This regression, as it is exposed by the test added in this commit,
first became noticable with this commit:
commit d182f2797922a305fbd1ef6a483cc39a56b43e02
Date: Mon Mar 8 07:27:57 2021 -0700
Convert c-exp.y to use operations
But, this commit only added converted the C expression parser to make
use of code that was added in this commit:
commit a00b7254fb614af557de7ae7cc0eb39a0ce0e408
Date: Mon Mar 8 07:27:57 2021 -0700
Implement function call operations
And it was this second commit that actually introduced the bugs (there
are two).
In structop_base_operation::evaluate_funcall we build up an argument
list in the vector vals. Later in this function the argument list
might be passed to value_struct_elt.
Prior to commit a00b7254fb614 the vals vector (or argvec as it used to
be called) stored the value for the function callee in the argvec at
index 0. This 'callee' value is what ends up being passed to
evaluate_subexp_do_call, and represents the function to be called, the
value contents are the address of the function, and the value type is
the function signature. The remaining items held in the argvec were
the values to pass to the function. For a non-static member function
the `this' pointer would be at index 1 in the array.
After commit a00b7254fb614 this callee value is now held in a separate
variable, not the vals array. So, for non-static member functions,
the `this' pointer is now at index 0, with any other arguments after
that.
What this means is that previous, when we called value_struct_elt we
would pass the address of argvec[1] as this was the first argument.
But now we should be passing the address of vals[0]. Unfortunately,
we are still passing vals[1], effectively skipping the first
argument.
The second issue is that, prior to commit a00b7254fb614, the argvec
array was NULL terminated. This is required as value_struct_elt
calls search_struct_method, which calls typecmp, and typecmp requires
that the array have a NULL at the end.
After commit a00b7254fb614 this NULL has been lost, and we are
therefore violating the API requirements of typecmp.
This commit fixes both of these regressions. I also extended the
header comments on search_struct_method and value_struct_elt to make
it clearer that the array required a NULL marker at the end.
You will notice in the test attached to this commit that I test
calling a non-static member function, but not calling a static member
function. The reason for this is that calling static member functions
is currently broken due to a different bug. That will be fixed in a
later patch in this series, at which time I'll add a test for calling
a static member function.
gdb/ChangeLog:
PR gdb/27994
* eval.c (structop_base_operation::evaluate_funcall): Add a
nullptr to the end of the args array, which should not be included
in the argument array_view. Pass all the arguments through to
value_struct_elt.
* valops.c (search_struct_method): Update header comment.
(value_struct_elt): Likewise.
gdb/testsuite/ChangeLog:
PR gdb/27994
* gdb.cp/method-call-in-c.cc: New file.
* gdb.cp/method-call-in-c.exp: New file.
When GCC emits .debug_aranges, it adds padding to align the contents
to two times the address size. GCC has done this for many years --
but there is nothing in the DWARF standard that says this should be
done, and LLVM does not seem to add this padding.
It's simple to detect if the padding exists, though: if the contents
of one .debug_aranges CU (excluding the header) are not a multiple of
the alignment that GCC uses, then anything extra must be padding.
This patch changes gdb to correctly read both styles. It removes the
requirement that the padding bytes be zero, as this seemed
unnecessarily pedantic to me.
gdb/ChangeLog
2021-06-25 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (create_addrmap_from_aranges): Change padding
logic.
gdb/testsuite/ChangeLog
2021-06-25 Tom Tromey <tom@tromey.com>
* lib/gdb.exp (add_gdb_index, ensure_gdb_index): Add "style"
parameter.
* gdb.rust/dwindex.exp: New file.
* gdb.rust/dwindex.rs: New file.
dwarf2_cu has a 'language' value, but dwarf2_per_cu_data also holds a
value of this same type. There doesn't seem to be any reason to keep
two copies of this value. This patch removes the field from
dwarf2_cu, and arranges to set the value in the per-CU object instead.
Note that the value must still be set when expanding the full CU.
This is needed because the CUs will not be scanned when a DWARF index
is in use.
gdb/ChangeLog
2021-06-25 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (process_psymtab_comp_unit): Don't set 'lang'.
