I noticed this regression in the expression completer:
"(gdb) p std::[TAB]" => "(gdb) p std::std::"
obviously we should have not completed to "std::std::".
The problem is that in the earlier big completer rework, I missed
taking into account the fact that with expressions, the completion
word point is not always at the start of the symbol name (it is with
linespecs).
The fix is to run the common prefix / LCD string (what readline uses
to expand the input line) through make_completion_match_str too.
New testcase included, exercising both TAB completion and the complete
command.
gdb/ChangeLog:
2017-12-13 Pedro Alves <palves@redhat.com>
* completer.c (completion_tracker::maybe_add_completion): New
'text' and 'word' parameters. Use make_completion_match_str.
(completion_tracker::add_completion): New 'text' and 'word'
parameters. Pass down.
(completion_tracker::recompute_lowest_common_denominator): Change
parameter type to gdb::unique_xmalloc_ptr rval ref. Adjust.
* completer.h (completion_tracker::add_completion): New 'text' and
'word' parameters.
(completion_tracker::recompute_lowest_common_denominator): Change
parameter type to gdb::unique_xmalloc_ptr rval ref.
(completion_tracker::recompute_lowest_common_denominator): Change
parameter type to gdb::unique_xmalloc_ptr rval ref.
* symtab.c (completion_list_add_name): Pass down 'text' and 'word'
as well.
gdb/testsuite/ChangeLog:
2017-12-13 Pedro Alves <palves@redhat.com>
* gdb.cp/cpcompletion.exp: Load completion-support.exp.
("expression with namespace"): New set of tests.
* gdb.cp/pr9594.cc (Test_NS::foo, Test_NS::bar)
(Nested::Test_NS::qux): New.
* lib/completion-support.exp (test_gdb_complete_cmd_multiple): Add
defaults to 'start_quote_char' and 'end_quote_char' parameters.
We have several places doing essentially the same thing; factor them
out to a central place. Some of the places overallocate for no good
reason, or use strcat unnecessarily. The centralized version is more
precise and to the point.
(I considered making the gdb::unique_xmalloc_ptr overload version of
make_completer_match_str try to realloc (not xrealloc) probably
avoiding an allocation in most cases, but that'd be probably overdoing
it, and also, now that I'm writing this I thought I'd try to see how
could we ever get to filename_completer with "text != word", but I
couldn't figure it out. Running the testsuite with 'gdb_assert (text
== word);' never tripped on the assertion either. So post gdb 8.1,
I'll probably propose a patch to simplify filename_completer a bit,
and the gdb::unique_xmalloc_str overload can be removed then.)
gdb/ChangeLog:
2017-12-13 Pedro Alves <palves@redhat.com>
* cli/cli-decode.c (complete_on_cmdlist, complete_on_enum): Use
make_completion_match_str.
* completer.c: Use gdb::unique_xmalloc_ptr and
make_completion_match_str.
(make_completion_match_str_1): New.
(make_completion_match_str(const char *, const char *,
const char *)): New.
(make_completion_match_str(gdb::unique_xmalloc_ptr<char> &&,
const char *, const char *)): New.
* completer.h (make_completion_match_str(const char *,
const char *, const char *)): New.
(make_completion_match_str(gdb::unique_xmalloc_ptr<char> &&,
const char *, const char *)): New.
* interps.c (interpreter_completer): Use make_completion_match_str.
* symtab.c (completion_list_add_name, add_filename_to_list): Use
make_completion_match_str.
I find the documentation of the gdb.Breakpoint constructor hard to read
and not very informative, especially since we have added the new
linespec parameters. There are multiple problems (some are subjective):
- It's not clear that you should use either the spec string or the
explicit arguments, not both.
- It's not clear what combination of parameters you can use.
- The big block of text describing the arguments is hard to read.
- Currently, it seems like the "spec" argument is mandatory, even though
it is not (if you use explicit linespec).
- The square bracket nesting
[arg1 [, arg2[, arg3]]]
makes it seems like if you specify arg3, you must specify arg1 and
arg2 (it's not the case here).
This patch tries to address these problems.
gdb/doc/ChangeLog:
* python.texi (Manipulating breakpoints using Python): Split doc
of Breakpoint.__init__ in two, split text in multiple
paragraphs, don't nest parameter square brackets.
R_AARCH64_ABS64, R_AARCH64_ABS32 and R_AARCH64_ABS16 are data relocations
supported in AArch64 elf ABI.
R_AARCH64_ABS64 under LP64 is allowed in shared object and a dynamic relocation entry
will be generated. This allows the dynamic linker to do further symbol resolution.
R_AARCH64_ABS32 likewise is allowed in shared object, however under ILP32 abi.
The original behavior for R_AARCH64_ABS32 under LP64 is that, it's allowed
in shared object and silently resolved at static linking time.
No dynamic relocation entry is generate for it.
R_AARCH64_ABS16 is allowed in shared object under both L64 and ILP32.
It's resolved at static linking time as well.
Under LP64, the address should be 64-bit. R_AARCH64_ABS32 relocation indicates
an address that is only sized 32 bits which is meaningless in LP64 shared object.
It's useful to error out.
I have checked glibc dynamic linker code, R_AARCH64_ABS16 is not supported at all. So
R_AARCH64_ABS16 should be reject in shared object completely.
In this patch, R_AARCH64_ABS32 is rejected under LP64 in constant section of shared object.
