While digging into a different memory corruption error, I happened to
notice one coming from the linetable code. In a couple of spots, the
wrong termination condition was used in a loop, leading gdb to read
one element past the end of the linetable.
Built and regtested on x86-64 Fedora 18. Also verified using
valgrind. I'm checking this in.
2013-11-14 Tom Tromey <tromey@redhat.com>
* python/py-linetable.c (ltpy_has_line)
(ltpy_get_all_source_lines): Fix loop termination condition.
Frontend sometimes need to evaluate expressions that are
language-specific. For instance, Eclipse uses the following
expression to determine the size of an address on the target:
-data-evaluate-expression "sizeof (void*)"
Unfortunately, if the main of the program being debugged is not C,
this may not work. For instance, if the main is in Ada, you get...
-data-evaluate-expression "sizeof (void*)"
^error,msg="No definition of \"sizeof\" in current context."
... and apparently decides to stop the debugging session as a result.
The recommendation sent was to specifically set the language to C
before trying to evaluate the expression. Something such as:
1. save current language
2. set language c
3. -data-evaluate-expression "sizeof (void*)"
4. Restore language
This has the same disadvantages as the ones outlined in the "Context
Management" section of the GDB/MI documentation regarding setting
the current thread or the current frame, thus recommending the use of
general command-line switches such as --frame, or --thread instead.
This patch follows the same steps for the language, adding a similar
new command option: --language LANG. Example of use:
-data-evaluate-expression --language c "sizeof (void*)"
^done,value="4"
gdb/ChangeLog:
* mi/mi-parse.h (struct mi_parse) <language>: New field.
* mi/mi-main.c (mi_cmd_execute): Temporarily set language to
PARSE->LANGUAGE during command execution, if set.
* mi/mi-parse.c: Add "language.h" #include.
(mi_parse): Add parsing of "--language" command option.
* NEWS: Add entry mentioning the new "--language" command option.
gdb/testsuite/ChangeLog:
* gdb.mi/mi-language.exp: New file.
gdb/doc/ChangeLog:
* gdb.texinfo (Show): Add xref anchor for "show language" command.
(Context management): Place current subsection text into its own
subsubsection. Add new subsubsection describing the "--language"
command option.
This function provides the exact same functionality as extract_arg,
except that it takes a "const char**" instead of a "char **".
It allows us also to re-implement extract_arg almost as a simple
wrapper around the new function.
gdb/ChangeLog:
Pedro Alves <palves@redhat.com>
Joel Brobecker <brobecker@adacore.com>
* cli/cli-utils.h (extract_arg_const): Add declaration.
* cli/cli-utils.c (extract_arg_const): New function.
(extract_arg): Reimplement using extract_arg_const.
In addition to the fact that language.h depends on a number of struct
types declared in symtab.h, language.h also depends on an enumerated
type (domain_enum). So language.h should #include "symtab.h".
gdb/ChangeLog:
* language.h: Add "symtab.h" #include.
Rather than having -list-features report support for the GDB/MI
commands providing access to Ada exception catchpoints with one entry,
and the GDB/MI command providing the list of Ada exceptions with
a second entry, this patch merges it all within one single entry.
This is OK, because all these commands were added within a short
amount of time, and within the same release cycle; and it reduces
a bit the size of the output.
gdb/ChangeLog:
* mi/mi-main.c (mi_cmd_list_features): Replace "info-ada-exceptions"
entry with "ada-exceptions".
gdb/doc/ChangeLog:
* gdb.texinfo (GDB/MI Miscellaneous Commands): Delete
the documentation of "info-ada-exceptions" in the output
of the "-list-features" command. Add the documentation
of the "ada-exception" entry instead.
This patch aims at fixing the following problem, where the user:
. debugs its program
. makes a modification and rebuilds it *without exiting the debugger*
. returns to its debugging session and restarts the inferior
In that situation, the debugger notices that the underlying executable
has changed and that re-reading its symbols is needed. Shortly after
displaying a message informing the user of the situation, GDB crashes:
(gdb) run
[...]