(scan_partial_symbols, partial_die_parent_scope)
(add_partial_symbol, add_partial_subprogram)
(compute_delayed_physnames, rust_union_quirks)
(process_full_comp_unit, process_full_type_unit)
(process_imported_unit_die, process_die, dw2_linkage_name)
(dwarf2_compute_name, dwarf2_physname, read_import_statement)
(read_file_scope, queue_and_load_dwo_tu, read_func_scope)
(read_variable, dwarf2_get_subprogram_pc_bounds)
(dwarf2_attach_fields_to_type, dwarf2_add_member_fn)
(dwarf2_attach_fn_fields_to_type)
(quirk_ada_thick_pointer_struct, read_structure_type)
(handle_struct_member_die, process_structure_scope)
(read_array_type, read_array_order, prototyped_function_p)
(read_subroutine_type, dwarf2_init_complex_target_type)
(read_base_type, read_subrange_type, read_unspecified_type)
(load_partial_dies, partial_die_info::fixup, set_cu_language)
(new_symbol, need_gnat_info, determine_prefix, typename_concat)
(dwarf2_canonicalize_name, follow_die_offset)
(prepare_one_comp_unit): Update.
* dwarf2/cu.c (dwarf2_cu::start_symtab): Update.
The DWARF reader currently sets the CU's language in two different
spots. It is primarily done in prepare_one_comp_unit, but
read_file_scope also checks the producer and may change the language
based on the result.
This patch consolidates all language-setting into
prepare_one_comp_unit. set_cu_language is renamed and changed not to
set language_defn; instead that is done in prepare_one_comp_unit after
the correct language enum value is chosen.
This fixes a minor latent bug, which is that read_file_scope could set
the language enum value to language_opencl, but then neglected to
reset language_defn in this case.
gdb/ChangeLog
2021-06-25 Tom Tromey <tom@tromey.com>
* dwarf2/read.c (read_file_scope): Don't call set_cu_language.
(dwarf_lang_to_enum_language): Rename from set_cu_language. Don't
set language_defn. Handle DW_LANG_OpenCL.
(prepare_one_comp_unit): Check producer and set language_defn.
This commit adds initial support for catchpoints to the python
breakpoint API.
This commit adds a BP_CATCHPOINT constant which corresponds to
GDB's internal bp_catchpoint. The new constant is documented in the
manual.
The user can't create breakpoints with type BP_CATCHPOINT after this
commit, but breakpoints that already exist, obtained with the
`gdb.breakpoints` function, can now have this type. Additionally,
when a stop event is reported for hitting a catchpoint, GDB will now
report a BreakpointEvent with the attached breakpoint being of type
BP_CATCHPOINT - previously GDB would report a generic StopEvent in
this situation.
gdb/ChangeLog:
* NEWS: Mention Python BP_CATCHPOINT feature.
* python/py-breakpoint.c (pybp_codes): Add bp_catchpoint support.
(bppy_init): Likewise.
(gdbpy_breakpoint_created): Likewise.
gdb/doc/ChangeLog:
* python.texinfo (Breakpoints In Python): Add BP_CATCHPOINT
description.
gdb/testsuite/ChangeLog:
* gdb.python/py-breakpoint.c (do_throw): New function.
(main): Call do_throw.
* gdb.python/py-breakpoint.exp (test_catchpoints): New proc.
This commit adds initial support for catchpoints to the guile
breakpoint API.
This commit adds a BP_CATCHPOINT constant which corresponds to
GDB's internal bp_catchpoint. The new constant is documented in the
manual.
The user can't create breakpoints with type BP_CATCHPOINT after this
commit, but breakpoints that already exist, obtained with
the (breakpoints) function, can now have this type.
gdb/ChangeLog:
* guile/scm-breakpoint.c (bpscm_type_to_string): Handle
bp_catchpoint.
(bpscm_want_scm_wrapper_p): Likewise.
(gdbscm_make_breakpoint): Likewise.
(breakpoint_integer_constants): Likewise.
gdb/doc/ChangeLog:
* guile.texinfo (Breakpoints In Guile): Add BP_CATCHPOINT
description.
gdb/testsuite/ChangeLog:
* gdb.guile/scm-breakpoint.exp (test_catchpoints): New proc.
When creating a breakpoint using the guile API, if an invalid
breakpoint type number was used then the error would report the wrong
argument position, like this:
(gdb) guile (define wp2 (make-breakpoint "result" #:wp-class WP_WRITE #:type 999))
ERROR: In procedure make-breakpoint:
ERROR: In procedure gdbscm_make_breakpoint: Out of range: invalid breakpoint type in position 3: 999
Error while executing Scheme code.
(gdb)
The 'position 3' here is actually pointing at WP_WRITE, when it should
say 'position 5' and point to the 999. This commit fixes this.
However, you also get errors like this:
(gdb) guile (define wp2 (make-breakpoint "result" #:wp-class WP_WRITE #:type BP_NONE))
ERROR: In procedure make-breakpoint:
ERROR: In procedure gdbscm_make_breakpoint: Out of range: invalid breakpoint type in position 3: 0
Error while executing Scheme code.
