This patch is a fix to PR gdb/17235. The bug is about an unused
variable that got declared and set during one of the parsing phases of
an SDT probe's argument. I took the opportunity to rewrite some of the
code to improve the parsing. The bug was actually a thinko, because
what I wanted to do in the code was to discard the number on the string
being parsed.
During this portion, the code identifies that it is dealing with an
expression that begins with a sign ('+', '-' or '~'). This means that
the expression could be:
- a numeric literal (e.g., '+5')
- a register displacement (e.g., '-4(%rsp)')
- a subexpression (e.g., '-(2*3)')
So, after saving the sign and moving forward 1 char, now the code needs
to know if there is a digit followed by a register displacement prefix
operand (e.g., '(' on x86_64). If yes, then it is a register
operation. If not, then it will be handled recursively, and the code
will later apply the requested operation on the result (either a '+', a
'-' or a '~').
With the bug, the code was correctly discarding the digit (though using
strtol unnecessarily), but it wasn't properly dealing with
subexpressions when the register indirection prefix was '(', like on
x86_64. This patch also fixes this bug, and includes a testcase. It
passes on x86_64 Fedora 20.
valgrind caught that parse_number reads uninitialized memory when we
parse literal "0":
$ valgrind ./gdb -q -nx -ex "set height 0"
(...)
==10378== Conditional jump or move depends on uninitialised value(s)
==10378== at 0x548A10: parse_number (c-exp.y:1828)
==10378== by 0x54A340: lex_one_token (c-exp.y:2638)
==10378== by 0x54B4BB: c_lex (c-exp.y:3089)
==10378== by 0x544951: c_parse_internal (c-exp.c:2208)
==10378== by 0x54BF8C: c_parse (c-exp.y:3260)
==10378== by 0x6502E7: parse_exp_in_context_1 (parse.c:1221)
==10378== by 0x650064: parse_exp_in_context (parse.c:1122)
==10378== by 0x65001F: parse_exp_1 (parse.c:1114)
==10378== by 0x650421: parse_expression (parse.c:1266)
==10378== by 0x5A74B7: parse_and_eval_long (eval.c:92)
==10378== by 0x501ABD: do_set_command (cli-setshow.c:302)
==10378== by 0x721059: execute_command (top.c:452)
==10378==
(gdb)
I've pushed the obvious fix.
Tested on x86_64 Fedora 20.
gdb/ChangeLog:
* c-exp.y (parse_number): Skip handling base-switching prefixes if
the input is only one character long.
This commit fixes the PR mentioned in $subject. It is about a set but
unused variable that refers to the output format of integer values
printed in Fortran.
This was probably a thinko (like most set-but-unused-vars), but it
could cause an internal error depending on the scenario. I am sending
a testcase which triggers this error as well.
gdb/ChangeLog:
2014-09-04 Sergio Durigan Junior <sergiodj@redhat.com>
PR fortran/17237
* f-valprint.c (f_val_print): Specify the correct print option to
use when printing integer values.
gdb/testsuite/ChangeLog:
2014-09-04 Sergio Durigan Junior <sergiodj@redhat.com>
PR fortran/17237
* gdb.fortran/print-formatted.exp: New file.
* gdb.fortran/print-formatted.f90: Likewise.
Historically the Linux x86 watchpoint code did not cope with multi-
threaded processes and LWP IDs were passed to it wrapped as PIDs.
Not all entry points were converted when the Linux x86 watchpoint
code was made multi-thread-aware, so a handler was left in place to
cope with wrapped LWPs. Since then all such entry points have been
converted to pass regular LWPs and the handler is now redundant.
This commit removes the handler and adds assertions to ensure no
wrapped LWPs are passed in future.
gdb/ChangeLog:
* x86-linux-nat.c (x86_linux_dr_get, x86_linux_dr_set):
Remove code to cope with LWPs wrapped as PIDs.