`/[...]/dest' has changed; re-reading symbols.
zsh: 13434922 segmentation fault (core dumped)
The crash occurs while trying to allocate some memory on the bfd_bfd
obstack. But, at some point in time, the whole obstack data gets
corrupted, nullified. So the memory allocation fails trying to call
a function at a NULL address. (side note: when debugging GDB in GDB,
top-gdb reports a SIGILL, while the shell makes it look like it was
a SIGSEGV - the discrepancy is not critical to the investigation
and therefore was not explored)
The corruption occurred because the region where the per_bfd data
got free'ed nearly after it got allocated! This is what happens,
in chronological order (see reread_symbols):
1. GDB notices that the executable has changed, decides to
re-read its symbols.
2. Opens a new bfd, unrefs the old one
3. Calls set_objfile_per_bfd (objfile);
4. Re-initializes the objfile's obstack:
obstack_init (&objfile->objfile_obstack);
I think that the normal behavior for set_objfile_per_bfd would
be to search for already-allocated shared per_bfd data, and
allocate new one if not found. The critical difference between
a platform such as x86_64-linuxe where it works, and ppc-aix,
where it doesn't lies in the fact that bfd-data sharing is not
activated on ppc-aix, and as a result, the per-bfd data gets
allocated on the objfile's obstack instead of in the bfd objalloc:
/* If the object requires gdb to do relocations, we simply fall
back to not sharing data across users. These cases are rare
enough that this seems reasonable. */
if (abfd != NULL && !gdb_bfd_requires_relocations (abfd))
{
storage = bfd_zalloc (abfd, sizeof (struct objfile_per_bfd_storage));
set_bfd_data (abfd, objfiles_bfd_data, storage);
}
else
storage = OBSTACK_ZALLOC (&objfile->objfile_obstack,
struct objfile_per_bfd_storage);
Allocating that per_bfd storage is of course nearly useless since
we end up free-ing right after in step (4) above. Eventually,
the memory region ends up being re-used, hence the corruption
leading to the crash.
This fix was simply to move the call to set_objfile_per_bfd after
the objfile's obstack re-initialization.
gdb/ChangeLog:
* symfile.c (reread_symbols): Move call to set_objfile_per_bfd
after re-initialization of OBJFILE's obstack.
Upstream GCC's new pass '-fisolate-erroneous-paths' may introduce
traps at places where GCC has determined undefined behavior, e.g. when
passing a NULL pointer to a function that defines this argument as
__attribute__(__nonnull__(...)). In particular this applies to
uniquify_strings(), because it invokes qsort() with NULL when the
'strings' vector is empty. I hit this problem on s390x when trying to
execute "break main" on a C program.
gdb/
2013-11-12 Andreas Arnez <arnez@linux.vnet.ibm.com>
* objc-lang.c (uniquify_strings): Prevent invoking qsort with
NULL.
* dwarf2read.c (read_index_from_section): Update comment.
(struct dw2_symtab_iterator): New member global_seen.
(dw2_symtab_iter_init): Initialize it.
(dw2_symtab_iter_next): Skip duplicate global symbols.
(dw2_expand_symtabs_matching): Ditto.
This patch adds a new command "info exceptions" whose purpose is to
provide the list of exceptions currently defined in the inferior.
The usage is:
(gdb) info exceptions [REGEXP]
Without argument, the command lists all exceptions. Otherwise,
only those whose name match REGEXP are listed.
For instance:
(gdb) info exceptions
All defined Ada exceptions:
constraint_error: 0x613dc0
program_error: 0x613d40
storage_error: 0x613d00
tasking_error: 0x613cc0
global_exceptions.a_global_exception: 0x613a80
global_exceptions.a_private_exception: 0x613ac0
The name of the command, as well as its output is part of a legacy
I inherited long ago. It's output being parsed by frontends such as
GPS, I cannot easily change it. Same for the command name.
