The /FMT specification on the print command currently breaks command
completion, so:
(gdb) p var.<TAB><TAB>
.... list of fields in var .....
But,
(gdb) p/d var.<TAB><TAB>
..... list of all symbols .....
After this commit this issue is now resolved.
There are some other details around tab-completion and /FMT which
hopefully this commit improves. So, before:
(gdb) p/<TAB><TAB>
.... lists all symbols .....
After:
(gdb) p/<TAB><TAB> # Nothing changes...
The thinking here is that after a / the user must type a FMT, but we
don't offer tab completion on FMT characters. Placing a symbol
directly after a / will not do what the user expects, so offering that
seems wrong.
Similarly, before we had:
(gdb) p/d<TAB><TAB>
... lists all symbols starting with 'd' ....
But afterwards:
(gdb) p/d<TAB><TAB> # Adds a single space, so we get:
(gdb) p/d <CURSOR>
As before, typing a symbol where FMT is expected will not do what the
user expects. If the user has added a FMT string then upon tab
completion GDB assumes the FMT string is complete and prepares the
user to type an expression.
In this commit I have also added completion functions for the 'x' and
'display' commands. These commands also support /FMT specifiers and
so share some code with 'print'.
gdb/ChangeLog:
* printcmd.c: Include 'safe-ctype.c'.
(skip_over_slash_fmt): New function.
(print_command_completer): Call skip_over_slash_fmt.
(display_and_x_command_completer): New function.
(_initialize_printcmd): Add command completion for 'x' and
'display'.
gdb/testsuite/ChangeLog:
* gdb.base/completion.exp: Add new tests.
Extract FLAGM (Condition flag manipulation) feature from Armv8.4-A.
Please note that FLAGM stays a Armv8.4-A feature but now can be
assigned to other architectures or CPUs.
New -march option +flagm is added to enable independently this
feature.
Tom de Vries pointed out that the series to handle
-fgnat-encodings=minimal regressed data-loc.exp:
https://sourceware.org/pipermail/gdb-patches/2020-November/173035.html
This was my oversight. Looking at the test, it mimics what GNAT
emits. The patch series in question changed the approach that gdb
takes to such arrays -- now they are transformed back into "old" style
structures. (This is perhaps a step backward in a semantics sense,
but on the other hand it more accurately reflects the underlying
reality, and lets gdb continue to create objects of this type.)
This patch updates the test case, by adjusting the expected output.
It also removes the tests done in the C language, because they are now
incorrect, and because (IMO) they don't add much value.
gdb/testsuite/ChangeLog
2020-11-16 Tom Tromey <tromey@adacore.com>
* gdb.dwarf2/data-loc.exp: Update expected output. Remove C
tests.
The recent commit to make scoped_restore_current_thread's cdtors
exception free regressed gdb.base/eh_return.exp:
Breakpoint 1, 0x00000000004012bb in eh2 (gdb/frame.c:641: internal-error: frame_id get_frame_id(frame_info*): Assertion `stashed' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) FAIL: gdb.base/eh_return.exp: hit breakpoint (GDB internal error)
That testcase uses __builtin_eh_return and, before the regression, the
backtrace at eh2 looked like this:
(gdb) bt
#0 0x00000000004006eb in eh2 (p=0x4006ec <continuation>) at src/gdb/testsuite/gdb.base/eh_return.c:54
Backtrace stopped: previous frame identical to this frame (corrupt stack?)
That "previous frame identical to this frame" is caught by the cycle
detection based on frame id.
The assertion failing is this one:
638 /* Since this is the first frame in the chain, this should
639 always succeed. */
640 bool stashed = frame_stash_add (fi);
641 gdb_assert (stashed);
originally added by
commit f245535cf583ae4ca13b10d47b3c7d3334593ece
Author: Pedro Alves <palves@redhat.com>
AuthorDate: Mon Sep 5 18:41:38 2016 +0100
Fix PR19927: Avoid unwinder recursion if sniffer uses calls parse_and_eval
The assertion is failing because frame #1's frame id was stashed
before the id of frame #0 is stashed. The frame id of frame #1 was
stashed here:
(top-gdb) bt
#0 frame_stash_add (frame=0x1e24c90) at src/gdb/frame.c:276
#1 0x0000000000669c1b in get_prev_frame_if_no_cycle (this_frame=0x19f8370) at src/gdb/frame.c:2120
#2 0x000000000066a339 in get_prev_frame_always_1 (this_frame=0x19f8370) at src/gdb/frame.c:2303
#3 0x000000000066a360 in get_prev_frame_always (this_frame=0x19f8370) at src/gdb/frame.c:2319