Add assertions to ensure no wrapped LWPs are passed.
Git 9a0dc9e3 regressed gdb.dwarf2/pieces-optimized-out.exp, visible on
i686 (the test doesn't run on x86_64):
(gdb) p s
-$1 = {a = 5, b = <optimized out>, c = <optimized out>, d = <optimized out>}
+$1 = {a = 5, b = <optimized out>, c = 0, d = 0}
-(gdb) PASS: gdb.dwarf2/pieces-optimized-out.exp: print s
+(gdb) FAIL: gdb.dwarf2/pieces-optimized-out.exp: print s
The regression was caused by this removal in cp-valprint.c:
@@ -293,12 +293,6 @@ cp_print_value_fields (struct type *type, struct type *real_type,
{
fputs_filtered (_("<synthetic pointer>"), stream);
}
- else if (!value_bits_valid (val,
- TYPE_FIELD_BITPOS (type, i),
- TYPE_FIELD_BITSIZE (type, i)))
- {
- val_print_optimized_out (val, stream);
- }
else
{
struct value_print_options opts = *options;
The idea was that we'd just fallback to calling value_field_bitfield,
which handles unavailable values (in unpack_value_bits_as_long_1) so
should be able to handle optimized out values too. Alas, it doesn't.
This is currently a bit too messy. Instead of teaching
unpack_value_bits_as_long_1 about optimized out bits, let's bite the
bullet and teach the value code to handle partially optimized out
bitfield, by having it unpack a bitfield and then propagate the range
metadata. Turns out the resulting code looks simpler and clearer.
Tested on x86_64 Fedora 20, -m64/-m32.
gdb/ChangeLog:
* value.c (value_ranges_copy_adjusted): New function, factored out
from ...
(value_contents_copy_raw): ... here.
(unpack_value_bits_as_long_1): Rename back to ...
(unpack_bits_as_long): ... this. Remove 'original_value' and
'result' parameters. Change return type to LONGEST.
(unpack_value_bits_as_long): Delete.
(unpack_value_field_as_long_1): Delete.
(unpack_value_field_as_long, unpack_field_as_long): Reimplement.
(unpack_value_bitfield): New function.
(value_field_bitfield): Reimplement using unpack_value_bitfield.
(value_fetch_lazy): Use unpack_value_bitfield.
* value.h (unpack_value_bits_as_long): Delete declaration.
gdb/doc/ChangeLog:
* python.texi (Types In Python): Type.template_argument(n) returns a
gdb.Value or a gdb.Type and throws an exception if n is out of
range.
The ability to read registers is needed to use Frame Filter API to
display the frames created by JIT compilers.
gdb/ChangeLog:
2014-08-29 Sasha Smundak <asmundak@google.com>
* python/py-frame.c (frapy_read_register): New function.
gdb/doc/ChangeLog:
2014-08-26 Sasha Smundak <asmundak@google.com>
* python.texi (Frames in Python): Add read_register description.
gdb/testsuite/ChangeLog:
2014-08-26 Sasha Smundak <asmundak@google.com>
* gdb.python/py-frame.exp: Test Frame.read_register.
PTRACE_PEEKUSER can return -1, which is usually used to determine whether
a system call has reported an error, so errno must be used alone to
determine whether an error occurred. However errno isn't modified by a
successful system call so it must be reset to a known value (0) before the
syscall call.
Add the missing errno reset when reading the DSP_CONTROL register in the
native MIPS Linux backend and the MIPS gdbserver backend.
gdb/:
* mips-linux-nat.c (mips_linux_read_description): Reset errno to 0
prior to reading DSP_CONTROL with PTRACE_PEEKUSER ptrace call.
gdb/gdbserver/:
* linux-mips-low.c (mips_read_description): Reset errno to 0 prior
to reading DSP_CONTROL with PTRACE_PEEKUSER ptrace call.