The implementation is mostly self-contained, and is written in a way
that should make it easy to implement the GDB/MI equivalent. The
careful reviewer will notice that the code added in ada-lang.h could
normally be made private inside ada-lang.c. But these will be used
by the GDB/MI implementation. Rather than making those private now,
only to move them later, I've made them public right away.
gdb/ChangeLog:
* ada-lang.h: #include "vec.h".
(struct ada_exc_info): New.
(ada_exc_info): New typedef.
(DEF_VEC_O(ada_exc_info)): New vector.
(ada_exceptions_list): Add declaration.
* ada-lang.c (ada_is_exception_sym)
(ada_is_non_standard_exception_sym, compare_ada_exception_info)
(sort_remove_dups_ada_exceptions_list)
(ada_exc_search_name_matches, ada_add_standard_exceptions)
(ada_add_exceptions_from_frame, ada_add_global_exceptions)
(ada_exceptions_list_1, ada_exceptions_list)
(info_exceptions_command): New function.
(_initialize_ada_language): Add "info exception" command.
gdb/testsuite/ChangeLog:
* gdb.ada/info_exc: New testcase.
When using the GDB/MI commands to insert a catchpoint on a specific
Ada exception, any re-evaluation of that catchpoint (for instance
a re-evaluation performed after a shared library got mapped by the
inferior) fails. For instance, with any Ada program:
(gdb)
-catch-exception -e program_error
^done,bkptno="1",bkpt={[...]}
(gdb)
-exec-run
=thread-group-started,id="i1",pid="28315"
=thread-created,id="1",group-id="i1"
^running
*running,thread-id="all"
(gdb)
=library-loaded,[...]
&"warning: failed to reevaluate internal exception condition for catchpoint 1: No definition of \"exec\" in current context.\n"
&"warning: failed to reevaluate internal exception condition for catchpoint 1: No definition of \"exec\" in current context.\n"
[...]
The same is true if using an Ada exception catchpoint.
The problem comes from the fact that that we deallocate the strings
given as arguments to create_ada_exception_catchpoint, while the latter
just makes shallow copies of those strings, thus creating dandling
pointers.
This patch fixes the issue by passing freshly allocated strings to
create_ada_exception_catchpoint, while at the same time updating
create_ada_exception_catchpoint's documentation to make it clear
that deallocating the strings is no longer the responsibility of
the caller.
gdb/ChangeLog:
* ada-lang.c (create_ada_exception_catchpoint): Enhance
the documentation of fields "except_string" and "condition".
* mi/mi-cmd-catch.c (mi_cmd_catch_assert): Reallocate
CONDITION on the heap before passing it to
create_ada_exception_catchpoint.
(mi_cmd_catch_exception): Likewise for EXCEPTION_NAME and
CONDITION.
gdb/testsuite/ChangeLog:
* gdb.ada/mi_ex_cond: New testcase.
Tested on x86_64-linux. The "-break-list" test FAILs without
this patch.
An earlier patch removed the check for "syscall" since the results
were not used in the C code. However, the result was used, via the
cache variable, elsewhere in configure.
This patch fixes the problem by checking for "syscall" at the point at
which HAVE_TKILL_SYSCALL is defined.
2013-11-11 Tom Tromey <tromey@redhat.com>
* config.in, configure: Rebuild.
* configure.ac (HAVE_TKILL_SYSCALL): Check for "syscall".
* dwarf2read.c (dwarf2_read_debug): Change to unsigned int.
(create_debug_types_hash_table): Only print debugging messages for
each TU if dwarf2-read >= 2.
(process_queue): Ditto.
(_initialize_dwarf2_read): Make "set debug dwarf2-read" a zuinteger.
Update doc string.
doc/
* gdb.texinfo (Debugging Output): Update text for
"set debug dwarf2-read".
My grepping around showed that HAVE_MULTIPLE_PROC_FDS is only ever
mentioned in a comment in configure.ac. Since the macro is long dead,
let's remove the last mention.