#4 0x000000000066b56c in get_frame_unwind_stop_reason (frame=0x19f8370) at src/gdb/frame.c:3028
#5 0x000000000059f929 in dwarf2_frame_cfa (this_frame=0x19f8370) at src/gdb/dwarf2/frame.c:1462
#6 0x00000000005ce434 in dwarf_evaluate_loc_desc::get_frame_cfa (this=0x7fffffffc070) at src/gdb/dwarf2/loc.c:666
#7 0x00000000005989a9 in dwarf_expr_context::execute_stack_op (this=0x7fffffffc070, op_ptr=0x1b2a053 "\364\003", op_end=0x1b2a053 "\364\003") at src/gdb/dwarf2/expr.c:1161
#8 0x0000000000596af6 in dwarf_expr_context::eval (this=0x7fffffffc070, addr=0x1b2a052 "\234\364\003", len=1) at src/gdb/dwarf2/expr.c:303
#9 0x0000000000597b4e in dwarf_expr_context::execute_stack_op (this=0x7fffffffc070, op_ptr=0x1b2a063 "", op_end=0x1b2a063 "") at src/gdb/dwarf2/expr.c:865
#10 0x0000000000596af6 in dwarf_expr_context::eval (this=0x7fffffffc070, addr=0x1b2a061 "\221X", len=2) at src/gdb/dwarf2/expr.c:303
#11 0x00000000005c8b5a in dwarf2_evaluate_loc_desc_full (type=0x1b564d0, frame=0x19f8370, data=0x1b2a061 "\221X", size=2, per_cu=0x1b28760, per_objfile=0x1a84930, subobj_type=0x1b564d0, subobj_byte_offset=0) at src/gdb/dwarf2/loc.c:2260
#12 0x00000000005c9243 in dwarf2_evaluate_loc_desc (type=0x1b564d0, frame=0x19f8370, data=0x1b2a061 "\221X", size=2, per_cu=0x1b28760, per_objfile=0x1a84930) at src/gdb/dwarf2/loc.c:2444
#13 0x00000000005cb769 in locexpr_read_variable (symbol=0x1b59840, frame=0x19f8370) at src/gdb/dwarf2/loc.c:3687
#14 0x0000000000663137 in language_defn::read_var_value (this=0x122ea60 <c_language_defn>, var=0x1b59840, var_block=0x0, frame=0x19f8370) at src/gdb/findvar.c:618
#15 0x0000000000663c3b in read_var_value (var=0x1b59840, var_block=0x0, frame=0x19f8370) at src/gdb/findvar.c:822
#16 0x00000000008c7d9f in read_frame_arg (fp_opts=..., sym=0x1b59840, frame=0x19f8370, argp=0x7fffffffc470, entryargp=0x7fffffffc490) at src/gdb/stack.c:542
#17 0x00000000008c89cd in print_frame_args (fp_opts=..., func=0x1b597c0, frame=0x19f8370, num=-1, stream=0x1aba860) at src/gdb/stack.c:890
#18 0x00000000008c9bf8 in print_frame (fp_opts=..., frame=0x19f8370, print_level=0, print_what=SRC_AND_LOC, print_args=1, sal=...) at src/gdb/stack.c:1394
#19 0x00000000008c92b9 in print_frame_info (fp_opts=..., frame=0x19f8370, print_level=0, print_what=SRC_AND_LOC, print_args=1, set_current_sal=1) at src/gdb/stack.c:1119
#20 0x00000000008c75f0 in print_stack_frame (frame=0x19f8370, print_level=0, print_what=SRC_AND_LOC, set_current_sal=1) at src/gdb/stack.c:366
#21 0x000000000070250b in print_stop_location (ws=0x7fffffffc9e0) at src/gdb/infrun.c:8110
#22 0x0000000000702569 in print_stop_event (uiout=0x1a8b9e0, displays=true) at src/gdb/infrun.c:8126
#23 0x000000000096d04b in tui_on_normal_stop (bs=0x1bcd1c0, print_frame=1) at src/gdb/tui/tui-interp.c:98
...
Before the commit to make scoped_restore_current_thread's cdtors
exception free, scoped_restore_current_thread's dtor would call
get_frame_id on the selected frame, and we use
scoped_restore_current_thread pervasively. That had the side effect
of stashing the frame id of frame #0 before reaching the path shown in
the backtrace. I.e., the frame id of frame #0 happened to be stashed
before the frame id of frame #1. But that was by chance, not by
design.
This commit:
commit 256ae5dbc73d1348850f86ee77a0dc3b04bc7cc0
Author: Kevin Buettner <kevinb@redhat.com>
AuthorDate: Mon Oct 31 12:47:42 2016 -0700
Stash frame id of current frame before stashing frame id for previous frame
Fixed a similar problem, by making sure get_prev_frame computes the
frame id of the current frame before unwinding the previous frame, so
that the cycle detection works properly. That fix misses the scenario
we're now running against, because if you notice, the backtrace above
shows that frame #4 calls get_prev_frame_always, not get_prev_frame.
I.e., nothing is calling get_frame_id on the current frame.
The fix here is to move Kevin's fix down from get_prev_frame to
get_prev_frame_always. Or actually, a bit further down to
get_prev_frame_always_1 -- note that inline_frame_this_id calls
get_prev_frame_always, so we need to be careful to avoid recursion in
that scenario.
gdb/ChangeLog:
* frame.c (get_prev_frame): Move get_frame_id call from here ...
(get_prev_frame_always_1): ... to here.
* inline-frame.c (inline_frame_this_id): Mention
get_prev_frame_always_1 in comment.