R_AARCH64_ABS16 is rejected in constant section of shared object in both ABI.
This will sometimes provide useful information for buggy code.
Fix a few issues not using the gettext _() wrapper and issues where
we are using %p directly instead of the dedicated host/target functions.
gdb/ChangeLog:
yyyy-mm-dd Stafford Horne <shorne@gmail.com>
* or1k-tdep.c (or1k_analyse_inst): Use _() wrapper for message
strings.
(or1k_unwind_pc): Use paddress() instead of %p.
(or1k_unwind_sp): Likewise.
(or1k_frame_cache): Use host_address_to_string()/paddress()
instead of %p and use _() wrapper for message strings.
opcodes * disassemble.c: Enable disassembler_needs_relocs for PRU.
gas * testsuite/gas/pru/extern.s: New test for print of U16_PMEMM
relocation.
* testsuite/gas/pru/extern.d: New test driver.
This fixes these failures on 64 bit which currently occur when running
the Binutils testsuite with a default PIE compiler.
< FAIL: Build rdynamic-1
< FAIL: Build dynamic-1
< FAIL: Build pr22269-1
bfd/ChangeLog:
2017-12-13 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
* elf64-s390.c (elf_s390_adjust_dynamic_symbol): Use
UNDEFWEAK_NO_DYNAMIC_RELOC.
(allocate_dynrelocs): Likewise.
(elf_s390_relocate_section): Check resolved_to_zero.
(elf_s390_finish_dynamic_symbol): Don't generate runtime reloc if
UNDEFWEAK_NO_DYNAMIC_RELOC.
The "server" command prefix no longer turns confirmation queries off.
We can reproduce this with any program by tring to delete all breakpoints,
for instance:
(gdb) break main
Breakpoint 1 at 0x40049b: file /[...]/break-fun-addr1.c, line 21.
(gdb) server delete breakpoints
Delete all breakpoints? (y or n)
GDB should not be asking "Delete all breakpoints? (y or n)", but
instead just delete all breakpoints without asking for confirmation.
Looking at utils.c::defaulted_query gives a glimpse of how this feature
is expected to work:
/* Automatically answer the default value if the user did not want
prompts or the command was issued with the server prefix. */
if (!confirm || server_command)
return def_value;
So, it relies on the server_command global to be set when the "server "
command prefix is used, which is no longer the case since the following
commit:
commit b69d38afde
Date: Wed Mar 9 18:25:00 2016 +0000
Subject: Command line input handling TLC
The patch was simplifying the handling for the command line, and
I believe there was just a small oversight of removing the setting
of the server_command global.
This patch restores that, and adds a testcase to make sure we test
that feature.
gdb/ChangeLog:
* event-top.c (handle_line_of_input): Set server_command.
gdb/testsuite/ChangeLog:
* gdb.base/server-del-break.c: New file.
* gdb.base/server-del-break.exp: New file.
Tested on x86_64-linux, no regression.
This is the testsuite for the or1k sim, it tests running many of the
basic architecture instructions on the openrisc sim.
sim/testsuite/sim/or1k/ChangeLog:
2017-12-12 Peter Gavin <pgavin@gmail.com>
Stafford Horne <shorne@gmail.com>
* add.S: New file.
* alltests.exp: New file.
* and.S: New file.
* basic.S: New file.
* div.S: New file.
* ext.S: New file.
* find.S: New file.
* flag.S: New file.
* fpu.S: New file.
* jump.S: New file.
* load.S: New file.
* mac.S: New file.
* mfspr.S: New file.
* mul.S: New file.
* or.S: New file.
* or1k-asm-test-env.h: New file.
* or1k-asm-test-helpers.h: New file.
* or1k-asm-test.h: New file.
* or1k-asm.h: New file.
* or1k-test.ld: New file.
* ror.S: New file.
* shift.S: New file.
* spr-defs.h: New file.
* sub.S: New file.
* xor.S: New file.
sim/testsuite/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Peter Gavin <pgavin@gmail.com>
* configure: Regenerated.
These are separted out to make the patch easier to read and smaller.
sim/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Peter Gavin <pgavin@gmail.com>
* configure: Regenerated.
* or1k/aclocal.m4: Generated.
* or1k/config.in: Generated.
* or1k/configure: Generated.
These are the simulator files generated by cgen. These are split out
from the main sim patch to make the patch easier to review and smaller.
sim/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Peter Gavin <pgavin@gmail.com>
* or1k/arch.c: Generated.
* or1k/arch.h: Generated.
* or1k/cpu.c: Generated.
* or1k/cpu.h: Generated.
* or1k/cpuall.h: Generated.
* or1k/decode.c: Generated.
* or1k/decode.h: Generated.
* or1k/model.c: Generated.
* or1k/sem-switch.c: Generated.
* or1k/sem.c: Generated.
This adds the OpenRISC 32-bit sim target. The OpenRISC sim is a CGEN
based sim so the bulk of the code is generated from the .cpu files by
CGEN. The engine decode and execute logic in mloop uses scache with
pseudo-basic-block extraction and supports both full and fast (switch)
modes.
The sim does not implement an mmu at the moment. The sim does implement
fpu instructions via the common sim-fpu implementation.
sim/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Peter Gavin <pgavin@gmail.com>
* configure.tgt: Add or1k sim.
* or1k/README: New file.
* or1k/Makefile.in: New file.