The BP_NONE is a valid breakpoint type, it's just not valid for
creating breakpoints through the 'make-breakpoint' API. The use of
'0' in the error message (which is the value of BP_NONE) is not
great. This commit changes the error in this case to:
(gdb) guile (define wp2 (make-breakpoint "result" #:wp-class WP_WRITE #:type BP_NONE))
ERROR: In procedure make-breakpoint:
ERROR: In procedure gdbscm_make_breakpoint: unsupported breakpoint type in position 5: "BP_NONE"
Error while executing Scheme code.
Which seems better; we now use the name of the type, and report that
this type is unsupported.
gdb/ChangeLog:
* guile/scm-breakpoint.c (gdbscm_make_breakpoint): Split the error
for invalid breakpoint numbers, and unsupported breakpoint
numbers.
gdb/testsuite/ChangeLog:
* gdb.guile/scm-breakpoint.exp (test_watchpoints): Add new tests.
This patch adds a file with the ISA 3.1 check. The ISA 3.1 check is
added to the test to ensure the test is only run on ISA 3.1 or newer.
gdb/testsuite/ChangeLog
2021-06-25 Carl Love <cel@us.ibm.com>
* gdb.arch/powerpc-plxv-norel.exp: Add call to skip_power_isa_3_1_tests.
* lib/gdb.exp(skip_power_isa_3_1_tests): New gdb_caching_proc test.
While working on the DWARF psymtab replacement, I needed
addrmap_foreach to accept a gdb::function_view. This seemed like a
worthwhile change on its own, so I've written it separately for
submission.
Regression tested on x86-64 Fedora 32.
gdb/ChangeLog
2021-06-25 Tom Tromey <tom@tromey.com>
* dwarf2/index-write.c (struct addrmap_index_data): Add
initializers.
<operator()>: Declare.
(addrmap_index_data::operator()): Rename from
add_address_entry_worker. Remove 'datap' parameter.
(write_address_map): Update.
* psymtab.c (struct dump_psymtab_addrmap_data): Remove
(dump_psymtab_addrmap_1): Remove 'data' parameter, add other
parameters.
(dump_psymtab_addrmap): Update.
* addrmap.c (struct addrmap_funcs) <foreach>: Remove 'data'
parameter.
(addrmap_foreach, addrmap_fixed_foreach): Likewise.
(struct mutable_foreach_data): Remove.
(addrmap_mutable_foreach_worker): Update.
(addrmap_mutable_foreach): Remove 'data' parameter.
* addrmap.h (addrmap_foreach_fn): Use gdb::function_view.
(addrmap_foreach): Remove 'data' parameter.
GNAT emits encoded type names, but these aren't usually of interest to
users. The Ada language code in gdb hides this oddity -- but the
Python layer does not. This patch changes the Python code to use the
decoded Ada type name, when appropriate.
I looked at decoding Ada type names during construction, as that would
be cleaner. However, the Ada support in gdb relies on the encodings
at various points, so this isn't really doable right now.
2021-06-25 Tom Tromey <tromey@adacore.com>
* python/py-type.c (typy_get_name): Decode an Ada type name.
gdb/testsuite/ChangeLog
2021-06-25 Tom Tromey <tromey@adacore.com>
* gdb.ada/py_range.exp: Add type name test cases.
When ada_decode encounters a name that it cannot decode, it simply
wraps it in <...>, which is used elsewhere in the Ada code to indicate
that a verbatim match should be done.
A subequent patch needed the ability to suppress this wrapping, so
this patch adds a new mode to ada_decode.
2021-06-25 Tom Tromey <tromey@adacore.com>
* ada-lang.c (ada_decode): Add wrap parameter.
* ada-lang.h (ada_decode): Add wrap parameter.
At the moment, the core target has its own gdbarch (m_core_gdbarch), and that
gets set from the core_bfd on the core target's constructor.
That gdbarch doesn't contain a target description because it is constructed
before we get a chance to fetch the target description.
As a result, some hooks that depend on the target description being set are
not set, and that leads to problems. One of the examples is
gdbarch_report_signal_info, which is used to show AArch64 tag violation
information.
Fix this by reading the target description before fetching the core file's
gdbarch.
gdb/ChangeLog:
2021-06-25 Luis Machado <luis.machado@linaro.org>
* corelow.c (core_target::core_target) Update to read target
description.
When running test-case gdb.base/info-macros.exp, I run into:
...
