gas: Change to "swym 0" as canonical nop insn for MMIX
While "set $0, $0" works, that's not the documented instruction to do
nothing for MMIX. However, I'm not changing it for "nop_type 5" as
seen in gas.exp and org-1.s, because "set $0, $0" seems like it could
be re-used there, for some future ISA.
gas/
* config/tc-mmix.h (md_single_noop_insn): Change to "swym 0".
Tom Tromey [Thu, 31 Dec 2020 21:55:27 +0000 (14:55 -0700)]
Update gdb.rust tests for Rust 1.49
Rust 1.49 was released today, and it includes some library changes
which caused some gdb.rust tests to fail. This patch adapts the test
suite to the new output. I also verified that this continues to work
with Rust 1.48.
gdb/testsuite/ChangeLog
2020-12-31 Tom Tromey <tom@tromey.com>
* gdb.rust/simple.exp: Update output for Rust 1.49.
Alan Modra [Wed, 30 Dec 2020 11:30:57 +0000 (22:00 +1030)]
PR27128, nm -P portable output format regression
binutils/
PR 27128
* nm.c (print_symname): Append version string to symbol name
before printing the lot under control of "form". Append version
to demangled names too.
ld/
PR 27128
* testsuite/ld-elf/pr27128.s: New file.
* testsuite/ld-elf/pr27128.t: Likewise.
* testsuite/ld-elf/pr27128a.d: Likewise.
* testsuite/ld-elf/pr27128b.d: Likewise.
* testsuite/ld-elf/pr27128c.d: Likewise.
* testsuite/ld-elf/pr27128d.d: Likewise.
* testsuite/ld-elf/pr27128e.d: Likewise.
Tom Tromey [Sun, 27 Dec 2020 20:36:55 +0000 (13:36 -0700)]
Simplify MULTI_SUBSCRIPT implementation
The MULTI_SUBSCRIPT code in evaluate_subexp_standard has a comment
saying that perhaps the EVAL_SKIP handling is incorrect. This patch
simplifies this code. In particular, it precomputes all the indices
in a separate loop and removes some complicated flow-control.
Tested using the gdb.modula2 and gdb.dlang test suites, as these are
the only parsers that emit MULTI_SUBSCRIPT.
gdb/ChangeLog
2020-12-27 Tom Tromey <tom@tromey.com>
H.J. Lu [Fri, 25 Dec 2020 13:56:36 +0000 (05:56 -0800)]
gas: Update 80387 floating point 's' suffix
Update 80387 floating point 's' suffix to read:
* Integer constructors are '.word', '.long' or '.int', and '.quad'
for the 16-, 32-, and 64-bit integer formats. The corresponding
instruction mnemonic suffixes are 's' (short), 'l' (long), and 'q'
(quad).
instead of 's' (single).
PR gas/27106
* doc/c-i386.texi: Update 80387 floating point 's' suffix
Andrew Burgess [Fri, 18 Dec 2020 11:59:54 +0000 (11:59 +0000)]
gdb: avoid resolving dynamic properties for non-allocated arrays
In PR gdb/27059 an issue was discovered where GDB would sometimes
trigger undefined behaviour in the form of signed integer overflow.
The problem here is that GDB was reading random garbage from the
inferior memory space, assuming this data was valid, and performing
arithmetic on it.
This bug raises an interesting general problem with GDB's DWARF
expression evaluator, which is this:
We currently assume that the DWARF expressions being evaluated are
well formed, and well behaving. As an example, this is the expression
that the bug was running into problems on, this was used as the
expression for a DW_AT_byte_stride of a DW_TAG_subrange_type:
Two values are read from the inferior and multiplied together. GDB
should not assume that any value read from the inferior is in any way
sane, as such the implementation of DW_OP_mul should be guarding
against overflow and doing something semi-sane here.
However, it turns out that the original bug PR gdb/27059, is hitting a
more specific case, which doesn't require changes to the DWARF
expression evaluator, so I'm going to leave the above issue for
another day.
In the test mentioned in the bug GDB is actually trying to resolve the
dynamic type of a Fortran array that is NOT allocated. A
non-allocated Fortran array is one that does not have any data
allocated for it yet, and even the upper and lower bounds of the array
are not yet known.
It turns out that, at least for gfortran compiled code, the data
fields that describe the byte-stride are not initialised until the
array is allocated.
This leads me to the following conclusion: GDB should not try to
resolve the bounds, or stride information for an array that is not
allocated (or not associated, a similar, but slightly different
Fortran feature). Instead, each of these properties should be set to
undefined if the array is not allocated (or associated).
That is what this commit does. There's a new flag that is passed
around during the dynamic array resolution. When this flag is true
the dynamic properties are resolved using the DWARF expressions as
they currently are, but when this flag is false the expressions are
not evaluated, and instead the properties are set to undefined.
gdb/ChangeLog:
PR gdb/27059
* eval.c (evaluate_subexp_for_sizeof): Handle not allocated and
not associated arrays.