Change-Id: Id960c98ab2d072c48a436c3eb160cc4b2a5cfd1d
The purpose of the test is to look for entries of "foo" and
"bar" in the ".got". The old "objdump -s ..." has been replaced
with "objdump -D ..." to inspect the final executable. A sample
output looks like:
------------------------8<------------------------
$ arc-elf32-objdump -D -j .got tls_ie-01
tls_ie-01: file format elf32-littlearc
Disassembly of section .got:
00002110 <_GLOBAL_OFFSET_TABLE_>:
...
211c: 08 00 00 00 .word 0x00000008
2120: 0c 00 00 00 .word 0x0000000c
------------------------>8------------------------
ld/
* testsuite/ld-arc/tls_ie-01.d: Update the dump command.
This patch adds support for binary comparison operators with
fixed-point type values.
gdb/ChangeLog:
* valarith.c (fixed_point_binop): Add BINOP_EQUAL and BINOP_LESS
handling.
(value_less): Add fixed-point handling.
gdb/testsuite/ChangeLog:
* gdb.ada/fixed_cmp.exp: Add -fgnat-encodings=minimal testing.
* gdb.dwarf2/dw2-fixed-point.c (pck__fp1_var2): New global.
(main): Add reference to pck__fp1_var2.
* gdb.dwarf2/dw2-fixed-point.exp: Add comparison operator testing.
This patch adds support for binary operations on fixed-point values,
as well as for the negative unary operator.
gdb/ChangeLog:
* eval.c (binop_promote): Add fixed-point type handling.
* valarith.c (fixed_point_binop): New function.
(scalar_binop): Add fixed-point type handling.
(value_neg): Add fixed-point type handling.
* valops.c (value_cast_to_fixed_point): New function.
(value_cast): Add fixed-point type handling.
gdb/testsuite/ChangeLog:
* gdb.dwarf2/dw2-fixed-point.exp: Add arithmetic tests.
Note that the ptype information printed for types described
via pure DWARF debug info is slightly less informative as
the one printed when the information is encoded in the type's
name, via the GNAT encoding. As a result, the output in
the case of DWARF-described fixed point types is slightly
different. In pratice, this is no real loss because the information
not available in DWARF has no bearing on how the type is actually
stored in memory.
gdb/ChangeLog:
* ada-typeprint.c (ada_print_type): Add handing of fixed-point
range types.
* c-typeprint.c (c_type_print_varspec_prefix)
(c_type_print_varspec_suffix, c_type_print_base_1): Add
TYPE_CODE_FIXED_POINT handling.
* p-typeprint.c (pascal_type_print_varspec_prefix)
(pascal_type_print_varspec_suffix): Likewise.
* typeprint.c (print_type_fixed_point): New function.
* typeprint.h (print_type_fixed_point): Add declaration.
gdb/testsuite/ChangeLog:
* gdb.ada/fixed_points.exp: Add ptype tests.
* gdb.dwarf2/dw2-fixed-point.exp: Likewise.
Consider a fixed-point type such the scaling factor is 1/16,
as the following Ada code snippet would create:
type FP1_Type is delta 0.1 range -1.0 .. +1.0;
FP1_Var : FP1_Type := 0.25;
Printing the value of this variable with a format modifier yields
the wrong value. E.g.:
(gdb) p /x fp1_var
$6 = 0x4
Since the real value is 0.25, we therefore expected...
(gdb) p /x fp1_var
$6 = 0x0
What happens, in this case, is that the value being printed is
actually the "raw" value of our object, before the scaling factor
gets applied.
This commit fixes the issue by using approach as for float values,
where we convert the value into an integer value, prior to printing,
knowing that the conversion takes the scaling factor into account.
gdb/ChangeLog:
* printcmd.c (print_scalar_formatted): Add fixed-point type
handling when options->format is set.
gdb/testsuite/ChangeLog:
* gdb.dwarf2/dw2-fixed-point.exp: Add "print /x" tests.
This commit introduces a new kind of type, meant to describe
fixed-point types, using a new code added specifically for
this purpose (TYPE_CODE_FIXED_POINT).
It then adds handling of fixed-point base types in the DWARF reader.
And finally, as a first step, this commit adds support for printing
the value of fixed-point type objects.
Note that this commit has a known issue: Trying to print the value
of a fixed-point object with a format letter (e.g. "print /x NAME")
causes the wrong value to be printed because the scaling factor
is not applied. Since the fix for this issue is isolated, and
this is not a regression, the fix will be made in a pach of its own.
This is meant to simplify review and archeology.
Also, other functionalities related to fixed-point type handling
(ptype, arithmetics, etc), will be added piecemeal as well, for
the same reasons (faciliate reviews and archeology). Related to this,
the testcase gdb.ada/fixed_cmp.exp is adjusted to compile the test
program with -fgnat-encodings=all, so as to force the use of GNAT
encodings, rather than rely on the compiler's default to use them.
The intent is to enhance this testcase to also test the pure DWARF
approach using -fgnat-encodings=minimal as soon as the corresponding
suport gets added in. Thus, the modification to the testcase is made
in a way that it prepares this testcase to be tested in both modes.
gdb/ChangeLog:
* ada-valprint.c (ada_value_print_1): Add fixed-point type handling.