2013-11-08 Tom Tromey <tromey@redhat.com>
* configure: Rebuild.
* configure.ac: Remove mentions of HAVE_MULTIPLE_PROC_FDS.
Now that the configury needed for the "common" and "target"
directories is in common.m4, some code in gdb's configure.ac is
redundant.
I ran this script after making an "ID" file using mkid:
sed -n 's/^.*\(HAVE_[A-Z0-9_]*\).*$/\1/p' config.in |
while read x; do
echo ===== $x
gid $x | egrep -v '^(testsuite|gnulib|common|target|gdbserver)/'
done
This finds all the spots using HAVE_ defines, and, more importantly,
makes it clear which defines aren't used in the main parts of gdb.
From this I came up with this patch to remove all the unused bits.
There are a few that are subtly used -- for example the configure
script sometimes checks internal configure cache variables, meaning
some checks cannot be removed.
2013-11-08 Tom Tromey <tromey@redhat.com>
* configure, config.in: Rebuild.
* configure.ac: Remove unused configury.
m32c-tdep.c is the last user of HAVE_STRING_H in gdb proper. It
really ought to be using gdb_string.h instead, as the rest of gdb
does.
2013-11-08 Tom Tromey <tromey@redhat.com>
* m32c-tdep.c: Use gdb_string.h.
The removal of solib-sunos.c also removed the last user of various
macros defined by configure.
This patch removes the corresponding configure code.
2013-11-08 Tom Tromey <tromey@redhat.com>
* configure, config.in: Rebuild.
* configure.ac: Remove all link.h-related checks.
It has bothered me for a while that files in common/ use macros
defined via autoconf checks, but rely on each configure.ac doing the
proper checks independently.
This patch introduces common/common.m4 which consolidates the checks
assumed by code in common.
The rule I propose is that if something is needed or used by common,
it should be checked for by common.m4. However, if the check is also
needed by gdb or gdbserver, then it should be duplicated there.
Built and regtested on x86-64 Fedora 18 (though this is hardly the
most strenuous case) and using the Fedora 18 mingw cross compilers. I
also examined the config.in diffs to ensure that symbols did not go
missing.
2013-11-08 Tom Tromey <tromey@redhat.com>
* acinclude.m4: Include common.m4.
* common/common.m4: New file.
* configure, config.in: Rebuild.
* configure.ac: Use GDB_AC_COMMON.
2013-11-08 Tom Tromey <tromey@redhat.com>
* acinclude.m4: Include common.m4, codeset.m4.
* configure, config.in: Rebuild.
* configure.ac: Use GDB_AC_COMMON.
This patch constifies the target_ops method to_detach.
This is a small cleanup, but also, I think, a bug-prevention fix,
since gdb already acts as if the "args" argument here was const.
In particular, top.c:quit_force calls kill_or_detach via
iterate_over_inferiors. kill_or_detach calls target_detach, passing
the same argument each time. So, if one of these methods was not
const-correct, then kill_or_detach would change its behavior in a
strange way.
I could not build every target I modified in this patch. I've
inspected them all by hand, though. Many targets do not use the
"args" parameter; a couple pass it to atoi; and a few pass it on to
the to_detach method of the target beneath. The only code that
required a real change was in linux-nat.c, and that only needed the
introduction of a temporary variable for const-correctness.
2013-11-08 Tom Tromey <tromey@redhat.com>
* aix-thread.c (aix_thread_detach): Update.
* corelow.c (core_detach): Update.
* darwin-nat.c (darwin_detach): Update.
* dec-thread.c (dec_thread_detach): Update.
* gnu-nat.c (gnu_detach): Update.
* go32-nat.c (go32_detach): Update.
* inf-ptrace.c (inf_ptrace_detach): Update.
* inf-ttrace.c (inf_ttrace_detach): Update.
* linux-fork.c (linux_fork_detach): Update.
* linux-fork.h (linux_fork_detach): Update.
