This patch is part of a series that adds support for Armv8.6-A
(Matrix Multiply and BFloat16 extensions) to binutils.
This patch introduces BFloat16 instructions to the arm backend.
The following BFloat16 instructions are added: vdot, vfma{l/t},
vmmla, vfmal{t/b}, vcvt, vcvt{t/b}.
* config/tc-arm.c (arm_archs): Add armv8.6-a option.
(cpu_arch_ver): Add TAG_CPU_ARCH_V8 tag for Armv8.6-a.
* doc/c-arm.texi (-march): New armv8.6-a arch.
* config/tc-arm.c (arm_ext_bf16): New feature set.
(enum neon_el_type): Add NT_bfloat value.
(B_MNEM_vfmat, B_MNEM_vfmab): New bfloat16 encoder
helpers.
(BAD_BF16): New message.
(parse_neon_type): Add bf16 type specifier.
(enum neon_type_mask): Add N_BF16 type.
(type_chk_of_el_type): Account for NT_bfloat.
(el_type_of_type_chk): Account for N_BF16.
(neon_three_args): Split out from neon_three_same.
(neon_three_same): Part split out into neon_three_args.
(CVT_FLAVOUR_VAR): Add bf16_f32 cvt flavour.
(do_neon_cvt_1): Account for vcvt.bf16.f32.
(do_bfloat_vmla): New.
(do_mve_vfma): New function to deal with the mnemonic clash between the BF16
vfmat and the MVE vfma in a VPT block with a 't'rue condition.
(do_neon_cvttb_1): Account for vcvt{t,b}.bf16.f32.
(do_vdot): New
(do_vmmla): New
(insns): Add vdot and vmmla mnemonics.
(arm_extensions): Add "bf16" extension.
* doc/c-arm.texi: Document "bf16" extension.
* testsuite/gas/arm/attr-march-armv8_6-a.d: New test.
* testsuite/gas/arm/bfloat16-bad.d: New test.
* testsuite/gas/arm/bfloat16-bad.l: New test.
* testsuite/gas/arm/bfloat16-bad.s: New test.
* testsuite/gas/arm/bfloat16-cmdline-bad-2.d: New test.
* testsuite/gas/arm/bfloat16-cmdline-bad-3.d: New test.
* testsuite/gas/arm/bfloat16-cmdline-bad.d: New test.
* testsuite/gas/arm/bfloat16-neon.s: New test.
* testsuite/gas/arm/bfloat16-non-neon.s: New test.
* testsuite/gas/arm/bfloat16-thumb-bad.d: New test.
* testsuite/gas/arm/bfloat16-thumb-bad.l: New test.
* testsuite/gas/arm/bfloat16-thumb.d: New test.
* testsuite/gas/arm/bfloat16-vfp.d: New test.
* testsuite/gas/arm/bfloat16.d: New test.
* testsuite/gas/arm/bfloat16.s: New test.
* arm-dis.c (select_arm_features): Update bfd_march_arm_8 with
Armv8.6-A.
(coprocessor_opcodes): Add bfloat16 vcvt{t,b}.
(neon_opcodes): Add bfloat SIMD instructions.
(print_insn_coprocessor): Add new control character %b to print
condition code without checking cp_num.
(print_insn_neon): Account for BFloat16 instructions that have no
special top-byte handling.
[Patch][binutils][arm] Create a new generic coprocessor array [3/10]
Hi,
This patch is part of a series that adds support for Armv8.6-A
(Matrix Multiply and BFloat16 extensions) to binutils.
Some generic instructions match a large range of encoding space (e.g.
stc, mcr, mrc).
Currently these instructions are in the coprocessor_opcodes array, which
means they are checked before many other instructions when disassembling
arm and thumb32 codes.
This patch moves the generic instructions into a separate array to be
checked later on.
This is done in order to avoid instruction conflict between the generic
instructions and newer ones -- this has already been seen with MVE, and
is also a problem with BFloat16.
One way to avoid the conflict could be to swap the search order between
coprocessor_opcodes and neon_opcodes.
We avoid this since it's a larger change that may introduce extra bugs
(that aren't caught by the testsuite).
We have decided against searching the generic array after searching the
arm specific and thumb32 specific arrays with a similar reasoning about
keeping the change small.
* arm-dis.c (print_insn_coprocessor,
print_insn_generic_coprocessor): Create wrapper functions around
the implementation of the print_insn_coprocessor control codes.
(print_insn_coprocessor_1): Original print_insn_coprocessor
function that now takes which array to look at as an argument.
(print_insn_arm): Use both print_insn_coprocessor and
print_insn_generic_coprocessor.
(print_insn_thumb32): As above.
* config/tc-aarch64.c (vectype_to_qualifier): Special case the
S_2H operand qualifier.
* doc/c-aarch64.texi: Document bf16 and bf16mmla4 extensions.
* testsuite/gas/aarch64/bfloat16.d: New test.
* testsuite/gas/aarch64/bfloat16.s: New test.
* testsuite/gas/aarch64/illegal-bfloat16.d: New test.
* testsuite/gas/aarch64/illegal-bfloat16.l: New test.
* testsuite/gas/aarch64/illegal-bfloat16.s: New test.
* testsuite/gas/aarch64/sve-bfloat-movprfx.s: New test.
* testsuite/gas/aarch64/sve-bfloat-movprfx.d: New test.
* opcode/aarch64.h (AARCH64_FEATURE_BFLOAT16): New feature macros.
