Alan Modra [Tue, 29 Oct 2013 06:23:25 +0000 (16:53 +1030)]
Report overflow on PowerPC64 @h and @ha relocations.
This changes the behaviour of @h and @ha on PowerPC64 to report errors
on 32-bit overflow. The motivation for this change is that on
PowerPC64, most uses of @h and @ha modifiers and their corresponding
relocations are to build up 32-bit offsets. We'd like to know when
such offsets overflow. Only rarely do people use @h or @ha with the
high 32-bit modifiers to build a 64-bit constant. Those uses will now
need to use two new modifiers, @high and @higha, if the constant isn't
known at assembly time. For now, we won't report overflow at assembly
time..
This also fixes an error when applying some of the HIGHER and HIGHEST
relocations.
include/elf/
* ppc64.h (R_PPC64_ADDR16_HIGH, R_PPC64_ADDR16_HIGHA,
R_PPC64_TPREL16_HIGH, R_PPC64_TPREL16_HIGHA,
R_PPC64_DTPREL16_HIGH, R_PPC64_DTPREL16_HIGHA): New.
(IS_PPC64_TLS_RELOC): Match new tls relocs.
bfd/
* reloc.c (BFD_RELOC_PPC64_ADDR16_HIGH, BFD_RELOC_PPC64_ADDR16_HIGHA,
BFD_RELOC_PPC64_TPREL16_HIGH, BFD_RELOC_PPC64_TPREL16_HIGHA,
BFD_RELOC_PPC64_DTPREL16_HIGH, BFD_RELOC_PPC64_DTPREL16_HIGHA): New.
* elf64-ppc.c (ppc64_elf_howto_raw): Add entries for new relocs.
Make all _HA and _HI relocs report signed overflow.
(ppc64_elf_reloc_type_lookup): Handle new relocs.
(must_be_dyn_reloc, ppc64_elf_check_relocs): Likewise.
(dec_dynrel_count, ppc64_elf_relocate_section): Likewise.
(ppc64_elf_relocate_section): Don't apply 0x8000 adjust to
R_PPC64_TPREL16_HIGHER, R_PPC64_TPREL16_HIGHEST,
R_PPC64_DTPREL16_HIGHER, and R_PPC64_DTPREL16_HIGHEST.
* libbfd.h: Regenerate.
* bfd-in2.h: Regenerate.
gas/
* config/tc-ppc.c (SEX16): Don't mask.
(REPORT_OVERFLOW_HI): Define as zero.
(ppc_elf_suffix): Support @high, @higha, @dtprel@high, @dtprel@higha,
@tprel@high, and @tprel@higha modifiers.
(md_assemble): Ignore X_unsigned when applying 16-bit insn fields.
Add (disabled) code to check @h and @ha reloc overflow for powerpc64.
Handle new relocs.
(md_apply_fix): Similarly.
elfcpp/
* powerpc.h (R_PPC64_ADDR16_HIGH, R_PPC64_ADDR16_HIGHA,
R_PPC64_TPREL16_HIGH, R_PPC64_TPREL16_HIGHA,
R_PPC64_DTPREL16_HIGH, R_PPC64_DTPREL16_HIGHA): Define.
gold/
* powerpc.cc (Target_powerpc::Scan::check_non_pic): Handle new relocs.
(Target_powerpc::Scan::global, local): Likewise.
(Target_powerpc::Relocate::relocate): Likewise. Check for overflow
on all ppc64 @h and @ha relocs.
Jan Beulich [Tue, 29 Oct 2013 16:52:24 +0000 (17:52 +0100)]
ld/ELF: refine fake STT_FILE symbol emission
There's no real need to emit these always: They're needed only if we
indeed want to emit a localized symbol. Hence defer emission until we
at least did the basic early checks that would lead to no such symbol
getting emitted. This in particular avoids emitting such a symbol in
the majority of (if not all) "ld -r" cases.
I hope my set of cross build tests caught all the test cases needing
adjustment - please forgive if I missed a few.
bfd/
2013-10-29 Jan Beulich <jbeulich@suse.com>
* elflink.c (struct elf_outext_info): Add field file_sym_done.
(bfd_elf_final_link): Initialize new field. Move fake STT_FILE
symbol emission from here ...
(elf_link_output_extsym): ... to here.
gas/testsuite/
2013-10-29 Jan Beulich <jbeulich@suse.com>
* gas/microblaze/relax_size.elf: Drop expectation of no longer
present STT_FILE symbol.
