This patch does some basic cleanups on the SystemTap SDT probes API. It
removes spurious newlines, brackets, reindents some code, and do
explicit checks for NULL, NUL, and 0 where applicable.
2013-12-23 Sergio Durigan JUnior <sergiodj@redhat.com>
* stap-probe.c (struct stap_probe) <args_parsed>: Add comment.
(stap_is_generic_prefix): Delete extra brackets. Reindent.
(stap_parse_register_operand): Remove spurious newlines. Simplify
code to parse special token.
(stap_parse_argument_conditionally): Add gdb_assert.
(stap_parse_argument_1): Likewise. Explicitly check for NULL and
NUL.
(stap_parse_probe_arguments): Likewise.
(handle_stap_probe): Likewise. Reindent code.
(get_stap_base_address): Explicitly check for NULL.
(stap_get_probes): Likewise. Reindent code.
(stap_relocate): Explicitly check for 0.
(stap_gen_info_probes_table_values): Likewise.
PREFIX_ADDR isn't a prefix to opcode. This patch masks out PREFIX_ADDR
when adding prefix to opcode.
PR gdb/16305
* i386-tdep.c (i386_process_record): Mask out PREFIX_ADDR when
adding prefix to opcode.
64-bit mode doesn't use 16-bit address. We should always check SIB byte
for address in 64-bit mode.
PR gdb/16304
* i386-tdep.c (i386_record_lea_modrm_addr): Don't use 16-bit
address in 64-bit mode.
When there is ADDR32 prefix in 64-bit mode, we should zero-extend
address from 32-bit to 64-bit.
PR gdb/16304
* i386-tdep.c (i386_record_lea_modrm_addr): Zero-extend 32-bit
address to 64-bit in 64-bit mode.
X32 Linux system calls are diffferent from amd64 Linux system calls in
system call numbers as well as parameter types/values. This patch adds
amd64_x32_linux_record_tdep and amd64_x32_syscall for x32.
PR gdb/16304
* amd64-linux-tdep.c (amd64_canonicalize_syscall): Handle x32
system calls.
(amd64_x32_linux_record_tdep): New.
(amd64_linux_syscall_record_common): New function.
(amd64_linux_syscall_record): Call
amd64_linux_syscall_record_common with amd64_linux_record_tdep.
(amd64_x32_linux_syscall_record): Call
amd64_linux_syscall_record_common with
amd64_x32_linux_record_tdep.
(amd64_linux_init_abi_common): Move amd64_linux_record_tdep
initialization and tdep->i386_syscall_record setup to ...
(amd64_linux_init_abi): Here.
(amd64_x32_linux_init_abi): Initialize
amd64_x32_linux_record_tdep. Set tdep->i386_syscall_record to
amd64_x32_linux_syscall_record.
* amd64-linux-tdep.h (amd64_x32_syscall): New enum.
This patch extends the current generic parser for SystemTap SDT probe
arguments. It can be almost considered a cleanup, but the main point of
it is actually to allow the generic parser to accept multiple prefixes
and suffixes for the its operands (i.e., integers, register names, and
register indirection).
I have chosen to implement this as a list of const strings, and declare
this list as "static" inside each target's method used to initialize
gdbarch.
This patch is actually a preparation for an upcoming patch for ARM,
which implements the support for multiple integer prefixes (as defined
by ARM's asm spec). And AArch64 will also need this, for the same
reason.
This patch was regtested on all architectures that it touches (i.e.,
i386, x86_64, ARM, PPC/PPC64, s390x and IA-64). No regressions were found.
2013-12-19 Sergio Durigan Junior <sergiodj@redhat.com>
* amd64-tdep.c (amd64_init_abi): Declare SystemTap SDT probe
argument prefixes and suffixes. Initialize gdbarch with them.
* arm-linux-tdep.c (arm_linux_init_abi): Likewise.
* gdbarch.c: Regenerate.
* gdbarch.h: Regenerate.
* gdbarch.sh (stap_integer_prefix, stap_integer_suffix)
(stap_register_prefix, stap_register_suffix)
(stap_register_indirection_prefix)
(stap_register_indirection_suffix): Declare as "const char *const
*" instead of "const char *". Adjust printing function. Rename
all of the variables to the plural.
(pstring_list): New function.
* i386-tdep.c (i386_elf_init_abi): Declare SystemTap SDT probe
argument prefixes and suffixes. Initialize gdbarch with them.
* ia64-linux-tdep.c (ia64_linux_init_abi): Likewise.
* ppc-linux-tdep.c (ppc_linux_init_abi): Likewise.
* s390-linux-tdep.c (s390_gdbarch_init): Likewise.
* stap-probe.c (stap_is_generic_prefix): New function.
(stap_is_register_prefix): Likewise.
(stap_is_register_indirection_prefix): Likewise.
(stap_is_integer_prefix): Likewise.
(stap_generic_check_suffix): Likewise.