(AARCH64_ARCH_V8_6): Include BFloat16 feature macros.
(enum aarch64_opnd_qualifier): Introduce new operand qualifier
AARCH64_OPND_QLF_S_2H.
(enum aarch64_insn_class): Introduce new class "bfloat16".
(BFLOAT16_SVE_INSNC): New feature set for bfloat16
instructions to support the movprfx constraint.
* aarch64-asm.c (aarch64_ins_reglane): Use AARCH64_OPND_QLF_S_2H
in reglane special case.
* aarch64-dis-2.c (aarch64_opcode_lookup_1,
aarch64_find_next_opcode): Account for new instructions.
* aarch64-dis.c (aarch64_ext_reglane): Use AARCH64_OPND_QLF_S_2H
in reglane special case.
* aarch64-opc.c (struct operand_qualifier_data): Add data for
new AARCH64_OPND_QLF_S_2H qualifier.
* aarch64-tbl.h (QL_BFDOT QL_BFDOT64, QL_BFDOT64I, QL_BFMMLA2,
QL_BFCVT64, QL_BFCVTN64, QL_BFCVTN2_64): New qualifiers.
(aarch64_feature_bfloat16, aarch64_feature_bfloat16_sve,
aarch64_feature_bfloat16_bfmmla4): New feature sets.
(BFLOAT_SVE, BFLOAT): New feature set macros.
(BFLOAT_SVE_INSN, BFLOAT_BFMMLA4_INSN, BFLOAT_INSN): New macros
to define BFloat16 instructions.
(aarch64_opcode_table): Define new instructions bfdot,
bfmmla, bfcvt, bfcvtnt, bfdot, bfdot, bfcvtn, bfmlal[b/t]
bfcvtn2, bfcvt.
This patch is part of a series that adds support for Armv8.6-A
to binutils.
This first patch adds the Armv8.6-A flag to binutils.
No instructions are behind it at the moment.
Nick Clifton [Thu, 7 Nov 2019 11:56:54 +0000 (11:56 +0000)]
Allow the --output option of the "ar" prorgam to extract files to locations outside of the current directory.
* ar.c (open_output_file): Check for filename validity before
prefixing with output directory.
Display the constructed output filename if in verbose mode.
(extract_file): Let open_output_file display the filename.
Tom de Vries [Thu, 7 Nov 2019 09:49:56 +0000 (10:49 +0100)]
[gdb/contrib] Add words.sh script
Add a script that takes a list of files as arguments and output a list of
words from the C comments with their frequencies.
For:
...
$ ./gdb/contrib/words.sh $(find gdb -type f -name "*.c" -o -name "*.h")
...
it generates a list of ~15000 words prefixed with frequency.
This could be used to generate a dictionary that is kept as part of the
sources, against which new code can be checked, generating a warning or
error. The hope is that misspellings would trigger this frequently, and rare
words rarely, otherwise the burden of updating the dictionary would be too
much.
And for:
...
$ ./gdb/contrib/words.sh -f 1 $(find gdb -type f -name "*.c" -o -name "*.h")
...
it generates a list of ~5000 words with frequency 1.
This can be used to scan for misspellings manually.
Alan Modra [Thu, 7 Nov 2019 03:02:16 +0000 (13:32 +1030)]
Remove CR16C support
I think it is past time to remove CR16C support. CR16C was added in
2004, and only for ld. gas and binutils support is lacking, and there
have been no commits to bfd/elf32-cr16c.c other than warning fixes or
global maintainers making changes to all targets. I see no maintainer
listed for CR16C, and no commits from anyone at NSC supporting the
target. Furthermore, at the time the CR16 support was added in 2007,
config.sub was changed upstream to no longer recognise cr16c as a
valid cpu. That means the CR16C ld support is only available as a
secondary target by configuring with, for example,
--enable-targets=all or --enable-targets=cr16c-unknown-elf. No
testing of the CR16C target is possible.
Alan Modra [Wed, 6 Nov 2019 07:20:27 +0000 (17:50 +1030)]
Order targets in ld/configure.tgt
The target list was supposed to be more or less alphabetically sorted,
but this wasn't anywhere near the case. The comment about keeping
architecture variants together seems odd to me, and is no doubt the
reason why ix86 and x86_64 were grouped together, so I removed that
comment. The patch doesn't change order of entries for a given cpu.
Jan Beulich [Thu, 7 Nov 2019 08:28:20 +0000 (09:28 +0100)]
x86: adjust register names printed for MONITOR/MWAIT
As the comments (here: almost, in the opcode table: fully) correctly
state - all register operands except MONITOR's address one are fixed
at 32 bit size. Don't print 64-bit registers there.
Also adjust x86-64-suffix.d's name such that it wouldn't be identical to
x86-64-rep-suffix.d's, but instead resemble that of its sibling
x86-64-suffix-intel.d.
Tom Tromey [Sun, 27 Oct 2019 21:50:54 +0000 (15:50 -0600)]
Remove some includes of readline.h
I went through most of the spots that include readline.h and, when
appropriate, either removed the include or changed it to include
tilde.h.
Note that remote-sim.c and bsd-kvm.c could probably include tilde.h
instead, but I did not change these. I think I can't build the
latter, and I didn't want to set up a sim build for the former.
* tui/tui-interp.c: Don't include readline.h.
* tui/tui-hooks.c: Don't include readline.h.
* symmisc.c: Include tilde.h, not readline.h.