* gas/microblaze/relax_size2.elf: Likewise.
ld/testsuite/
2013-10-29 Jan Beulich <jbeulich@suse.com>
Tom Tromey [Wed, 16 Oct 2013 18:25:27 +0000 (12:25 -0600)]
undef reg in gdb_curses.h
I tried to build gdb on the AIX machine in the GCC compile farm
(gcc111), but it failed in a couple of spots because gdb uses "reg" as
a variable name and the AIX <curses.h> defines "reg" to "register".
I saw that we already had a workaround for this lurking in utils.c, so
I just moved that to gdb_curses.h.
This fixed the problem on AIX and still builds on x86-64 Fedora 18.
* infcmd.c (default_print_one_register_info): Use val_print to
print all values even optimized out or unavailable ones. Don't
try to print a raw form of optimized out or unavailable values.
Nicolas Blanc [Wed, 17 Jul 2013 09:44:26 +0000 (11:44 +0200)]
Test adding and removing a symbol file at runtime.
This test exercises the commands 'add-symbol-file'
and 'remove-symbol-file'.
2013-10-29 Nicolas Blanc <nicolas.blanc@intel.com>
gdb/testsuite
* gdb.base/sym-file-lib.c: New file.
* gdb.base/sym-file-loader.c: New file.
* gdb.base/sym-file-loader.h: New file.
* gdb.base/sym-file-main.c: New file.
* gdb.base/sym-file.exp: New file.
Signed-off-by: Nicolas Blanc <nicolas.blanc@intel.com>
Nicolas Blanc [Wed, 17 Jul 2013 09:33:45 +0000 (11:33 +0200)]
Create target sections for user-added symbol files.
Add the sections of the symbol files that are provided via
'add-symbol-file' to the set of current target sections.
User-added sections are removed upon notification of free_objfile
when their corresponding object file is deleted.
2013-10-29 Nicolas Blanc <nicolas.blanc@intel.com>
* exec.h (add_target_sections_of_objfile): New declaration.
* exec.c (add_target_sections_of_objfile): New function.
* symfile.c (add_symbol_file_command): Update current target sections.
(symfile_free_objfile): New function.
(_initialize_symfile): Register observer for free_objfile events.
Signed-off-by: Nicolas Blanc <nicolas.blanc@intel.com>
Nicolas Blanc [Fri, 5 Apr 2013 07:15:56 +0000 (09:15 +0200)]
Documentation for the remove-symbol-file command.
2013-10-29 Nicolas Blanc <nicolas.blanc@intel.com>
* NEWS: Add description of the remove-symbol-file command.
gdb/doc
* gdb.texinfo (Commands to Specify Files): Add description
of the remove-symbol-file command.
Signed-off-by: Nicolas Blanc <nicolas.blanc@intel.com>
Yao Qi [Thu, 24 Oct 2013 12:42:02 +0000 (20:42 +0800)]
Simplify REGISTRY cleanup usages
In registry.c:registry_clear_data, the registered data is iterated and
invoke each 'free' function with the data passed:
for (registration = data_registry->registrations, i = 0;
i < fields->num_data;
registration = registration->next, i++)
if (fields->data[i] != NULL && registration->data->free != NULL)
adaptor (registration->data->free, container, fields->data[i]);
we can see that data is passed to function 'free' and data is not NULL.
In each usage, we don't have to get the data again through key and
do NULL pointer checking. This patch is to simplify them.
gdb:
2013-10-29 Yao Qi <yao@codesourcery.com>
* auto-load.c (auto_load_pspace_data_cleanup): Get data from
parameter 'arg' instead of from program_space_data.
* objfiles.c (objfiles_pspace_data_cleanup): Likewise.
* solib-darwin.c (darwin_pspace_data_cleanup): Likewise.
* solib-dsbt.c (dsbt_pspace_data_cleanup): Likewise.
* solib-svr4.c (svr4_pspace_data_cleanup): Likewise.
* inflow.c (inflow_inferior_data_cleanup): Get data from
parameter 'arg' instead of inferior_data.
* registry.h: Add comments.
Pedro Alves [Mon, 28 Oct 2013 16:39:06 +0000 (16:39 +0000)]
infrun.c:handle_inferior_event: Make process_event_stop_test label a function.