* linux-nat.c (linux_nat_detach): Update. Introduce "tem"
local for const-correctness.
* linux-thread-db.c (thread_db_detach): Update.
* monitor.c (monitor_detach): Update.
* nto-procfs.c (procfs_detach): Update.
* procfs.c (procfs_detach): Update.
* record.c (record_detach): Update.
* record.h (record_detach): Update.
* remote-m32r-sdi.c (m32r_detach): Update.
* remote-mips.c (mips_detach): Update.
* remote-sim.c (gdbsim_detach): Update.
* remote.c (remote_detach_1, remote_detach)
(extended_remote_detach): Update.
* sol-thread.c (sol_thread_detach): Update.
* target.c (target_detach): Make "args" const.
(init_dummy_target): Update.
* target.h (struct target_ops) <to_detach>: Make argument const.
(target_detach): Likewise.
* windows-nat.c (windows_detach): Update.
* python/py-breakpoint.c (bppy_get_temporary): New function.
(bppy_init): New keyword: temporary. Parse it and set breakpoint
to temporary if True.
2013-11-07 Phil Muldoon <pmuldoon@redhat.com>
* gdb.python/py-breakpoint.exp: Add temporary breakpoint tests.
2013-11-07 Phil Muldoon <pmuldoon@redhat.com>
* gdb.texinfo (Breakpoints In Python): Document temporary
option in breakpoint constructor, and add documentation to the
temporary attribute.
This patch does some cleanups, removing some language-related stuff.
Note that mi_cmd_var_info_expression uses varobj_language_string,
which is redundant, because we can get language name from
lang->la_natural_name.
varobj_language_string doesn't have "Ada", which looks like a bug to
me. With this patch applied, this problem doesn't exist, because the
language name is got from the same place (field la_natural_name).
gdb:
2013-11-07 Yao Qi <yao@codesourcery.com>
* mi/mi-cmd-var.c: Include "language.h".
(mi_cmd_var_info_expression): Get language name from
language_defn.
* varobj.c (varobj_language_string): Remove.
(variable_language): Remove declaration.
(languages): Remove.
(varobj_get_language): Change the type of return value.
(variable_language): Remove.
* varobj.h (enum varobj_languages): Remove.
(varobj_language_string): Remove declaration.
(varobj_get_language): Update declaration.
gdb/doc:
2013-11-07 Yao Qi <yao@codesourcery.com>
* gdb.texinfo (GDB/MI Variable Objects): Update doc about the
output of "-var-info-expression".
Hi,
When I add another name of language, I find field 'la_name' can be
'const char *'. This patch is to constify it.
gdb:
2013-11-07 Yao Qi <yao@codesourcery.com>
* language.c (language_str): Return const char *.
(add_language): Add const to 'language_names'
* language.h (struct language_defn) <la_name>: Add const.
(language_str: Update declaration.
When checking for the presence of the TDB regset, the current code
interprets ENODATA from PTRACE_GETREGSET as an indication that the TDB
regset *could* occur on this system, but the inferior stopped outside
a transaction. However, the Linux kernel actually reports ENODATA
even on systems without the transactional execution facility. Thus
the logic is now changed to check the TE field in the HWCAP as well.
This version also checks the existence of the TDB regset -- just to be
on the safe side when running on TE-enabled hardware with a kernel
that does not offer the TDB regset for some reason.
gdb/
* s390-linux-nat.c (s390_read_description): Consider the TE field
in the HWCAP for determining 'have_regset_tdb'.
gdbserver/
* linux-s390-low.c (HWCAP_S390_TE): New define.
(s390_arch_setup): Consider the TE field in the HWCAP for
determining 'have_regset_tdb'.
When reading objects with corrupt debug information it is possible that
the sibling chain can form a loop, which leads to an infinite loop and
memory exhaustion.
