections_1): Set SEC_KEEP on
create_object_symbols_section.
* testsuite/ld-elf/pr22319.d: Don't xfail dlx.
Sudakshina Das [Wed, 20 Mar 2019 18:00:07 +0000 (18:00 +0000)]
[BFD, AArch64] Define elf_backend_fixup_gnu_properties in AArch64
This patch add support for elf_backend_fixup_gnu_properties for GNU
property support for AArch64. The new AArch64 specific definition
_bfd_aarch64_elf_link_fixup_gnu_properties goes through the property
list to find AArch64 type properties and removes the properties that
are marked as "property_remove".
Tom Tromey [Wed, 20 Mar 2019 14:12:38 +0000 (08:12 -0600)]
Use @defvar to document gdb.pretty_printers
While referencing the manual, I noticed that gdb.pretty_printers
wasn't documented using @defvar. This made it more difficult to find
in the info pages. This patch adds the @defvar and also an
introductory paragraph in that node.
gdb/doc/ChangeLog
2019-03-20 Tom Tromey <tromey@adacore.com>
* python.texi (Selecting Pretty-Printers): Use @defvar for
gdb.pretty_printers.
Tom Tromey [Wed, 20 Mar 2019 13:57:09 +0000 (07:57 -0600)]
Merge handle_inferior_event and handle_inferior_event_1
I noticed that handle_inferior_event is just a small wrapper that
frees the value chain. This patch replaces it with a
scoped_value_mark, reducing the number of lines of code here.
Regression tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-03-20 Tom Tromey <tromey@adacore.com>
* infrun.c (handle_inferior_event): Rename from
handle_inferior_event_1. Create a scoped_value_mark.
(handle_inferior_event): Remove.
However, I think the first call to print_stop_event is intended just
to emit the MI *stopped notification, which explains why the source
line does not show up two times.
This patch fixes the bug by changing print_stop_event to only call
do_displays for non-MI-like ui-outs.
Tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-03-19 Tom Tromey <tromey@adacore.com>
* mi/mi-interp.c (mi_on_normal_stop_1): Only show displays once.
* infrun.h (print_stop_event): Add "displays" parameter.
* infrun.c (print_stop_event): Add "displays" parameter.
gdb/testsuite/ChangeLog
2019-03-19 Tom Tromey <tromey@adacore.com>
* gdb.mi/mi2-cli-display.c: New file.
* gdb.mi/mi2-cli-display.exp: New file.
Pedro Alves [Tue, 19 Mar 2019 18:08:27 +0000 (18:08 +0000)]
Add comments describing tui_ui_out and its fields, cleanup a bit
This commit add comments describing tui_ui_out and its fields, and
cleans up the code a little bit.
Also switch to using in-class initialization so that the initial
values can be seen alongside the comments.
I see no reason for initializing m_line as -1 instead of 0, since all
the checks in the .c file are of the form "> 0". AFAICS there's no
practical difference between -1 and 0. So it seems simpler to
initialize it as 0.
There's a bit of redundancy in tui_ui_out::do_field_string, which is
fixed by this commit.
gdb/ChangeLog:
2019-03-19 Pedro Alves <palves@redhat.com>
* tui/tui-out.c (tui_ui_out::do_field_string): Simplify.
(tui_ui_out::do_text): Add comments. Reset M_LINE to 0 instead of
to -1. Fix TABs vs spaces.
(tui_ui_out::tui_ui_out): Don't initialize fields here.
* tui/tui-out.h (tui_ui_out) Add intro comments.
<m_line, m_start_of_line>: In-class initialize, and add describing
comment.
H.J. Lu [Tue, 19 Mar 2019 13:12:47 +0000 (21:12 +0800)]
x86: Correct EVEX vector load/store optimization
Update EVEX vector load/store optimization:
1. There is no need to check AVX since AVX2 is required for AVX512F.
2. We need to check both operands for ZMM register since AT&T syntax
may not set zmmword on the first operand.
3. Update Opcode_SIMD_IntD check and set.
4. Since the VEX prefix has 2 or 3 bytes, the EVEX prefix has 4 bytes,
EVEX Disp8 has 1 byte and VEX Disp32 has 4 bytes, we choose EVEX Disp8
over VEX Disp32.