(stap_check_integer_suffix): Likewise.
(stap_check_register_suffix): Likewise.
(stap_check_register_indirection_suffix): Likewise.
(stap_parse_register_operand): Remove unecessary declarations for
variables holding prefix and suffix information. Use the new
functions listed above for checking for prefixes and suffixes.
(stap_parse_single_operand): Likewise.
Code rationale
==============
by: Gabriel Krisman Bertazi
This is a fix for bug 16297. The problem occurs when the user attempts
to catch any syscall 0 (such as syscall read on Linux/x86_64). GDB was
not able to catch the syscall and was missing the breakpoint.
Now, breakpoint_hit_catch_syscall returns immediately when it finds the
correct syscall number, avoiding a following check for the end of the
search vector, that returns a no hit if the syscall number was zero.
Testcase rationale
==================
by: Sergio Durigan Junior
This testcase is a little difficult to write. By doing a quick
inspection at the Linux source, one can see that, in many targets, the
syscall number 0 is restart_syscall, which is forbidden to be called
from userspace. Therefore, on many targets, there's just no way to test
this safely.
My decision was to take the simpler route and just adds the "read"
syscall on the default test. Its number on x86_64 is zero, which is
"good enough" since many people here do their tests on x86_64 anyway and
it is a popular architecture.
However, there was another little gotcha. When using "read" passing 0
as the third parameter (i.e., asking it to read 0 bytes), current libc
implementations could choose not to effectively call the syscall.
Therefore, the best solution was to create a temporary pipe, write 1
byte into it, and then read this byte from it.
gdb/ChangeLog
2013-12-19 Gabriel Krisman Bertazi <gabriel@krisman.be>
PR breakpoints/16297
* breakpoint.c (breakpoint_hit_catch_syscall): Return immediately
when expected syscall is hit.
gdb/testsuite/ChangeLog
2013-12-19 Sergio Durigan Junior <sergiodj@redhat.com>
PR breakpoints/16297
* gdb.base/catch-syscall.c (read_syscall, pipe_syscall)
(write_syscall): New variables.
(main): Create a pipe, write 1 byte in it, and read 1 byte from
it.
* gdb.base/catch-syscall.exp (all_syscalls): Include "pipe,
"write" and "read" syscalls.
(fill_all_syscalls_numbers): Improve the way to obtain syscalls
numbers.
Now that struct serial_ops is const everywhere, we can easily turn the
instances into globals. This patch implements this idea.
On the one hand I think this is nicer since it makes a bit more data
readonly and slightly reduces allocations. On the other hand it
reduces readability somewhat.
If the readability is a concern to anyone I was thinking I could write
a macro that conditionally uses GCC's designated initializer
extension.
Tested by rebuilding on x86-64 Fedora 18, both natively and using the
mingw cross tools.
2013-12-19 Tom Tromey <tromey@redhat.com>
* ser-unix.c (hardwire_ops): New global.
(_initialize_ser_hardwire): Use it.
* ser-tcp.c (tcp_ops): New global.
(_initialize_ser_tcp): Use it.
* ser-pipe.c (pipe_ops): New global.
(_initialize_ser_pipe): Use it.
* ser-mingw.c (hardwire_ops, tty_ops, pipe_ops, tcp_ops): New
globals.
(_initialize_ser_windows): Use them.
I noticed that the serial_ops vtable is not const, but really it ought
to be.
This patch constifies it, removing the only mutable field in the
process.
Tested by rebuilding on x86-64 Fedora 18, both natively and using the
mingw cross tools.
2013-12-19 Tom Tromey <tromey@redhat.com>
* serial.c (serial_ops_p): New typedef.
(serial_ops_list): Now a VEC.
(serial_interface_lookup): Return const. Use VEC_iterate.
(serial_add_interface): Make parameter const.
(serial_open): Update.
(serial_fdopen_ops): Make 'ops' const.
(serial_pipe): Update.
* ser-tcp.c (_initialize_ser_tcp): Update.
* ser-pipe.c (_initialize_ser_pipe): Update.
* ser-unix.c (_initialize_ser_hardwire): Update.
* ser-mingw.c (_initialize_ser_windows): Update.
* ser-go32.c (dos_ops): Now const. Update.
* serial.h (struct serial) <ops>: Now const.
(struct serial_ops) <next>: Remove.
(serial_add_interface): Make parameter const.
* aarch64-linux-nat.c (aarch64_linux_set_debug_regs): Set
iov.iov_len with the real length in use.
gdb/gdbserver/
* linux-aarch64-low.c (aarch64_linux_set_debug_regs): Set
iov.iov_len with the real length in use.
This patch adds a typedef target_xfer_partial_ftype. When we change
the signature of xfer_partial functions (for example, adding a new
parameter), we don't have to modify all of their declarations.
