Recent commit b2829fcf9b ("[gdb] Fix rethrow exception slicing in
insert_bp_location") introduced macro RETHROW_ON_TARGET_CLOSE_ERROR.
I wrote this as a macro in order to have the rethrowing throw be part of the
same function as the catch, but as it turns out that's not necessary.
Rewrite into a function.
Tested on x86_64-linux.
Commit f34652de0b ("internal_error: remove need to pass
__FILE__/__LINE__") renamed the internal_error function to
internal_error_loc. Change gdb-gdb.gdb.in accordingly.
Change-Id: I876e1623607b6becf74ade53d102ead53a74ed86
The R_RISCV_DELETE relocations are no longer deleted at another relax pass,
so we should reset 'again' flag to true for _bfd_riscv_relax_pc, while the
deleted bytes are marked as R_RISCV_DELETE.
bfd/
* elfnn-riscv.c (_bfd_riscv_relax_pc): Set `again' to true while the
deleted bytes are marked as R_RISCV_DELETE.
The riscv port does deletion and symbol table update for each relocation
while relaxing, so we are moving section bytes and scanning symbol table once
for each relocation. Compared to microblaze port, they record the relaxation
changes into a table, then do the deletion and symbol table update once per
section, rather than per relocation. Therefore, they should have better link
time complexity than us.
To improve the link time complexity, this patch try to make the deletion in
linear time. Compared to record the relaxation changes into a table, we
replace the unused relocation with R_RISCV_DELETE for the deleted bytes, and
then resolve them at the end of the section. Assuming the number of
R_RISCV_DELETE is m, and the number of symbols is n, the total link complexity
should be O(m) for moving section bytes, and O(m*n^2) for symbol table update.
If we record the relaxation changes into the table, and then sort the symbol
table by values, then probably can reduce the time complexity to O(m*n*log(n))
for updating symbol table, but it doesn't seem worth it for now.
bfd/
* elfnn-riscv.c (_riscv_relax_delete_bytes): Renamed from
riscv_relax_delete_bytes, updated to reduce the tiem complexity to O(m)
for memmove.
(typedef relax_delete_t): Function pointer declaration of delete functions.
(riscv_relax_delete_bytes): Can choose to use _riscv_relax_delete_piecewise
or _riscv_relax_delete_immediate for deletion.
(_riscv_relax_delete_piecewise): Just mark the deleted bytes as R_RISCV_DELETE.
(_riscv_relax_delete_immediate): Delete some bytes from a section while
relaxing.
(riscv_relax_resolve_delete_relocs): Delete the bytes for R_RISCV_DELETE
relocations from a section, at the end of _bfd_riscv_relax_section.
(_bfd_riscv_relax_call): Mark deleted bytes as R_RISCV_DELETE by reusing
R_RISCV_RELAX.
(_bfd_riscv_relax_lui): Likewise, but reuse R_RISCV_HI20 for lui, and reuse
R_RISCV_RELAX for c.lui.
(_bfd_riscv_relax_tls_le): Likewise, but resue R_RISCV_TPREL_HI20 and
R_RISCV_TPREL_ADD.
(_bfd_riscv_relax_pc): Likewise, but resue R_RISCV_PCREL_HI20 for auipc.
(_bfd_riscv_relax_align): Updated, don't need to resue relocation since
calling _riscv_relax_delete_immediate.
(_bfd_riscv_relax_delete): Removed.
(_bfd_riscv_relax_section): Set riscv_relax_delete_bytes for each relax_func,
to delete bytes immediately or later. Call riscv_relax_resolve_delete_relocs
to delete bytes for DELETE relocations from a section.
In the lm32 simulator, I was seeing some warnings about missing
function declarations.
The lm32 simulator has a weird header structure, in order to pull in
the full cpu.h header we need to define WANT_CPU_LM32BF. This is done
in some files, but not in others. Critically, it's not done in some
files that then use functions declared in cpu.h
In this commit I added the missing #define so that the full cpu.h can
be included.