* or1k/configure.ac: New file.
* or1k/mloop.in: New file.
* or1k/or1k-sim.h: New file.
* or1k/or1k.c: New file.
* or1k/sim-if.c: New file.
* or1k/sim-main.h: New file.
* or1k/traps.c: New file.
The test case requires adding a nop instruction. For or1k the
instruction is `l.nop`. This change uses the correct operation.
gdb/testsuite/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
* gdb.base/bp-permanent.c: Define nop of or1k.
This patch prepares the current GDB port of the OpenRISC processor from
https://github.com/openrisc/binutils-gdb for upstream merging.
Testing has been done with a cgen sim provided in a separate patch. This
has been tested with 2 toolchains. GCC [1] 5.4.0 from the OpenRISC
project with Newlib [2] and GCC 5.4.0 with Musl [3] 1.1.4.
It supports or1knd (no delay slot target).
The default target is or1k (with delay slot).
You can change the target arch with:
(gdb) set architecture or1knd
The target architecture is assumed to be or1knd
[1] https://github.com/openrisc/or1k-gcc
[2] https://github.com/openrisc/newlib
[3] https://github.com/openrisc/musl-cross
gdb/doc/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Stefan Wallentowitz <stefan@wallentowitz.de>
Franck Jullien <franck.jullien@gmail.com>
Jeremy Bennett <jeremy.bennett@embecosm.com>
* gdb.texinfo: Add OpenRISC documentation.
gdb/ChangeLog:
2017-12-12 Stafford Horne <shorne@gmail.com>
Stefan Wallentowitz <stefan@wallentowitz.de>
Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Franck Jullien <franck.jullien@gmail.com>
Jeremy Bennett <jeremy.bennett@embecosm.com>
* configure.tgt: Add targets for or1k and or1knd.
* or1k-tdep.c: New file.
* or1k-tdep.h: New file.
* features/Makefile: Add or1k.xml to build.
* features/or1k.xml: New file.
* features/or1k-core.xml: New file.
* features/or1k.c: Generated.
The TOC pointer register, r2, on powerpc64 is generally not mentioned
in debug info. It is saved and restored by call linkage code, and
set to the callee value either by call stub code (ELFv1) or in the
callee global entry point code (ELFv2). A call stub uses the caller
TOC pointer to access the PLT. So for gdb to read the correct PLT
entry in order to determine the destination of the trampoline, gdb
needs to know the caller r2. When skipping over trampolines in the
normal forward direction, the caller r2 is simply the current value of
r2 (at the start of the trampoline). However, when reversing over
trampolines the current value of r2 is that for the callee. Using
that value results in wild reads of memory rather than the correct PLT
entry.
This patch corrects the value of r2 by using the value saved on the
stack for reverse execution. Note that in reverse execution mode it
isn't really necessary for skip_trampoline_code to return the actual
destination, so we're doing a little more work than needed here. Any
non-zero return value would do (and it would be nicer if the interface
was changed to return the start of the stub).
PR tdep/22576
* ppc64-tdep.c (ppc64_plt_entry_point): Rewrite to take TOC-relative
PLT offset, and retrieve r2 from stack when executing in reverse.
(ppc64_standard_linkage1_target): Drop pc param. Calculate offset
rather than PLT address.
(ppc64_standard_linkage2_target): Likewise.
(ppc64_standard_linkage3_target): Likewise.
(ppc64_standard_linkage4_target): Likewise.
(ppc64_skip_trampoline_code_1): Adjust to suit.
Commit
remote: C++ify thread_item and threads_listing_context
21fe1c752e
broke the test gdb.threads/names.exp. The problem is that since we now
use an std::string to hold the extra_info, an empty string is returned
by target_extra_thread_info to print_thread_info_1 when the remote stub
didn't send any extra info, instead of NULL before. Because of that,
print_thread_info_1 prints the extra info between parentheses, which
results in some spurious empty parentheses.
Expected: * 1 Thread 22752.22752 "main" all_threads_ready () at ...
Actual : * 1 Thread 22752.22752 "main" () all_threads_ready () a ...
Since the bug was introduced by a behavior change in the remote target,
I chose to fix it on the remote target side by making it return NULL
when the extra string is empty. This will avoid possibly changing the
behavior of the common code and affecting other targets.
The name field has the same problem. If a remote stub returns no thread
names, remote_thread_name will return an empty string instead of NULL,
so print_thread_info_1 will show empty quotes ("") instead of nothing.
gdb/ChangeLog:
PR gdb/22556
* remote.c (remote_thread_name): Return NULL if name is empty.
(remote_threads_extra_info): Return NULL if extra info is empty.
Gold plugins may wish to further process an input file added by a plugin. For
example, the plugin may need to assign a unique segment for sections in a
plugin-generated input file. This patch adds a plugin callback that the linker
will call when reading symbols from a new input file added after the
all_symbols_read event (i.e. an input file added by a plugin).
2017-12-11 Stephen Crane <sjc@immunant.com>
* plugin-api.h: Add new plugin hook to allow processing of input
files added by a plugin.
(ld_plugin_new_input_handler): New function hook type.
(ld_plugin_register_new_input): New interface.
(LDPT_REGISTER_NEW_INPUT_HOOK): New enum val.
(tv_register_new_input): New member.
* plugin.cc (Plugin::load): Include hooks for register_new_input
in transfer vector.
(Plugin::new_input): New function.