* symfile.c: Include tilde.h, not readline.h.
* source.c: Include tilde.h, not readline.h.
* solib.c: Include tilde.h, not readline.h.
* psymtab.c: Include tilde.h, not readline.h.
* exec.c: Include tilde.h, not readline.h.
* corelow.c: Include tilde.h, not readline.h.
* cli/cli-dump.c: Include tilde.h, not readline.h.
* cli/cli-cmds.c: Don't include readline.h.
Fix an off-by-one error in the IN_RANGE macro used by readelf. Add another use of the macro.
* readelf.c (IN_RANGE): Rename parameter OFF to NELEM. Add
comment. Catch potential integer overflow and fix off by one
error whilst checking reloc location against section size.
(apply_relocations): Use IN_RANGE macro.
Tom Tromey [Thu, 3 Oct 2019 22:59:17 +0000 (16:59 -0600)]
Remove la_get_string member
The la_get_string member of struct language_defn was intended to
provide a way to fetch string data from a "string" object in a
language-dependent way. However, it turned out that this was never
needed, and was only ever implemented for C. This patch removes the
language hook entirely.
Tom Tromey [Mon, 21 Oct 2019 17:21:14 +0000 (11:21 -0600)]
Style disassembly in the TUI
This patch changes the TUI disassembly window to style its contents.
The styling should be identical to what is seen in the CLI. This
involved a bit of rearrangement, so that the source and disassembly
windows could share both the copy_source_line utility function, and
the ability to react to changes in "set style enabled".
This version introduces a new function to strip the styling from the
address string when computing the length. As a byproduct, it also
removes the unused "insn_size" computation from
tui_disasm_window::set_contents.
* tui/tui-source.h (struct tui_source_window): Inline
constructor. Remove destructor.
<style_changed, m_observable>: Move to superclass.
* tui/tui-winsource.h (tui_copy_source_line): Declare.
(struct tui_source_window_base): Move private members to end.
<style_changed, m_observable>: Move from tui_source_window.
* tui/tui-winsource.c (tui_copy_source_line): Move from
tui-source.c. Rename from copy_source_line. Add special handling
for negative line number.
(tui_source_window_base::style_changed): Move from
tui_source_window.
(tui_source_window_base): Register observer.
(~tui_source_window_base): New.
* tui/tui-source.c (copy_source_line): Move to tui-winsource.c;
rename.
(tui_source_window::set_contents): Use tui_copy_source_line.
(tui_source_window::tui_source_window): Move to tui-source.h.
(tui_source_window::~tui_source_window): Remove.
(tui_source_window::style_changed): Move to superclass.
* tui/tui-disasm.c (tui_disassemble): Create string file with
styling, when possible. Add "addr_size" parameter.
(tui_disasm_window::set_contents): Use tui_copy_source_line.
Don't compute maximum size.
(len_without_escapes): New function
Tim R?hsen [Tue, 5 Nov 2019 16:03:07 +0000 (16:03 +0000)]
Fix memory allocation and release problems in the bfd documentation processor.
* doc/chew.c (add_to_definition): Use correct type when
calculating size of array reallocation.
(nextword): Always initialise the word return parameter.
(compile): Check return value of nextword().
Jan Beulich [Tue, 5 Nov 2019 08:18:23 +0000 (09:18 +0100)]
x86: consolidate disassembler enum naming a little
The original idea looks to have been for names to be composed in the
order that decoding gets done, which helps both reading and modifying
the code. Switch (back) to this model for some of the affected non-
vector insn enumerators.
Alan Modra [Fri, 1 Nov 2019 03:15:25 +0000 (13:45 +1030)]
GENERATE_SHLIB_SCRIPT vs. EMBEDDED.
A target that sets EMBEDDED non-empty is one that doesn't want to put
the ELF file header or program header in a memory image. Likely the
target isn't interested in supporting dynamically loaded executables,
shared libraries, or PIEs, because loaders for such binaries generally
require program headers to be present. This isn't 100% true though.
The target might be one where the loader accesses the file system in
order to retrieve headers.
Also, note that not all "shared libraries" require gcc -shared or the
shared library support in the linker. I believe one implementation of
shared libraries on uclinux is like this.
So, this patch removes GENERATE_SHLIB_SCRIPT and GENERATE_PIE_SCRIPT
in most emulparams files where EMBEDDED is set, restoring the shared
lib and pie support in emulparams files that unset EMBEDDED after
including a file where the support is removed.
Exceptions to the general rule that EMBEDDED disables shared libs are:
arm*-*-symbianelf*, where the OS wants shared library support
without ELF program headers in the image, and
sh*-*-uclinux*, where I've left things as they were, ie. both
EMBEDDED and GENERATE_SHLIB_SCRIPT because I'm unsure as to the
shared library scheme.
Right now it doesn't remove any code, but obviates the need to test on
that ancient platform. Besides, some of the patches I have in my queue
would require different solutions for Solaris 10 and 11.
There are a few comment-only references that I've kept since they are
still correct as is, even when GDB doesn't support Solaris 10 any
longer. The only code fragment I've left in is support for
/proc/<pid/path/a.out in procfs.c (procfs_target::pid_to_exec_file):
while current Solaris 11 updates provide /proc/<pid>/execname, that
wasn't present in Solaris 11.0 and still isn't in current Illumos and I
didn't want to make live harder for them.