After doing this there are still a few functions that are used
undeclared, these functions appear to be missing any declarations at
all, so I've added some to cpu.h.
With this done all the warnings when compiling lm32 are resolved for
both gcc and clang, so I've removed the SIM_WERROR_CFLAGS line from
Makefile.in, this allows lm32 to build with -Werror.
There are two places in the h8300 simulator where we assign a variable
to itself. Clang gives a warning for this, which is converted into an
error by -Werror.
Silence the warning by removing the self assignments. As these
assignments were in a complex if/then/else tree, rather than try to
adjust all the conditions, I've just replaced the self assignments
with a comment and an empty statement.
In the ppc simulator, clang was warning about some code like this:
busy_ptr->nr_writebacks = 1 + (PPC_ONE_BIT_SET_P(out_vmask)) ? 1 : 2;
The warning was:
operator '?:' has lower precedence than '+'; '+' will be evaluated first
I suspect that this is not the original authors intention.
PPC_ONE_BIT_SET_P is going to be 0 or 1, so if we evaluate the '+'
first, the condition will always be non-zero, so true. The whole
expression could then be simplified to just '1', which doesn't make
much sense.
I suspect the answer the author was expecting was either 2 or 3. Why
they didn't just write:
busy_ptr->nr_writebacks = (PPC_ONE_BIT_SET_P(out_vmask)) ? 2 : 3;
I have no clue, however, to keep the structure of the code unchanged,
I've updated things to:
busy_ptr->nr_writebacks = 1 + (PPC_ONE_BIT_SET_P (out_vmask) ? 1 : 2);
which silences the warning from clang, and is, I am guessing, what the
original author intended.
In the ppc simulator's do_fstat function, which provides the fstat
call for the simulator, if the fstat is going to fail then we
currently write an uninitialized buffer into the simulated target.
In theory, I think this is fine, we also write the error status into
the simulated target, so, given that the fstat has failed, the target
shouldn't be relying on the buffer contents.
However, writing an uninitialized buffer means we might leak simulator
private data into the simulated target, which is probably a bad thing.
Plus it probably makes life easier if something consistent, like all
zeros, is written rather than random junk, which might look like a
successful call (except for the error code).
So, in this commit, I initialize the stat buffer to zero before
it is potentially used. If the stat call is not made then the buffer
will be left initialized as all zeros.
The ppc simulator, in sim_create_inferior, tries to print the function
local entry_point variable before the variable is initialized.
In this commit, I defer the debug print line until the variable has
been initialized.
The preferred way of rethrowing an exception is by using throw without
expression, because it avoids object slicing of the exception [1].
Fix this in insert_bp_location.
Tested on x86_64-linux.
[1] https://en.cppreference.com/w/cpp/language/throw
Approved-By: Andrew Burgess <aburgess@redhat.com>
The preferred way of rethrowing an exception is by using throw without
expression, because it avoids object slicing of the exception [1].
Fix this in gdb_pretty_print_disassembler::pretty_print_insn.
Tested on x86_64-linux.
[1] https://en.cppreference.com/w/cpp/language/throw
Approved-By: Andrew Burgess <aburgess@redhat.com>
If a custom arm-elf target is disabling the shared support, a lot of
failures are reported by the testsuite.
Moreover, some tests try to access libraries which have been explicitly
skipped earlier (eg mixed-lib.so).
ld/ChangeLog:
* testsuite/ld-arm/arm-elf.exp: Separate tests needing shared
lib support.
There's little point in having Intel syntax disassembler tests when the
purpose of a test is assembler functionality: Drop all
*avx512*_vpclmulqdq-wig1-intel.
For *avx512*_vpclmulqdq-wig1 share source with *avx512*_vpclmulqdq.
Finally put in place similar tests for -mvexwig=1.
There's little point in having Intel syntax disassembler tests when the
purpose of a test is assembler functionality: Drop all
*avx512*_vaes-wig1-intel.
For *avx512*_vaes-wig1 share source with *avx512*_vaes. This in
particular makes sure that the 32-bit VL test actually tests any EVEX
encodings in the first place.