* f-lang.c (fortran_adjust_dynamic_array_base_address_hack): Don't
adjust arrays that are not allocated/associated.
* gdbtypes.c (resolve_dynamic_range): Update header comment. Add
new parameter which is used to sometimes set dynamic properties to
undefined.
(resolve_dynamic_array_or_string): Update header comment. Add new
parameter which is used to guard evaluating dynamic properties.
Resolve allocated/associated properties first.
gdb/testsuite/ChangeLog:
PR gdb/27059
* gdb.dwarf2/dyn-type-unallocated.c: New file.
* gdb.dwarf2/dyn-type-unallocated.exp: New file.
Lancelot SIX [Thu, 24 Dec 2020 16:01:21 +0000 (11:01 -0500)]
gdb/gdbtypes.h: Fix comparison of uninitialized values
When called with an array type of unknown dimensions,
is_scalar_type_recursive ended up comparing uninitialized values.
This was picked up by the following compiler warning:
CXX gdbtypes.o
/binutils-gdb/gdb/gdbtypes.c: In function int is_scalar_type_recursive(type*):
/binutils-gdb/gdb/gdbtypes.c:3670:38: warning: high_bound may be used uninitialized in this function [-Wmaybe-uninitialized]
3670 | return high_bound == low_bound && is_scalar_type_recursive (elt_type);
| ~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/binutils-gdb/gdb/gdbtypes.c:3670:38: warning: low_bound may be used uninitialized in this function [-Wmaybe-uninitialized]
This patch makes sure that when dealing with an array of unknown size
(or an array of more than 1 element), is_scalar_type_recursive returns
false.
gdb/ChangeLog:
* gdbtypes.c (is_scalar_type_recursive): Prevent comparison
between uninitialized values.
Alan Modra [Thu, 24 Dec 2020 05:41:03 +0000 (16:11 +1030)]
asan: print_vms_time signed integer overflow
I really don't think anyone cares about underflow of vms time values,
but the potential segfault on a gmtime failure is worth fixing.
* readelf.c (INT64_MIN): Define if not already defined.
(print_vms_time): Catch 64-bit overflow when converting from
vms time to posix time. Don't segfault if gmtime returns NULL.
Alan Modra [Thu, 24 Dec 2020 05:26:23 +0000 (15:56 +1030)]
PR27100, final link failed: bad value
The failure on this PR is due to using the same bfd section for
multiple output sections. Commit 21401fc7bf67 managed to create
duplicate linker script output section statements, but not the actual
bfd sections.
PR 27100
* ldlang.h (lang_output_section_statement_type): Add dup_output.
* ldlang.c (lang_output_section_statement_lookup): Set dup_output.
(init_os): Test dup_output rather than constraint.
* testsuite/ld-scripts/pr27100.d,
* testsuite/ld-scripts/pr27100.s,
* testsuite/ld-scripts/pr27100.t: New test.
* testsuite/ld-scripts/data.exp: Run it. Don't exclude aout here.
* testsuite/ld-scripts/data.d: Do so here instead.
* testsuite/ld-scripts/fill.d: Likewise.
* testsuite/ld-scripts/fill16.d: Likewise.
Alex Bennée [Wed, 23 Dec 2020 21:36:16 +0000 (16:36 -0500)]
Clarify language for the '?' packet
Both QEMU and kgdb make the assumption that the '?' packet is only
sent during the initial setup of a gdbstub connection. Both use that
knowledge to reset breakpoints and ensure the gdbstub is in a
clean-state on a resumed connection. This can cause confusion for
others implementing clients that speak to gdbstub devices. To avoid
that make the language clearer that this is a start-up query packet
that you only expect to see once.
H.J. Lu [Wed, 23 Dec 2020 21:00:39 +0000 (13:00 -0800)]
x86-64: Add Intel LAM property support
Add Intel Linear Address Masking (LAM) property support. LAM modifies
the checking that is applied to 64-bit linear addresses, allowing
software to use of the untranslated address bits for metadata.
bfd/
* elf-linker-x86.h (elf_x86_cet_report): Renamed to ...
(elf_x86_prop_report): This.
(elf_linker_x86_params): Add lam_u48, lam_u57, lam_u48_report
and lam_u57_report.
* elfxx-x86.c (_bfd_x86_elf_link_setup_gnu_properties): Support
GNU_PROPERTY_X86_FEATURE_1_LAM_U48 and
GNU_PROPERTY_X86_FEATURE_1_LAM_U57.
(_bfd_x86_elf_link_fixup_gnu_properties): Keep LAM features only
for 64-bit output.
binutils/
* NEWS: Mention LAM_U48 and LAM_U57 support.
* elfedit.c (elf_x86_feature): Support lam_u48 and lam_u57.
(usage): Add lam_u48 and lam_u57.