* dwarf2/read.c (get_dwarf2_rational_constant)
(get_dwarf2_unsigned_rational_constant, finish_fixed_point_type)
(has_zero_over_zero_small_attribute): New functions.
read_base_type, set_die_type): Add fixed-point type handling.
* gdb-gdb.py.in: Add fixed-point type handling.
* gdbtypes.c: #include "gmp-utils.h".
(create_range_type, set_type_code): Add fixed-point type handling.
(init_fixed_point_type): New function.
(is_integral_type, is_scalar_type): Add fixed-point type handling.
(print_fixed_point_type_info): New function.
(recursive_dump_type, copy_type_recursive): Add fixed-point type
handling.
(fixed_point_type_storage): New typedef.
(fixed_point_objfile_key): New static global.
(allocate_fixed_point_type_info, is_fixed_point_type): New functions.
(fixed_point_type_base_type, fixed_point_scaling_factor): New
functions.
* gdbtypes.h: #include "gmp-utils.h".
(enum type_code) <TYPE_SPECIFIC_FIXED_POINT>: New enum.
(union type_specific) <fixed_point_info>: New field.
(struct fixed_point_type_info): New struct.
(INIT_FIXED_POINT_SPECIFIC, TYPE_FIXED_POINT_INFO): New macros.
(init_fixed_point_type, is_fixed_point_type)
(fixed_point_type_base_type, fixed_point_scaling_factor)
(allocate_fixed_point_type_info): Add declarations.
* valprint.c (generic_val_print_fixed_point): New function.
(generic_value_print): Add fixed-point type handling.
* value.c (value_as_address, unpack_long): Add fixed-point type
handling.
gdb/testsuite/ChangeLog:
* gdb.ada/fixed_cmp.exp: Force compilation to use -fgnat-encodings=all.
* gdb.ada/fixed_points.exp: Add fixed-point variables printing tests.
* gdb.ada/fixed_points/pck.ads, gdb.ada/fixed_points/pck.adb:
New files.
* gdb.ada/fixed_points/fixed_points.adb: Add use of package Pck.
* gdb.dwarf2/dw2-fixed-point.c, gdb.dwarf2/dw2-fixed-point.exp:
New files.
This is a generic function which I would like to use in a followup
patch adding support for fixed-point types. So this commit moves it
out of valarith.c into util.c, and makes it non-static.
gdb/ChangeLog:
* utils.h (uinteger_pow): Add declaration.
* utils.c (uinteger_pow): Moved here (without changes)...
* valarith.c (uinteger_pow): ... from here.
This API was motivated by a number of reasons:
- GMP's API does not handle "long long" and "unsigned long long",
so using LONGEST and ULONGEST is not straightforward;
- Automate the need to initialize GMP objects before use, and
clear them when no longer used.
However, this API grew also to help with similar matter such
as formatting to a string, and also reading/writing fixed-point
values from byte buffers.
Dedicated unit testing is also added.
gdb/ChangeLog:
* gmp-utils.h, gmp-utils.h: New file.
* unittests/gmp-utils-selftests.c: New file.
* Makefile.in (SUBDIR_UNITTESTS_SRCS): Add
unittests/gmp-utils-selftests.c.
(COMMON_SFILES) Add gmp-utils.c.
(HFILES_NO_SRCDIR): Add gmp-utils.h.
This commit modifies gdb's configure script to trigger an error
if we cannot find a usable libgmp.
For the record, making this a requirement was discussed in March 2018:
https://sourceware.org/pipermail/gdb-patches/2018-March/147373.html
gdb/ChangeLog:
* configure.ac: Generate an error if a usable GMP library
could not be found.
* configure: Regenerate.
This patch allows a user to tell gdb's configure script where
his GMP library is installed.
gdb/ChangeLog:
* configure.ac: Add support for --with-libgmp-prefix.
* Makefile.in (LIBGMP): New variable.
(CLIBS): Include $(LIBGMP).
* configure, config.in: Regenerate
Add support for tab-completion on Fortran field names. Consider this
test case:
program test
type my_type
integer :: field_a
integer :: other_field
integer :: last_field
end type my_type
type(my_type) :: var
print *, var
end program test
And the GDB session before this patch:
(gdb) start
...
(gdb) p var% <- Trigger TAB completion here.
Display all 200 possibilities? (y or n) n
(gdb) p var%
And the GDB session with this patch:
(gdb) start
...
(gdb) p var% <- Trigger TAB completion here.
field_a last_field other_field
(gdb) p var%
The implementation for this is basically copied from c-exp.y, I
tweaked the parser patterns to be appropriate for Fortran, and it
"just worked".
gdb/ChangeLog:
PR cli/26879
* f-exp.y (COMPLETE): New token.
(exp): Two new rules for tab-completion.
(saw_name_at_eof): New static global.
(last_was_structop): Likewise.
(yylex): Set new variables, and return COMPLETE token at the end
of the input stream in some cases.
gdb/testsuite/ChangeLog:
PR cli/26879
* gdb.fortran/completion.exp: New file.
* gdb.fortran/completion.f90: New file.
A while back I noticed that fetch_inferior_event used "int" for
should_stop, whereas it can be bool. The method it is assigned from:
should_stop = thread_fsm->should_stop (thr);
... already returns bool.