(register_new_input): New function.
(Plugin_manager::claim_file): Call Plugin::new_input if in
replacement phase.
* plugin.h (Plugin::set_new_input_handler): New function.
* testsuite/plugin_new_section_layout.c: New plugin to test
new_input plugin API.
* testsuite/plugin_final_layout.sh: Add new input test.
* testsuite/Makefile.am (plugin_layout_new_file): New test case.
* testsuite/Makefile.in: Regenerate.
When checking mapping symbols backwardly, the section which defines the symbol
is not considerted. This patch fixes this by moving the section checking code
into get_sym_code_type () function which is shared by forward and backword
mapping symbol searching.
opcodes/
2017-12-11 Petr Pavlu <petr.pavlu@arm.com>
Renlin Li <renlin.li@arm.com>
* aarch64-dis.c (print_insn_aarch64): Move symbol section check ...
(get_sym_code_type): Here.
binutils/
2017-12-11 Renlin Li <renlin.li@arm.com>
* testsuite/binutils-all/aarch64/objdump.d: New.
* testsuite/binutils-all/aarch64/objdump.s: New.
As reported at
<https://sourceware.org/ml/gdb-patches/2017-12/msg00229.html>, this
commit:
~~~~
commit abccd1e7b7
Author: Jan Kratochvil <jan.kratochvil@redhat.com>
AuthorDate: Fri Dec 8 22:44:11 2017 +0000
Change dwarf2_initialize_objfile's return value
dwarf2_initialize_objfile was returning boolean whether it is psymtabs
or .gdb_index while now it needs to return also whether it is
.debug_names.
~~~~
breaks non-ELF-target builds:
dwarf2read.o: In function `dwarf2_initialize_objfile(objfile*)':
/home/yao.qi/SourceCode/gnu/binutils-gdb/gdb/dwarf2read.c:6486:
undefined reference to `elf_sym_fns_gdb_index'
/home/yao.qi/SourceCode/gnu/binutils-gdb/gdb/dwarf2read.c:6490:
undefined reference to `elf_sym_fns_debug_names'
/home/yao.qi/SourceCode/gnu/binutils-gdb/gdb/dwarf2read.c:6495:
undefined reference to `elf_sym_fns_lazy_psyms'
collect2: error: ld returned 1 exit status
Makefile:1920: recipe for target 'gdb' failed
because gdb/elfread.c is not included in the gdb build unless bfd also
includes elf support.
Fix this by reverting the patch mentioned above and at the same time
re-adding .debug_names support by adding a new output parameter to
dwarf2_initialize_objfile to indicate the index variant in use. We
can reuse the new dw_index_kind enum in dwarf2read.c for that.
gdb/ChangeLog:
2017-12-11 Pedro Alves <palves@redhat.com>
* defs.h (elf_sym_fns_lazy_psyms, elf_sym_fns_gdb_index)
(elf_sym_fns_debug_names): Move to elfread.c.
* dwarf2read.c (dwarf2_initialize_objfile): Return a boolean
instead of a sym_fns and add 'index_kind' output parameter. Fill
the latter in with the index variant kind if using an index.
(enum dw_index_kind): Moved to symfile.h.
* elfread.c (elf_sym_fns_gdb_index, elf_sym_fns_debug_names)
(elf_sym_fns_lazy_psyms): Move from defs.h.
(elf_symfile_read): Adjust to new dwarf2_initialize_objfile
interface.
* symfile.h (enum class dw_index_kind): New, moved from
dwarf2read.c.
(dwarf2_initialize_objfile): Change prototype.
When converting parts of the mantissa to MPFR, we need to make sure to do
an *unsigned* conversion. Since we convert at most 32 bits at a time,
stored in an unsigned long, this doesn't matter on systems where "long"
is larger than 32 bits. But on systems where it is 32 bits, we can get
conversion errors.
gdb/ChangeLog
2017-12-11 Ulrich Weigand <uweigand@de.ibm.com>
* target-float.c (mpfr_float_ops::from_target): Use mpfr_set_ui
instead of mpfr_set_si to convert mantissa bits.
Recent versions of GNAT are capable of reordering record components
to make their access for efficient. This patch adapts this test to
accept both cases (reordered or not).
gdb/testsuite/ChangeLog:
* gdb.ada/variant_record_packed_array.exp: Adapt test to accept
output with components being reordered.
Tested on x86_64-linux.
There was a difference between C++ dispatch table and Ada's in the
way the Offset_To_Top field is used to determined the base address
of an object:
* in C++ it is a negative offset, so converting abstract interface to
deriving object requires adding this offset to “this”;
* in Ada, it was a positive offset, so the same conversion required
subtracting the offset value.
So in ada, the base address for a tagged type was computed using this formula:
base_address = value_address (obj) - offset_to_top;
The offset_to_top value was previously set to 0 or a positive value.
With recent version of AdaCore's GNAT compiler, the offset has been
changed to match C++, which means it's set to zero or a negative value
As a result, the new formula has to be:
base_address = value_address (obj) + offset_to_top;
Because we want to support old code compiled before GNAT compiler change
done in 19.0w (20171023-64) with this version and future versions of gdb,
then we change the sign of the offset_to_top if required. Required here
means if offset_to_top is positive since it indicates that the code has
been compiled with an old GNAT compiler.
This patch changes the formula as described above.