* readelf.c (decode_x86_feature_1): Support LAM_U48 and LAM_U57.
* doc/binutils.texi: Update elfedit with lam_u48 and lam_u57
support.
* testsuite/binutils-all/x86-64/lam-u48.d: New file.
* testsuite/binutils-all/x86-64/lam-u48.s: Likewise.
* testsuite/binutils-all/x86-64/lam-u57.d: Likewise.
* testsuite/binutils-all/x86-64/lam-u57.s: Likewise.
Andrew Burgess [Thu, 19 Nov 2020 17:52:00 +0000 (17:52 +0000)]
gdb: remove some uses of LA_PRINT_STRING
This commit removes some, but not all, uses of LA_PRINT_STRING. In
this commit I've removed those uses where there is an obvious language
object on which I can instead call the printstr method.
In the remaining 3 uses it is harder to know if the correct thing is
to call printstr on the current language, or on a specific language.
Currently obviously, we always call on the current language (as that's
what LA_PRINT_STRING does), and clearly this behaviour is good enough
right now, but is it "right"? I've left them for now and will give
them more thought in the future.
Andrew Burgess [Fri, 30 Oct 2020 15:09:35 +0000 (15:09 +0000)]
gdb: move rust_language into rust-lang.h
Move the rust_language class declaration into the rust-lang.h header
file. This allows for the function implementations called directly in
rust-lang.c and rust-exp.y without the need for trampoline functions.
There should be no user visible changes after this commit.
gdb/ChangeLog:
* rust-exp.y (rust_parse): Rename to...
(rust_language::parser): ...this.
* rust-lang.c (-rust_printstr): Rename to...
(rust_language::printstr): ...this.
(rust_value_print_inner): Delete declaration.
(val_print_struct): Rename to...
(rust_language::val_print_struct): ...this. Update calls to
member functions.
(rust_print_enum): Rename to...
(rust_language::print_enum): ...this. Update calls to member
functions.
(rust_value_print_inner): Rename to...
(rust_language::value_print_inner): ...this. Update calls to
member functions.
(exp_descriptor_rust): Rename to...
(rust_language::exp_descriptor_tab): ...this.
(class rust_language): Move to rust-lang.h.
(rust_language::language_arch_info): Implementation moved to here
from class declaration.
(rust_language::print_type): Likewise.
(rust_language::emitchar): Likewise.
(rust_language::is_string_type_p): Likewise.
* rust-lang.h: Add 'demangle.h', 'language.h', 'value.h', and
'c-lang.h' includes.
(rust_parse): Delete declaration.
(class rust_language): Class declaration moved here from
rust-lang.c.
Andrew Burgess [Wed, 28 Oct 2020 18:10:01 +0000 (18:10 +0000)]
gdb: move pascal_language into p-lang.h
Move the pascal_language class declaration into the p-lang.h header
file. This allows for the function implementations to be spread over
the different p-*.c files without the need for global trampoline
functions.
As a consequence of this change many of the Pascal value and type
printing helper functions have become member functions within the
pascal_language class.
There should be no user visible changes after this commit.
gdb/ChangeLog:
* p-exp.y (exp): Update call to pascal_is_string_type.
(pascal_parse): Rename to...
(pascal_language::parser): ...this.
* p-lang.c (is_pascal_string_type): Rename to...
(pascal_is_string_type): ...this.
(pascal_one_char): Rename to...
(pascal_language::print_one_char): ...this.
(pascal_printchar): Rename to...
(pascal_language::printchar): ...this. Update call to
print_one_char member function.
(pascal_op_print_tab): Rename to...
(pascal_language::op_print_tab): ...this.
(class pascal_language): Moved to p-lang.h.
(pascal_language::language_arch_info): Function implementation
moved out of class declaration.
(pascal_language::printstr): Likewise.
* p-lang.h (pascal_parse): Delete declaration.
(pascal_is_string_type): Declare.
(pascal_print_type): Delete declaration.
(pascal_print_typedef): Delete declaration.
(pascal_value_print_inner): Delete declaration.
(pascal_value_print): Delete declaration.
(pascal_type_print_method_args): Delete declaration.
(is_pascal_string_type): Delete declaration.
(pascal_printchar): Delete declaration.
(pascal_builtin_types): Delete declaration.
(pascal_type_print_base): Delete declaration.
(pascal_type_print_varspec_prefix): Delete declaration.
(class pascal_language): Moved here from p-lang.c.
* p-typeprint.c (pascal_type_print_varspec_suffix): Delete
declaration.
(pascal_type_print_derivation_info): Delete declaration.
(pascal_print_type): Rename to...
(pascal_language::print_type): ...this. Update calls to member
functions.
(pascal_print_typedef): Rename to...
(pascal_language::print_typedef): ...this. Update calls to member
functions.
(pascal_type_print_derivation_info): Rename to...
(pascal_language::type_print_derivation_info): ...this.