Finally put in place similar tests for -mvexwig=1.
When no AVX512-specific functionality is in use, the disassembly of
AVX512VL insns is indistinguishable from their AVX counterparts (if such
exist). Emit the {evex} pseudo-prefix in such cases.
Where applicable drop stray uses of PREFIX_OPCODE from table entries.
I did a gdb build without python support, and during testing ran into FAILs in
test-case gdb.python/tui-window-names.exp.
Fix this by adding the missing skip_python_test.
Tested on x86_64-linux.
Now that we run `check/foo.exp` instead of `check/./foo.exp`,
update the config/ & lib/ exceptions to cover both paths.
Bug: https://sourceware.org/PR29596
Make sure we invoke runtest with the same exp filenames when running in
parallel as it will find when run single threaded. When `runtest` finds
files itself, it will use paths like "aarch64/allinsn.exp". When we run
`find .` with the %p option, it produces "./aarch64/allinsn.exp". Switch
to %P to get "aarch64/allinsn.exp".
Bug: https://sourceware.org/PR29596
These configure scripts check $target and change behavior. They
shouldn't be doing that, but until we can rework the sim to change
behavior based on the input ELF, restore AC_CANONICAL_SYSTEM to
these so that $target is correctly populated.
This was lost in the d3562f83a7
("sim: unify toolchain probing logic") refactor as the logic was
hoisted up to the common code. But the fact the vars weren't
passed down to the sub-configure scripts was missed.
Bug: https://sourceware.org/PR29439
This test sends my CI in an infinite loop of failures. We expect to
have a handful of iterations (5 on my development machine, where the
test passes fine)but the log shows that it went up to 104340 iterations:
FAIL: gdb.base/unwind-on-each-insn.exp - instruction 104340: maint print frame-id
DUPLICATE: gdb.base/unwind-on-each-insn.exp - instruction 104340: maint print frame-id
FAIL: gdb.base/unwind-on-each-insn.exp - instruction 104340: [string equal $fid $main_fid]
FAIL: gdb.base/unwind-on-each-insn.exp - instruction 104340: get hexadecimal valueof "$pc"
Add a max instruction check, exit the loop if we reach 100 iterations.
This should allow the test to fail fast if there's a problem, but 100
iterations should be more than enough for when things are working.
Change-Id: I77978d593aca046068f9209272d82e1675ba17c2
- avoid "GDB proper" to refer to global locus, as object files and
program spaces are also GDB proper.
- gdb.register_unwinder does not accept locus=gdb.
- "a unwinder" -> "an unwinder"
Approved-by: Eli Zaretskii <eliz@gnu.org>
Change-Id: I98c1b1000e1063815238e945ca71ec6f37b5702e
Small cleanup to use std::vector iterators rather than raw pointers.
Approved-By: Tom Tromey <tom@tromey.com>
Change-Id: I8d50dbb3f2d8dad7ff94066a578d523f1f31b590
When building GDB with clang and --enable-ubsan, I get:
UNRESOLVED: gdb.dwarf2/frame-inlined-in-outer-frame.exp: starti prompt
The cause being:
$ ./gdb --data-directory=data-directory -nx -q -readnow testsuite/outputs/gdb.dwarf2/frame-inlined-in-outer-frame/frame-inlined-in-outer-frame
Reading symbols from testsuite/outputs/gdb.dwarf2/frame-inlined-in-outer-frame/frame-inlined-in-outer-frame...
Expanding full symbols from testsuite/outputs/gdb.dwarf2/frame-inlined-in-outer-frame/frame-inlined-in-outer-frame...
/home/simark/src/binutils-gdb/gdb/dwarf2/read.c:11954:47: runtime error: applying non-zero offset 8 to null pointer
I found this to happen with ld-linux on at least Arch Linux and Ubuntu
22.04:
$ ./gdb --data-directory=data-directory -nx -q -readnow -iex "set debuginfod enabled on" /lib64/ld-linux-x86-64.so.2
Reading symbols from /lib64/ld-linux-x86-64.so.2...