Tested by rebuilding.
gdb/ChangeLog
2020-11-14 Tom Tromey <tom@tromey.com>
* infrun.c (fetch_inferior_event): Use "bool" for should_stop.
I found another expression-related spot that could use constification.
This patch adds it. Tested by rebuilding.
gdb/ChangeLog
2020-11-14 Tom Tromey <tom@tromey.com>
* opencl-lang.c (opencl_component_ref): Make "comps" const.
This changes a few functions in c-lang.c to add "const" to parameters
and return types. Tested by rebuilding.
gdb/ChangeLog
2020-11-13 Tom Tromey <tom@tromey.com>
* c-lang.c (convert_ucn, convert_octal, convert_hex)
(convert_escape, parse_one_string): Constify.
bfdlink.h has
/* Symbol will be converted from absolute to section-relative. Set for
symbols defined by a script from "dot" (also SEGMENT_START or ORIGIN)
outside of an output section statement. */
unsigned int rel_from_abs : 1;
linker.c has
.{* Return TRUE if the symbol described by a linker hash entry H
. is going to be absolute. Linker-script defined symbols can be
. converted from absolute to section-relative ones late in the
. link. Use this macro to correctly determine whether the symbol
. will actually end up absolute in output. *}
.#define bfd_is_abs_symbol(H) \
. (((H)->type == bfd_link_hash_defined \
. || (H)->type == bfd_link_hash_defweak) \
. && bfd_is_abs_section ((H)->u.def.section) \
. && !(H)->rel_from_abs)
.
Set rel_from_abs to 1 for __ehdr_start which will be converted from
absolute to section-relative in assign_file_positions_for_load_sections.
PR ld/26869
* ldelf.c (ldelf_before_allocation): Set rel_from_abs to 1 for
__ehdr_start.
* testsuite/ld-i386/i386.exp: Run pr26869.
* testsuite/ld-i386/pr26869.d: New file.
* testsuite/ld-i386/pr26869.s: Likewise.
When GDB loads an ELF file, it will warn when a section is not located
in an ELF segment:
$ ./gdb -q -iex "set build-id-verbose 0" --ex "b systemctl_main" -ex "r" -batch --args systemctl kexec
Breakpoint 1 at 0xc24d: file ../src/systemctl/systemctl.c, line 8752.
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libgcc_s.so.1
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libcap.so.2
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libacl.so.1
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libcryptsetup.so.12
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libgcrypt.so.20
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libip4tc.so.2
[snip]
This has feature has also been reported by various users, most notably
the Fedora-EOL'd bug 1553086.
Mark Wielaard explains the issue quite nicely in
https://sourceware.org/bugzilla/show_bug.cgi?id=24717#c2
The short of it is, the ELF program headers for debuginfo files are
not suited to this particular use case. Consequently, the warning
generated above really is useless and should be ignored.
This patch follows the same heuristic that BFD itself uses.
gdb/ChangeLog
2020-11-13 Keith Seitz <keiths@redhat.com>
https://bugzilla.redhat.com/show_bug.cgi?id=1553086
* elfread.c (elf_symfile_segments): Omit "Loadable section ...
outside of ELF segments" warning for debugin
As requested in gdb/23034, I would like to extend the warning message
GDB displays when it detects a loadable section that is outside any
ELF segment.
Before:
$ gdb -q --ex "b systemctl_main" -ex "r" -batch --args systemctl kexec
Breakpoint 1 at 0xc24d: file ../src/systemctl/systemctl.c, line 8752.
warning: Loadable section ".note.gnu.property" outside of ELF segments
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
warning: Loadable section ".note.gnu.property" outside of ELF segments
warning: Loadable section ".note.gnu.property" outside of ELF segments
warning: Loadable section ".note.gnu.property" outside of ELF segments
warning: Loadable section ".note.gnu.property" outside of ELF segments
[snip]
Breakpoint 1, systemctl_main (argv=0x7fffffffd348, argc=2) at ../src/systemctl/systemctl.c:8752
8752 r = systemctl_main(argc, argv);
After:
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libcap.so.2
warning: Loadable section ".note.gnu.property" outside of ELF segments
in .gnu_debugdata for /lib64/libacl.so.1
[snip]
I think this is eminently more useful output.
gdb/ChangeLog
2020-11-13 Keith Seitz <keiths@redhat.com>
PR gdb/23034
* elfread.c (elf_symfile_segments): Output a BFD file name
for the "Loadable section ... outside of ELF segments" warning.
When loading the code file provided in PR 26828 and GDB is build with
UBSan, we get:
Core was generated by `./Foo'.
Program terminated with signal SIGABRT, Aborted.
#0 0xb6c3809c in pthread_cond_wait () from /home/simark/build/binutils-gdb/gdb/repo/lib/libpthread.so.0
[Current thread is 1 (LWP 29367)]
(gdb) bt
/home/simark/src/binutils-gdb/gdb/arm-tdep.c:1551:30: runtime error: shift exponent 32 is too large for 32-bit type 'unsigned int'
The sequence of instructions at pthread_cond_wait, in the
libpthread.so.0 library, contains this instruction with an immediate
constant with a "rotate amount" of 0:
e24dd044 sub sp, sp, #68 ; 0x44
Since arm_analyze_prologue shifts by "32 - rotate amount", it does a 32
bit shift of a 32 bit type, which is caught by UBSan.