(pascal_type_print_method_args): Rename to...
(pascal_language::type_print_method_args): ...this.
(pascal_type_print_varspec_prefix): Rename to...
(pascal_language::type_print_varspec_prefix): ...this. Update
calls to member functions.
(pascal_print_func_args): Rename to...
(pascal_language::print_func_args): ...this. Update calls to
member functions.
(pascal_type_print_func_varspec_suffix): Rename to...
(pascal_language::type_print_func_varspec_suffix): ...this.
Update calls to member functions.
(pascal_type_print_varspec_suffix): Rename to...
(pascal_language::type_print_varspec_suffix): ...this. Update
calls to member functions.
(pascal_type_print_base): Rename to...
(pascal_language::type_print_base): ...this. Update calls to
member functions.
* p-valprint.c (pascal_value_print_inner): Rename to...
(pascal_language::value_print_inner): ...this. Update calls to
member functions.
(pascal_value_print): Rename to...
(pascal_language::value_print): ...this. Update calls to member
functions.
Andrew Burgess [Tue, 20 Oct 2020 13:03:42 +0000 (14:03 +0100)]
gdb: move go_language class declaration into header file
Move the go_language class into go-lang.h, this allows us to have
member functions implemented directly in the different go-*.c files
instead of having to trampoline out to global functions.
There should be no user visible changes after this commit.
gdb/ChangeLog:
* go-exp.y (go_parse): Rename to...
(go_language::parser): ...this.
* go-lang.c (go_demangle): Rename to...
(go_language::demangle_symbol): ...this.
(go_language::expression_ops): Implementation moved here out of
class declaration.
(go_op_print_tab): Rename to...
(go_language::op_print_tab): ...this, update comment.
(class go_language): Declaration moved to go-lang.h.
(go_language::language_arch_info): Implementation moved here out
of class declaration.
* go-lang.h (go_parse): Delete declaration.
(go_demangle): Delete declaration.
(go_print_type): Delete declaration.
(go_value_print_inner): Delete declaration.
(class go_language): Declaration moved here from go-lang.c.
* go-typeprint.c (go_print_type): Rename to...
(go_language::print_type): ...this.
* go-valprint.c (go_value_print_inner): Rename to...
(go_language::value_print_inner): ...this.
* symtab.c (demangle_for_lookup): Call demangle_symbol method on
the go_language object.
Andrew Burgess [Tue, 20 Oct 2020 12:41:18 +0000 (13:41 +0100)]
gdb: remove LA_EMIT_CHAR macro
Now that every use of the LA_EMIT_CHAR macro is within a language_defn
member function we can simply call the emitchar member function
directly instead of using the LA_EMIT_CHAR macro.
If we are ever inside a language object, for example, cplus_language,
while current_language points at something other than cplus_language
then this commit will result in a change in behaviour. However, I
believe if we did have such a difference then this would be a bug in
GDB. AS such I'm going to claim there _should_ be no user visible
changes from this commit.
Andrew Burgess [Tue, 20 Oct 2020 11:38:42 +0000 (12:38 +0100)]
gdb: avoid accessing global C++ language implementation functions
The function c_printchar is called from two places; it provides the
implementation of language_defn::printchar and it is called from
dwarf2_compute_name.
It would be nice to rename c_printchar as language_defn::printchar and
so avoid the trampoline.
To achieve this, instead of calling c_printchar directly from the
DWARF code, I lookup the C++ language object and call the printchar
member function.
In a later commit I can then rename c_printchar.
There should be no user visible changes after this commit.
gdb/ChangeLog:
* dwarf2/read.c (dwarf2_compute_name): Call methods on C++
language object instead of calling global functions directly.
Samuel Thibault [Mon, 21 Dec 2020 16:01:07 +0000 (16:01 +0000)]
hurd: Fix getting VM_MIN/MAX_ADDRESS
gnu-nat.c was getting the inclusion of vm_param.h only by luck. We need
to explicitly include it to be sure to get the definitions of
VM_MIN/MAX_ADDRESS.
Shahab Vahedi [Tue, 10 Nov 2020 18:34:57 +0000 (19:34 +0100)]
arc: Take into account the REGNUM in supply/collect gdb hooks
All the arc_linux_supply_*() target operations and the
arc_linux_collect_v2_regset() in arc-linux-tdep.c were
supplying/collecting all the registers in regcache as if the
REGNUM was set to -1.
The more efficient behavior is to examine the REGNUM and act
accordingly. That is what this patch does.
Anton Kolesov [Thu, 22 Dec 2016 18:52:16 +0000 (21:52 +0300)]
arc: Add support for signal frames for Linux targets
Implement functions needed to unwind signal frames on ARC Linux targets.
gdb/ChangeLog
* arc-linux-tdep.c (arc_linux_sc_reg_offsets): New static variable.
(arc_linux_is_sigtramp): New function.
(arc_linux_sigcontext_addr): Likewise.