Avoid this situation by disregarding and DW_AT_sibling values that point
to a lower address than the current entry.
gdb/ChangeLog:
2013-11-06 Will Newton <will.newton@linaro.org>
PR gdb/12866
* dwarf2read.c (skip_one_die): Sanity check DW_AT_sibling
values. (read_partial_die): Likewise.
We add a new dir gdb.perf in testsuite for all performance tests.
However, current 'make check' logic will either run dejagnu in
directory testsuite or iterate all gdb.* directories which has *.exp
files. Both of them will run tests in gdb.perf. We want to achieve:
1) typical 'make check' should not run performance tests. In each perf
test case, GDB_PERFTEST_MODE is checked. If it doesn't exist, return.
2) run perf tests easily. We add a new makefile target 'check-perf'.
gdb:
2013-11-06 Yao Qi <yao@codesourcery.com>
* Makefile.in (check-perf): New target.
gdb/testsuite:
2013-11-06 Yao Qi <yao@codesourcery.com>
* Makefile.in (check-perf): New target.
* configure.ac (AC_OUTPUT): Output Makefile in gdb.perf.
* configure: Re-generated.
* gdb.perf/Makefile.in: New.
I noticed a large (100MB) restore took hours to complete. The problem
is memory_xfer_partial repeatedly mallocs and memcpys the entire
100MB buffer for breakpoint shadow handling only to find a small
portion of it is actually written.
The testcase that originally took hours now takes 50 seconds.
gdb/
2013-07-29 Anton Blanchard <anton@samba.org>
* target.c (memory_xfer_partial): Cap write to 4KB.
As discussed on the GDB ML[1], libc probes for longjmp were not being
loaded if a custom <arch>_get_longjmp_target function was not
implemented.
This is trivially fixed by moving the 'if (!gdbarch_get_longjmp_target_p
(gdbarch))' down, just bellow libc probe code and above the per-objfile
cache lookup.
While the condition could also be removed altogether with no
side-effects, it is in fact an optimization to avoid searching for
symbols if the arch doesn't provide support for get_longjmp_target().
This has been tested on PPC and PPC64.
[1] https://sourceware.org/ml/gdb/2013-10/msg00191.html
gdb/
2013-11-01 Tiago Stürmer Daitx <tdaitx@linux.vnet.ibm.com>
* breakpoint.c (create_longjmp_master_breakpoint): Allow libc
probe scan even when the arch provides no get_longjmp_target.
IMO, it doesn't make sense to map random syscall, fork, etc. events to
GDB_SIGNAL_TRAP, and possible have the debuggee see that trap. This
just seems conceptually wrong to me - these aren't real signals a
debuggee would ever see. In fact, when stopped for those events, on
Linux, the debuggee isn't in a signal-stop -- there's no way to
resume-and-deliver-signal at that point, for example. E.g., when
stopped at a fork event:
(gdb) catch fork
Catchpoint 2 (fork)
(gdb) c
Continuing.
Catchpoint 2 (forked process 4570), 0x000000323d4ba7c4 in __libc_fork () at ../nptl/sysdeps/unix/sysv/linux/fork.c:131
131 pid = ARCH_FORK ();
(gdb) set debug infrun 1
(gdb) signal SIGTRAP
Continuing with signal SIGTRAP.
infrun: clear_proceed_status_thread (process 4566)
infrun: proceed (addr=0xffffffffffffffff, signal=5, step=0)
infrun: resume (step=0, signal=5), trap_expected=0, current thread [process 4566] at 0x323d4ba7c4
infrun: wait_for_inferior ()
infrun: target_wait (-1, status) =
infrun: 4566 [process 4566],
infrun: status->kind = exited, status = 0
infrun: infwait_normal_state
infrun: TARGET_WAITKIND_EXITED
[Inferior 1 (process 4566) exited normally]
infrun: stop_stepping
(gdb)
Note the signal went nowhere. It was swallowed.
Resuming with a SIGTRAP from a syscall event does queue the signal,
but doesn't deliver it immediately, like "signal SIGTRAP" from a real
signal would. It's still an artificial SIGTRAP:
(gdb) catch syscall
Catchpoint 2 (any syscall)
(gdb) c
Continuing.