Reading symbols from /home/simark/.cache/debuginfod_client/22bd7a2c03d8cfc05ef7092bfae5932223189bc1/debuginfo...
Expanding full symbols from /home/simark/.cache/debuginfod_client/22bd7a2c03d8cfc05ef7092bfae5932223189bc1/debuginfo...
/home/simark/src/binutils-gdb/gdb/dwarf2/read.c:11954:47: runtime error: applying non-zero offset 8 to null pointer
The problem happens when doing this:
sect_offset *offsetp = offsets.data () + 1
When `offsets` is an empty vector, `offsets.data ()` returns nullptr.
Fix it by wrapping that in a `!offsets.empty ()` check.
Change-Id: I6d29ba2fe80ba4308f68effd9c57d4ee8d67c29f
Approved-By: Tom Tromey <tom@tromey.com>
PR symtab/29694 points out a regression caused by the new DWARF
scanner when the cc-with-gdb-index target board is used.
What happens here is that an older version of gdb will make an index
describing the "A" type as:
[737] A: 1 [global, type]
whereas the new gdb says:
[1008] A: 0 [global, type]
Here the old one is correct because the A in CU 0 is just a
declaration without a size:
<1><45>: Abbrev Number: 10 (DW_TAG_structure_type)
<46> DW_AT_name : A
<48> DW_AT_declaration : 1
<48> DW_AT_sibling : <0x6d>
This patch fixes the problem by introducing the idea of a "type
declaration". I think gdb still needs to recurse into these types,
searching for methods, but by marking the type itself as a
declaration, gdb can skip this type during lookups and when writing
the index.
Regression tested on x86-64 using the cc-with-gdb-index board.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29694
A user noticed that gdb would crash when printing a packed array after
doing "set lang c". Packed arrays don't exist in C, but it's
occasionally useful to print things in C mode when working in a non-C
language -- this lets you see under the hood a little bit.
The bug here is that generic value printing does not handle packed
arrays at all. This patch fixes the bug by introducing a new function
to extract a value from a bit offset and width.
The new function includes a hack to avoid problems with some existing
test cases when using -fgnat-encodings=all. Cleaning up this code
looked difficult, and since "all" is effectively deprecated, I thought
it made sense to simply work around the problems.
A user found a bug where an array of packed arrays was printed
incorrectly. The bug here is that the packed array has a bit stride,
but the outer array does not -- and should not. However,
update_static_array_size does not distinguish between an array of
packed arrays and a multi-dimensional packed array, and for the
latter, only the innermost array will end up with a stride.
This patch fixes the problem by adding a flag to indicate whether a
given array type is a constituent of a multi-dimensional array.
PR 29654
* ld.h (struct ld_config_type): Add no_warnings field.
* ldlex.h (enum option_values): Add OPTION_NO_WARNINGS.
* lexsup.c (ld_options): Add --no-warnings.
(parse_args): Add support for -w and --no-warnings.
* ldmisc.c (vfinfo): Return early if the message is a warning and
-w has been enabled.
* ld.texi (options): Document new command line option.
* NEWS: Mention the new feature.
Currently, when using GDB to do reverse debugging, if we try to use the
command "reverse next" to skip a recursive function, instead of skipping
all of the recursive calls and stopping in the previous line, we stop at
the second to last recursive call, and need to manually step backwards
until we leave the first call. This is well documented in PR gdb/16678.
This bug happens because when GDB notices that a reverse step has
entered into a function, GDB will add a step_resume_breakpoint at the
start of the function, then single step out of the prologue once that
breakpoint is hit. The problem was happening because GDB wouldn't give
that step_resume_breakpoint a frame-id, so the first time the breakpoint
was hit, the inferior would be stopped. This is fixed by giving the
current frame-id to the breakpoint.
This commit also changes gdb.reverse/step-reverse.c to contain a
recursive function and attempt to both, skip it altogether, and to skip
the second call from inside the first call, as this setup broke a
previous version of the patch.