Now that all ecs->random_signal handing is always done before the
'process_event_stop_test' label, we can easily make that a real
function and actually give it a describing comment that somewhat makes
sense.
Reindenting the new function will be handled in a follow up patch.
2013-10-28 Pedro Alves <palves@redhat.com>
* infrun.c (process_event_stop_test): New function, factored out
from handle_inferior_event.
(handle_inferior_event): 'process_event_stop_test' is now a
function instead of a goto label -- adjust.
We only ever call "goto process_event_stop_test;" right after checking
that ecs->random_signal is clear. The code at the
process_event_stop_test label looks like:
/* For the program's own signals, act according to
the signal handling tables. */
if (ecs->random_signal)
{
... random signal handling ...
return;
}
else
{
... the stop tests that actually matter for the goto callers.
}
So this moves the label into the else branch. It'll make converting
process_event_stop_test into a function a bit clearer.
gdb/
2013-10-28 Pedro Alves <palves@redhat.com>
* infrun.c (handle_inferior_event): Move process_event_stop_test
goto label to the else branch of the ecs->random_signal check,
along with FRAME and GDBARCH re-fetching.
Pedro Alves [Mon, 28 Oct 2013 16:39:05 +0000 (16:39 +0000)]
infrun.c:handle_inferior_event: Put all ecs->random_signal tests together.
I recently added a new ecs->random_signal test after the "switch back to
stepped thread" code, and before the stepping tests. Looking at
making process_event_stop_test a proper function, I realized it'd be
better to keep ecs->random_signal related code together. To do that,
I needed to factor out the "switch back to stepped thread" code to a new
function, and call it in both the "random signal" and "not random
signal" paths.
gdb/
2013-10-28 Pedro Alves <palves@redhat.com>
* infrun.c (switch_back_to_stepped_thread): New function, factored
out from handle_inferior_event.
(handle_inferior_event): Adjust to call
switch_back_to_stepped_thread. Call it also at the tail of the
random signal handling, and return, instead of also handling
random signals just before the stepping tests.
Pedro Alves [Mon, 28 Oct 2013 16:39:05 +0000 (16:39 +0000)]
infrun.c:handle_inferior_event: Remove some more dead code.
'ecs' is always memset before being passed to handle_inferior_event.
The stop func is only filled in later in the flow. And since "Remove
dead sets/clears of ecs->random signal", nothing ever sets
ecs->random_signal before this part is reached either.
(Also tested with some added assertions in place.)
Yao Qi [Sat, 26 Oct 2013 07:27:13 +0000 (15:27 +0800)]
Rename field 'lang' to 'lang_ops'.
On 10/25/2013 11:34 AM, Joel Brobecker wrote:
> Also, as a followup, I think it would be beneficial if we renamed
> field "lang" in the varobj_root into "lang_ops". I think it's more
> descriptive, especially since "lang" is used elsewhere with different
> meanings (and types).
Here is the patch to rename 'lang' to 'lang_ops'. Committed as obvious.
Yao Qi [Thu, 17 Oct 2013 13:15:21 +0000 (21:15 +0800)]
New field la_varobj_ops in struct language_defn
This is a follow-up series to move language stuff out of varobj.c.
This patch adds a new field la_varobj_ops in struct language_defn so
that each language has varobj-related options. Not every language
supports varobj, and the operations are identical to operations of c
languages.
'struct language_defn' is the ideal place to save all language-related
operations. After this patch, some cleanups can be done in patch 2/2,
which removes language-related stuff completely from varobj.c.
Regression tested on x86_64-linux.
gdb:
2013-10-25 Yao Qi <yao@codesourcery.com>
* language.h (struct lang_varobj_ops): Declare.
(struct language_defn) <la_varobj_ops>: New field.
* ada-lang.c: Include "varobj.h"
(defn ada_language_defn): Initialize field 'la_varobj_ops' by
ada_varobj_ops.
* c-lang.c: Include "varobj.h"
(c_language_defn): Initialize field 'la_varobj_ops' by
c_varobj_ops.
(cplus_language_defn): Initialize field 'la_varobj_ops' by
cplus_varobj_ops.
(asm_language_defn): Initialize field 'la_varobj_ops' by
default_varobj_ops.
(minimal_language_defn): Likewise.
* d-lang.c (d_language_defn): Likewise.
* f-lang.c (f_language_defn): Likewise.
* go-lang.c (go_language_defn): Likewise.