Fix it by factoring out the decoding of immediates in a new function,
arm_expand_immediate.
I added a selftest for arm_analyze_prologue that replicates the
instruction sequence. Without the fix, it crashes GDB if it is build
with --enable-ubsan.
I initially wanted to re-use the abstract_memory_reader class already in
arm-tdep.c, used to make arm_process_record testable. However,
arm_process_record and arm_analyze_prologue don't use the same kind of
memory reading functions. arm_process_record uses a function that
returns an error status on failure while arm_analyze_prologue uses one
that throws an exception. Since i didn't want to introduce any other
behavior change, I decided to just introduce a separate interface
(arm_instruction_reader). It is derived from
abstract_instruction_reader in aarch64-tdep.c.
gdb/ChangeLog:
PR gdb/26835
* arm-tdep.c (class arm_instruction_reader): New.
(target_arm_instruction_reader): New.
(arm_analyze_prologue): Add instruction reader parameter and use
it. Use arm_expand_immediate.
(class target_arm_instruction_reader): Adjust.
(arm_skip_prologue): Adjust.
(arm_expand_immediate): New.
(arm_scan_prologue): Adjust.
(arm_analyze_prologue_test): New.
(class test_arm_instruction_reader): New.
Change-Id: Ieb1c1799bd66f8c7421384f44f5c2777b578ff8d
GCC for Windows target produces executables called foo.exe when given
"-o foo". (More specifically, it's done that for native compilers for
a long time, and for cross compilers to Windows target since GCC
commit 5bc86b599054f494ec0a45e49b82749320eaa9c4, in GCC 8 and later.)
This causes problems for many GDB tests expecting a program to have
the exact file name passed to -o.
Fix this by checking for the case where only the .exe exists in
gdb_file_cmd and adjusting the name passed to the file command
accordingly. There may well be other places with this issue in the
GDB testsuite, but this fix allows many tests to succeed that
previously fell over.
2020-11-12 Joseph Myers <joseph@codesourcery.com>
* lib/gdb.exp (gdb_file_cmd): Check for case where $arg.exe exists
but $arg does not.
PR 26829
* dwarf.c (struct dwo_info): Add cu_offset field.
(add_dwo_info): Add cu_offset parameter. Record in new dwo_info
struct.
(add_dwo_name): Add cu_offset field.
(add_dwo_dir): Add cu_offset field.
(add_dwo_id): Add cu_offset field.
(read_and_display_attr_value): Pass cu_offset to dwo recording
functions.
(load_separate_debug_files): Accumulate name, dir and id values
and display once for each CU.
* testsuite/binutils-all/dwo.sL Use a separate CU for the second
dwo link.
* testsuite/binutils-all/readelf.k2: Update expected output.
One declaration in f-lang.h is for a function that doesn't even exist,
another is for a function that is only used within f-lang.c.
One declaration is deleted, the other function I make static in
f-lang.c.
gdb/ChangeLog:
* f-lang.c (fortran_argument_convert): Add declaration. Add
header comment, taken from f-lang.h. Make static.
* f-lang.h (f77_get_dynamic_array_length): Delete declaration.
(fortran_argument_convert): Delete declaration.
This patch fixes a mistake when enabling MVE instructions that disabled support
for single precision vmla and vmul for arm mode.
gas/ChangeLog:
2020-11-12 Andre Vieira <andre.simoesdiasvieira@arm.com>
PR 26858
* config/tc-arm.c (asm_opcode insns): Fix vmul and vmla's ARM_VARIANT.
* testsuite/gas/arm/pr26858.s: New test.
* testsuite/gas/arm/pr26858.d: New test.
Consider the following GDB session:
$ gdb
(gdb) set language c
(gdb) ptype void
type = void
(gdb) set language fortran
(gdb) ptype void
No symbol table is loaded. Use the "file" command.
(gdb)
With no symbol file loaded GDB and the language set to C GDB knows
about the type void, while when the language is set to Fortran GDB
doesn't know about the void, why is that?
In f-lang.c, f_language::language_arch_info, we do have this line:
lai->primitive_type_vector [f_primitive_type_void]
= builtin->builtin_void;
where we add the void type to the list of primitive types that GDB
should always know about, so what's going wrong?
It turns out that the primitive types are stored in a C style array,
indexed by an enum, so Fortran uses `enum f_primitive_types'. The
array is allocated and populated in each languages language_arch_info
member function. The array is allocated with an extra entry at the
end which is left as a NULL value, and this indicates the end of the
array of types.
Unfortunately for Fortran, a type is not assigned for each element in
the enum. As a result the final populated array has gaps in it, gaps
which are initialised to NULL, and so every time we iterate over the
list (for Fortran) we stop early, and never reach the void type.
This has been the case since 2007 when this functionality was added to
GDB in commit cad351d11d6c3f6487cd.
Obviously I could just fix Fortran by ensuring that either the enum is
trimmed, or we create types for the missing types. However, I think a
better approach would be to move to C++ data structures and removed
the fixed enum indexing into the array approach.