This patch also updates the type of parameters of target_xfer_partial
from "void *" to "gdb_byte *".
gdb:
2013-12-18 Yao Qi <yao@codesourcery.com>
* target.h (target_xfer_partial_ftype): New typedef.
(target_xfer_partial): Update declaration.
* auxv.h (memory_xfer_auxv): Likewise.
* ia64-hpux-nat.c (super_xfer_partial): Likewise.
* ia64-linux-nat.c (super_xfer_partial): Likewise.
* linux-nat.c (super_xfer_partial): Likewise.
* procfs.c (procfs_xfer_partial): Likewise.
* record-full.c (record_full_beneath_to_xfer_partial):
(tmp_to_xfer_partial): Likewise.
* sparc-nat.c (inf_ptrace_xfer_partial): Likewise.
* target.c (default_xfer_partial): Likewise.
(current_xfer_partial): Likewise.
(target_xfer_partial): Change parameter type to 'gdb_byte *'.
I notice that two lines of a recent changelog entry are not prefixed
with tab. They are prefixed with a space and a tab. This patch
is to remove the space.
gdb:
2013-12-18 Yao Qi <yao@codesourcery.com>
* ChangeLog: Fix the format of one entry.
Like when stepping, the current stack frame location is expected to be
printed as result of tfind command, if that results in moving to a
different function. In tfind_1 we see:
if (from_tty
&& (has_stack_frames () || traceframe_number >= 0))
{
enum print_what print_what;
/* NOTE: in imitation of the step command, try to determine
whether we have made a transition from one function to
another. If so, we'll print the "stack frame" (ie. the new
function and it's arguments) -- otherwise we'll just show the
new source line. */
if (frame_id_eq (old_frame_id,
get_frame_id (get_current_frame ())))
print_what = SRC_LINE;
else
print_what = SRC_AND_LOC;
print_stack_frame (get_selected_frame (NULL), 1, print_what, 1);
do_displays ();
}
However, when we haven't collected any registers in the tracepoint
(collect $regs), that doesn't actually work:
(gdb) tstart
(gdb) info tracepoints
Num Type Disp Enb Address What
1 tracepoint keep y 0x080483b7 in func0
at ../.././../git/gdb/testsuite/gdb.trace/circ.c:28
collect testload
installed on target
2 tracepoint keep y 0x080483bc in func1
at ../.././../git/gdb/testsuite/gdb.trace/circ.c:32
collect testload
installed on target
(gdb) c
Continuing.
Breakpoint 3, end () at ../.././../git/gdb/testsuite/gdb.trace/circ.c:72
72 }
(gdb) tstop
(gdb) tfind start
Found trace frame 0, tracepoint 1
#0 func0 () at ../.././../git/gdb/testsuite/gdb.trace/circ.c:28
28 }
(gdb) tfind
Found trace frame 1, tracepoint 2
32 }
(gdb)
When we don't have info about the stack available
(UNWIND_UNAVAILABLE), frames end up with outer_frame_id as frame ID.
And in the scenario above, the issue is that both frames before and
after the second tfind (the frames for func0 an func1) have the same
id (outer_frame_id), so the frame_id_eq check returns false, even
though the frames were of different functions. GDB knows that,
because the PC is inferred from the tracepoint's address, even if no
registers were collected.
To fix this, this patch adds support for frame ids with a valid code
address, but <unavailable> stack address, and then makes the unwinders
use that instead of the catch-all outer_frame_id for such frames. The
frame_id_eq check in tfind_1 then automatically does the right thing
as expected.
I tested with --directory=gdb.trace/ , before/after the patch, and
compared the resulting gdb.logs, then adjusted the tests to expect the
extra output that came out. Turns out that was only circ.exp, the
original test that actually brought this issue to light.
Tested on x86_64 Fedora 17, native and gdbserver.
gdb/
2013-12-17 Pedro Alves <palves@redhat.com>
* frame.h (enum frame_id_stack_status): New enum.
(struct frame_id) <stack_addr>: Adjust comment.
<stack_addr_p>: Delete field, replaced with ...
<stack_status>: ... this new field.
(frame_id_build_unavailable_stack): Declare.
* frame.c (frame_addr_hash, fprint_field, outer_frame_id)
(frame_id_build_special): Adjust.
(frame_id_build_unavailable_stack): New function.
(frame_id_build, frame_id_build_wild): Adjust.
(frame_id_p, frame_id_eq, frame_id_inner): Adjust to take into
account frames with unavailable stack.
* amd64-tdep.c (amd64_frame_this_id)
(amd64_sigtramp_frame_this_id, amd64_epilogue_frame_this_id): Use
frame_id_build_unavailable_stack.
* dwarf2-frame.c (dwarf2_frame_this_id): Likewise.