(arc_linux_init_osabi): Use them.
Anton Kolesov [Mon, 22 Aug 2016 16:39:46 +0000 (19:39 +0300)]
arc: Add support for signal handlers
This patch adds the necessary infrastructure to handle signal frames for
ARC architecture. It is fairly similar to what any other architecture
would have. Linux specific parts will be in a separate patch.
v2 [1]:
- Make the logic of "arc_sigtramp_frame_sniffer ()" simpler.
[1] Tom's remark for the first version
https://sourceware.org/pipermail/gdb-patches/2020-November/173221.html
gdb/ChangeLog:
* arc-tdep.c (arc_make_sigtramp_frame_cache): New function.
(arc_sigtramp_frame_this_id): Likewise.
(arc_sigtramp_frame_prev_register): Likewise.
(arc_sigtramp_frame_sniffer): Likewise.
(arc_siftramp_frame_unwind): New global variable.
(arc_gdbarch_init): Use sigtramp capabilities.
(arc_dump_tdep): Print sigtramp fields.
* arc-tdep.h (gdbarch_tdep): Add sigtramp fields.
Tom Tromey [Sat, 21 Nov 2020 15:49:10 +0000 (08:49 -0700)]
Move enum noside earlier in expression.h
For the expression rewrite series, I needed to move enum noside
earlier in expression.h. Because this is a pure move, and because it
seems harmless and uncontroversial to move an enum definition earlier
in a file, I'm pushing it in early, to reduce the size of that series.
Tested by rebuilding.
gdb/ChangeLog
2020-12-21 Tom Tromey <tom@tromey.com>
Peter Waller [Mon, 21 Dec 2020 14:38:07 +0000 (09:38 -0500)]
Preserve gdb_std{out, err, log, targ, targerr} across interpreter_exec_cmd
Calls through interpreter_exec_cmd can cause the output state to be modified in
a way which doesn't get back after the execution.
It looks like the intent is that interp::resume should put things back how they
should be, however, mi_interp::resume modifies gdb_stdout and nothing currently
restores it to the previous state.
Prior to this patch, on a terminal environment, the first backtrace is
coloured, and the second backtrace is not. The reason is that
stdio_file::can_emit_style_escape becomes false, because the gdb_stdout gets
overwritten in mi_interp::resume and not replaced.
Markus Metzger [Fri, 4 Dec 2020 09:29:06 +0000 (10:29 +0100)]
testsuite, gdb.btrace: fix 32-bit PIE false positives
For 32-bit position independent executables, GCC generates an extra call to
__x86.get_pc_thunk.<reg>
which appears in the function call history. It is correct to appear there
but this confuses the tests, which check for an expected sequence of
functions.
Build with nopie to avoid this complication.
gdb/testsuite/ChangeLog:
2020-12-04 Markus Metzger <markus.t.metzger@intel.com>
Markus Metzger [Thu, 3 Dec 2020 16:04:29 +0000 (17:04 +0100)]
gdb, record: rephrase the 'not recording' error message
When trying to use one of the record commands without having enabled
recording first, GDB gives the error message:
(gdb) record function-call-history
No record target is currently active.
Use one of the "target record-<TAB><TAB>" commands first.
In the record help, however, we say:
(gdb) help record
record, rec
Start recording.
List of record subcommands:
record btrace, record b -- Start branch trace recording.
record delete, record del, record d -- Delete the rest of execution log and start recording it anew.
record full -- Start full execution recording.
record function-call-history -- Prints the execution history at function granularity.
record goto -- Restore the program to its state at instruction number N.
record instruction-history -- Print disassembled instructions stored in the execution log.
record save -- Save the execution log to a file.
record stop, record s -- Stop the record/replay target.
Change the above error message to
(gdb) record function-call-history
No recording is currently active.
Use the "record full" or "record btrace" command first.
to align with the help text.
gdb/ChangeLog:
2020-12-03 Markus Metzger <markus.t.metzger@intel.com>
Tom de Vries [Sun, 20 Dec 2020 08:35:36 +0000 (09:35 +0100)]
[gdb/testsuite] Add save_target_board_info
Add a proc save_target_board_info, similar to save_vars, such that we can do:
...
save_target_board_info { multilib_flags } {
global board
set board [target_info name]
unset_board_info multilib_flags
set_board_info multilib_flags "$override_multilib_flags"
...
}
...
and use it in gdb_compile_shlib.
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2020-12-20 Tom de Vries <tdevries@suse.de>
* lib/gdb.exp (save_target_board_info): New proc.
(gdb_compile_shlib): Use save_target_board_info.
Thread 1 received signal SIGSEGV, Segmentation fault.
[Switching to Thread 6184.0x1cc4]
check_typedef (type=0x0) at C:/src/repos/binutils-gdb.git/gdb/gdbtypes.c:2745
2745 while (type->code () == TYPE_CODE_TYPEDEF)
PR exp/27070
* gdb.python/compare-enum-type-a.c: New test.