* m2-lang.c (m2_language_defn): Likewise.
* objc-lang.c (objc_language_defn): Likewise.
* opencl-lang.c (opencl_language_defn): Likewise.
* p-lang.c (pascal_language_defn): Likewise.
* language.c (unknown_language_defn): Likewise.
(auto_language_defn): Likewise.
(local_language_defn): Likewise.
* jv-lang.c (java_language_defn): Initialize field
'la_varobj_ops' by java_varobj_ops.
* varobj.c (varobj_create): Update.
* varobj.h (default_varobj_ops): Define macro.
Anton Kolesov [Wed, 9 Oct 2013 08:58:14 +0000 (12:58 +0400)]
testsuite: Fix gdb.base/bang.exp for remote stubs without exit
Some remote stubs do not have a proper exit() function implementation.
gdb.base/bang.exp was failing on those targets due to timeout. With
this patch bang.exp uses already defined library procedures to handle
this situation gracefully without breaking native targets.
Tested with x86_64 (unix, native-gdbserver) and with arc-*-elf32.
gdb/testsuite/ChangeLog:
2013-10-25 Anton Kolesov <Anton.Kolesov@synopsys.com> (tiny change)
* gdb.base/bang.exp: Use gdb_continue_to_end to properly support
remote stubs where exit() behaviour is unreliable.
Pedro Alves [Fri, 25 Oct 2013 10:37:13 +0000 (11:37 +0100)]
Print nonexisting/optimized out static fields gracefully.
With:
struct static_struct { static int aaa; };
struct static_struct sss;
int main () { return 0; }
We get:
(gdb) p sss
$1 = {static aaa = <optimized out>}
(gdb) p sss.aaa
field aaa is nonexistent or has been optimized out
Note that the "field aaa ..." message is an error being thrown.
GDB is graceful everywhere else when printing optimized out values.
IOW it usually prints an <optimized out> value and puts that in the
value history. I see no reason for here to be different, more so that
when the print the whole "containing" object (well, it's a static
field, so it's not really a container), we already print <optimized
out>.
After the patch:
(gdb) p sss
$1 = {static aaa = <optimized out>}
(gdb) p sss.aaa
$2 = <optimized out>
The value_entirely_optimized_out checks are there to preserve
behavior. Without those, if the static field is a struct/union, GDB
would go and print its fields one by one (and print <optimized out>
for each).
Tested on x86_64 Fedora 17.
gdb/
2013-10-25 Pedro Alves <palves@redhat.com>
* cp-valprint.c (cp_print_value_fields): No longer handle a NULL
static field value.
(cp_print_static_field): If the value is entirely optimized out,
print <optimized out> here.
* jv-valprint.c (java_print_value_fields): No longer handle a NULL
static field value.
* p-valprint.c (pascal_object_print_static_field): If the value is
entirely optimized out, print <optimized out> here.
* valops.c (do_search_struct_field)
(value_struct_elt_for_reference): No longer handle a NULL static
field value.
* value.c (value_static_field): Return an optimized out value
instead of NULL.
gdb/testsuite/
2013-10-25 Pedro Alves <palves@redhat.com>
* gdb.cp/m-static.exp: Adjust expected output of printing a
nonexistent or optimized out static field. Also test printing the
the "container" object.
Nick Clifton [Fri, 25 Oct 2013 08:33:59 +0000 (09:33 +0100)]
PR binutils/15994
* dwarf2.c (lookup_address_in_line_info_table): Change return type
to be the range of addresses covered by the table.
(comp_unit_find_nearest_line): Likewise.
(find_line): Search all CUs. Select the one that matches and
covers the smallest address range.
Nick Clifton [Fri, 25 Oct 2013 08:32:43 +0000 (09:32 +0100)]
PR binutils/15994
* dwarf2.c (lookup_address_in_line_info_table): Change return type
to be the range of addresses covered by the table.
(comp_unit_find_nearest_line): Likewise.
(find_line): Search all CUs. Select the one that matches and
covers the smallest address range.
Yao Qi [Thu, 24 Oct 2013 01:28:33 +0000 (09:28 +0800)]
Send qXfer:traceframe-info:read when traceframe is selected.
When I do 'si', I find many 'qXfer:traceframe-info:read' packets are sent,
which is not necessary. It slows down the single step.