* i386-tdep.c (i386_frame_this_id, i386_epilogue_frame_this_id)
(i386_sigtramp_frame_this_id): Likewise.
gdb/testsuite/
2013-12-17 Pedro Alves <palves@redhat.com>
* gdb.trace/circ.exp: Expect frame info to be printed when
switching between frames with unavailable stack, but different
functions.
https://sourceware.org/ml/gdb-patches/2013-12/msg00144.html
The vector of unavailable parts of a value is currently byte based. Given
that we can model a value down to the bit level, we can potentially loose
information with the current implementation. After this patch we model the
unavailable information in bits.
gdb/ChangeLog
* dwarf2loc.c (read_pieced_value): Mark bits, not bytes
unavailable, use correct bit length.
* value.c (struct value): Extend comment on unavailable to
indicate that it is bit based.
(value_bits_available): New function.
(value_bytes_available): Call value_bits_available.
(value_entirely_available): Check against the bit length, not byte
length.
(mark_value_bits_unavailable): New function.
(mark_value_bytes_unavailable): Move contents to
mark_value_bits_unavailable, call to same.
(memcmp_with_bit_offsets): New function.
(value_available_contents_bits_eq): New function, takes the
functionality from value_available_contents_eq but uses
memcmp_with_bit_offsets now, and is bit not byte based.
(value_available_contents_eq): Move implementation into
value_available_contents_bits_eq, call to same.
(value_contents_copy_raw): Work on bits, not bytes.
(unpack_value_bits_as_long_1): Check availability in bits, not
bytes.
* value.h (value_bits_available): Declare new function.
(mark_value_bits_unavailable): Declare new function.
gdb/testsuite/ChangeLog
* gdb.trace/unavailable-dwarf-piece.c: New file.
* gdb.trace/unavailable-dwarf-piece.exp: New file.
Like on GNU/Linux (linux-thread-db.c), the Solaris solaris-threads
target (handles libthread_db.so) shouldn't be pushed when remote
debugging.
This uses the same predicate used by linux-thread-db.c.
gdb/
2013-12-16 Pedro Alves <palves@redhat.com>
PR 16329
* sol-thread.c (check_for_thread_db): If the target can't run or
isn't a core, return without pushing.
This reverts commit 07293be44859c607a36c313e51bec2dcdcd3c243, as it
causes an unintended change of behavior with GDB/MI's =library-loaded
events: The host-name="<path>" part of the event is now showing the
target-side path instead of the host-side path.
This revert affects Darwin and AIX systems, however, where the BFD
is either artificial or icomplete, leading to the outputt of
"info shared" not containing the information we'd like. For instance,
on Darwin, we would see:
(top-gdb) info shared
From To Syms Read Shared Object Library
0x00007fff8d060de4 0x00007fff8d09ce1f Yes (*) i386:x86-64
0x00007fff8af08b10 0x00007fff8b1c6f73 Yes (*) i386:x86-64
To compensate for that, we overwrite the filename of the associated bfd.
gdb/ChangeLog:
Revert the following commit:
* solib.c (solib_map_sections): Remove code overwriting
SO->SO_NAME with the bfd's filename.
Make the following changes required after the revert above:
* solib-aix.c (solib_aix_bfd_open): Set the filename of the
returned bfd to a copy of the synthetic pathname.
* solib-darwin.c (darwin_bfd_open): Set the filename of the
returned bfd to a copy of PATHNAME.
This function has the following code:
elt_type = type;
for (i = n; i > 1; i--)
elt_type = TYPE_TARGET_TYPE (type);
For multi-dimension arrays, the code above tries to find the array
type corresponding to the dimension we're trying to inspect.
The problem is that, past the second dimension, the loop does
nothing other than repeat the first iteration. There is a little
thinko where it got the TYPE_TARGET_TYPE of TYPE instead of ELT_TYPE!
To my surprise, I was unable to produce an Ada exemple that demonstrated
the problem. That's because the examples I created all trigger a parallel
___XA type which we then use in place of the ELT_TYPE in order to
determine the bounds - see the code that immediately follows our
loop above:
index_type_desc = ada_find_parallel_type (type, "___XA");
ada_fixup_array_indexes_type (index_type_desc);
if (index_type_desc != NULL)
[...]
So, in order to avoid depending on an Ada example where the compiler
can potentially decide one way or the other, I decided to use an
artificial example, written in C. With ...
int multi[1][2][3];
... forcing the language to Ada, and trying to print the 'last,
we get:
(gdb) p multi'last(1)
$1 = 0
(gdb) p multi'last(2)
$2 = 1
(gdb) p multi'last(3)
$3 = 1 <<<--- This should be 2!
Additionally, I noticed that a couple of check_typedef's were missing.
This patch adds them. And since the variable in question only gets
used within an "else" block, I moved the variable declaration and
use inside that block - making it clear what the scope of the variable
is.
gdb/ChangeLog:
* ada-lang.c (ada_array_bound_from_type): Move the declaration
and assignment of variable "elt_type" inside the else block
where it is used. Add two missing check_typedef calls.