* gdb.python/compare-enum-type-b.c: New test.
* gdb.python/compare-enum-type.exp: New file.
* gdb.python/compare-enum-type.h: New test.
Hannes Domani [Fri, 18 Dec 2020 15:17:46 +0000 (16:17 +0100)]
Fix accessing a method's fields from Python
Considering this example:
struct C
{
int func() { return 1; }
} c;
int main()
{
return c.func();
}
Accessing the fields of C::func, when requesting the function by its
type, works:
(gdb) py print(gdb.parse_and_eval('C::func').type.fields()[0].type)
C * const
But when trying to do the same via a class instance, it fails:
(gdb) py print(gdb.parse_and_eval('c')['func'].type.fields()[0].type)
Traceback (most recent call last):
File "<string>", line 1, in <module>
TypeError: Type is not a structure, union, enum, or function type.
Error while executing Python code.
The difference is that in the former the function type is TYPE_CODE_FUNC:
Jameson Nash [Fri, 18 Dec 2020 19:08:53 +0000 (14:08 -0500)]
gdb: define COFF file offsets with file_ptr
The arguments to these functions are file_ptr, so these declarations
were accidentally implicitly down-casting them to signed int. This
allows for reading files between 2 and 4 GB in size in my testing (I
don't have a larger dll currently to test). These may not be natively
supported by Windows, but can appear when using split-dwarf information.
This solves a "can't get string table" error resulting from attempting
to pass a negative offset to bfd_seek. I encountered this occuring while
trying to use a debug file for libLLVM.dll, but searching online reveals
at least one other person may have run into a similar problem with
Firefox?
Tom Tromey [Fri, 18 Dec 2020 15:48:36 +0000 (08:48 -0700)]
Run fixed_points.exp with -fgnat-encodings=minimal
This changes the test case gdb.ada/fixed_points.exp to also be run
with -fgnat-encodings=minimal. This change pointed out that the test
case had a few incorrect expected outputs; these are fixed as well.
Note that the Overprecise_Object test only uses the non-legacy output
with GCC trunk.
gdb/testsuite/ChangeLog
2020-12-18 Tom Tromey <tromey@adacore.com>
* gdb.ada/fixed_points.exp: Also run with
-fgnat-encodings=minimal. Update expected output.
Alan Modra [Thu, 17 Dec 2020 12:51:11 +0000 (23:21 +1030)]
Remove some static buffers
Fixes possible overflow of a static buffer for powerpc with translated
messages, and on v850 when symbol names are large.
* archive.c (_bfd_ar_spacepad, _bfd_ar_sizepad): Use auto buf.
* coff-mcore.c (coff_mcore_relocate_section): Likewise.
* elf32-ppc.c (ppc_elf_unhandled_reloc): Use asprintf in place
of fixed size and possibly too small buf for translated message.
* elf64-ppc.c (ppc64_elf_unhandled_reloc): Likewise.
* elf32-v850.c (v850_elf_check_relocs): Likewise.
* ecoff.c (ecoff_type_to_string): Pass in return string buff rather
than using static buffer2. Delete dead code. Remove unnecessary
parentheses.
(_bfd_ecoff_print_symbol): Pass auto buff to ecoff_type_to_string.
* elf32-rx.c (describe_flags): Pass in return string buf rather
than using static buf.
(rx_elf_merge_private_bfd_data): Pass buf to describe_flags.
(rx_elf_print_private_bfd_data): Likewise.
* mach-o.c (cpusubtype): Pass in return string buffer rather than
using static buffer.
(bfd_mach_o_bfd_print_private_bfd_data): Pass buff to cpusubtype.
* opncls.c (separate_debug_file_exists): Make buffer an auto var.
(bfd_fill_in_gnu_debuglink_section): Likewise.
* peXXigen.c (rsrc_resource_name): Pass in return string buffer
rather than using static buffer.
(rsrc_sort_entries): Pass buff to rsrc_resource_name.
* vms-alpha.c (_bfd_vms_write_emh): Pass tbuf to get_vms_time_string.
* vms-misc.c (get_vms_time_string): Pass in return string tbuf
rather than using static tbuf.
* vms.h (get_vms_time_string): Update prototype.
Alan Modra [Wed, 16 Dec 2020 21:14:48 +0000 (07:44 +1030)]
Statically initialise target common sections
This tidies initialisation of target common sections, doing so using a
static initialiser rather than via code and deleting unnecessary
symbol_ptr_ptr variables (the one in asection is used instead).
The patch also initialises ecoff.c:bfd_debug_section using
BFD_FAKE_SECTION. That does change bfd_debug_section slightly,
output_section was NULL now bfd_debug_section, and symbol_ptr_ptr
was NULL now &bfd_debug_section.symbol, but I believe those changes
are safe.