(gdb) si
Sending packet: $qTStatus#49...Packet received: T0;tnotrun:0;tframes:0;tcreated:0;tfree:500000;tsize:500000;circular:0;disconn:0;starttime:0;stoptime:0;username:;notes::
Sending packet: $Z0,80483c7,1#b4...Packet received: OK
Sending packet: $Z0,4ce5b6b0,1#6e...Packet received: OK
Sending packet: $QPassSignals:e;10;14;17;1a;1b;1c;21;24;25;2c;4c;#5f...Packet received: OK
Sending packet: $vCont;s:p1b15.1b15;c#20...Packet received: T0505:44efffbf;04:44efffbf;08:d1830408;thread:p1b15.1b15;core:3;
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $mbfffef40,40#c0...Packet received: d183040878efffbf2e840408030000000000a040030000000500000070efffbf07000000010000004984040807000000030000000500000000000000b396e84c
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $z0,80483c7,1#d4...Packet received: OK
Sending packet: $z0,4ce5b6b0,1#8e...Packet received: OK
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
Sending packet: $qXfer:traceframe-info:read::0,fff#0b...Packet received: E01
This problem was introduced by this patch
(https://sourceware.org/ml/gdb-patches/2013-04/msg00000.html), in
which get_traceframe_number is not checked before calling
traceframe_available_memory. This patch moves the check to
remote_traceframe_info, say, if GDB doesn't have traceframe selected, GDB
doesn't need to send qXfer:traceframe-info:read packets.
With this patch applied, there is no qXfer:traceframe-info:read sent
out and single step is speed up a little bit.
Yao Qi [Thu, 24 Oct 2013 00:46:07 +0000 (08:46 +0800)]
Remove global traceframe_fun and traceframe_sal
I happen to see traceframe_fun and traceframe_sal are static variables,
which are not necessary to me. They are only used in set_traceframe_context,
and they are not stateful. This patch is to remove them.
gdb:
2013-10-24 Yao Qi <yao@codesourcery.com>
* tracepoint.c (traceframe_fun): Remove.
(traceframe_sal): Remove.
(set_traceframe_context): Add local variables.
* lib/gdb.exp (gdb_finish): Send a kill request to `gdbserver'
if in the persistent mode.
* gdb.trace/disconnected-tracing.exp: Reconnect before completion.
Pedro Alves [Thu, 24 Oct 2013 15:10:05 +0000 (16:10 +0100)]
Make STARTUP_WITH_SHELL a runtime toggle -- add new "set/show startup-with-shell" option.
Occasionaly we hear about people having problems with GDB not being
able to start programs (with "run"/"start"). GDB spawns a shell to
start the program, and most often, it'll be the case that the problem
is actually with the user's shell setup.
GDB has code to disable the use of the shell to start programs.
That's the STARTUP_WITH_SHELL macro that native targets could set to 0
in their nm.h file (though no target actually uses it nowadays).
This patch makes that setting a run-time knob instead. This will be
useful to quickly diagnose such shell issues, and might also come in
handy at other times (such as when debugging the shell itself, if you
don't have a different shell handy).
gdb/
2013-10-24 Pedro Alves <palves@redhat.com>
* NEWS (New options): Mention set/show startup-with-shell.
* config/alpha/nm-osf3.h (START_INFERIOR_TRAPS_EXPECTED): Set to 2
instead of 3.
* fork-child.c (fork_inferior, startup_inferior): Handle 'set
startup-with-shell'.
(show_startup_with_shell): New function.
(_initialize_fork_child): Register the set/show startup-with-shell
commands.
* inf-ptrace.c (inf_ptrace_create_inferior): Remove comment.
* inf-ttrace.c (inf_ttrace_him): Remove comment.
* procfs.c (procfs_init_inferior): Remove comment.
* infcmd.c (startup_with_shell): New global.
* inferior.h (startup_with_shell): Declare global.
(STARTUP_WITH_SHELL): Delete.
(START_INFERIOR_TRAPS_EXPECTED): Set to 1 by default instead of 2.
gdb/doc/
2013-10-24 Pedro Alves <palves@redhat.com>
Pedro Alves [Wed, 23 Oct 2013 15:14:53 +0000 (16:14 +0100)]
infrun debug output: print enum gdb_signal symbol names instead of POSIX signal names.