Fix bug where we got TYPE's TYPE_TARGET_TYPE, where in fact
we really wanted to get ELT_TYPE's TYPE_TARGET_TYPE.
gdb/testsuite/ChangeLog:
* gdb.ada/arraydim: New testcase.
PR python/16113
* NEWS (Python Scripting): Add entry for the new feature and the
new attribute of gdb.Field objects.
* python/py-type.c (gdbpy_is_field): New function
(convert_field): Add 'parent_type' attribute to gdb.Field
objects.
* python/py-value.c (valpy_getitem): Allow subscript value to be
a gdb.Field object.
(value_has_field): New function
(get_field_flag): New function
* python/python-internal.h (gdbpy_is_field): Add declaration.
testsuite/
* gdb.python/py-value-cc.cc: Improve test case.
* gdb.python/py-value-cc.exp: Add new tests to test usage of
gdb.Field objects as subscripts on gdb.Value objects.
doc/
* gdb.texinfo (Values From Inferior): Add a note about using
gdb.Field objects as subscripts on gdb.Value objects.
(Types In Python): Add description about the new attribute
"parent_type" of gdb.Field objects.
I have a case that could use an exception for "unsupported feature".
I found UNSUPPORTED_ERROR, but looking deeper, I think as is, reusing
it for other things would be fragile. E.g., if the Python script
sourced by source_script_from_stream triggers any other missing
functionality that would result in UNSUPPORTED_ERROR being propagated
out to source_script_from_stream, that would confuse the error for
Python not being built into GDB.
This patch thus redoes things a little. Instead of using an exception
for the "No Python" scenario, check whether Python is configured in
before actually trying to source the file. It adds a new function
instead of using #ifdef HAVE_PYTHON directly, as that is better at
avoiding bitrot, as both Python and !Python paths are visible to the
compiler this way.
Tested on Fedora 17, with and without Python.
gdb/
2013-12-12 Pedro Alves <palves@redhat.com>
* cli/cli-cmds.c (source_script_from_stream) Use have_python
instead of catching UNSUPPORTED_ERROR.
* exceptions.h (UNSUPPORTED_ERROR): Delete.
* python/python.c (source_python_script) [!HAVE_PYTHON]: Internal
error if called.
* python/python.h (have_python): New static inline function.
This patch sanitizes the access to gdbarch made by various functions of
the SDT probe API. Before this patch, gdbarch was being accessed via
the probe's objfile; however, this proved to cause a bug on 32-bit ARM
targets because during the parsing of the probe's arguments the code
needed to access some pseudo-registers of the architecture, and this
information is not fully correct on the objfile's gdbarch.
Basically, the approach taken was to instead pass the current/selected
frame to the parsing and evaluation functions, so that they can extract
the gdbarch directly from the frame. It solved the ARM bug reported
above, and also contributed to make the API cleaner.
Tested on x86_64 and 32-bit ARM.
2013-12-11 Sergio Durigan Junior <sergiodj@redhat.com>
* break-catch-throw.c (fetch_probe_arguments): Pass selected frame
to get_probe_argument_count and evaluate_probe_argument.
* probe.c (get_probe_argument_count): Adjust declaration to accept
frame. Pass frame to probe_ops's get_probe_argument_count.
(evaluate_probe_argument): Likewise, for evaluate_probe_argument.
(probe_safe_evaluate_at_pc): Pass frame to
get_probe_argument_count and evaluate_probe_argument.
* probe.h (struct probe_ops) <get_probe_argument_count,
evaluate_probe_argument>: Adjust declarations to accept frame.
(get_probe_argument_count, evaluate_probe_argument): Likewise.
* solib-svr4.c (solib_event_probe_action): Get current frame.
Pass it to get_probe_argument_count.
(svr4_handle_solib_event): Get current frame. Pass it to
get_probe_argument_count and evaluate_probe_argument.
* stap-probe.c (stap_parse_probe_arguments): Adjust declaration to
accept gdbarch. Do not obtain it from the probe's objfile.
(stap_get_probe_argument_count): Adjust declaration to accept
frame. Obtain gdbarch from the frame. Call generic
can_evaluate_probe_arguments. Pass gdbarch to
stap_parse_probe_arguments.
(stap_get_arg): Adjust declaration to accept gdbarch. Pass it to
stap_parse_probe_arguments.
(stap_evaluate_probe_argument): Adjust declaration to accept
frame. Obtain gdbarch from the frame. Pass gdbarch to
stap_get_arg.
(stap_compile_to_ax): Pass agent_expr's gdbarch to stap_get_arg.
(compute_probe_arg): Obtain gdbarch from frame. Pass frame to
get_probe_argument_count and evaluate_probe_argument.
* c-lang.c (c_get_string): Ignore the declared size of the object
if a specific length is requested.
testsuite/
* gdb.python/py-value.c: #include stdlib.h, string.h.
(str): New struct.
(main): New local xstr.