Also, one side-effect of offset_to_top now being negative is that
we now have to worry about the sign when we read its value from the
inferior. Up to now, we have been reading its value using the data
address builtin type. But since addresses are not always signed, we
now need to make sure we use the proper type (type Storage_Offset
from System.Storage_Elements). Ideally, we would be looking this type
up from the inferior, and then use that type. However, it is not
guaranteed that this type always be described in the debugging
information, so this patch just builds our own, adding it to Ada's
list of primitive types.
gdb/ChangeLog:
* ada-lang.c (ada_tag_value_at_base_address): Change the way
tagged type base address is computed.
(enum ada_primitive_types) <ada_primitive_type_storage_offset>:
New enumerate.
(ada_language_arch_info): Set the ada_primitive_type_storage_offset
element of lai->primitive_type_vector.
Tested on x86_64-linux. Fixes the following tests when using the newer
version of the compiler.
gdb.ada/iwide.exp: print My_Drawable
gdb.ada/iwide.exp: print d_access.all
gdb.ada/iwide.exp: print dp_access.all
gdb.ada/mi_interface.exp: create ggg1 varobj (unexpected output)
gdb.ada/mi_interface.exp: list ggg1's children (unexpected output)
gdb.mi/mi-var-rtti.exp: run to mi-var-rtti.cc:63 (set breakpoint) (unexpected output)
gdb.mi/mi-var-rtti.exp: run to mi-var-rtti.cc:63 (set breakpoint)
Recent versions of GNAT are capable of reordering record components
to make their access for efficient. This patch adapts this test to
accept both cases (reordered or not).
gdb/testsuite/ChangeLog:
* gdb.ada/pkd_arr_elem.exp: Adapt "print test" test to accept
output with components being reordered.
Now that dw2_expand_symtabs_matching_symbol works with the abstract
mapped_index_base, we can make mock_mapped_index inherit
mapped_index_base too instead of having it pretend to be a real
.gdb_index table.
gdb/ChangeLog:
2017-12-08 Pedro Alves <palves@redhat.com>
* dwarf2read.c (mock_mapped_index): Reimplement as an extension of
mapped_index_base.
(check_match): Adjust to use mock_index directly.
(check_find_bounds_finds)
(test_mapped_index_find_name_component_bounds): Adjust to work
with a mapped_index_base.
dw2_debug_names_expand_symtabs_matching currently doesn't support
symbol_name_match_type::WILD, it always matches symbol names fully.
The .gdb_index code supports via dw2_expand_symtabs_matching_symbol,
which builds the mapped_index::name_components table on demand, and
then binary searches that table.
The .debug_names names index is pretty much the same as the .gdb_index
names index, i.e., a list of fully-qualified names with no
parameter/overload info. (There's no
what-is-the-language-of-symbol-name info in .debug_names either,
unfortunately.)
So this fixes .debug_names by factoring out the related .gdb_index
code out of the mapped_index class to a base class that is inherited
by both the .gdb_index (mapped_index) and .debug_names
(mapped_debug_names) map classes.
gdb/ChangeLog:
2017-12-08 Pedro Alves <palves@redhat.com>
* dwarf2read.c (struct mapped_index_base): New, partially factored
out from ...
(struct mapped_index): ... this. Inherit mapped_index_base.
(mapped_index::symbol_name_slot_invalid):
(mapped_index::symbol_name_at): Add override marker.
(mapped_index::symbol_name_count): New.
(struct mapped_debug_names): Inherit mapped_index_base.
(mapped_debug_names::symbol_name_at): New.
(mapped_debug_names::symbol_name_count): New.
(mapped_index::find_name_components_bounds): Rename to ...
(mapped_index_base::find_name_components_bounds): ... this.
(mapped_index::build_name_components): Rename to ...
(mapped_index_base::build_name_components): ... this. Adjust to
use mapped_index_base::symbol_name_count and
mapped_index_base::symbol_name_slot_invalid.
(dw2_expand_symtabs_matching_symbol): Take a mapped_index_base
instead of a mapped_index. Use
dw2_expand_symtabs_matching_symbol.
This replaces a couple ptr+size pairs with gdb::array_view in the
.gdb_index code, and simplifies things by using an aggregate for the
type of the symbol table hash bucket instead of having to consider the
distinction between size of table vs number of slots and access name
vs vec by index.
gdb/ChangeLog:
2017-12-08 Pedro Alves <palves@redhat.com>
* dwarf2read.c (mapped_index::symbol_table_slot): New.
(mapped_index::address_table): Now a gdb::array_view of const
gdb_byte.
(mapped_index::symbol_table): Now a gdb::array_view of
symbol_table_slot.
(mapped_index::address_table_size)
(mapped_index::symbol_table_slots): Delete.
(create_addrmap_from_index): Adjust.
(find_slot_in_mapped_hash): Adjust.
(read_index_from_section): Adjust.
(dwarf2_read_index): Adjust.
Some testcases needed to be updated as they were missing
.debug_aranges. While that does not matter for no-index (as GDB
builds the mapping internally during dwarf2_build_psymtabs_hard) and
neither for .gdb_index (as GDB uses that internally built mapping
which it stores into .gdb_index) it does matter for .debug_names as
that simply assumes existing .debug_aranges from GCC.
gdb/ChangeLog:
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
Pedro Alves <palves@redhat.com>
* defs.h (elf_sym_fns_debug_names): New declaration.
* dwarf2read.c: Include "hash_enum.h".