This PR came from a Red Hat bug that was filed recently. I checked and
it still exists on HEAD, so here's a proposed fix. Although this is
marked as a Python backend bug, this is really about the completion
mechanism used by GDB. Since this code reminds me of my first attempt
to make a good noodle, it took me quite some time to fix it in a
non-intrusive way.
The problem is triggered when one registers a completion method inside a
class in a Python script, rather than registering the command using a
completer class directly. For example, consider the following script:
class MyFirstCommand(gdb.Command):
def __init__(self):
gdb.Command.__init__(self,'myfirstcommand',gdb.COMMAND_USER,gdb.COMPLETE_FILENAME)
def invoke(self,argument,from_tty):
raise gdb.GdbError('not implemented')
class MySecondCommand(gdb.Command):
def __init__(self):
gdb.Command.__init__(self,'mysecondcommand',gdb.COMMAND_USER)
def invoke(self,argument,from_tty):
raise gdb.GdbError('not implemented')
def complete(self,text,word):
return gdb.COMPLETE_FILENAME
MyFirstCommand ()
MySecondCommand ()
When one loads this into GDB and tries to complete filenames for both
myfirstcommand and mysecondcommand, she gets:
(gdb) myfirstcommand /hom<TAB>
(gdb) myfirstcommand /home/
^
...
(gdb) mysecondcommand /hom<TAB>
(gdb) mysecondcommand /home
^
(The "^" marks the final position of the cursor after the TAB).
So we see that myfirstcommand honors the COMPLETE_FILENAME class (as
specified in the command creation), but mysecondcommand does not. After
some investigation, I found that the problem lies with the set of word
break characters that is used for each case. The set should be the same
for both commands, but it is not.
During the process of deciding which type of completion should be used,
the code in gdb/completer.c:complete_line_internal analyses the command
that requested the completion and tries to determine the type of
completion wanted by checking which completion function will be called
(e.g., filename_completer for filenames, location_completer for
locations, etc.).
This all works fine for myfirstcommand, because immediately after the
command registration the Python backend already sets its completion
function to filename_completer (which then causes the
complete_line_internal function to choose the right set of word break
chars). However, for mysecondcommand, this decision is postponed to
when the completer function is evaluated, and the Python backend uses an
internal completer (called cmdpy_completer). complete_line_internal
doesn't know about this internal completer, and can't choose the right
set of word break chars in time, which then leads to a bad decision when
completing the "/hom" word.
So, after a few attempts, I decided to create another callback in
"struct cmd_list_element" that will be responsible for handling the case
when there is an unknown completer function for complete_line_internal
to work with. So far, only the Python backend uses this callback, and
only when the user provides a completer method instead of registering
the command directly with a completer class. I think this is the best
option because it not very intrusive (all the other commands will still
work normally), but especially because the whole completion code is so
messy that it would be hard to fix this without having to redesign
things.
I have regtested this on Fedora 18 x86_64, without regressions. I also
included a testcase.
gdb/ChangeLog:
2014-09-03 Sergio Durigan Junior <sergiodj@redhat.com>
PR python/16699
* cli/cli-decode.c (set_cmd_completer_handle_brkchars): New
function.
(add_cmd): Set "completer_handle_brkchars" to NULL.
* cli/cli-decode.h (struct cmd_list_element)
<completer_handle_brkchars>: New field.
* command.h (completer_ftype_void): New typedef.
(set_cmd_completer_handle_brkchars): New prototype.
* completer.c (set_gdb_completion_word_break_characters): New
function.
(complete_line_internal): Call "completer_handle_brkchars"
callback from command.
* completer.h: Include "command.h".
(set_gdb_completion_word_break_characters): New prototype.
* python/py-cmd.c (cmdpy_completer_helper): New function.
(cmdpy_completer_handle_brkchars): New function.
(cmdpy_completer): Adjust to use cmdpy_completer_helper.