* gdb.python/py-value.exp (test_value_in_inferior): Add test to
fetch a value as a string with a length beyond the declared length
of the array.
GDB is able to cache memory accesses requested in target_read_code,
so target_read_code is more efficient than general target_read_memory.
This patch uses target_read_code and its variants to read target
memory in the functions related to i386_skip_prologue. It improves
the performance when doing 'b foo' (foo is a function) in remote
debugging.
Nowadays, when we set a breakpoint on function f1, GDB will fetch the
code in f1 to determine the start of the function body (say skip the
prologue), it requests read from target many times. With this patch
applied, the number of RSP 'm' packets are reduced.
gdb:
2013-12-10 Yao Qi <yao@codesourcery.com>
* corefile.c (read_code): New function.
(read_code_integer): New function.
(read_code_unsigned_integer): New function.
* gdbcore.h (read_code): Declare.
(read_code_integer): Declare.
(read_code_unsigned_integer): Declare.
* i386-tdep.c (i386_follow_jump): Call target_read_code instead
of target_read_memory. Call read_code_unsigned_integer instead
of read_memory_unsigned_integer.
(i386_analyze_struct_return): Likewise.
(i386_skip_probe): Likewise.
(i386_analyze_stack_align): Likewise.
(i386_match_pattern): Likewise.
(i386_skip_noop): Likewise.
(i386_analyze_frame_setup): Likewise.
(i386_analyze_register_saves): Likewise.
(i386_skip_prologue): Likewise.
(i386_skip_main_prologue): Likewise.
(i386_frame_cache_1): Likewise.
This helps with the following issue: Given an Ada program defining
a global variable:
package Pck is
Watch : Integer := 1974;
end Pck;
When printing the address of this variable, GDB also tries to print
the associated symbol name:
(gdb) p watch'address
$1 = (access integer) 0x6139d8 <pck__watch>
^^
||
The problem is that GDB prints the variable's linkage name, instead
of its natural name. This is because the language of the associated
minimal symbol never really gets set.
This patch adds handling for Ada symbols in symbol_find_demangled_name.
After this patch, we now get:
(gdb) p watch'address
$1 = (access integer) 0x6139d8 <pck.watch>
^
|
gdb/ChangeLog:
* symtab.c (symbol_find_demangled_name): Add handling of
Ada symbols.
gdb/testsuite/ChangeLog:
* gdb.ada/int_deref.exp: Add test verifying that we print
the decoded symbol name when printing the address of Ada
symbols.
This adds "exec-run-start-option" in the output of the -list-features
commands, allowing front-ends to easily determine whether -exec-run
supports the --start option.
gdb/ChangeLog:
* mi/mi-main.c (mi_cmd_list_features): add "exec-run-start-option".
* NEWS: Expand the entry documenting the new -exec-run --start
option to mention the corresponding new entry in the output of
"-list-features".
gdb/doc/ChangeLog:
* gdb.texinfo (GDB/MI Miscellaneous Commands): Document the new
"exec-run-start-option" entry in the output of the "-list-features"
command.
gdb/testsuite/ChangeLog:
* gdb.mi/mi-start.exp: Add test verifying that -list-features
contains "exec-run-start-option".
We observed on Windows 2012 that we were unable to unwind past
exception handlers. For instance, with any Ada program raising
an exception that does not get handled:
% gnatmake -g a -bargs -shared
% gdb a
(gdb) start
(gdb) catch exception unhandled
Catchpoint 2: unhandled Ada exceptions
(gdb) c
Catchpoint 2, unhandled CONSTRAINT_ERROR at <__gnat_unhandled_exception> (
e=0x645ff820 <constraint_error>) at s-excdeb.adb:53
53 s-excdeb.adb: No such file or directory.
At this point, we can already see that something went wrong, since
the frame selected by the debugger corresponds to a runtime function
rather than the function in the user code that caused the exception
to be raised (in our case procedure A).
This is further confirmed by the fact that we are unable to unwind
all the way to procedure A:
(gdb) bt
#0 <__gnat_unhandled_exception> (e=0x645ff820 <constraint_error>)
at s-excdeb.adb:53
#1 0x000000006444e9a3 in <__gnat_notify_unhandled_exception> (excep=0x284d2
+0)
at a-exextr.adb:144
#2 0x00000000645f106a in __gnat_personality_imp ()
from C:\[...]\libgnat-7.3.dll
#3 0x000000006144d1b7 in _GCC_specific_handler (ms_exc=0x242fab0,
this_frame=0x242fe60, ms_orig_context=0x242f5c0, ms_disp=0x242ef70,
gcc_per=0x645f0960 <__gnat_personality_imp>)
at ../../../src/libgcc/unwind-seh.c:289
#4 0x00000000645f1211 in __gnat_personality_seh0 ()
from C:\[...]\libgnat-7.3.dll
#5 0x000007fad3879f4d in ?? ()
Backtrace stopped: previous frame inner to this frame (corrupt stack?)