(mapped_debug_names): New.
(struct dwarf2_per_objfile): Add debug_names, debug_aranges and
debug_names_table.
(dwarf2_elf_names): Add ".debug_names" and ".debug_aranges".
(struct dwz_file): Add debug_names.
(dwarf2_per_objfile::locate_sections): Handle debug_names and
debug_aranges.
(locate_dwz_sections): Handle debug_names.
(create_signatured_type_table_from_debug_names)
(create_addrmap_from_aranges): New.
(dwarf2_read_index): Update function comment.
(dwarf5_augmentation): Moved up.
(read_debug_names_from_section, create_cus_from_debug_names_list)
(create_cus_from_debug_names, dwarf2_read_debug_names): New.
(dwarf5_djb_hash): Moved up.
(dw2_debug_names_iterator): New.
(read_indirect_string_at_offset): New declaration.
(mapped_debug_names::namei_to_name)
(dw2_debug_names_iterator::find_vec_in_debug_names)
(dw2_debug_names_iterator::next, dw2_debug_names_lookup_symbol)
(dw2_debug_names_dump, dw2_debug_names_expand_symtabs_for_function)
(dw2_debug_names_expand_symtabs_matching, dwarf2_debug_names_functions):
New.
(dwarf2_initialize_objfile): Return also elf_sym_fns_debug_names.
(debug_names::djb_hash): Rename it to dwarf5_djb_hash.
(debug_names::build): Update djb_hash caller.
(write_debug_names): Move out and rename augmentation to
dwarf5_augmentation.
* elfread.c (elf_sym_fns_debug_names): New.
* psymtab.h (dwarf2_debug_names_functions): New declaration.
* symfile.h (struct dwarf2_debug_sections): Add debug_names and
debug_aranges.
* xcoffread.c (dwarf2_xcoff_names): Add debug_names and debug_aranges.
gdb/testsuite/ChangeLog:
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
Pedro Alves <palves@redhat.com>
* gdb.base/maint.exp (check for .gdb_index): Check also for
.debug_names.
* gdb.dlang/watch-loc.c (.debug_aranges): New.
* gdb.dwarf2/dw2-case-insensitive-debug.S: Likewise.
* gdb.dwarf2/gdb-index.exp (check if index present, .gdb_index used)
(.gdb_index used after symbol reloading): Support also .debug_names.
* gdb.mi/dw2-ref-missing-frame-func.c (.debug_aranges): New.
The DWARF-5 .debug_names consumer patch will want to use an
std::unordered_map with an enum as key type, like:
std::unordered_map<sect_offset, dwarf2_per_cu_data*>
That doesn't work in C++11 in non-recent compilers due to a language
defect:
http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#2148
~~~
In file included from /usr/include/c++/5.3.1/bits/hashtable.h:35:0,
from /usr/include/c++/5.3.1/unordered_set:47,
from src/gdb/dwarf2read.c:79:
/usr/include/c++/5.3.1/bits/hashtable_policy.h: In instantiation of ‘struct std::__detail::__is_noexcept_hash<sect_offset, std::hash<sect_offset> >’:
/usr/include/c++/5.3.1/type_traits:137:12: required from ‘struct std::__and_<std::__is_fast_hash<std::hash<sect_offset> >, std::__detail::__is_noexcept_hash<sect_offset, std::hash<sect_offset> > >’
/usr/include/c++/5.3.1/type_traits:148:38: required from ‘struct std::__not_<std::__and_<std::__is_fast_hash<std::hash<sect_offset> >, std::__detail::__is_noexcept_hash<sect_offset, std::hash<sect_offset> > > >’
/usr/include/c++/5.3.1/bits/unordered_map.h💯66: required from ‘class std::unordered_map<sect_offset, dwarf2_per_cu_data*>’
src/gdb/dwarf2read.c:3260:30: required from here
/usr/include/c++/5.3.1/bits/hashtable_policy.h:85:34: error: no match for call to ‘(const std::hash<sect_offset>) (const sect_offset&)’
noexcept(declval<const _Hash&>()(declval<const _Key&>()))>
^
In file included from /usr/include/c++/5.3.1/bits/move.h:57:0,
from /usr/include/c++/5.3.1/bits/stl_pair.h:59,
from /usr/include/c++/5.3.1/bits/stl_algobase.h:64,
from /usr/include/c++/5.3.1/bits/char_traits.h:39,
from /usr/include/c++/5.3.1/string:40,
from /home/pedro/gdb/mygit/src/gdb/common/common-utils.h:23,
from /home/pedro/gdb/mygit/src/gdb/common/common-defs.h:78,
from /home/pedro/gdb/mygit/src/gdb/defs.h:28,
from /home/pedro/gdb/mygit/src/gdb/dwarf2read.c:31:
~~~
This commits adds a helper replacement.
gdb/ChangeLog:
2017-12-08 Pedro Alves <palves@redhat.com>
* common/hash_enum.h: New file.
Preparation for the next patch.
gdb/ChangeLog
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* dwarf2read.c (create_cu_from_index_list): New from ...
(create_cus_from_index_list): ... this function, use it.
(dw_expand_symtabs_matching_file_matcher)
(dw2_expand_symtabs_matching_one): New from ...