Tested on i386-pc-solaris2.10 (obsolete configuration rejected) and
x86_64-pc-linux-gnu x sparc64-solaris2.10 (likewise)
resp. x86_64-pc-linux-gnu x sparcv9-solaris2.11 (still builds; I'm using
the sparcv9 form for 64-bit SPARC customary on Solaris in the
MAINTAINERS file now).
* NEWS (Changes since GDB 8.3): Document Solaris 10 removal.
* configure.host: Mark *-*-solaris2.10* obsolete.
* configure.tgt: Mark Solaris < 11 obsolete.
* MAINTAINERS (Target Instruction Set Architectures) <sparc>:
Update target triplet.
Jan Beulich [Mon, 4 Nov 2019 14:48:38 +0000 (15:48 +0100)]
x86: re-arrange process_operands()
Alter the sequence of conditions evaluated, without affecting the
overall result. This is going to help subsequent changes (and as a nice
side effect also slightly reduces overall indentation depth).
While doing this take the liberty of simplifying the calculation of the
operand index of the register operand in ShortForm handling.
Nick Clifton [Mon, 4 Nov 2019 12:02:20 +0000 (12:02 +0000)]
Fix potential array overruns when disassembling corrupt v850 binaries.
* v850-dis.c (get_v850_sreg_name): New function. Returns the name
of a v850 system register. Move the v850_sreg_names array into
this function.
(get_v850_reg_name): Likewise for ordinary register names.
(get_v850_vreg_name): Likewise for vector register names.
(get_v850_cc_name): Likewise for condition codes.
* get_v850_float_cc_name): Likewise for floating point condition
codes.
(get_v850_cacheop_name): Likewise for cache-ops.
(get_v850_prefop_name): Likewise for pref-ops.
(disassemble): Use the new accessor functions.
As a result, ARM targets only read the mapping symbols the first time we
load a symbol file. When reloaded, the speed optimization above will
cause an early return and gdbarch_record_special_symbol won't be called to
save mapping symbol data, which in turn affects disassembling of thumb
instructions.
First load and correct disassemble output:
Dump of assembler code for function main:
0x0000821c <+0>: bx pc
0x0000821e <+2>: nop
0x00008220 <+4>: mov r0, #0
0x00008224 <+8>: bx lr
Second load and incorrect disassemble output:
Dump of assembler code for function main:
0x0000821c <+0>: bx pc
0x0000821e <+2>: nop
0x00008220 <+4>: movs r0, r0
0x00008222 <+6>: b.n 0x8966
0x00008224 <+8>: vrhadd.u16 d14, d14, d31
This happens because the mapping symbol data is stored in an objfile_key-based
container, and that data isn't preserved across the two symbol loading
operations.
The following patch fixes this by storing the mapping symbol data in a
bfd_key-based container, which doesn't change as long as the bfd is the same.
I've also added a new test to verify the correct disassemble output.
Andrew Burgess [Tue, 24 Sep 2019 22:01:14 +0000 (23:01 +0100)]
gdb: Don't print a newline in language la_print_typedef methods
When calling the language la_print_typedef method, don't include a
newline at the end, instead print the newline from the users of
la_print_typedef.
This change will be useful in a later commit when the output from
la_print_typedef will be placed into an MI output field, in which case
the trailing newline is not required.
There should be no user visible changes after this commit.
gdb/ChangeLog:
* ada-typeprint.c (ada_print_typedef): Don't print newline at the
end.
* c-typeprint.c (c_print_typedef): Likewise.
* f-typeprint.c (f_print_typedef): Likewise.
* m2-typeprint.c (m2_print_typedef): Likewise.
* p-typeprint.c (pascal_print_typedef): Likewise.
* rust-lang.c (rust_print_typedef): Likewise.
* symtab.c (print_symbol_info): Print a newline after calling
typedef_print.
Andrew Burgess [Wed, 10 Jul 2019 14:54:03 +0000 (15:54 +0100)]
gdb: Add new commands to list module variables and functions
This patch adds two new commands "info module functions" and "info
module variables". These commands list all of the functions and
variables grouped by module and then by file.
For example:
(gdb) info module functions
All functions in all modules:
The new commands take set of flags that allow the output to be
filtered, the user can filter by variable/function name, type, or
containing module.
As GDB doesn't currently track the relationship between a module and
the variables or functions within it in the symbol table, so I filter
based on the module prefix in order to find the functions or variables
in each module. What this makes clear is that a user could get this
same information using "info variables" and simply provide the prefix
themselves, for example:
(gdb) info module functions -m mod1 _a
All functions matching regular expression "_a",
in all modules matching regular expression "mod1":
The benefits I see for a separate command are that the user doesn't
have to think (or know) about the module prefix format, nor worry
about building a proper regexp. The user can also easily scan across
modules without having to build complex regexps.
The new function search_module_symbols is extern in this patch despite
only being used within symtab.c, this is because a later patch in this
series will also be using this function from outside symtab.c.
This patch is a new implementation of an idea originally worked on by
Mark O'Connor, Chris January, David Lecomber, and Xavier Oro from ARM.
gdb/ChangeLog:
* symtab.c (info_module_cmdlist): New variable.
(info_module_command): New function.
(search_module_symbols): New function.
(info_module_subcommand): New function.
(struct info_modules_var_func_options): New struct.
(info_modules_var_func_options_defs): New variable.
(make_info_modules_var_func_options_def_group): New function.
(info_module_functions_command): New function.
(info_module_variables_command): New function.