PASS: gdb.base/info-macros.exp: info macro --
PASS: gdb.base/info-macros.exp: info macro --
DUPLICATE: gdb.base/info-macros.exp: info macro --
PASS: gdb.base/info-macros.exp: info macro --
...
These messages come from gdb_test calls using the following commands:
- "info macro --"
- "info macro -- "
- "info macro -- ".
Apparantly the test names get stripped of trailing whitespace, and the first
two end up identical.
Fix this by explicitly specifying an <EOL> after the trailing whitespace in
the test name, such that we have:
...
PASS: gdb.base/info-macros.exp: info macro --
PASS: gdb.base/info-macros.exp: info macro -- <EOL>
PASS: gdb.base/info-macros.exp: info macro -- <EOL>
...
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2021-06-24 Tom de Vries <tdevries@suse.de>
* gdb.base/info-macros.exp: Add <EOL> after trailing whitespace in
test names.
I found the following duplicates in gdb.base/argv0-symlink.exp:
...
DUPLICATE: gdb.base/argv0-symlink.exp: set print repeats 10000
DUPLICATE: gdb.base/argv0-symlink.exp: set print elements 10000
...
Fix these by using with_test_prefix "file symlink" / "dir symlink".
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2021-06-24 Tom de Vries <tdevries@suse.de>
* gdb.base/argv0-symlink.exp: Use with_test_prefix.
Power10 prefixed instructions must not cross 64-byte boundaries.
This is already handled.
However, on XCOFF, the csect must be updated to match the new
alignment.
* config/tc-ppc.c (md_assemble): Update ppc_current_csect
alignment when finding prefixed instructions.
If code tries to send a signal to itself, the callback layer ignores
it and forces the caller to handle it. This allows the sim to turn
that into an engine halt rather than actually killing the sim.
This will make it easier to emulate the syscall. If the kill target
is the sim itself, don't do anything. This forces the higher layers
to make a decision as to how to handle this event: like halting the
overall engine process.
We rewrite srcdir in subdir Makefiles that we generate from the common
parent dir since it points to the parent dir. Since @srcdir@ can be a
variety of formats (relative & absolute), switch to @abs_srcdir@ which
is a lot easier to adjust. Our use of srcdir in here should handle it.
On Ubuntu 20.04, when the debug info package for libc is not installed,
I get:
FAIL: gdb.base/info-types-c++.exp: info types
FAIL: gdb.base/info-types-c.exp: info types
The reason is that the output of info types is exactly:
(gdb) info types
All defined types:
File /home/smarchi/src/binutils-gdb/gdb/testsuite/gdb.base/info-types.c:
52: typedef enum {...} anon_enum_t;
45: typedef struct {...} anon_struct_t;
68: typedef union {...} anon_union_t;
28: typedef struct baz_t baz;
31: typedef struct baz_t * baz_ptr;
21: struct baz_t;
double
33: enum enum_t;
float
int
38: typedef enum enum_t my_enum_t;
17: typedef float my_float_t;
16: typedef int my_int_t;
54: typedef enum {...} nested_anon_enum_t;
47: typedef struct {...} nested_anon_struct_t;
70: typedef union {...} nested_anon_union_t;
30: typedef struct baz_t nested_baz;
29: typedef struct baz_t nested_baz_t;
39: typedef enum enum_t nested_enum_t;
19: typedef float nested_float_t;
18: typedef int nested_int_t;
62: typedef union union_t nested_union_t;
56: union union_t;
unsigned int
(gdb)
The lines we expect in the test contain an empty line at the end:
...
"62:\[\t \]+typedef union union_t nested_union_t;" \
"56:\[\t \]+union union_t;" \
"--optional" "\[\t \]+unsigned int" \
""]
This is written with the supposition that other files will be listed, so
an empty line will be included to separate the symbols from this file
from the next one. This empty line is not included when info-types.c is
the only file listed.
Fix this by rewriting gdb_test_lines to accept a single, plain tcl multiline
regexp, such that we can write:
...
"62:\[\t \]+typedef union union_t nested_union_t;" \
"56:\[\t \]+union union_t;(" \
"\[\t \]+unsigned int)?" \
"($|\r\n.*)"]
...
Tested affected test-cases:
- gdb.base/info-types-c.exp
- gdb.base/info-types-c++.exp
- gdb.base/info-macros.exp
- gdb.cp/cplusfuncs.exp
on x86_64-linux (openSUSE Leap 15.2), both with check and check-read1.
Also tested the first two with gcc-4.8.