After this commit the primitive types are pushed into a vector, and
GDB just iterates over the vector in the obvious way when it needs to
hunt for a type. After this commit all the currently defined
primitive types can be found when the language is set to Fortran, for
example:
$ gdb
(gdb) set language fortran
(gdb) ptype void
type = void
(gdb)
A new test checks this functionality.
I didn't see any other languages with similar issues, but I could have
missed something.
gdb/ChangeLog:
* ada-exp.y (find_primitive_type): Make parameter const.
* ada-lang.c (enum ada_primitive_types): Delete.
(ada_language::language_arch_info): Update.
* c-lang.c (enum c_primitive_types): Delete.
(c_language_arch_info): Update.
(enum cplus_primitive_types): Delete.
(cplus_language::language_arch_info): Update.
* d-lang.c (enum d_primitive_types): Delete.
(d_language::language_arch_info): Update.
* f-lang.c (enum f_primitive_types): Delete.
(f_language::language_arch_info): Update.
* go-lang.c (enum go_primitive_types): Delete.
(go_language::language_arch_info): Update.
* language.c (auto_or_unknown_language::language_arch_info):
Update.
(language_gdbarch_post_init): Use obstack_new, use array indexing.
(language_string_char_type): Add header comment, call function in
language_arch_info.
(language_bool_type): Likewise
(language_arch_info::bool_type): Define.
(language_lookup_primitive_type_1): Delete.
(language_lookup_primitive_type): Rewrite as a templated function
to call function in language_arch_info, then instantiate twice.
(language_arch_info::type_and_symbol::alloc_type_symbol): Define.
(language_arch_info::lookup_primitive_type_and_symbol): Define.
(language_arch_info::lookup_primitive_type): Define twice with
different signatures.
(language_arch_info::lookup_primitive_type_as_symbol): Define.
(language_lookup_primitive_type_as_symbol): Rewrite to call a
member function in language_arch_info.
* language.h (language_arch_info): Complete rewrite.
(language_lookup_primitive_type): Make templated.
* m2-lang.c (enum m2_primitive_types): Delete.
(m2_language::language_arch_info): Update.
* opencl-lang.c (OCL_P_TYPE): Delete.
(enum opencl_primitive_types): Delete.
(opencl_type_data): Delete.
(builtin_opencl_type): Delete.
(lookup_opencl_vector_type): Update.
(opencl_language::language_arch_info): Update, lots of content
moved from...
(build_opencl_types): ...here. This function is now deleted.
(_initialize_opencl_language): Delete.
* p-lang.c (enum pascal_primitive_types): Delete.
(pascal_language::language_arch_info): Update.
* rust-lang.c (enum rust_primitive_types): Delete.
(rust_language::language_arch_info): Update.
gdb/testsuite/ChangeLog:
* gdb.fortran/types.exp: Add more tests.
It took me a while to understand why that would even compile: it looks
like we pass a type name as a pointer, that makes no sense. By looking
at the DWARF, I understood that the compiler actually interprets it as a
function declaration. So the statement was doing nothing, no
dwarf2_queue_guard was instantiated. Fix it by passing the right
variable name.
gdb/ChangeLog:
* dwarf2/read.c (dw2_do_instantiate_symtab): Fix call to
dwarf2_queue_guard.
Change-Id: I3a7bdead9e8c39f8342a471f10181b85b8f0d801
Binutils support for LMBD instruction was merged [1]. So add it also
to simulator.
LMBD instruction does left-most-bit-detection. It returns 32 if
the given bit value is not found in the provided word value.
[1] https://sourceware.org/pipermail/binutils/2020-October/113901.html
sim/pru/ChangeLog:
* pru.h (RS1SEL): New macro.
(RS1_WIDTH): New macro.
* pru.isa: Describe the LMBD instruction.
sim/testsuite/sim/pru/ChangeLog:
* lmbd.s: New test.
Add dwarf_read_debug_printf and dwarf_read_debug_printf_v macros and use
them throughout dwarf2/read.c. The second one is used for "verbose"
prints, when the value of "set debug dwarf-read" is >= 2.
gdb/ChangeLog:
* dwarf2/read.c (dwarf_read_debug_printf,
dwarf_read_debug_printf_v): New macros, use throughout the file.
Change-Id: I694da69da2e1f2caa4c27a421a975790636411e2
The redundant -md option was removed in e4ae357fe8, but it is required
for backwards compatibility with GCC 10, which passes it to the
assembler implicitly in certain situations.
It is now silently ignored.
gas/ChangeLog:
* config/tc-msp430.c (OPTION_MOVE_DATA): Define.
(md_parse_option): Ignore OPTION_MOVE_DATA.
(md_longopts): Handle -md option.
* testsuite/gas/msp430/msp430.exp: Run new test.
* testsuite/gas/msp430/empty.s: New test.
* testsuite/gas/msp430/ignore-md.d: New test.
In collect_register() function of arc-linux-tdep.c, the "eret"
(exception return) register value was not being reported correctly.
This patch fixes that.