H.J. Lu [Tue, 19 Mar 2019 13:10:21 +0000 (21:10 +0800)]
x86: Correct EVEX to 128-bit EVEX optimization
Since not all AVX512F processors support AVX512VL, we can optimize
512-bit EVEX to 128-bit EVEX encoding for upper 16 vector registers
only when AVX512VL is enabled explicitly at command-line or via
".arch .avx512vl" directive.
PR gas/24352
* config/tc-i386.c (optimize_encoding): Check only
cpu_arch_flags.bitfield.cpuavx512vl.
* testsuite/gas/i386/i386.exp: Run x86-64-optimize-2b.
* testsuite/gas/i386/x86-64-optimize-2.d: Revert the last
change.
* testsuite/gas/i386/x86-64-optimize-2b.d: New file.
* testsuite/gas/i386/x86-64-optimize-2b.s: Likewise.
Pedro Alves [Mon, 18 Mar 2019 18:32:42 +0000 (18:32 +0000)]
Fix first time you type UP or DOWN in TUI's command window
The first time you type UP or DOWN arrow in the command window, GDB
should scroll the source window, but instead it displays the line
number and the file name in the command window(?).
What happens there is that the first time we call
tui_ui_out::do_field_int, it doesn't initialize m_line, because
m_start_of_line is -1, as set by the constructor; and then the
following call to tui_ui_out::do_field_string falls back to
cli_ui_out::do_field_string because m_line is zero.
The problem is caused by a typo in the C++ification of tui_ui_out,
commit 112e8700a6f, where m_line and m_start_of_line's initial values
were swapped from what they used to be:
Eli Zaretskii [Mon, 18 Mar 2019 18:04:40 +0000 (20:04 +0200)]
Fix gdb/TUI behavior in response to [Enter] keypress
gdb/ChangeLog:
2019-03-18 Eli Zaretskii <eliz@gnu.org>
* tui/tui-io.c (gdb_wgetch): Don't echo CR.
(tui_getc): When gdb_wgetch returns a CR, behave the same as when
it returns a newline. This fixes a regression in TUI mode, whereby
the next line is output on the same screen line as the user input.
Tom Tromey [Mon, 18 Mar 2019 15:32:09 +0000 (09:32 -0600)]
Fix regression caused by minimal symbol changes
The earlier patch to change minimal symbol allocations to use xmalloc
erroneously left a call to obstack_blank in
minimal_symbol_reader::install. Because obstack_blank does not finish
the object allocation on an obstack, this in turn could cause invalid
memory reads in some situations.
This patch fixes the problem by removing the call. Tested on x86-64
Fedora 29; also verified with valgrind.
gdb/ChangeLog
2019-03-18 Tom Tromey <tromey@adacore.com>
* minsyms.c (minimal_symbol_reader::install): Remove call to
obstack_blank.
Pedro Alves [Mon, 18 Mar 2019 14:26:00 +0000 (14:26 +0000)]
Improve/fix the TUI's current source line highlight
With styling enabled, I think the way we display the TUI's
highlighted/current line is very ugly and distracting. The problem in
my view is that we reverse foreground/background in colored text as
well, leading to rainbow of background colors.
This patch changes that to something that I find much more sensible --
only reverse the default foreground/background colors, leave styled
text colors alone. If the foreground color is not the default
(because the text was styled), leave the foreground color as is. If
e.g., the terminal is fg=BLACK, and bg=WHITE, and the style wants to
print text in RED, reverse the background color (print in BLACK), but
still print the text in RED.
Note: The new ui_file_style::set_fg method isn't called set_foreground
instead, because set_foreground is a macro in /usr/lib/term.h (ncurses).
gdb/ChangeLog:
2019-03-18 Pedro Alves <palves@redhat.com>
* tui/tui-io.c (reverse_mode_p, reverse_save_bg, reverse_save_fg):
New globals.
(apply_style): New, factored out from ...
(apply_ansi_escape): ... this. Handle reverse video mode.
(tui_set_reverse_mode): New function.
* tui/tui-io.h (tui_set_reverse_mode): New declaration.
* tui/tui-winsource.c (tui_show_source_line): Use
tui_set_reverse_mode instead of setting A_STANDOUT.
* ui-style.h (struct ui_file_style) <set_reverse, set_fg, set_bg>:
New setter methods.
Hannes Domani [Mon, 18 Mar 2019 14:25:59 +0000 (14:25 +0000)]
Fix scrolling right in the TUI
This commit fixes two issues in scrolling right in the TUI:
#1 - Scrolling right with the arrow keys, the first keypress doesn't
do anything. The problem is that copy_source_line() checks if
(column < first_col), and because of the ++column directly before, it
basically starts with 1 instead of 0.