Also tested on ubuntu 18.04.
gdb/testsuite/ChangeLog:
2021-06-23 Tom de Vries <tdevries@suse.de>
* lib/gdb.exp (gdb_test_lines): Rewrite to accept single
multiline tcl regexp.
* gdb.base/info-types.exp.tcl: Update. Make empty line at end of
regexp optional.
* gdb.base/info-macros.exp: Update.
* gdb.cp/cplusfuncs.exp: Update.
There's been a prototype for this forever, but the implementation was
missing. Probably because there weren't any callers, but we'll start
using it to implement the kill function.
These ports only use the pieces that have been unified, so we can
merge them into the common configure script and get rid of their
unique one entirely.
We still compile & link separate run programs, and have dedicated
subdir Makefiles, but the configure script portion is merged.
do_target_wait has a wait_ptid parameter, to filter what ptid we wait
on. The sole caller of do_target_wait passes minus_one_ptid, meaning
"all ptids". So in practice, this parameter is not needed, remove it.
gdb/ChangeLog:
* infrun.c (do_target_wait): Remove wait_ptid parameter.
(fetch_inferior_event): Adjust.
Change-Id: I54119beb43db678e4b2081dc490f89e7ff878e74
If we have multiple registered unwinders, this will helps identify which
unwinder was chosen and make it easier to track down potential problems.
Unwinders have a mandatory name argument, which we can use in the
message.
First, make gdb._execute_unwinders return a tuple containing the name,
in addition to the UnwindInfo. Then, make pyuw_sniffer include the name
in the debug message.
I moved the debug message earlier. I think it's good to print it as
early as possible, so that we see it in case an assert is hit in the
loop below, for example.
gdb/ChangeLog:
* python/lib/gdb/__init__.py (_execute_unwinders): Return tuple
with name of chosen unwinder.
* python/py-unwind.c (pyuw_sniffer): Print name of chosen
unwinder in debug message.
Change-Id: Id603545b44a97df2a39dd1872fe1f38ad5059f03
Without that it is impossible to debug on riscv64.
gdb/
PR symtab/27999
* dwarf2/loc.c (decode_debug_loclists_addresses): Support
DW_LLE_start_end.
gdb/testsuite/
PR symtab/27999
* lib/dwarf.exp (start_end): New proc inside loclists.
* gdb.dwarf2/loclists-start-end.exp: New file.
* gdb.dwarf2/loclists-start-end.c: New file.
This test-case is intended to excercise this code in process_imported_unit_die:
...
/* We're importing a C++ compilation unit with tag DW_TAG_compile_unit
into another compilation unit, at root level. Regard this as a hint,
and ignore it. */
if (die->parent && die->parent->parent == NULL
&& per_cu->unit_type == DW_UT_compile
&& per_cu->lang == language_cplus)
return;
...
in the sense that the test-case should fail if the
"per_cu->lang == language_cplus" clause is removed.
Tested on x86_64-linux.
gdb/testsuite/ChangeLog:
2021-06-22 Tom de Vries <tdevries@suse.de>
* gdb.dwarf2/imported-unit-c.exp: New file.
The original discussion was here,
https://github.com/riscv/riscv-elf-psabi-doc/issues/184
After discussing with Kito Cheng, I think the addends of %pcrel_hi
and %pcrel_lo are both allowed in GNU toolchain. However, both of
the them mean the offset of symbols, rather than the pc address.
But the addends of %got_pcrel_hi and it's %pcrel_lo do not look
reasonable. I believe gcc won't generate the got patterns with
addends, so linker should report dangerous relocation errors,
in case the assembly code use them.
Another issue was here,
https://sourceware.org/pipermail/binutils/2021-June/116983.html
At the beginnig, I suppose %pcrel_hi and %pcrel_lo are valid only
when they are in the same input section. But Jim Wilson points out
that gcc may generate %hi and %lo in the different input sections,
when -freorder-blocks-and-partition option is used. So that a memory
references for a loop may have the %hi outside the loop, but the %lo
remain in the loop. However, it is hard to create the testcases,
to see if %pcrel_hi and %pcrel_lo have the same behavior.
Unfortunately, I notice that the current pcrel resolver cannot
work for the above case. For now we build a hash table for pcrel
at the start of riscv_elf_relocate_section, and then free the hash
at the end. But riscv_elf_relocate_section only handles an input
section at a time, so that means we can only resolve the %pcrel_hi
and %pcrel_lo which are in the same input section. Otherwise, like
the above case, we will report "%pcrel_lo missing matching %pcrel_hi"
for them. I have no plan to improve this in the short-term, so maybe
we can wait until someone meets the problem before we deal with it.
bfd/
* elfnn-riscv.c (riscv_pcrel_hi_reloc): Added field to store
the original relocation type, in case the type is converted to
R_RISCV_HI20.