(dw2_expand_symtabs_matching): ... this function, use them.
dwarf2_initialize_objfile was returning boolean whether it is psymtabs
or .gdb_index while now it needs to return also whether it is
.debug_names.
gdb/ChangeLog
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* defs.h (elf_sym_fns_lazy_psyms, elf_sym_fns_gdb_index): Move here
declarations from elfread.c.
(dwarf2_initialize_objfile): Change return value.
* elfread.c (elf_sym_fns_lazy_psyms, elf_sym_fns_gdb_index): Move these
declarations to defs.h.
(elf_symfile_read): Adjust dwarf2_initialize_objfile caller.
* symfile.h (dwarf2_initialize_objfile): Change return type.
This adds a new "-dwarf-5" switch to "save gdb-index" that makes it
generate index files with DWARF-5 .debug_names/.debug_str sections
instead of GDB's own .gdb_index.
We should probably add a command line option to
contrib/gdb-add-index.sh (incl. cc-with-tweaks.sh) for the new
-dwarf-5 GDB option, and a new target board to make it more convenient
to exercise this. To be done later.
gdb/ChangeLog
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
Pedro Alves <palves@redhat.com>
* contrib/gdb-add-index.sh (index): Rename to ...
(index4): ... here.
(index5, debugstr, debugstrmerge, debugstrerr): New variables.
Support also .debug_names and .debug_str.
* dwarf2read.c: Include cmath, set, list.
(INDEX_SUFFIX): Rename to ...
(INDEX4_SUFFIX): ... here.
(INDEX5_SUFFIX, DEBUG_STR_SUFFIX): New.
(file_write(FILE *, const void *, size_t)): New.
(file_write(FILE *, const std::vector<Elem, Alloc> &)): New.
(data_buf::append_unsigned_leb128, data_buf::empty): New.
(data_buf::file_write): Use ::file_write.
(data_buf::c_str, dwarf5_djb_hash, debug_names)
(check_dwarf64_offsets): New.
(psyms_seen_size, write_gdbindex): New from
write_psymtabs_to_index code.
(dwarf5_gdb_augmentation, write_debug_names, assert_file_size)
(enum dw_index_kind): New.
(write_psymtabs_to_index): New parameter index_kind. Support
filename_str and out_file_str. Move code to write_gdbindex,
possibly call write_debug_names.
(save_gdb_index_command): New parameter -dwarf-5.
(_initialize_dwarf2_read): Document the new parameter -dwarf-5.
gdb/doc/ChangeLog
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
* gdb.texinfo (Index Files): Document .debug_names and -dwarf-5.
--
gdb/contrib/gdb-add-index.sh | 53 ++
gdb/doc/gdb.texinfo | 24 +
gdb/dwarf2read.c | 919 ++++++++++++++++++++++++++++++++++++++++--
3 files changed, 935 insertions(+), 61 deletions(-)
With DWARF-5 .debug_names, the commands to add the index to the symbol
file are more complicated, as now also .debug_str needs to be
modified.
Currently, contrib/cc-with-tweaks.sh calls objcopy to handle the '-i'
option instead of using contrib/gdb-add-index.sh which basically does
the same. To help with .debug_names, this commit makes
contrib/cc-with-tweaks.sh reuse contrib/gdb-add-index.sh instead.
A problem this ran into is whether contrib/cc-with-tweaks.sh should
fail or not when no index is produced.
Currently, contrib/cc-with-tweaks.sh is more quiet (=successful) than
contrib/gdb-add-index.sh and so with no further changes testsuite runs
with an index would "regress". This commit tries to keep the behavior
unchanged. Some cases still error with:
Ada is not currently supported by the index
But some cases (such as some trivial gdb.dwarf2/ testcases with no DWARF data
to index) produce no index while the testcases still PASS now instead of:
-PASS: gdb.arch/i386-bp_permanent.exp: stack pointer value matches
+gdb compile failed, gdb-add-index.sh: No index was created for gdb/testsuite.unix.-m64/outputs/gdb.arch/i386-bp_permanent/i386-bp_permanent
+gdb-add-index.sh: [Was there no debuginfo? Was there already an index?]
+UNTESTED: gdb.arch/i386-bp_permanent.exp: failed to compile
gdb/ChangeLog
2017-12-08 Jan Kratochvil <jan.kratochvil@redhat.com>
Pedro Alves <palves@redhat.com>
* contrib/cc-with-tweaks.sh: Change interpreter to bash, incl. initial
comment.
(GDB_ADD_INDEX): New variable.
<$want_index>: Call $GDB_ADD_INDEX.
While investigating PR gdb/22557 ("Regression:
gdb.base/dtrace-probe.exp"), I noticed that the code is wrongly
declaring a new "expression_up" variable inside the TRY block in
"dtrace_process_dof_probe". This causes the outter "expr" variable to
be empty, which may have an impact later when evaluating the
expression.
This commit fixes that. Unfortunately the script used to test DTrace
probes (gdb/testsuite/lib/pdtrace.in) is not very reliable so I cannot
say whether this commit fixes the PR mentioned above. Nonetheless,
it's an obvious fix and should go in.
gdb/ChangeLog:
2017-12-08 Sergio Durigan Junior <sergiodj@redhat.com>
* dtrace-probe.c (dtrace_process_dof_probe): Do not declare a new
"expression_up" inside the TRY block.