#2 - Scrolling right handles TABS and escaped characters as single
characters, which just looks weird. The problem is that there's a
spot that misses handling TABS.
gdb/ChangeLog:
2019-03-18 Hannes Domani <ssbssa@yahoo.de>
* tui/tui-source.c (copy_source_line): Fix handling of 'column'.
Handle tabs.
Tom Tromey [Wed, 13 Mar 2019 14:40:25 +0000 (08:40 -0600)]
Fix Ada "ptype" bug with array types
Using ptype on an array type in Ada can sometimes show an incorrect
high bound. This happens because ada_evaluate_subexp will create an
array with an incorrect upper bound in the EVAL_AVOID_SIDE_EFFECTS
case.
This patch fixes the problem by arranging to always create such an
array with valid bounds.
Tested on x86-64 Fedora 29.
gdb/ChangeLog
2019-03-18 Tom Tromey <tromey@adacore.com>
gdb/testsuite/ChangeLog
2019-03-18 Joel Brobecker <brobecker@adacore.com>
Tom Tromey <tromey@adacore.com>
* gdb.ada/ptype_array/pck.adb: New file.
* gdb.ada/ptype_array/pck.ads: New file.
* gdb.ada/ptype_array/foo.adb: New file.
* gdb.ada/ptype_array.exp: New file.
Alan Modra [Mon, 18 Mar 2019 11:46:24 +0000 (22:16 +1030)]
Fix MRI mode testsuite failures
These fails were introduced by git commit 2469b3c584 with the
inroduction of "input_from_string":
m68k-linux FAIL: MRI structured for
m68k-linux FAIL: MRI structured if
m68k-linux FAIL: MRI structured repeat
m68k-linux FAIL: MRI structured while
Since the m68k parser called expression() without setting
input_from_string, get_symbol_name rejected FAKE_LABEL_CHAR in names.
* config/m68k-parse.y (yylex): Use temp_ilp and restore_ilp.
Alan Modra [Mon, 18 Mar 2019 11:08:36 +0000 (21:38 +1030)]
PR24355, segmentation fault in function called from ppc_finish_symbols
This one looks to be a bug going back to 2009, git commit e054468f6c
"STT_GNU_IFUNC support for PowerPC". That bug was carried over with
git commit 49c09209d0 "Rearrange PLT reloc output on powerpc".
If the refcount for an ifunc local sym plt entry was zero,
ppc_elf_size_dynamic_sections would correctly set plt.offset to -1 but
leave glink_offset uninitialized. That leads to occasional segfaults
(which can be made solid with MALLOC_PERTURB_=1 when using glibc).
So, guard the write_glink_stub call with plt.offset != -1. Also,
remove the totally ineffective attempt at writing multiple-use glink
stubs only once.
PR 24355
* elf32-ppc.c (ppc_finish_symbols): Don't call write_glink_stub
for local iplt syms with ent->plt.offset == -1. Remove ineffective
attempt at writing glink stubs only once.
Alan Modra [Mon, 18 Mar 2019 00:41:39 +0000 (11:11 +1030)]
csky ld testsuite fixes
* testsuite/ld-elf/merge.d: Remove csky from xfails, add moxie.
* testsuite/ld-elf/pr21884.d: Remove csky from xfails.
* testsuite/ld-elf/shared.exp: Add csky to list not xfailing pr22374.
* testsuite/ld-unique/pr21529.d: Remove csky from xfails
The build failure was noticed by Helmut Jarausch in
https://bugs.gentoo.org/680232:
$ ./configure CXXFLAGS='-std=c++17 -Os'
...
CXXLD gdb
ld: init.o: in function `initialize_all_files()':
init.c:(.text+0x113): undefined reference to `_initialize_string_view_selftests()'
It happens because '_initialize_string_view_selftests()' is
conditionally defined based on C++ default.
The change defines '_initialize_string_view_selftests()'
unconditionally and leaves implementation a no-op on c++17
compilers.
gdb/ChangeLog:
2019-03-17 Sergei Trofimovich <siarheit@google.com>
H.J. Lu [Mon, 18 Mar 2019 00:56:10 +0000 (08:56 +0800)]
x86: Optimize EVEX vector load/store instructions
When there is no write mask, we can encode lower 16 128-bit/256-bit
EVEX vector register load and store instructions as VEX vector register
load and store instructions with -O1.