Catchpoint 2 (call to syscall clone), 0x000000323d4ba7c4 in __libc_fork () at ../nptl/sysdeps/unix/sysv/linux/fork.c:131
131 pid = ARCH_FORK ();
(gdb) set debug infrun 1
(gdb) signal SIGTRAP
Continuing with signal SIGTRAP.
infrun: clear_proceed_status_thread (process 4622)
infrun: proceed (addr=0xffffffffffffffff, signal=5, step=0)
infrun: resume (step=0, signal=5), trap_expected=0, current thread [process 4622] at 0x323d4ba7c4
infrun: wait_for_inferior ()
infrun: target_wait (-1, status) =
infrun: 4622 [process 4622],
infrun: status->kind = exited syscall
infrun: infwait_normal_state
infrun: TARGET_WAITKIND_SYSCALL_RETURN
infrun: syscall number = '56'
infrun: BPSTAT_WHAT_STOP_NOISY
infrun: stop_stepping
Catchpoint 2 (returned from syscall clone), 0x000000323d4ba7c4 in __libc_fork () at ../nptl/sysdeps/unix/sysv/linux/fork.c:131
131 pid = ARCH_FORK ();
(gdb) c
Continuing.
infrun: clear_proceed_status_thread (process 4622)
infrun: proceed (addr=0xffffffffffffffff, signal=144, step=0)
infrun: resume (step=0, signal=0), trap_expected=0, current thread [process 4622] at 0x323d4ba7c4
infrun: wait_for_inferior ()
infrun: target_wait (-1, status) =
infrun: 4622 [process 4622],
infrun: status->kind = stopped, signal = SIGTRAP
infrun: infwait_normal_state
infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x323d4ba7c4
infrun: random signal 5
Program received signal SIGTRAP, Trace/breakpoint trap.
infrun: stop_stepping
0x000000323d4ba7c4 in __libc_fork () at ../nptl/sysdeps/unix/sysv/linux/fork.c:131
131 pid = ARCH_FORK ();
(gdb)
In all the above, I used 'signal SIGTRAP' to emulate 'handle SIGTRAP
pass'. As described in "keep_going", 'handle SIGTRAP pass' does have
its place:
/* Do not deliver GDB_SIGNAL_TRAP (except when the user
explicitly specifies that such a signal should be delivered
to the target program). Typically, that would occur when a
user is debugging a target monitor on a simulator: the target
monitor sets a breakpoint; the simulator encounters this
breakpoint and halts the simulation handing control to GDB;
GDB, noting that the stop address doesn't map to any known
breakpoint, returns control back to the simulator; the
simulator then delivers the hardware equivalent of a
GDB_SIGNAL_TRAP to the program being debugged. */
... and I've made use of that myself when implementing/debugging
stubs/monitors. But in these cases, treating these events as SIGTRAP
possibly injects signals in the debuggee they'd never see otherwise,
because you need to use ptrace to enable these special events, which
aren't real signals.