(info_module_var_func_command_completer): New function.
(_initialize_symtab): Register new 'info module functions' and
'info module variables' commands.
* symtab.h (typedef symbol_search_in_module): New typedef.
(search_module_symbols): Declare new function.
* NEWS: Mention new commands.
gdb/doc/ChangeLog:
* gdb.texinfo (Symbols): Document new 'info module variables' and
'info module functions' commands.
gdb/testsuite/ChangeLog:
* gdb.fortran/info-modules.exp: Update expected results, and add
additional tests for 'info module functinos', and 'info module
variables'.
* gdb.fortran/info-types.exp: Update expected results.
* gdb.fortran/info-types.f90: Extend testcase with additional
module variables and functions.
Andrew Burgess [Tue, 9 Jul 2019 20:38:59 +0000 (21:38 +0100)]
gdb/fortran: Add new 'info modules' command
Add a new command 'info modules' that lists all of the modules GDB
knows about from the debug information.
A module is a debugging entity in the DWARF defined with
DW_TAG_module, currently Fortran is known to use this tag for its
modules. I'm not aware of any other language that currently makes use
of DW_TAG_module.
The output style is similar to the 'info type' output:
(gdb) info modules
All defined modules:
File info-types.f90:
16: mod1
24: mod2
(gdb)
Where the user is told the file the module is defined in and, on the
left hand side, the line number at which the module is defined along
with the name of the module.
This patch is a new implementation of an idea originally worked on by
Mark O'Connor, Chris January, David Lecomber, and Xavier Oro from ARM.
gdb/ChangeLog:
* dwarf2read.c (dw2_symtab_iter_next): Handle MODULE_DOMAIN.
(dw2_expand_marked_cus): Handle MODULES_DOMAIN.
(dw2_debug_names_iterator::next): Handle MODULE_DOMAIN and
MODULES_DOMAIN.
(scan_partial_symbols): Only create partial module symbols for non
declarations.
* psymtab.c (recursively_search_psymtabs): Handle MODULE_DOMAIN
and MODULES_DOMAIN.
* symtab.c (search_domain_name): Likewise.
(search_symbols): Likewise.
(print_symbol_info): Likewise.
(symtab_symbol_info): Likewise.
(info_modules_command): New function.
(_initialize_symtab): Register 'info modules' command.
* symtab.h (enum search_domain): Add MODULES_DOMAIN.
* NEWS: Mention new 'info modules' command.
gdb/doc/ChangeLog:
* gdb.texinfo (Symbols): Document new 'info modules' command.
gdb/testsuite/ChangeLog:
* gdb.fortran/info-modules.exp: New file.
* gdb.fortran/info-types.exp: Build with new file.
* gdb.fortran/info-types.f90: Include and use new module.
* gdb.fortran/info-types-2.f90: New file.
Test the convenience functions $_gdb_setting and $_gdb_setting_str.
gdb/testsuite/ChangeLog
2019-10-31 Philippe Waroquiers <[email protected]>
* gdb.base/setshow.exp: Test $_gdb_setting and $_gdb_setting_str.
* gdb.base/settings.exp: Test all settings types using
$_gdb_maint_setting and $_gdb_maint_setting_str in proc_show_setting,
that now verifies that the value of "maint show" is the same as
returned by the settings functions. Test the type of the
maintenance settings.
* gdb.base/default.exp: Update show_conv_list.
Implement convenience functions to examine GDB settings.
The new convenience functions $_gdb_setting and $_gdb_setting_str
provide access to the GDB settings in user-defined commands.
Similarly, $_gdb_maint_setting and $_gdb_maint_setting_str
provide access to the GDB maintenance settings.
The patch was developed following a comment of Eli about the
'set may-call-functions'. Eli said that user-defined functions
should have a way to change their behavior according to this setting.
Rather than have a specialized $_may_call_functions, this patch
implements a general way to access any GDB setting.
Compared to doing such access via Python 'gdb.parameter' and/or
'gdb.execute("set somesetting tosomevalue"):
* The 'with' command is much better than the above python usage:
if the user types C-c or an error happens between the set pagination off
and the python "set pagination on", the above python
does not restore the original setting.
* Effectively, with the "gdb.parameter" python one liner, it is possible to do
simple 'if' conditions, such as set and restore pagination.
But mixing the "python if" within canned
sequence of commands is cumbersome for non trivial combinations.
E.g. if several commands have to be done for a certain condition
accessed from python, I guess something like will be needed:
python if __some_setting: gdb.execute("some command")
python if __some_setting: gdb.execute("some other command")
python if __some_setting: gdb.execute("some different command")
(without speaking about nested "if-s").
With the convenience function:
if $_gdb_setting("some_setting")
some command
some other command
some different command
end
Integer settings (for example print elements) will also be more difficult
to use.
For example, a user defined function that scans and prints a linked list
might want to use the value of "set print elements" to stop printing
the linked list.
Doing that by mixing python expression/if is likely doable, but seems
not easy with the above one liners.
So, in summary, the $_gdb_setting and $_gdb_setting_str avoids to have the
heterogeneous mix of python and GDB commands in one single script
(and of course, it works even if python is not configured, but that
must be an unusual setup I guess).
Since safe_strerror is in gdbsupport, gdbserver also needs to
check for strerror_r, although it's less critical since gdbserver
does not use threads as much.
* agent.c (set_can_use_agent): When the setting is turned on,
look up agent symbols if we don't have them yet.