(cmdpy_init): Set completer_handle_brkchars to
cmdpy_completer_handle_brkchars.
gdb/testsuite/ChangeLog:
2014-09-03 Sergio Durigan Junior <sergiodj@redhat.com>
PR python/16699
* gdb.python/py-completion.exp: New file.
* gdb.python/py-completion.py: Likewise.
The loop macro ALL_DEBUG_REGISTERS does not iterate over the status or
control registers, so its name is misleading. This commit renames it
as ALL_DEBUG_ADDRESS_REGISTERS and updates all uses. This commit also
updates its loop conditions to an equivalent but better form, and
makes two functions use it that had previously hardwired the loop.
A comment on a related field in the x86_debug_reg_state structure is
also updated to reflect that the field refers specifically to address
registers only.
gdb/ChangeLog:
* nat/x86-dregs.h (ALL_DEBUG_REGISTERS): Renamed as...
(ALL_DEBUG_ADDRESS_REGISTERS): New macro. All uses updated.
Loop conditions changed to equivalent form.
(struct x86_debug_reg_state): Updated dr_ref_count comment.
* x86-linux-nat.c (x86_linux_prepare_to_resume): Use
ALL_DEBUG_ADDRESS_REGISTERS.
gdb/gdbserver/ChangeLog:
* linux-x86-low.c (x86_linux_prepare_to_resume): Use
ALL_DEBUG_ADDRESS_REGISTERS.
This commit renames nine files that contain code used by both 32- and
64-bit Intel ports such that their names are prefixed with "x86"
rather than "i386". All types, functions and variables within these
files are likewise renamed such that their names are prefixed with
"x86" rather than "i386". This makes GDB follow the convention used
by gdbserver such that 32-bit Intel code lives in files called
"i386-*", 64-bit Intel code lives in files called "amd64-*", and code
for both 32- and 64-bit Intel lives in files called "x86-*".
This commit only renames OS-independent files. The Linux ports of
both GDB and gdbserver now follow the i386/amd64/x86 convention fully.
Some ports still use the old convention where "i386" in file/function/
type/variable names can mean "32-bit only" or "32- and 64-bit" but I
don't want to touch ports I can't fully test except where absolutely
necessary.
gdb/ChangeLog:
* i386-nat.h: Renamed as...
* x86-nat.h: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
* i386-nat.c: Renamed as...
* x86-nat.c: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
* common/i386-xstate.h: Renamed as...
* common/x86-xstate.h: New file. All type, function and variable
name prefixes changed from "i386_" to "x86_". All references
updated.
* nat/i386-cpuid.h: Renamed as...
* nat/x86-cpuid.h: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
* nat/i386-gcc-cpuid.h: Renamed as...
* nat/x86-gcc-cpuid.h: New file. All type, function and variable
name prefixes changed from "i386_" to "x86_". All references
updated.
* nat/i386-dregs.h: Renamed as...
* nat/x86-dregs.h: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
* nat/i386-dregs.c: Renamed as...
* nat/x86-dregs.c: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
gdb/gdbserver/ChangeLog:
* i386-low.h: Renamed as...
* x86-low.h: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
* i386-low.c: Renamed as...
* x86-low.c: New file. All type, function and variable name
prefixes changed from "i386_" to "x86_". All references updated.
This commit replaces two uses of xcalloc (1, ...) with XCNEW.
gdb/gdbserver/ChangeLog:
* linux-x86-low.c (x86_linux_new_process): Use XCNEW.
(x86_linux_new_thread): Likewise.
This commit replaces the hacky "exception" system in gdbserver with
the exceptions and cleanups subsystem from GDB.
Only the catch/cleanup code in what was "main" has been updated to
use the new system. Other parts of gdbserver can now be converted
to use TRY_CATCH and cleanups on an as-needed basis.
A side-effect of this commit is that some error messages will change
slightly, and in cases with multiple errors the error messages will
be printed in a different order.
gdb/gdbserver/ChangeLog:
* server.h (setjmp.h): Do not include.