Background:
When asked for the "pc" value, we have to update the "eret" register
with GDB's STOP_PC. The "eret" instructs the kernel code where to
jump back when an instruction has stopped due to a breakpoint. This
is how collect_register() was doing so:
--------------8<--------------
if (regnum == gdbarch_pc_regnum (gdbarch))
regnum = ARC_ERET_REGNUM;
regcache->raw_collect (regnum, buf + arc_linux_core_reg_offsets[regnum]);
-------------->8--------------
Root cause:
Although this is using the correct offset (ERET register's), it is also
changing the REGNUM itself. Therefore, raw_collect (regnum, ...) is
not reading from "pc" anymore.
v2:
- Fix a copy/paste issue as rightfully addressed by Tom [1].
[1]
https://sourceware.org/pipermail/gdb-patches/2020-November/173208.html
gdb/ChangeLog:
* arc-linux-tdep.c (collect_register): Populate "eret" by
"pc" value from the regcache when asked for "pc" value.
PR rust/26799 points out that a certain test case fails with -readnow.
This happens because, with -readnow, there are no partial symtabs; but
find_symbol_at_address requires these.
This patch fixes this problem by searching all of an objfile's
compunit symtabs if it does not have partial symbols.
Note that this test will still fail with .gdb_index. I don't think
that is readily fixable.
gdb/ChangeLog
2020-11-12 Tom Tromey <tom@tromey.com>
PR rust/26799:
* symtab.c (find_symbol_at_address): Search symtabs if no psymtabs
exist.
gdb/testsuite/ChangeLog
2020-11-12 Tom Tromey <tom@tromey.com>
PR rust/26799:
* gdb.rust/traits.exp: Remove kfails.
Clang fails to compile gdb.threads/tls-so_extern_main.c, giving the
following error:
/gdbtest/src/gdb/testsuite/gdb.threads/tls-so_extern_main.c:28:1:
warning: non-void function does not return a value [-Wreturn-type]
This commit adds a return statement to the offending function.
gdb/testsuite/ChangeLog:
* gdb.threads/tls-so_extern_main.c (tls_ptr): Add missing return
statement.
* m32r-sim.h (m32rbf_h_accum_get_handler): Always provide a
prototype for this function.
(m32rbf_h_accum_set_handler): Likewise.
(m32r2f_h_accums_get_handler): Prototype.
(m32r2f_h_accums_set_handler): Prototype.
PR 26850
* dwarf2dbg.c (dwarf2_gen_line_info_1): Do not record lines in
sections that are not executable or not loadable.
(out_debug_line): Move warning message into dwarf2_gen_line_info_1.
* testsuite/gas/elf/dwarf2-20.s: New test.
* testsuite/gas/elf/dwarf2-20.d: New test driver.
* testsuite/gas/elf/elf.exp: Run the new test.
* testsuite/gas/elf/warn-2.s: Use the .nop directive.
GDB has two approaches to generating the target descriptions found in
gdb/features/, the whole description approach, where the XML file
contains a complete target description which is then used to generate
a single C file that builds that target description. Or, the split
feature approach, where the XML files contain a single target feature,
each feature results in a single C file to create that one feature,
and then a manually written C file is used to build a complete target
description from individual features.
There's a Makefile, gdb/features/Makefile, which is responsible for
managing the regeneration of the C files from the XML files.
However, some of the logic that selects between the whole description
approach, or the split feature approach, is actually hard-coded into
GDB, inside target-descriptions.c:maint_print_c_tdesc_cmd we check the
path to the incoming XML file and use this to choose which type of C
file we should generate.
This commit removes this hard coding from GDB, and makes the Makefile
entirely responsible for choosing the approach. This makes sense as
the Makefile already has the XML files partitioned based on which
approach they should use.
In order to allow this change the 'maint print c-tdesc' command is
given a new command option '-single-feature', which tells GDB which
type of C file should be created. The makefile now supplies this flag
to GDB.
This did reveal a bug in features/Makefile, the rx.xml file was in the
wrong list, this didn't matter previously as the actual choice of
which approach to use was done in GDB. Now the Makefile decides, so
placing each XML file in the correct list is critical.
Tested this by doing 'make GDB=/path/to/gdb clean-cfiles cfiles' to
regenerate all the C files from their XML source. There are no
changes after this commit.
gdb/ChangeLog:
* features/Makefile (XMLTOC): Add rx.xml.
(FEATURE_XMLFILES): Remove rx.xml.
(FEATURE_CFILES rule): Pass '-single-feature' flag.
* features/rx.c: Regenerate.
* features/rx.xml: Wrap in `target` tags, and reindent.
* target-descriptions.c (struct maint_print_c_tdesc_options): New
structure.
(maint_print_c_tdesc_opt_def): New typedef.
(maint_print_c_tdesc_opt_defs): New static global.
(make_maint_print_c_tdesc_options_def_group): New function.
(maint_print_c_tdesc_cmd): Make use of command line flags, only
print single feature C file for target descriptions containing a
single feature.
(maint_print_c_tdesc_cmd_completer): New function.
(_initialize_target_descriptions): Update call to register command
completer, and include command line flag in help text.
gdb/doc/ChangeLog:
* gdb.texinfo (Maintenance Commands): Update description of 'maint
print c-tdesc'.