(agent_new_objfile): Don't look up agent symbols when the agent
setting is off.
Also stores the result in a thread-local static variable and
changes the return value to a const char*.
This is already important because Guile creates threads and
Python can create threads, but with the patch series here:
https://gnutoolchain-gerrit.osci.io/r/c/binutils-gdb/+/176
GDB itself will create threads, too.
* configure: Regenerate.
* configure.ac: Check for strerror_r.
* gdbsupport/common-utils.h (safe_strerror): Change return value
to const char * and document that this function is now threadsafe.
* gdbsupport/posix-strerror.c (safe_strerror): Make buf
thread_local and call strerror_r, if available.
* utils.c (perror_string): Update.
(print_sys_errmsg): Update.
Luis Machado [Thu, 31 Oct 2019 19:30:44 +0000 (16:30 -0300)]
[ARM] Store exception handling information per-bfd instead of per-objfile
Based on feedback from Tromey, update the use of objfile_key in gdb/arm-tdep.c
to use bfd_key instead. That way we don't have to re-create the exception
handling data all over again if it was done before for the same BFD.
* arm-tdep.c (arm_exidx_data_key): Use bfd_key instead of
objfile_key.
(arm_exidx_new_objfile): Adjust to use objfile->obfd instead of
objfile to fetch per-bfd data.
(arm_find_exidx_entry): Likewise.
* gdbsupport/agent.c (debug_agent): Change type to bool.
(use_agent): Likewise.
(all_agent_symbols_look_up): Likewise.
(agent_loaded_p): Change return value to bool.
(agent_look_up_symbols): Update.
(agent_capability_check): Change return value to bool.
* gdbsupport/agent.h (agent_loaded_p): Likewise.
(debug_agent): Change type to bool.
(use_agent): Likewise.
(agent_capability_check): Change return value to bool.
Mihail Ionescu [Thu, 31 Oct 2019 11:22:58 +0000 (11:22 +0000)]
Add support for context sensitive '.arch_extension' to the ARM assembler.
If the extension is not found in the context sensitive table, the legacy
tables are still checked as a fallback. This is particularly useful for
Armv8.1-M as it enables the use of '.arch_extension' with the 'mve' and
'mve.fp' extensions which are not part of the legacy table.
* config/tc-arm.c (selected_ctx_ext_table) New static variable.
(arm_parse_arch): Set context sensitive extension table based on the
chosen base architecture.
(s_arm_arch_extension): Change to lookup extensions in the new context
sensitive tables.
* gas/testsuite/gas/arm/mve-ext.s: New.
* gas/testsuite/gas/arm/mve-ext.d: New.
* gas/testsuite/gas/arm/mvefp-ext.s: New.
* gas/testsuite/gas/arm/mvefp-ext.d: New.
Tom de Vries [Wed, 30 Oct 2019 16:41:03 +0000 (17:41 +0100)]
[gdb/testsuite] Add -early pattern flag for gdb_test_multiple
Proc gdb_test_multiple builds up and executes a gdb_expect expression with
pattern/action clauses. The clauses are either implicit (added by
gdb_test_multiple) or explicit (passed via the gdb_test_multiple parameter
user_code).
However, there are a few implicit clauses which are inserted before the
explicit ones, making sure those take precedence.
Add an -early pattern flag for a gdb_test_multiple user_code clause to specify
that the clause needs to be inserted before any implicit clause.
Using this pattern flag, we can f.i. setup a kfail for an assertion failure
<assert> during gdb_continue_to_breakpoint by the rewrite:
...
gdb_continue_to_breakpoint <msg> <pattern>
...
into:
...
set breakpoint_pattern "(?:Breakpoint|Temporary breakpoint) .* (at|in)"
gdb_test_multiple "continue" "continue to breakpoint: <msg>" {
-early -re "internal-error: <assert>" {
setup_kfail gdb/nnnnn "*-*-*"
exp_continue
}
-re "$breakpoint_pattern <pattern>\r\n$gdb_prompt $" {
pass $gdb_test_name
}
}
* minsyms.c (clear_minimal_symbol_hash_tables): New function.
(build_minimal_symbol_hash_tables): Code to clear the table moved
to clear_minimal_symbol_hash_tables.
(minimal_symbol_reader::install): Call clear_minimal_symbol_hash_tables
when needed.
Delia Burduv [Wed, 30 Oct 2019 13:23:35 +0000 (13:23 +0000)]
Modify the ARNM assembler to accept the omission of the immediate argument for the writeback form of the LDRAA and LDRAB mnemonics
This is a shorthand for the immediate argument being 0, as described here:
https://developer.arm.com/docs/ddi0596/latest/base-instructions-alphabetic-order/ldraa-ldrab-load-register-with-pointer-authentication
This is because the instructions still have a use with an immediate
argument of 0, unlike loads without the PAC functionality. Currently,
the mnemonics are
gas * config/tc-aarch64.c (parse_address_main): Accept the omission of
the immediate argument for ldraa and ldrab as a shorthand for the
immediate being 0.
* testsuite/gas/aarch64/ldraa-ldrab-no-offset.d: New test.
* testsuite/gas/aarch64/ldraa-ldrab-no-offset.s: New test.
* testsuite/gas/aarch64/illegal-ldraa.s: Modified to accept the
writeback form with no offset.
* testsuite/gas/aarch64/illegal-ldraa.s: Removed missing offset
error.
opcodes * aarch64-opc.c (print_immediate_offset_address): Don't print the
immediate for the writeback form of ldraa/ldrab if it is 0.