(toplevel): Do not declare.
(common-exceptions.h): Include.
(cleanups.h): Likewise.
* server.c (toplevel): Do not define.
(exit_code): New static global.
(detach_or_kill_for_exit_cleanup): New function.
(main): New function. Original main renamed to...
(captured_main): New function.
* utils.c (verror) [!IN_PROCESS_AGENT]: Use throw_verror.
This commit creates a new file, common/gdb_setjmp.h, to hold some
portability macros for setjmp/longjmp et al. that are used by the
exceptions subsystem and by the demangler crash catcher.
gdb/ChangeLog:
* common/gdb_setjmp.h: New file.
* Makefile.in (HFILES_NO_SRCDIR): Add common/gdb_setjmp.h.
* configure.ac: Move sigsetjmp check...
* common/common.m4: ...here.
* configure: Regenerate.
* cp-support.c (SIGJMP_BUF): Delete.
(SIGSETJMP): Likewise.
(SIGLONGJMP): Likewise.
* exceptions.h (gdb_setjmp.h): Include.
(setjmp.h): Do not include.
(EXCEPTIONS_SIGJMP_BUF): Delete.
(EXCEPTIONS_SIGSETJMP): Likewise.
(EXCEPTIONS_SIGLONGJMP): Likewise.
Replace all uses of EXCEPTIONS_SIG* macros with SIG* macros
from gdb_setjmp.h.
* exceptions.c: Likewise.
gdb/gdbserver/ChangeLog:
* config.in: Regenerate.
* configure: Likewise.
This commit moves cleanups.[ch] into gdb/common/. The only change to
the content of the files is that cleanups.c's include list was altered
to match its new location.
gdb/ChangeLog:
* cleanups.h: Moved to...
* common/cleanups.h: New file.
* cleanups.c: Moved to...
* common/cleanups.c: New file. Include common-defs.h and
cleanups.h. Do not include defs.h.
* Makefile.in (SFILES): Replace cleanups.c with common/cleanups.c.
(HFILES_NO_SRCDIR): Replace cleanups.h with common/cleanups.h.
(cleanups.o): New rule.
gdb/gdbserver/ChangeLog:
* Makefile.in (SFILES): Add common/cleanups.c.
(OBS): cleanups.o.
(cleanups.o): New rule.
Various warnings are emitted during startup and option-parsing using
fprintf_unfiltered. One warning is prefixed with the command name,
the others are not. This commit replaces these hardwired warnings
with calls to warning. It also sets warning_pre_print to prefix all
warnings with the command name until option parsing is complete.
gdb/ChangeLog:
* main.c (captured_main): Use warning during startup.
Prefix startup warning messages with command name.
This commit replaces the hardwired fprintf/exit error handlers
for usage errors with calls to error.
gdb/ChangeLog:
* main.c (captured_main): Handle usage errors with error.
go32_create_inferior invokes a hardwired fprintf/exit error handler
if v2loadimage fails. I could find no reason for this other than that
the block seems to have been copy-and-pasted from v2loadimage's
manpage. This commit replaces the hardwired handler with a call to
error.
gdb/ChangeLog:
* go32-nat.c (go32_create_inferior): Replace a fprintf/
exit pair with a call to error. Wrap the message with _().