It turns out that the unwinder has been doing its job flawlessly
up until frame #5. The address in frame #5 is correct, but GDB
is not able to associate it with any symbol or unwind record.
And this is because this address is inside ntdll.dll, and when
we received the LOAD_DLL_DEBUG_EVENT for that DLL, the system
was not able to tell us the name of the library, thus causing us
to silently ignoring the event. Because GDB does not know about
ntdll.dll, it is unable to access the unwind information from it.
And because the function at that address does not use a frame
pointer, the unwinding becomes impossible.
This patch helps recovering ntdll.dll at the end of the "run/attach"
phase, simply by trying to locate that specific DLL again.
In terms of our medium to long term planning, it seems to me that
we should be able to simplify the code by ignoring LOAD_DLL_DEBUG_EVENT
during the startup phase, and modify windows_ensure_ntdll_loaded
to then detect and report all shared libraries after we've finished
inferior creation. But for a change just before 7.7 branch creation,
I thought it was safest to just handle ntdll.dll specifically. This
is less intrusive, and ntdll is the only DLL affected by the problem
I know so far.
gdb/ChangeLog:
* windows-nat.c (handle_load_dll): Add comments.
(windows_ensure_ntdll_loaded): New function.
(do_initial_windows_stuff): Use windows_ensure_ntdll_loaded.
Add FIXME comment.
Hi,
I see such warning below on one compiler I am using.
cc1: warnings being treated as errors
../../workspace/gdb/stack.c: In function 'frame_info':
../../workspace/gdb/stack.c:1519:20: error: 'caller_pc' may be used uninitialized in this function
Go through the gdb-patches archives and find the "canonical" way to
fix this warning is to initialize the variable.
gdb:
2013-12-08 Yao Qi <yao@codesourcery.com>
* stack.c (frame_info): Initialize variable caller_pc.
Doing "info frame" in the outermost frame, when that was indicated by
the next frame saying the unwound PC is undefined/not saved, results
in error and incomplete output:
(gdb) bt
#0 thread_function0 (arg=0x0) at threads.c:63
#1 0x00000034cf407d14 in start_thread (arg=0x7ffff7fcb700) at pthread_create.c:309
#2 0x000000323d4f168d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:115
(gdb) frame 2
#2 0x000000323d4f168d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:115
115 call *%rax
(gdb) info frame
Stack level 2, frame at 0x0:
rip = 0x323d4f168d in clone (../sysdeps/unix/sysv/linux/x86_64/clone.S:115); saved rip Register 16 was not saved
(gdb)
Not saved register values are treated as optimized out values
internally throughout. stack.c:frame_info is handing unvailable
values, but not optimized out ones. The patch deletes the
frame_unwind_caller_pc_if_available wrapper function and instead lets
errors propagate to frame_info (it's only user).
As frame_unwind_pc now needs to be able to handle and cache two
different error scenarios, the prev_pc.p variable is replaced with an
enumeration.
(FWIW, I looked into making gdbarch_unwind_pc or a variant return
struct value's instead, but it results in lots of boxing and unboxing
for no real gain -- e.g., the mips and arm implementations need to do
computation on the unboxed PC value. Might as well throw an error on
first attempt to get at invalid contents.)
After the patch, we get:
(gdb) info frame
Stack level 2, frame at 0x0:
rip = 0x323d4f168d in clone (../sysdeps/unix/sysv/linux/x86_64/clone.S:115); saved rip = <not saved>
Outermost frame: outermost
caller of frame at 0x7ffff7fcafc0
source language asm.
Arglist at 0x7ffff7fcafb8, args:
Locals at 0x7ffff7fcafb8, Previous frame's sp is 0x7ffff7fcafc8
(gdb)
A new test is added. It's based off dw2-reg-undefined.exp, and tweaked to
mark the return address (rip) of "stop_frame" as undefined.
Tested on x86_64 Fedora 17.
gdb/
2013-12-06 Pedro Alves <palves@redhat.com>
* frame.c (enum cached_copy_status): New enum.
(struct frame_info) <prev_pc.p>: Change type to enum
cached_copy_status.
(fprint_frame): Handle not saved and unavailable prev_pc values.
(frame_unwind_pc_if_available): Delete and merge contents into ...
(frame_unwind_pc): ... here. Handle OPTIMIZED_OUT_ERROR. Adjust
to use enum cached_copy_status.
(frame_unwind_caller_pc_if_available): Delete.
(create_new_frame): Adjust.
* frame.h (frame_unwind_caller_pc_if_available): Delete
declaration.
* stack.c (frame_info): Use frame_unwind_caller_pc instead of
frame_unwind_caller_pc_if_available, and handle
NOT_AVAILABLE_ERROR and OPTIMIZED_OUT_ERROR errors.
* valprint.c (val_print_optimized_out): Use val_print_not_saved.