Nowadays, GDB can't set watchpoint on tagged address on AArch64,
(gdb) p p2
$1 = (int *) 0xf000fffffffff474
(gdb) watch *((int *) 0xf000fffffffff474)
Hardware watchpoint 2: *((int *) 0xf000fffffffff474)
(gdb) c
Continuing.
main () at
binutils-gdb/gdb/testsuite/gdb.arch/aarch64-tagged-pointer.c:45
45 void (*func_ptr) (void) = foo;
Unexpected error setting hardware debug registers
This patch is about setting watchpoint on a tagged address. Unlike
breakpoint, watchpoint record the expression rather than the address, and
when a watchpoint is fired, GDB checks the expression value changed
instead of matching address, so we can mask the watchpoint address by
getting rid of non-significant bits of address.
gdb:
2017-12-08 Yao Qi <yao.qi@linaro.org>
* breakpoint.c (update_watchpoint): Call
address_significant.
gdb/testsuite:
2017-12-08 Yao Qi <yao.qi@linaro.org>
* gdb.arch/aarch64-tagged-pointer.c (main): Update.
* gdb.arch/aarch64-tagged-pointer.exp: Add tests for watchpoint.
Tag in tagged address on AArch64 is treated as a non-significant bits of
address, which can be got by gdbarch method significant_addr_bit, and gdb
can clear these bits.
With this patch, when user sets a breakpoint on tagged address on AArch64,
GDB will drop the top byte of address, and put breakpoint at the new place,
as shown below,
(gdb) hbreak *func_ptr
warning: Breakpoint address adjusted from 0xf000000000400690 to 0x00400690.
Hardware assisted breakpoint 2 at 0x400690
(gdb) break *func_ptr
warning: Breakpoint address adjusted from 0xf000000000400690 to 0x00400690.
Breakpoint 3 at 0x400690
When program hits a breakpoint, the stopped pc reported by Linux kernel is
the address *without* tag, so it is better the address recorded in
breakpoint location is the one without tag too, so we can still match
breakpoint location address and stopped pc reported by Linux kernel, by
simple compare.
gdb:
2017-12-08 Yao Qi <yao.qi@linaro.org>
* breakpoint.c (adjust_breakpoint_address): Call
address_significant.
gdb/testsuite:
2017-12-08 Yao Qi <yao.qi@linaro.org>
* gdb.arch/aarch64-tagged-pointer.c (main): Update.
* gdb.arch/aarch64-tagged-pointer.exp: Add test for breakpoint.
ARMv8 supports tagged address, that is, the top one byte in address
is ignored. It is always enabled on aarch64-linux. See
https://www.kernel.org/doc/Documentation/arm64/tagged-pointers.txt
The tag in the tagged address is modeled as non-significant bits in
address, so this patch adds a new gdbarch method significant_addr_bit and
clear the non-significant bits (the top byte in ARMv8) of the virtual
address at the point before passing address to target cache layer. IOW,
the address used in the target cache layer is already cleared.
Before this patch,
(gdb) x/x 0x0000000000411030
0x411030 <global>: 0x00000000
(gdb) x/x 0xf000000000411030
0xf000000000411030: Cannot access memory at address 0xf000000000411030
After this patch,
(gdb) x/x 0x0000000000411030
0x411030 <global>: 0x00000000
(gdb) x/x 0xf000000000411030
0xf000000000411030: 0x00000000
Note that I used address_significant in paddress, but it causes a
regression gdb.base/long_long.exp, because gdb clears the non-significant
bits in address, but test still expects them.
p/a val.oct^M
$24 = 0x2ee53977053977^M
(gdb) FAIL: gdb.base/long_long.exp: p/a val.oct
so I defer the change there.
gdb:
2017-12-08 Yao Qi <yao.qi@linaro.org>
* aarch64-tdep.c (aarch64_gdbarch_init): Install gdbarch
significant_addr_bit.
* gdbarch.sh (significant_addr_bit): New.
* gdbarch.c, gdbarch.h: Re-generated.
* target.c (memory_xfer_partial): Call address_significant.
* utils.c (address_significant): New function.
* utils.h (address_significant): Declare.
2017-12-08 Yao Qi <yao.qi@linaro.org>
gdb/testsuite:
* gdb.arch/aarch64-tagged-pointer.c: New file.
* gdb.arch/aarch64-tagged-pointer.exp: New file.
This replaces parse_format_string with a class, removing some
constructors along the way. While doing this, I found that one
argument to gen_printf is unused, so I removed it.
Also, I am not completely sure, but the use of `release' in
maint_agent_printf_command and parse_cmd_to_aexpr seems like it may
leak expressions.
Regression tested by the buildbot.
ChangeLog
2017-12-08 Tom Tromey <tom@tromey.com>
* printcmd.c (ui_printf): Update. Use std::vector.
* common/format.h (struct format_piece): Add constructor.
<string>: Now const.
(class format_pieces): New class.
(parse_format_string, free_format_pieces)
(free_format_pieces_cleanup): Remove.
* common/format.c (format_pieces::format_pieces): Rename from
parse_format_string. Update.
(free_format_pieces, free_format_pieces_cleanup): Remove.
* breakpoint.c (parse_cmd_to_aexpr): Update. Use std::vector.
* ax-gdb.h (gen_printf): Remove argument.
* ax-gdb.c (gen_printf): Remove "frags" argument.
(maint_agent_printf_command): Update. Use std::vector.
gdbserver/ChangeLog
2017-12-08 Tom Tromey <tom@tromey.com>
* ax.c (ax_printf): Update.