If the requested interpreter cannot be set captured_main reports the
error with a hardwired fprintf/exit pair. A fprintf/exit pair on the
previous line was replaced with a call to error in March 2003
(https://sourceware.org/ml/gdb-patches/2003-03/msg00444.html) but I
found no documentation as to why this particular hardwired handler
was left untouched. I was also unable to come up with a situation
where error would not be suitable, so I have replaced it with a call
to error.
gdb/ChangeLog:
* main.c (captured_main): Replace a fprintf/exit
pair with a call to error. Wrap the message with _().
tui_initialize_io contains a pair of hardwired fprintf/exit error
handlers. I was unable to find any documentation as to why they're
hardwired (the code appeared in a monolithic block back in 2001:
https://sourceware.org/ml/gdb-patches/2001-07/msg00490.html) and I
was also unable to come up with a situation where error would not
be suitable, so I have replaced both handlers with calls to error.
gdb/ChangeLog:
* tui/tui-io.c (tui_initialize_io): Replace two fprintf/exit
pairs with calls to error. Wrap the message with _().
warning will crash if called before the first call to set_width. This
commit makes the warning usable from the moment gdb_stderr is set up.
gdb/ChangeLog:
* utils.c (vwarning): Protect calls to target_terminal_ours
and wrap_here.
error (and other exception-throwing functions) are callable from the
first line of captured_main, but the exception printing code will
crash if called before the first call to set_width. This commit makes
the exception printing code usable from the moment gdb_stderr is set
up.
gdb/ChangeLog:
* exceptions.c (print_flush): Protect calls to
target_terminal_ours and wrap_here.
internal_vproblem can be called (via malloc_failure) from almost the
first line of captured_main, but it will crash if called before the
first call to set_width. This commit makes internal_vproblem work
at any time.
There are two parts to this. If called before gdb_stderr is set up,
internal_vproblem will fall back to printing the message on regular
stderr and aborting. If called after gdb_stderr is set up but before
filtered printing is set up, internal_vproblem will operate as usual
except that it can not query whether to quit and/or dump core so it
defaults to doing both.
gdb/ChangeLog:
* utils.h (filtered_printing_initialized): New declaration.
* utils.c (abort_with_message): New function.
(internal_vproblem): Use abort_with_message for first level
recursive internal problems, and if gdb_stderr is not set up.
Protect calls to target_terminal_ours, begin_line and query.
clang was using eax to construct %0 here:
asm ("mov %%eax, 0(%0)\n\t"
"mov %%ebx, 4(%0)\n\t"
"mov %%ecx, 8(%0)\n\t"
"mov %%edx, 12(%0)\n\t"
"mov %%esi, 16(%0)\n\t"
"mov %%edi, 20(%0)\n\t"
: /* no output operands */
: "r" (data)
: "eax", "ebx", "ecx", "edx", "esi", "edi");
which caused amd64-word.exp (and others similarly) to fail.
It's a perfectly legit thing for clang to do given the available data.
The patch fixes this by marking the registers as live from the
time of the preceding breakpoint.
gdb/testsuite/ChangeLog:
* gdb.arch/amd64-pseudo.c (main): Rewrite to better specify when
eax,etc. are live with values set by gdb and thus the compiler can't
use them.
* gdb.arch/i386-pseudo.c (main): Ditto.
This commit removes the now-unused fatal function and prototype.
gdb/gdbserver/ChangeLog:
* utils.h (fatal): Remove declaration.
* utils.c (fatal): Remove function.
This commit converts four calls to fatal into calls to
perror_with_name. perror_with_name calls error, which
in IPA terminates with exit (1) rather than longjmp, so
there is no functional change here.
gdb/gdbserver/ChangeLog:
* tracepoint.c (gdb_agent_init): Replace fatal with
perror_with_name.
(initialize_tracepoint): Likewise.
This commit converts a call to fatal in remote_prepare with a call to
error. remote_prepare is called precisely once, from main, at a point
where jumping to toplevel will call exit (1), so error and fatal are
functionally equivalent at this point. Note that remote_prepare calls
perror_with_name (which calls error) so callers of remote_prepare must
already handle the fact that it may exit via longjmp.
gdb/gdbserver/ChangeLog:
* remote-utils.c (remote_prepare): Replace fatal with error.
This commit downgrades a fatal error to a warning in linux_async.
linux_async is called from two different places in gdbserver:
Via target_async from handle_accept_event. The argument
is always zero, so the warning will never be printed here.