The other day while debugging something related to random signals, I
got confused with "set debug infrun 1" output, for it said:
infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x323d4e8b94
infrun: random signal 20
On GNU/Linux, 20 is SIGTSTP. For some reason, it took me a few
minutes to realize that 20 is actually a GDB signal number, not a
target signal number (duh!). In any case, I propose making GDB's
output clearer here:
One way would be to use gdb_signal_to_name, like already used
elsewhere:
infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x323d4e8b94
infrun: random signal SIGCHLD (20)
but I think that might confuse someone too ("20? Why does GDB believe
SIGCHLD is 20?"). So I thought of printing the enum string instead:
infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x323d4e8b94
infrun: random signal GDB_SIGNAL_CHLD (20)
Looking at a more complete infrun debug log, we had actually printed
the (POSIX) signal name name a bit before:
infrun: target_wait (-1, status) =
infrun: 9300 [Thread 0x7ffff7fcb740 (LWP 9300)],
infrun: status->kind = stopped, signal = SIGCHLD
...
infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x323d4e8b94
infrun: random signal 20
So I'm now thinking that it'd be even better to make infrun output
consistently use the enum symbol string, like so:
GDB's signal numbers are public and hardcoded (see
include/gdb/signals.h), so there's really no need to clutter the
output with numeric values in some places while others not. Replacing
the magic "144" with GDB_SIGNAL_DEFAULT in "proceed"'s debug output
(see above) I think is quite nice.
I posit that all this makes it clearer to newcomers that GDB has its
own signal numbering (and that there must be some mapping going on).
Tested on x86_64 Fedora 17.
gdb/
2013-10-23 Pedro Alves <palves@redhat.com>
* common/gdb_signals.h (gdb_signal_to_symbol_string): Declare.
* common/signals.c: Include "gdb_assert.h".
(signals): New field 'symbol'.
(SET): Use the 'symbol' parameter.
(gdb_signal_to_symbol_string): New function.
* infrun.c (handle_inferior_event) <random signal>: In debug
output, print the random signal enum as string in addition to its
number.
* target/waitstatus.c (target_waitstatus_to_string): Print the
signal's enum value as string instead of the (POSIX) signal name.
Gary Benson [Wed, 23 Oct 2013 12:58:26 +0000 (13:58 +0100)]
Fix off-by-one errors in *scanf format strings.
In the first hunk, the format string was off-by-one for cmd, and cmd
itself was larger than the maximum size required. cmd was reduced in
size and the format string adjusted.
In the second hunk, the format string was off-by-one for local_address,
remote_address and extra, although the buffers for the two addresses
were large enough for this not to matter. The specifiers for the two
addresses was corrected, and a number of unused variables including
extra were suppressed from parsing.
In the third hunk, the format string was off-by-one for name,
dependencies and status. This code was rewritten using strtok since
dependencies can be arbitrarily long.
gdb/
2013-10-23 Gary Benson <gbenson@redhat.com>
PR 16013
* common/linux-osdata.c (command_from_pid): Reduced size of cmd
from 32 to 18. Adjusted fscanf format string accordingly.
(Avoids leaving cmd unterminated.)
(print_sockets): Do not parse tlen, inode, sl, timeout, txq, rxq,
trun, retn or extra. (Avoids leaving extra unterminated.) Check
that local_address and remote_address will not overflow.
(linux_xfer_osdata_modules): Parse lines using strtok to avoid
leaving dependencies unterminated. Parse size as "%u" to match
definition.
Pedro Alves [Tue, 22 Oct 2013 14:11:28 +0000 (15:11 +0100)]
Remove dead sets/clears of ecs->random signal.
'*ecs' is always memset by handle_inferior_event's callers, so all
these clears are unnecessary. There's one place that sets the flag to
true, but, afterwards, before ecs->random_signal is ever read, we
reach the part of handle_inferior_even that clears ecs->random_signal,
among other things:
Pedro Alves [Tue, 22 Oct 2013 14:11:27 +0000 (15:11 +0100)]
infrun.c:keep_going: update comments.
This function still has comments referring back to when it was a goto
label in wait_for_inferior, eons ago. Looking closer, actually most
of its comments could use a facelift (contents/formatting/typos).
That's what this patch does.
Pedro Alves [Tue, 22 Oct 2013 14:02:28 +0000 (15:02 +0100)]
remote: Map invalid signal numbers to GDB_SIGNAL_UNKNOWN.