When GDB is stopped at a ret instruction and no debug information is
available for unwinding, GDB defaults to the amd64 epilogue unwinder, to
be able to generate a decent backtrace. However, when calculating the
frame id, the epilogue unwinder generates information as if the return
instruction was the whole frame.
This was an issue especially when attempting to reverse debug, as GDB
would place a step_resume_breakpoint from the epilogue of a function if
we were to attempt to skip that function, and this breakpoint should
ideally have the current function's frame_id to avoid other problems
such as PR record/16678.
This commit changes the frame_id calculation for the amd64 epilogue,
so that it is always the same as the dwarf2 unwinder's frame_id.
It also adds a test to confirm that the frame_id will be the same,
regardless of using the epilogue unwinder or not, thanks to Andrew
Burgess.
Co-Authored-By: Andrew Burgess <aburgess@redhat.com>
Similarly to booleans and following the fix for PR python/29217 make
`gdb.parameter' accept `None' for `unlimited' with parameters of the
PARAM_UINTEGER, PARAM_INTEGER, and PARAM_ZUINTEGER_UNLIMITED types, as
`None' is already returned by parameters of the two former types, so
one might expect to be able to feed it back. It also makes it possible
to avoid the need to know what the internal integer representation is
for the special setting of `unlimited'.
Expand the testsuite accordingly.
Approved-By: Simon Marchi <simon.marchi@polymtl.ca>
Also verify PARAM_UINTEGER, PARAM_INTEGER, and PARAM_ZINTEGER parameter
types, in addition to PARAM_ZUINTEGER and PARAM_ZUINTEGER_UNLIMITED
already covered, and verify a choice of existing GDB parameters. Add
verification for reading parameters via `<parameter>.value' in addition
to `gdb.parameter('<parameter>')' as this covers different code paths.
Approved-By: Simon Marchi <simon.marchi@polymtl.ca>
Do not assert that a value intended for an integer parameter, of either
the PARAM_UINTEGER or the PARAM_ZUINTEGER_UNLIMITED type, is boolean,
causing error messages such as:
ERROR: In procedure make-parameter:
ERROR: In procedure gdbscm_make_parameter: Wrong type argument in position 15 (expecting integer or #:unlimited): 3
Error while executing Scheme code.
when initialization with a number is attempted. Instead assert that it
is integer. Keep matching `#:unlimited' keyword as an alternative. Add
suitable test cases.
Approved-By: Simon Marchi <simon.marchi@polymtl.ca>
With gcc 7.5.0 and test-case gdb.base/rtld-step.exp, I run into:
...
gdb compile failed, gcc: error: unrecognized command line option \
'-static-pie'; did you mean '-static'?
...
Silence this by checking in the test-case that -static-pie is supported, and
emitting instead:
...
UNTESTED: gdb.base/rtld-step.exp: \
failed to compile (-static-pie not supported or static libc missing)
...
Tested on x86_64-linux, with:
- gcc 7.5.0: UNTESTED
- gcc 12.2.1 with static glibc not installed: UNTESTED
- gcc 12.2.1 with static glibc installed: PASS
Not only that sim/configure.ac does not AC_SUBST CXXFLAGS,
unless we need C++ compiler like CXX, substitution @CXXFLAGS@ is useless.
Because of this, this commit removes this substitution.
Since
commit 837e225ba1
Author: Jan Beulich <jbeulich@suse.com>
Date: Thu Oct 20 10:01:12 2022 +0200
x86: re-work AVX-VNNI support
moved AVX-VNNI after AVX512-VNNI, vector Disp8 is applied even when VEX
encoding is selected. Check VEX/EVEX encoding before checking vector
operands to avoid vector Disp8 with VEX encoding.
PR gas/29708
* config/tc-i386.c (match_template): Check VEX/EVEX encoding
before checking vector operands.
* testsuite/gas/i386/avx-vnni.d: Updated.
* testsuite/gas/i386/x86-64-avx-vnni.d: Likewise.
* testsuite/gas/i386/avx-vnni.s: Add a Disp32 test.
* testsuite/gas/i386/x86-64-avx-vnni.s: Likewise.