(riscv_pcrel_lo_reloc): Removed unused name field.
(riscv_pcrel_relocs): Added comments.
(riscv_zero_pcrel_hi_reloc): Removed unused input_bfd.
(riscv_record_pcrel_hi_reloc): Updated.
(riscv_record_pcrel_lo_reloc): Likewise.
(riscv_resolve_pcrel_lo_relocs): Likewise. Check the original
type of auipc, to make sure the %pcrel_lo without any addends.
Otherwise, report dangerous relocation error.
(riscv_elf_relocate_section): Updated above functions are changed.
For R_RISCV_GOT_HI20, report dangerous relocation error when addend
isn't zero.
ld/
* testsuite/ld-riscv-elf/ld-riscv-elf.exp: Updated.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3a.d: New testcase.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3a.s: Likewise.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3b.d: New testcase.
Should report error since the %pcrel_lo with addend refers to
%got_pcrel_hi.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3b.s: Likewise.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3c.d: New testcase.
Should report error since the %got_pcrel_hi with addend.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3c.s: Likewise.
* testsuite/ld-riscv-elf/pcrel-lo-addend-3.ld: Likewise.
I randomly hit a situation where gdbserver crashed immediately before
I issued a 'stepi' to GDB, it turns out that this causes GDB itself to
crash.
What happens is that as part of the stepi we try to insert some
breakpoints into the inferior, so from insert_breakpoints we figure
out what we want to insert, then, eventually, try to send some packets
to the remote to get the breakpoints inserted.
It is only at this point that GDB realises that the target has gone
away. This causes GDB to then enter this call stack:
unpush_and_perror
remote_unpush_target
generic_mourn_inferior
breakpoint_init_inferior
delete_breakpoint
update_global_location_list
So, we realise the target is gone and so delete the breakpoints
associated with that target.
GDB then throws a TARGET_CLOSE_ERROR from unpush_and_error.
This error is caught in insert_breakpoints where we then try to print
a nice error saying something like:
Cannot insert breakpoint %d: some error text here...
To fill in the '%d' we try to read properties of the breakpoint
object.
Which was deleted due to the delete_breakpoint call above.
And so GDB dies...
My proposal in this commit is that, should we catch a
TARGET_CLOSE_ERROR in insert_breakpoints, then we just rethrow the
error.
This will cause the main event loop to print something like:
Remote connection closed
Which I think is fine, I don't think the user will care much which
particular breakpoint GDB was operating on when the connection closed,
just knowing that the connection closed should be enough I think.
I initially added a test to 'gdb.server/server-kill.exp' for this
issue, however, my first attempt was not good enough, the test was
passing even without my fix.
Turns out that the server-kill.exp test actually kills the PID of the
inferior, not the PID of the server. This means that gdbserver is
actually able to send a packet to GDB saying that the inferior has
exited prior to gdbserver itself shutting down. This extra
information was enough to prevent the bug I was seeing manifest.
So, I have extended server-kill.exp to run all of the tests twice, the
first time we still kill the inferior. On the second run we hard kill
the gdbserver itself, this prevents the server from sending anything
to GDB before it exits.
My new test is only expected to fail in this second mode of
operation (killing gdbserver itself), and without my fix, that is what
I see.
gdb/ChangeLog:
* breakpoint.c (insert_bp_location): If we catch a
TARGET_CLOSE_ERROR just rethrow it, the breakpoints might have
been deleted.
gdb/testsuite/ChangeLog:
* gdb.server/server-kill.exp: Introduce global kill_pid_of, and
make use of this in prepare to select which pid we should kill.
Run all the tests twice with a different kill_pid_of value.
(prepare): Make use of kill_pid_of.
(test_stepi): New proc.
This commit provides a small performance improvement when starting up
CGEN based disassemblers by making use of __builtin_popcount.
The #if check used in this commit was copied from bfd/elf32-arm.c
where __builtin_popcount is also used.
I ran into this code while investigating some GDB tests that would
occasionally timeout. One of the reason these tests were having
problems is that the m16c and m32c disassemblers take so long to
initialise themselves. Speeding up count_decodable_bits helps, but is
not a total solution. Still, this felt like an easy win which added
minimal extra complexity, so I figure its worth doing.
opcodes/ChangeLog:
* cgen-dis.c (count_decodable_bits): Use __builtin_popcount when
available.