I realized that remote.c is not validating input here. Currently, if
a remote stub sends in an invalid signal number (or put another way,
if a future stub sends a new signal an old GDB doesn't know about),
GDB will do out of bounds accesses in the
signal_pass/signal_stop/signal_program arrays. It'll probably be a
long while before we add another signal number (and buggy stubs should
just be fixed), but can't hurt to be defensive.
Tested on x86_64 Fedora 17, native gdbserver.
gdb/
2013-10-22 Pedro Alves <palves@redhat.com>
* remote.c (remote_parse_stop_reply) <'T'/'S'/'X' replies>: Map
invalid signal numbers to GDB_SIGNAL_UNKNOWN.
Pedro Alves [Tue, 22 Oct 2013 11:16:51 +0000 (12:16 +0100)]
Fix up a couple oddities in GDB's signal names and strings.
- The Mach exception/signals escaped the TARGET_ -> GDB_ prefix change
done a while ago, but there's no real reason for that. I grepped
for TARGET_EXC and fixed all found, which unsurprisingly, means
darwin-nat.c needed fixing. I think the change there is as obvious
and trivial as it can get, so I'd be quite surprised if this broke
anything there somehow.
- GDB_SIGNAL_LAST's description string was unnecessarily inconsistent
with the enum name.
Built on x86_64 Fedora 17.
gdb/
2013-10-22 Pedro Alves <palves@redhat.com>
* include/gdb/signals.def (TARGET_EXC_BAD_ACCESS): Rename to
GDB_EXC_BAD_ACCESS.
(TARGET_EXC_BAD_INSTRUCTION): Rename to GDB_EXC_BAD_INSTRUCTION.
(TARGET_EXC_ARITHMETIC): Rename to GDB_EXC_ARITHMETIC.
(TARGET_EXC_EMULATION): Rename to GDB_EXC_EMULATION.
(TARGET_EXC_SOFTWARE): Rename to GDB_EXC_SOFTWARE.
(TARGET_EXC_BREAKPOINT): Rename to GDB_EXC_BREAKPOINT.
(GDB_SIGNAL_LAST): Change description string.
* common/signals.c (gdb_signal_from_host, do_gdb_signal_to_host):
Adjust to signal renaming.
* darwin-nat.c (darwin_decode_message): Likewise.
Jose E. Marchesi [Mon, 21 Oct 2013 12:42:02 +0000 (12:42 +0000)]
2013-10-21 Jose E. Marchesi <jose.marchesi@oracle.com>
PR gdb/15986
* gdb.base/run.c (main): gdb_get_line_number tag added for
commands.exp.
(factorial): Likewise.
* gdb.base/commands.exp (watchpoint_command_test): Use
gdb_get_line_number in order to determine the locations in run.c
where local_var is detected to go out of scope.
Chao-ying Fu [Fri, 18 Oct 2013 21:14:25 +0000 (21:14 +0000)]
2013-10-18 Chao-ying Fu <Chao-ying.Fu@imgtec.com>
* config/tc-mips.c (fpr_read_mask): Test MSA registers.
(fpr_write_mask): Test MSA registers.
(can_swap_branch_p): Check fpr write followed by fpr read.
* breakpoint.c (update_watchpoint): If hardware watchpoints are
forced off, downgrade them to software watchpoints if possible,
and error out if not possible.
(watch_command_1): Move watchpoint type selection closer to
watchpoint creation, and extend the comments.
gdb/testsuite/ChangeLog
* gdb.base/watchpoints.exp: Add test for setting software
watchpoints of different types before starting the inferior.
Notice how the third "si" didn't actually make progress.
Turning on infrun and lin-lwp debug, we see:
(gdb)
infrun: clear_proceed_status_thread (process 5252)
infrun: proceed (addr=0xffffffffffffffff, signal=144, step=1)
infrun: resume (step=1, signal=0), trap_expected=0, current thread [process 5252] at 0x323d4ba7c4
LLR: Preparing to step process 5252, 0, inferior_ptid process 5252
RC: Not resuming sibling process 5252 (not stopped)
LLR: PTRACE_SINGLESTEP process 5252, 0 (resume event thread)
sigchld
infrun: wait_for_inferior ()
linux_nat_wait: [process -1], []
LLW: enter
LNW: waitpid(-1, ...) returned 5252, No child processes
LLW: waitpid 5252 received Child exited (stopped)
LLW: Candidate event Child exited (stopped) in process 5252.