(val_print_not_saved): New function.
* valprint.h (val_print_not_saved): Declare.
gdb/testsuite/
2013-12-06 Pedro Alves <palves@redhat.com>
* gdb.dwarf2/dw2-undefined-ret-addr.S: New file.
* gdb.dwarf2/dw2-undefined-ret-addr.c: New file.
* gdb.dwarf2/dw2-undefined-ret-addr.exp: New file.
In order to catch <optimized out> errors like we catch <unavailable>
errors, this adds a new OPTIMIZED_OUT_ERROR error code, and throws it
in various places.
gdb/ChangeLog
2013-12-06 Andrew Burgess <aburgess@broadcom.com>
Pedro Alves <palves@redhat.com>
* exceptions.h (errors): Add OPTIMIZED_OUT_ERROR.
* dwarf2loc.c (write_pieced_value): Throw OPTIMIZED_OUT_ERROR.
* frame.c (frame_unwind_register): Throw OPTIMIZED_OUT_ERROR.
* spu-tdep.c (spu_software_single_step): Throw
OPTIMIZED_OUT_ERROR.
* valops.c (value_assign): Throw OPTIMIZED_OUT_ERROR.
The introductory comment to free_objfile is obsolete.
This patch fixes it by removing all the obsolete bits.
2013-12-06 Tom Tromey <tromey@redhat.com>
* objfiles.c (free_objfile): Update comment.
I happened to notice that nothing uses objfile_to_front.
This patch removes it.
2013-12-06 Tom Tromey <tromey@redhat.com>
* objfiles.h (objfile_to_front): Remove.
* objfiles.c (objfile_to_front): Remove.
This removes an unnecessary declaration from minsyms.c.
2013-12-06 Tom Tromey <tromey@redhat.com>
* minsyms.c (get_symbol_leading_char): Remove unnecessary
declaration.
This improves the packing of struct partial_symtab. I noticed with
pahole that were were a couple of holes. This consolidates the holes
without, I think, affecting readability -- it just moves the "user"
field a bit earlier in the struct. This change saves a small amount
of memory.
2013-12-06 Tom Tromey <tromey@redhat.com>
* psympriv.h (struct partial_symtab) <user>: Move earlier.
This fixes a couple of old "32x64" FIXME comments by using paddress
with current_gdbarch rather than hex_string and a cast to long.
2013-12-06 Tom Tromey <tromey@redhat.com>
* cli/cli-cmds.c (edit_command): Use paddress, not hex_string.
(list_command): Likewise.
This puts the psymtab filename in the filename bcache.
This saves a small amount of memory.
2013-12-06 Tom Tromey <tromey@redhat.com>
* psymtab.c (allocate_psymtab): Put the filename in the filename
bcache.
This makes symtab::dirname const and updates one spot to avoid an
intermediate constless result.
2013-12-06 Tom Tromey <tromey@redhat.com>
* buildsym.c (end_symtab_from_static_block): Use obstack_copy0.
* symtab.h (struct symtab) <dirname>: Now const.
This makes symtab::filename const and removes a newly unnecessary
cast.
2013-12-06 Tom Tromey <tromey@redhat.com>
* symfile.c (allocate_symtab): Remove cast.
* symtab.h (struct symtab) <filename>: Now const.
While looking into the probe API, it seemed to me that there were a
number of methods in sym_probe_fns that were not needed. This patch
removes them.
Specifically, it seems to me that sym_probe_fns ought to be concerned
with the API for constructing the probes. Any method relating to some
aspect of an individual probe can be handled via the probe's own
vtable. That is, the double indirection here doesn't seem useful --
it certainly isn't in fact used, but also I couldn't think of a
potential use.
2013-12-06 Tom Tromey <tromey@redhat.com>
* break-catch-throw.c (fetch_probe_arguments): Use
get_probe_argument_count and evaluate_probe_argument.
* elfread.c (elf_get_probe_argument_count)
(elf_can_evaluate_probe_arguments, elf_evaluate_probe_argument)
(elf_compile_to_ax): Remove.
(elf_probe_fns): Update.
* probe.c (get_probe_argument_count, can_evaluate_probe_arguments)
(evaluate_probe_argument): Call method on probe, not via sym
functions.
* stap-probe.c (compute_probe_arg): Use get_probe_argument_count,
evaluate_probe_argument.
(compile_probe_arg): Use get_probe_argument_count. Call method on
probe, not via sym functions.
* symfile-debug.c (debug_sym_get_probe_argument_count)
(debug_can_evaluate_probe_arguments)
(debug_sym_evaluate_probe_argument, debug_sym_compile_to_ax):
Remove.
(debug_sym_probe_fns): Remove.
* symfile.h (struct sym_probe_fns) <sym_get_probe_argument_count,
can_evaluate_probe_arguments, sym_evaluate_probe_argument,
sym_compile_to_ax>: Remove fields.