Commit 54758c3e398d made changes to the picojava support based on
https://sourceware.org/pipermail/binutils/2005-November/045136.html
An update from picojava to picojava II, I think. Unfortunately the
patch neglected any changes to the gas testsuite, resulting in
"FAIL: pj" since that date. This patch makes a few relatively simple
changes to cure the regression.
gas/
* config/tc-pj.c (md_apply_fix): Apply PJ_CODE_REL32 relocs.
* testsuite/gas/pj/ops.s: Update jsr, ret, getstatic,
putstatic, getfield, putfield, invokevirtual, invokespecial,
invokestatic, invokeinterface, goto_w, jsr_w assembly. Delete
version 1 picojava opcodes.
* testsuite/gas/pj/ops.d: Match expected output.
opcodes/
* pj-dis.c (print_insn_pj): Don't print trailing tab. Do
print separator for pcrel insns.
The sim-hardware configure option allows builders to select a set of
device models to enable. But this seems like unnecessary overkill:
the existence of individual device models doesn't affect performance
at all as they are only enabled at runtime if the config uses them,
and individually these are all <5KB a piece. Stripping off a total
of ~50KB from a ~1MB binary doesn't seem useful, and it's extremely
unlikely anyone will ever bother.
So let's simplify the configure/make logic by turning sim-hardware
into a boolean option like many of the other sim options. Any ports
that have unique device models will declare them in their Makefile
instead of at configure time. This will allow us to (eventually)
unify the setting into the common dir.
This commit adds support to RISC-V GDB for vector registers in the
incoming target description.
The vector registers should be described in a feature called
"org.gnu.gdb.riscv.vector", and should contain the register v0 to
v31. There's no restriction on the size or type of these registers,
so the target description can set these up as it requires.
However, if the target feature is present then all of the registers
must be present, and they must all be the same size, these
requirements are, I believe, inline with the RISC-V vector extension.
The DWARF register numbers for the vector registers have been added,
and the code to map between GDB's internal numbering and the DWARF
numbering has been updated.
I have not yet added a feature/riscv/*.xml file for the vector
extension, the consequence of this is that we can't, right now, detect
vector registers on a native target, this patch is all about
supporting vectors on a remote target.
It is worth noting that I don't actually have access to a RISC-V
target with vectors, so the only testing that this patch has had has
been done using 'set tdesc filename ....' to load a target description
to which I have manually added the vector feature. This has shown
that the vector register feature can be successfully parsed, and that
the registers show up in the expected register groups.
Additionally, the RISC-V vector extension is currently at v0.10, which
is also the v1.0 draft release. However, this extension is not yet
finalised. It is possible (but unlikely I think) that the register
set could change between now and the final release of the vector
extension. If this were to happen then we would potentially end up
changing the requirements for the new org.gnu.gdb.riscv.vector
feature. I really don't think it is likely that the register set will
change this late in the process, and even if it did, changing the
feature requirements will not be a problem as far as I am
concerned (when the alternative is GDB just continues without this
feature for now).
gdb/ChangeLog:
* NEWS: Mention new target feature name.
* arch/riscv.c (riscv_create_target_description): GDB doesn't
currently create target descriptions containing vector registers.
* arch/riscv.h (struct riscv_gdbarch_features) <vlen>: New member
variable.
<operator==>: Also compare vlen.
<hash>: Also include vlen.
* riscv-tdep.c (riscv_feature_name_vector): New static global.
(struct riscv_vector_feature): New struct.
(riscv_vector_feature): New static global.
(riscv_register_reggroup_p): Ensure vector registers are part of
the 'all' group, and part of the 'vector' group.
(riscv_dwarf_reg_to_regnum): Handle vector registers.
(riscv_gdbarch_init): Check vector register feature.
* riscv-tdep.h: Add vector registers to GDB's internal register
numbers, and to the DWARF register numbers.
gdb/doc/ChangeLog:
* gdb.texinfo (RISC-V Features): Mention vector register feature.
Add new methods to the PendingFrame and Frame classes to obtain the
stack frame level for each object.
The use of 'level' as the method name is consistent with the existing
attribute RecordFunctionSegment.level (though this is an attribute
rather than a method).
For Frame/PendingFrame I went with methods as these classes currently
only use methods, including for simple data like architecture, so I
want to be consistent with this interface.
gdb/ChangeLog:
* NEWS: Mention the two new methods.
* python/py-frame.c (frapy_level): New function.
(frame_object_methods): Register 'level' method.
* python/py-unwind.c (pending_framepy_level): New function.
(pending_frame_object_methods): Register 'level' method.
gdb/doc/ChangeLog:
* python.texi (Unwinding Frames in Python): Mention
PendingFrame.level.