There's more. Take this bit of handle_inferior_event, where we
determine whether a real signal (TARGET_WAITKIND_STOPPED) was random
or not:
if (ecs->event_thread->suspend.stop_signal == GDB_SIGNAL_TRAP)
ecs->random_signal
= !((bpstat_explains_signal (ecs->event_thread->control.stop_bpstat,
GDB_SIGNAL_TRAP)
!= BPSTAT_SIGNAL_NO)
|| stopped_by_watchpoint
|| ecs->event_thread->control.trap_expected
|| (ecs->event_thread->control.step_range_end
&& (ecs->event_thread->control.step_resume_breakpoint
== NULL)));
else
{
enum bpstat_signal_value sval;
sval = bpstat_explains_signal (ecs->event_thread->control.stop_bpstat,
ecs->event_thread->suspend.stop_signal);
ecs->random_signal = (sval == BPSTAT_SIGNAL_NO);
if (sval == BPSTAT_SIGNAL_HIDE)
ecs->event_thread->suspend.stop_signal = GDB_SIGNAL_TRAP;
}
Note that the
if (sval == BPSTAT_SIGNAL_HIDE)
ecs->event_thread->suspend.stop_signal = GDB_SIGNAL_TRAP;
bit is only reacheable for signals != GDB_SIGNAL_TRAP. AFAICS, sval
can only be BPSTAT_SIGNAL_HIDE if nothing in the bpstat returns
BPSTAT_SIGNAL_PASS. So that excludes a "catch signal" for the signal
in question in the bpstat. All other catchpoints that aren't based on
breakpoints behind the scenes call process_event_stop_test directly
(don't pass through here) (well, almost all: TARGET_WAITKIND_LOADED
does have a fall through, but only for STOP_QUIETLY or
STOP_QUIETLY_NO_SIGSTOP, which still return before this code is
reached). Catchpoints that are implemented as breakpoints behind the
scenes can only appear in the bpstat if the signal was GDB_SIGNAL_TRAP
(bkpt_breakpoint_hit returns false otherwise). So that leaves a
target reporting a hardware watchpoint hit with a signal other than
GDB_SIGNAL_TRAP. And even then it looks quite wrong to me to
magically convert the signal into a GDB_SIGNAL_TRAP here too -- if the
user has set SIGTRAP to "handle pass", the program will see a trap
that gdb invented, not one the program would ever see without gdb in
the picture.
Tested on x86_64 Fedora 17.
gdb/
2013-10-31 Pedro Alves <palves@redhat.com>
* infrun.c (handle_syscall_event): Don't set or clear stop_signal.
(handle_inferior_event) <TARGET_WAITKIND_FORKED,
TARGET_WAITKIND_VFORKED>: Don't set stop_signal to
GDB_SIGNAL_TRAP, or clear it. Pass GDB_SIGNAL_0 to
bpstat_explains signal, instead of GDB_SIGNAL_TRAP.
<bpstat handling>: If the bpstat chain wants the signal to be
hidden, then set stop_signal to GDB_SIGNAL_0 instead of
GDB_SIGNAL_TRAP.
As suggested before, rename the S/390-related source files (tdep and nat)
such that "-linux-" occurs in the file name, like with other GNU/Linux
targets. Since no other operating system is currently supported by GDB
on this architecture, this isn't strictly necessary. But the old names
sometimes caused GDB contributors to miss these files when performing a
change that affects all GNU/Linux targets. The latest such incident was
observed here:
https://sourceware.org/ml/gdb-patches/2013-09/msg00619.html
gdb/
2013-10-30 Andreas Arnez <arnez@linux.vnet.ibm.com>
* s390-tdep.h: Rename to...
* s390-linux-tdep.h: ...here.
* s390-tdep.c: Rename to...
* s390-linux-tdep.c: ...here. Adjust #include.
* s390-nat.c: Rename to...
* s390-linux-nat.c: ...here. Adjust #include.
* config/s390/s390.mh: Rename to...
* config/s390/linux.mh: ...here. Reflect rename s390-nat.o ->
s390-linux-nat.o.
* configure.host: Reflect host rename "s390" -> "linux".
* configure.tgt: Reflect rename s390-tdep.o -> s390-linux-tdep.o.
* Makefile.in (ALL_TARGET_OBS): Likewise.
(HFILES_NO_SRCDIR): Reflect rename s390-tdep.h ->
s390-linux-tdep.h.
(ALLDEPFILES): Reflect rename of .c files.
Tom Tromey
Joel Brobecker
Doug Evans
Keith Seitz
Andreas Arnez
Doug Evans
Phil Muldoon
Edjunior Barbosa Machado
Jose E. Marchesi
Yao Qi
Will Newton
Muhammad Bilal
Anton Blanchard
Tiago Stürmer Daitx
Pedro Alves
Andrew Burgess
Ulrich Weigand