Via start_non_stop from handle_general_set. This prints
its own error message to stderr on failure, which will
be preceded by the warning if it is emitted.
gdb/gdbserver/ChangeLog:
* linux-low.c (linux_async): Replace fatal with warning.
Tidy up and return.
(linux_start_non_stop): Return -1 if linux_async failed.
This commit converts if..fatal checks in both i386_dr_low_set_addr
implementations to gdb_asserts. It's not obvious from the context,
but the conditional in both cases is changed to match the equivalent
conditional in the i386_dr_low_get_addr implementations. Nothing
fundamental has changed because DR_FIRSTADDR is zero. This commit
also removes a vague comment in Linux i386_dr_low_get_addr. I could
have reworded the comment (and replicated it three times for the other
identical assertions) but I think the existence of specific functions
for the status and control registers makes it fairly obvious what is
going on.
gdb/gdbserver/ChangeLog:
* linux-x86-low.c (i386_dr_low_set_addr): Replace check with
gdb_assert.
(i386_dr_low_get_addr): Remove vague comment.
* win32-i386-low.c (i386_dr_low_set_addr): Replace check with
gdb_assert.
Hi,
This patch is to handle a software watchpoint case that program returns
to caller's epilogue, and it causes the fail in thumb mode,
finish^M
Run till exit from #0 func () at gdb/testsuite/gdb.base/watchpoint-cond-gone.c:26^M
0x000001f6 in jumper ()^M
(gdb) FAIL: gdb.base/watchpoint-cond-gone.exp: Catch the no longer valid watchpoint
In the test, jumper calls func, and programs returns from func to
jumper's epilogue, IOW, the branch instruction is the last instruction
of jumper's function body.
jumper:
.....
0x000001f2 <+10>: bl 0x200 [1] <---- indirect call to func
0x000001f6 <+14>: mov sp, r7 [2] <---- start of the epilogue
0x000001f8 <+16>: add sp, #8
0x000001fa <+18>: pop {r7}
0x000001fc <+20>: pop {r0}
0x000001fe <+22>: bx r0
When the inferior returns from func back to jumper, it is expected
that an expression of a software watchpoint becomes out-of-scope.
GDB validates the expression by checking the corresponding frame,
but this check is guarded by gdbarch_in_function_epilogue_p. See
breakpoint.c:watchpoint_check.
It doesn't work in this case, because program returns from func's
epilogue back to jumper's epilogue [2], GDB thinks the program is
still within the epilogue, but in fact it goes to a different one.
When PC points at [2], the sp-restore instruction is to be
executed, so the stack frame isn't destroyed yet and we can still
use the frame mechanism reliably.
Note that when PC points to the first instruction of restoring SP,
it is part of epilogue, but we still return zero. When goes to
the next instruction, the backward scan will still match the
epilogue sequence correctly. The reason for doing this is to
handle the "return-to-epilogue" case.
What this patch does is to restrict the epilogue matching that let
GDB think the first SP restore instruction isn't part of the epilogue,
and fall back to use frame mechanism. We set 'found_stack_adjust'
zero before backward scan, and we've done this for arm mode
counterpart (arm_in_function_epilogue_p) too.
The patch is tested in arm-none-eabi and arm-none-linux-gnueabi with
various multilibs. OK to apply?
gdb:
2014-08-28 Yao Qi <yao@codesourcery.com>
* arm-tdep.c (thumb_in_function_epilogue_p): Don't set
found_stack_adjust in forward scan. Remove condition check
on found_stack_adjust which is always true. Indent the code.