* aarch64-tbl.h: Updated the documentation for ADDR_SIMM10.
* aarch64-opc-2.c: Regenerated.
Keith Seitz [Wed, 30 Oct 2019 12:23:16 +0000 (12:23 +0000)]
Add the ability to the BFD library to read build-ids from core flies.
* elf-bfd.h (elf_backend_data) <elf_backend_core_find_build_id>:
New field.
(_bfd_elf32_core_find_build_id, _bfd_elf64_core_find_build_id):
New functions.
(elf_read_notes): Add declaration.
* elf.c (elf_read_notes): Move elf-bfd.h.
(_bfd_elf_core_find_build_id): New function.
(bfd_section_from_phdr): Scan core file PT_LOAD segments for
build-id if none is known.
(elf_parse_notes): For core files, scan for notes.
* elfcore.h (elf_core_file_matches_executable_p): If both
BFDs have identical build-ids, then they match.
(_bfd_elf_core_find_build_id): New function.
* elfxx-target.h (elf_backend_core_find_build_id): Define.
(elfNN_bed): Add elf_backend_core_find_build_id.
Fangrui Song [Wed, 30 Oct 2019 10:50:23 +0000 (10:50 +0000)]
Add a --output=<DIR> option to ar to allow the specifying of an output directory.
* ar.c (emum long option numbers): Declare. Use to provide
numerical values for long options.
(long_options): Add --output option.
(usage): Mention the --output option.
(open_output_file): New function. Create a filepath for an output
file and open it.
(extract_file): Use open_output_file().
(open_output_file):
* testsuite/binutils-all/ar.exp: Add a test of the new feature.
* doc/binutils.texi: Document the new feature.
* NEWS: Mention the new feature.
Jan Beulich [Wed, 30 Oct 2019 08:09:13 +0000 (09:09 +0100)]
x86: add tests to cover defaulting of operand sizes for ambiguous insns
Prior to changing the logic in the assembler install tests to make sure
the present defaulting of operand sizes won't get broken. There are a
few anomalies pointed out by this:
- arithmetic insns (add, sub, etc) allow defaulting when their immediate
fits in (signed) 8 bits, but they fail to assemble with larger values,
- mov, other than arithmetic insns, doesn't allow any defaulting,
- movsx/movzx default to byte sources (in AT&T mode), and their special
casing needs to be adjusted first
- bt and friends allow defaulting, while shl and friends don't,
- ambiguous AVX and AVX512 insns don't allow defaulting.
This should ultimately all become consistent (perhaps with the exception
some of the SIMD insns); respective tests will be added to the test
cases here as the issues get addressed.
Jan Beulich [Wed, 30 Oct 2019 08:07:40 +0000 (09:07 +0100)]
x86: re-do "shorthand" handling
Now that the opcode table gets preprocessed, undo parts of commit dc821c5f9a ("x86: replace Reg8, Reg16, Reg32, and Reg64"): Have the
preprocessor handle the expansion there, while making the expansions
explicit in i386-gen and the register table.
Jan Beulich [Wed, 30 Oct 2019 08:05:46 +0000 (09:05 +0100)]
x86: drop stray W
The flag is used to indicate opcodes which can be switched between byte
and word/dword/qword forms (in a "canonical" way). Obviously it's quite
odd then to see it on insns not allowing for byte operands in the first
place. As a result the opcode bytes need to be adjusted accordingly,
which includes comparisons done in optimize_encoding().
To make re-introduction of such issues less likely have i386-gen
diagnose it (in a generally non-fatal way for now).
Simon Marchi [Tue, 29 Oct 2019 21:52:24 +0000 (17:52 -0400)]
Clean up ada-lang.h
This patch cleans up ada-lang.h:
- Some functions just don't exist anymore, remove their declaration
- Some functions are implemented in ada-lang.c and only used there, make
them static to that file.
I moved some functions higher in the file to avoid having to
forward-declare them, but the implementations are unchanged.
gdb/ChangeLog:
* ada-lang.h (GROW_VECT): Move to ada-lang.c.
(grow_vect): Remove declaration.
(ada_type_of_array): Remove declaration.
(ada_update_initial_language): Remove declaration.
(ada_fold_name): Remove declaration.
(ada_fill_in_ada_prototype): Remove declaration.
(user_select_syms): Remove declaration.
(get_selections): Remove declaration.
(ada_tag_type): Remove declaration.
(ada_value_tag): Remove declaration.
(ada_is_others_clause): Remove declaration.
(ada_in_variant): Remove declaration.
(ada_value_struct_elt): Remove declaration.
(ada_attribute_name): Remove declaration.
(ada_system_address_type): Remove declaration.
* ada-lang.c (ada_watch_location_expression): Make static.
(GROW_VECT): Move here from ada-lang.h.
(grow_vect): Make static.
(ada_update_initial_language): Make static.
(ada_fold_name): Make static.
(ada_type_of_array): Make static.
(encoded_ordered_before): Move up.
(sort_choices): Move up.
(print_signatures): Move up.
(ada_print_symbol_signature): Move up.
(get_selections): Move up and make static.
(user_select_syms): Move up and make static.
(ada_value_struct_elt): Move up and make static.
(ada_tag_type): Make static.
(ada_value_tag): Make static.
(ada_is_others_clause): Make static.
(ada_in_variant): Make static.
(ada_attribute_name): Make static.