Tom Tromey [Thu, 17 Dec 2020 20:29:38 +0000 (13:29 -0700)]
Remove printfi_filtered and fprintfi_filtered
After seeing Simon's patch, I thought maybe it was finally time to
remove printfi_filtered and fprintfi_filtered, in favor of using the
"%*s" approach to indenting.
In this patch I took the straightforward approach of always adding a
leading "%*s", even when the format already started with "%s", to
avoid the trickier form of:
printf ("%*s", -indent, string)
Regression tested on x86-64 Fedora 32.
Let me know what you think.
gdb/ChangeLog
2020-12-17 Tom Tromey <tromey@adacore.com>
Tom Tromey [Wed, 16 Dec 2020 22:49:40 +0000 (15:49 -0700)]
Change parameters to language_defn::post_parser
In the expression rewrite, Ada type resolution will be done at parse
time rather than in a post-parse pass. At this point,
language_defn::post_parser will be removed. However, for this to
work, the information available to post_parser must be made available
during the actual parse.
This patch refactors this code slightly to make this possible. In
particular, "void_context_p" is passed to the parser_state
constructor, and the parser state is then passed to the post_parser
method.
gdb/ChangeLog
2020-12-16 Tom Tromey <tom@tromey.com>
Note that nothing in-tree passes true here. I don't know why this is,
but there is a use of this internally in AdaCore's tree. I will try
to submit that patch, if it is needed. (And if not, I will come back
around and remove this.)
gdb/ChangeLog
2020-12-16 Tom Tromey <tom@tromey.com>
Simon Marchi [Wed, 16 Dec 2020 21:43:18 +0000 (16:43 -0500)]
gdb/testsuite: make some tests in gdb.base enable non-stop using GDBFLAGS
For the same reason as explained in commit 7cb2893dfab1 ("gdb/testsuite:
gdb.mi/mi-nonstop-exit.exp: enable non-stop using GDBFLAGS").
Note that the use of
set GDBFLAGS "$GDBFLAGS ..."
instead of
append GDBFLAGS "..."
is intentional. "append" is silent when appending to a non-existent
variable. So if this code if moved to a proc (as is the case already
for step-sw-breakpoint-adjust-pc.exp) and we forget to add "global
GDBFLAGS", the flag won't be added to the global GDBFLAGS, and we won't
actually enable non-stop, and it might go unnoticed. Using the "set"
version will turn into an error if we forget the "global".
This makes these test work correctly with native-extended-gdbserver.
Some of them were silently failing because we runto_main is silent when
it fails.
H.J. Lu [Wed, 16 Dec 2020 21:40:42 +0000 (13:40 -0800)]
ld: Skip libdep plugin if not all plugin hooks are available
Skip plugin if not all required plugin hooks are available.
2020-12-16 Howard Chu <hyc@symas.com>
H.J. Lu <hongjiu.lu@intel.com>
PR ld/27081
* libdep_plugin.c (onload): Skip if not all required plugin hooks
are available.
* testsuite/config/default.exp (dep_plug_opt): New.
* testsuite/ld-elf/elf.exp: Pass $dep_plug_opt to nm.
* testsuite/ld-elf/pr26391.fd: New file.
Martin Liska [Wed, 16 Dec 2020 17:18:40 +0000 (18:18 +0100)]
[gdb] Print progress for debuginfod
Prints progress like:
Downloading 4.89 MB separate debug info for /usr/lib64/libgcrypt.so.20.
Downloading 1.10 MB separate debug info for /usr/lib64/liblzma.so.5.
Downloading 1.31 MB separate debug info for /usr/lib64/liblz4.so.1.
Downloading 0.96 MB separate debug info for /usr/lib64/libsmime3.so.
[### ]
Tested on x86_64-linux.
ChangeLog:
2020-12-16 Martin Liska <mliska@suse.cz>
Tom de Vries <tdevries@suse.de>
* gdb/debuginfod-support.c (struct user_data): Remove has_printed
field. Add meter field.
(progressfn): Print progress using meter.
Tom de Vries [Wed, 16 Dec 2020 17:18:40 +0000 (18:18 +0100)]
[gdb/testsuite] Fix shlib compilation with target board unix/-pie/-fPIE
When running test-case gdb.base/info-shared.exp with target board
unix/-pie/-fPIE, we run into:
...
spawn -ignore SIGHUP gcc -fno-stack-protector \
outputs/gdb.base/info-shared/info-shared-solib1.c.o \
-fdiagnostics-color=never -fPIC -shared -Wl,-soname,info-shared-solib1.so \
-lm -fPIE -pie -o outputs/gdb.base/info-shared/info-shared-solib1.so^M
ld: Scrt1.o: in function `_start':^M
start.S:104: undefined reference to `main'^M
collect2: error: ld returned 1 exit status^M
compiler exited with status 1
...
The intention of the -pie/-fPIE flags is to build and test PIE executables on
platforms where that is not the default. However, the flags clash with the
flags required to build shared libraries.