Hi,
dwarf_decode_lines is called in two functions,
dwarf2_build_include_psymtabs and handle_DW_AT_stmt_list, in which, 1
is passed to argument 'want_line_info' and 'want_line_info' is a
conditional variable in dwarf_decode_lines. We can simplify it by
removing 'want_line_info' and propagating the constant 1 into
dwarf_decode_lines. This is what this patch does. This patch also
remove one line comment about WANT_LINE_INFO in
handle_DW_AT_stmt_list, as handle_DW_AT_stmt_list doesn't have such
argument.
gdb:
2014-08-28 Yao Qi <yao@codesourcery.com>
* dwarf2read.c (dwarf_decode_lines): Update declaration.
(handle_DW_AT_stmt_list): Remove comment about WANT_LINE_INFO.
(dwarf_decode_lines): Remove argument
want_line_info. Remove condition check on want_line_info.
Callers update.
The TUI terminal state becomes corrupted (e.g. key sequences such as
Alt_F and Alt_B no longer work) when one attaches to an inferior process
(via "run" or "attach") from within TUI. This terminal corruption
remains until you switch out of TUI mode.
This happens because the terminal state is not properly saved when
switching to and out from TUI mode. Although the functions tui_enable()
and tui_disable() both call the function target_terminal_save_ours() to
save the terminal state, this function is a no-op unless GDB has already
attached to an inferior process. This is because only the "native"
target has a useful implementation of target_terminal_save_ours()
(namely child_terminal_save_ours()) and we only have the "native" target
in our target vector if GDB has already attached to an inferior process.
So without an inferior process, switching to and from TUI mode does not
actually save the terminal state. Therefore when you attach to an
inferior process from within TUI mode, the proper terminal state is not
restored (after swapping from the inferior's terminal back to the GDB
terminal).
To fix this we just have to ensure that the terminal state is always
being properly saved when switching from and to TUI mode. To achieve
this, this patch removes the polymorphic function
target_terminal_save_ours() and replaces it with a regular function
gdb_save_tty_state() that always saves the terminal state.
Tested on x86_64-unknown-linux-gnu by running "make check", no new
regressions.
gdb/ChangeLog:
* target.h (struct target_ops::to_terminal_save_ours): Remove
declaration.
(target_terminal_save_ours): Remove macro.
* target-delegates.c: Regenerate.
* inf-child.c (inf_child_target): Don't set the nonexistent
field to_terminal_save_ours.
* inferior.h (child_terminal_save_ours): Remove declaration.
* terminal.h (gdb_save_tty_state): New declaration.
* inflow.c (child_terminal_save_ours): Rename to ...
(gdb_save_tty_state): ... this.
* tui/tui.c: Include terminal.h.
(tui_enable): Use gdb_save_tty_state instead of
target_terminal_save_ours.
(tui_disable): Likewise.
I read comment of scan_partial_symbols about NEED_PC and how *LOWPC
and *HIGHPC are updated:
DW_AT_ranges). If NEED_PC is set, then this function will set
*LOWPC and *HIGHPC to the lowest and highest PC values found in CU
and record the covered ranges in the addrmap.
NEED_PC is only used in the callee of scan_partial_symbols,
add_partial_subprogram,
if (pdi->tag == DW_TAG_subprogram)
{
if (pdi->has_pc_info)
{
if (pdi->lowpc < *lowpc)
*lowpc = pdi->lowpc;
if (pdi->highpc > *highpc)
*highpc = pdi->highpc;
if (need_pc)
*LOWPC and *HIGHPC is updated regardless of NEED_PC. When NEED_PC is
true, addrmap is updated. It would be clear to rename NEED_PC to
SET_ADDRMAP. That is what this patch does. Beside this, this patch
also adjust comments in related functions.
gdb:
2014-08-24 Yao Qi <yao@codesourcery.com>
* dwarf2read.c (scan_partial_symbols): Update comments.
Rename argument 'need_pc' with 'set_addrmap'.
（add_partial_namespace): Rename argument 'need_pc' with
'set_addrmap'.
(add_partial_module): Likewise.
(add_partial_subprogram): Likewise. Update comments.
(dwarf2_name): Fix typo.