Change some arguments to gdb::string_view instead of name+len
Just some code cleanup. This change has a few benefits:
- Shorter argument list in the functions
- If the caller needs to calculate the string, they no longer
need to explicitly call strlen
- It is easy to pass std::string to this (done in one place
currently)
This also updates a couple of places that were passing 0/1 to
a bool parameter.
So this patch makes it so we only make the copy if the entry was
not found.
By auditing all callers of symbol_set_names, I found out that all cases
where the string may not be nullterminated already pass true for COPY_NAME.
So here, I am documenting that as a requirement and am removing the code
that relies on undefined behavior in symbol_set_names (it accessed the string
past the provided length to check for nulltermination). Note that the Ada
case at the beginning of symbol_set_names was already relying on this.
* symtab.h (symbol_set_names): Document that copy_name must be
set to true for non-nullterminated strings.
* symtab.c (symbol_set_names): Only make a nullterminated copy of
linkage_name if the entry was not found and we need to demangle.
Adds a configure option --with-system-gdbinit-dir to specify a directory
in which to look for gdbinit files. All files in this directory are
loaded on startup (subject to -n/-nx as usual) as long as the extension
matches a known and enabled scripting language (.gdb/.py/.scm).
This also changes get_ext_lang_of_file to support ".gdb" files, similar
to get_ext_lang_defn's handling of EXT_LANG_GDB.
* NEWS: Mention new --with-system-gdbinit-dir option.
* config.in: Regenerate.
* configure: Regenerate.
* configure.ac: Add new option --with-system-gdbinit-dir.
* extension.c (get_ext_lang_of_file): Return extension_language_gdb
for a ".gdb" suffix.
* main.c (get_init_files): Change system_gdbinit argument to
a vector and return the files in SYSTEM_GDBINIT_DIR in
addition to SYSTEM_GDBINIT.
(captured_main_1): Update.
(print_gdb_help): Update.
* top.c (print_gdb_configuration): Also print the value of
SYSTEM_GDBINIT_DIR.
* Makefile.in: Also set SYSTEM_GDBINIT_DIR for the info manual
generation.
* gdb.texinfo (many sections): Document new --with-system-gdbinit-dir
option.
Andrew Eikum [Tue, 29 Oct 2019 17:07:03 +0000 (17:07 +0000)]
When copying pe format files, copy the dos_message array, rather than re-initiialising it.
* libcoff-in.h (struct pe_tdata): Add dos_message field.
* libcoff.h: Regenerate.
* peXXigen.c (_bfd_XXi_only_swap_filehdr_out): Copy the
dos_message field rather than initialising it.
(_bfd_XX_bfd_copy_private_bfd_data_common): Copy the dos_message
field.
* peicode.h (pe_mkobject): Initialise the dos_message field.
(pe_mkobject_hook): Copy the dos_message field.
(pe_bfd_object_p): Copy the dos_message field.
Nick Clifton [Tue, 29 Oct 2019 09:17:39 +0000 (09:17 +0000)]
Fix array overruns in the S12Z disassembler.
* s12z-dis.c (opr_emit_disassembly): Check for illegal register
values.
(shift_size_table): Use a fixed size defined as S12Z_N_SIZES.
(print_insn_s12z): Check for illegal size values.
Makes sure that the string is longer than prefix, so that strncmp will
do the right thing even if the string is not null-terminated.
For use in my string_view conversion patch:
https://sourceware.org/ml/gdb-patches/2019-10/msg00030.html
https://gnutoolchain-gerrit.osci.io/r/c/binutils-gdb/+/125
Nick Clifton [Mon, 28 Oct 2019 16:15:34 +0000 (16:15 +0000)]
Fix buffer overrun in TIC30 disassembler.
* tic30-dis.c (OPERAND_BUFFER_LEN): Define. Use as length of
operand buffer. Set value to 15 not 13.
(get_register_operand): Use OPERAND_BUFFER_LEN.
(get_indirect_operand): Likewise.
(print_two_operand): Likewise.
(print_three_operand): Likewise.
(print_oar_insn): Likewise.
Nick Clifton [Mon, 28 Oct 2019 15:44:23 +0000 (15:44 +0000)]
Stop potential illegal memory access in the NS32K disassembler.
* ns32k-dis.c (bit_extract): Add sanitiy check of parameters.
(bit_extract_simple): Likewise.
(bit_copy): Likewise.
(pirnt_insn_ns32k): Ensure that uninitialised elements in the
index_offset array are not accessed.
Alan Modra [Sat, 26 Oct 2019 08:08:26 +0000 (18:38 +1030)]
Optimise away eh_frame advance_loc 0
These can be generated when multiple cfi directives are emitted for an
instruction and the insn frag is closed off between directives, as
happens when listings are enabled. No doubt the advance_loc of zero
could be avoided by backtracking over frags in dw2gencfi.c before
calling cfi_add_advance_loc, but that seems like more work than
cleaning up afterwards as this patch does.
Noticed when looking at the testcase in PR25125.
PR 25125
* dw2gencfi.c (output_cfi_insn): Don't output DW_CFA_advance_loc+0.
* ehopt.c (eh_frame_estimate_size_before_relax): Return -1 for
an advance_loc of zero.
(eh_frame_relax_frag): Translate fr_subtype of 7 to size -1.
(eh_frame_convert_frag): Handle fr_subtype of 7. Abort on
unexpected fr_subtype.
projects / binutils.git / log