(Frames In Python): Mention Frame.level.
gdb/testsuite/ChangeLog:
* gdb.python/py-frame.exp: Add Frame.level tests.
* gdb.python/py-pending-frame-level.c: New file.
* gdb.python/py-pending-frame-level.exp: New file.
* gdb.python/py-pending-frame-level.py: New file.
We already have two helper functions in py-utils.c:
gdb_py_object_from_longest (LONGEST l)
gdb_py_object_from_ulongest (ULONGEST l)
these wrap around calls to either PyLong_FromLongLong,
PyLong_FromLong, or PyInt_From_Long (if Python 2 is being used).
There is one place in gdb/python/* where a call to PyLong_FromLong was
added outside of the above utility functions, this was done in the
recent commit:
commit 55789354fcbaf879f3ca8475b647b2747dec486e
Date: Fri May 14 11:56:31 2021 +0200
gdb/python: add a 'connection_num' attribute to Inferior objects
In this commit I replace the direct use of PyLong_FromLong with a call
to gdb_py_object_from_longest. The only real change with this commit,
is that, for Python 2, we will now end up calling PyInt_FromLong
instead of PyLong_FromLong, but this should be invisible to the user.
For Python 3 there should be absolutely no change.
gdb/ChangeLog:
* python/py-inferior.c (infpy_get_connection_num): Call
gdb_py_object_from_longest instead of PyLong_FromLong directly.
This patch came about because I wanted to write a frame unwinder that
would corrupt the backtrace in a particular way. In order to achieve
what I wanted I ended up trying to write an unwinder like this:
class FrameId(object):
.... snip class definition ....
class TestUnwinder(Unwinder):
def __init__(self):
Unwinder.__init__(self, "some name")
def __call__(self, pending_frame):
pc_desc = pending_frame.architecture().registers().find("pc")
pc = pending_frame.read_register(pc_desc)
sp_desc = pending_frame.architecture().registers().find("sp")
sp = pending_frame.read_register(sp_desc)
# ... snip code to decide if this unwinder applies or not.
fid = FrameId(pc, sp)
unwinder = pending_frame.create_unwind_info(fid)
unwinder.add_saved_register(pc_desc, pc)
unwinder.add_saved_register(sp_desc, sp)
return unwinder
The important things here are the two calls:
unwinder.add_saved_register(pc_desc, pc)
unwinder.add_saved_register(sp_desc, sp)
On x86-64 these would fail with an assertion error:
gdb/regcache.c:168: internal-error: int register_size(gdbarch*, int): Assertion `regnum >= 0 && regnum < gdbarch_num_cooked_regs (gdbarch)' failed.
What happens is that in unwind_infopy_add_saved_register (py-unwind.c)
we call register_size, as register_size should only be called on
cooked (real or pseudo) registers, and 'pc' and 'sp' are implemented
as user registers (at least on x86-64), we trigger the assertion.
A simple fix would be to check in unwind_infopy_add_saved_register if
the register number we are handling is a cooked register or not, if
not we can throw a 'Bad register' error back to the Python code.
However, I think we can do better.
Consider that at the CLI we can do this:
(gdb) set $pc=0x1234
This works because GDB first evaluates '$pc' to get a register value,
then evaluates '0x1234' to create a value encapsulating the
immediate. The contents of the immediate value are then copied back
to the location of the register value representing '$pc'.
The value location for a user-register will (usually) be the location
of the real register that was accessed, so on x86-64 we'd expect this
to be $rip.
So, in this patch I propose that in the unwinder code, when
add_saved_register is called, if it is passed a
user-register (i.e. non-cooked) then we first fetch the register,
extract the real register number from the value's location, and use
that new register number when handling the add_saved_register call.
If either the value location that we get for the user-register is not
a cooked register then we can throw a 'Bad register' error back to the
Python code, but in most cases this will not happen.
gdb/ChangeLog:
* python/py-unwind.c (unwind_infopy_add_saved_register): Handle
saving user registers.
gdb/testsuite/ChangeLog:
* gdb.python/py-unwind-user-regs.c: New file.
* gdb.python/py-unwind-user-regs.exp: New file.
* gdb.python/py-unwind-user-regs.py: New file.
Inline the stats printf calls to avoid compiler warnings about
non-literal format strings. This in turn highlights bad type
sizes being passed in, so fix the strings to use the right size
type. This in turn highlights the rest of the func using casts
rather than the right type directly, so adjust all of those.
Finally, replace a few abort+sim_engine_halt calls with the
common sim_engine_abort. This provides good output while still
aborting as we want.