Fix this by filtering out PIE-related flags out of the multilib_flags settings
in compile_shared_lib.
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2020-12-16 Tom de Vries <tdevries@suse.de>
* lib/gdb.exp (gdb_compile_shlib_1): Factor out of ...
(gdb_compile_shlib): ... here. Filter out PIE-related flags.
Turns out we're not recording changes to the FPSR. The SIMD/FP data
instructions may set bits in the FPSR, so it needs to be recorded for
proper reverse operations.
gdb/ChangeLog:
2020-12-16 Luis Machado <luis.machado@linaro.org>
* aarch64-tdep.c (aarch64_record_data_proc_simd_fp): Record FPSR.
Luis Machado [Thu, 10 Dec 2020 19:51:20 +0000 (16:51 -0300)]
Fix TBI handling for watchpoints
When inserting hw watchpoints, we take care of masking off the top byte
of the address (and sign-extending it if needed). This guarantees we won't
pass tagged addresses to the kernel via ptrace.
However, from the kernel documentation on tagged pointers...
"Non-zero tags are not preserved when delivering signals. This means that
signal handlers in applications making use of tags cannot rely on the tag
information for user virtual addresses being maintained for fields inside
siginfo_t.
One exception to this rule is for signals raised in response to watchpoint
debug exceptions, where the tag information will be preserved."
So the stopped data address after a hw watchpoint hit can be potentially
tagged, and we don't handle this in GDB at the moment. This results in
GDB missing a hw watchpoint hit and attempting to step over an unsteppable
hw watchpoint, causing it to spin endlessly.
The following patch fixes this by adjusting the stopped data address and adds
some tests to expose the problem.
gdb/ChangeLog:
2020-12-16 Luis Machado <luis.machado@linaro.org>
* aarch64-linux-nat.c
(aarch64_linux_nat_target::stopped_data_address): Handle the TBI.
gdbserver/ChangeLog:
2020-12-16 Luis Machado <luis.machado@linaro.org>
* linux-aarch64-low.cc (address_significant): New function.
(aarch64_target::low_stopped_data_address): Handle the TBI.
gdb/testsuite/ChangeLog:
2020-12-16 Luis Machado <luis.machado@linaro.org>
* gdb.arch/aarch64-tagged-pointer.c (main): Add a few more
pointer-based memory accesses.
* gdb.arch/aarch64-tagged-pointer.exp: Exercise additional
hw watchpoint cases.
Alan Modra [Tue, 15 Dec 2020 14:12:21 +0000 (00:42 +1030)]
constify elfNN_bed
elfNN_bed was made writable as an expedient means of communicating
ld -z max-page-size and ld -z common-page-size values to BFD linker
code, and even for objcopy to communicate segment alignment between
copy_private_bfd_data, rewrite_elf_program_header and
assign_file_positions_for_load_sections. Some time later elfNN_bed
elf_osabi was written by gas. It turns out none of these
modifications to elfNN_bed was necessary, so make it const again.
include/
* bfdlink.h (struct bfd_link_info): Add maxpagesize and
commonpagesize.
bfd/
* elfxx-target.h (elfNN_bed): Constify.
* bfd.c (bfd_elf_set_pagesize): Delete.
(bfd_emul_set_maxpagesize, bfd_emul_set_commonpagesize): Delete.
* elf.c (get_program_header_size): Get commonpagesize from
link info.
(_bfd_elf_map_sections_to_segments): Get maxpagesize from link info.
(assign_file_positions_for_load_sections): Likewise.
(assign_file_positions_for_non_load_sections): Likewise.
(rewrite_elf_program_header): Add maxpagesize param. Set map_p_align.
(copy_private_bfd_data): Don't call bfd_elf_set_maxpagesize.
Instead pass maxpagesize to rewrite_elf_program_header.
* elf32-nds32.c (relax_range_measurement): Add link_info param.
Get maxpagesize from link_info. Adjust caller.
* bfd-in2.h: Regenerate.
gas/
* config/obj-elf.c (obj_elf_section): Don't set elf_osabi here.
(obj_elf_type): Likewise.
ld/
* ld.h (ld_config_type): Delete maxpagesize and commonpagesize.
* emultempl/elf.em: Use link_info rather than config
for maxpagesize and commonpagesize.
* emultempl/ppc32elf.em: Likewise.
* ldexp.c (fold_binary, fold_name): Likewise.
* ldemul.c (after_parse_default): Likewise.
(set_output_arch_default): Don't call bfd_emul_set_maxpagesize
or bfd_emul_set_commonpagesize.
Alan Modra [Tue, 15 Dec 2020 11:40:27 +0000 (22:10 +1030)]
XCOFF constify
There are occasions where it is reasonable to use a macro defining
function parameters, but this isn't one of them. Use typedefs
instead, which also simplifies declaring a const array of function
pointers.