SEL: Select single-step process 5252
LLW: exit
infrun: target_wait (-1, status) =
infrun: 5252 [process 5252],
infrun: status->kind = stopped, signal = SIGCHLD
infrun: infwait_normal_state
infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x323d4ba7c4
infrun: random signal 20
infrun: stepi/nexti
infrun: stop_stepping
So the inferior got a SIGCHLD (because the fork child exited while
we're doing 'si'), and since that signal is set to "nostop noprint
pass" (by default), it's considered a random signal, so it should not
cause a stop. But, it resulted in an immediate a stop_stepping call
anyway. So the single-step never really finished.
This is a regression caused by:
[[PATCH] Do not respawn signals, take 2.]
https://sourceware.org/ml/gdb-patches/2012-06/msg00702.html
Specifically, caused by this change (as mentioned in the "the lost
step issue first" part of that mail):
diff --git a/gdb/infrun.c b/gdb/infrun.c
index 53db335..3e8dbc8 100644
--- a/gdb/infrun.c
+++ b/gdb/infrun.c
@@ -4363,10 +4363,8 @@ process_event_stop_test:
(leaving the inferior at the step-resume-breakpoint without
actually executing it). Either way continue until the
breakpoint is really hit. */
- keep_going (ecs);
- return;
}
-
+ else
/* Handle cases caused by hitting a breakpoint. */
{
That made GDB fall through to the
> /* In all-stop mode, if we're currently stepping but have stopped in
> some other thread, we need to switch back to the stepped thread. */
> if (!non_stop)
part. However, if we don't have a stepped thread to get back to,
we'll now also fall through to all the "stepping" tests. For line
stepping, that'll turn out okay, as we'll just end up realizing the
thread is still in the stepping range, and needs to be re-stepped.
However, for stepi/nexti, we'll reach:
if (ecs->event_thread->control.step_range_end == 1)
{
/* It is stepi or nexti. We always want to stop stepping after
one instruction. */
if (debug_infrun)
fprintf_unfiltered (gdb_stdlog, "infrun: stepi/nexti\n");
ecs->event_thread->control.stop_step = 1;
print_end_stepping_range_reason ();
stop_stepping (ecs);
return;
}
and stop, even though the thread actually made no progress. The fix
is to restore the keep_going call, but put it after the "switch back
to the stepped thread" code, and before the stepping tests.
Tested on x86_64 Fedora 17, native and gdbserver. New test included.
gdb/
2013-10-18 Pedro Alves <palves@redhat.com>
PR gdb/16062
* infrun.c (handle_inferior_event): Keep going if we got a random
signal we should not stop for, instead of falling through to the
step tests.
gdb/testsuite/
2013-10-18 Pedro Alves <palves@redhat.com>
PR gdb/16062
* gdb.threads/stepi-random-signal.c: New file.
* gdb.threads/stepi-random-signal.exp: New file.
Tom Tromey [Thu, 17 Oct 2013 18:29:28 +0000 (18:29 +0000)]
fix for PR gdb/15995
This patch fixes PR gdb/15995.
The bug here is that gdb's printf command does not flush the output
stream. This makes a printf that is not newline-terminated interleave
incorrectly with other forms of output, such as that generated via a
call to an external program using "shell".
I note that the "output" command already does this flushing.
The fix is to call gdb_flush in printf_command.
Built and regtested on x86-64 Fedora 18.
New test case included.
This is a simple bug. target_disable_btrace and target_teardown_btrace,
both from gdb/target.c, do a "return" calling another function. But both
are marked as void. Despite the fact that the functions being called are
void as well, this is wrong. This patch fixes this by calling the functions
and then returning in the next line.
* create-version.sh: Align parameters to match those of
../../gdb/common/create-version.sh.
* Make-common.in (srcsim): New variable.
(version.c): Adjust call to create-version.sh as per above.
Chao-ying Fu [Mon, 14 Oct 2013 18:43:39 +0000 (18:43 +0000)]
2013-10-14 Chao-ying Fu <Chao-ying.Fu@imgtec.com>
* elfxx-mips.c (mips_elf_obj_tdata): Add abi_msa_bfd.
(mips_elf_merge_obj_attributes): Set abi_msa_bfd to the first object
file that has a Tag_GNU_MIPS_ABI_MSA attribute.
Merge Tag_GNU_MIPS_ABI_MSA attributes.
projects / thirdparty / binutils-gdb.git / log
