Pedro Alves [Fri, 10 Jan 2020 20:05:53 +0000 (20:05 +0000)]
Fix reconnecting to a gdbserver already debugging multiple processes, I
The multi-target patch will change the remote target's behavior when:
- the current inferior is connected to an extended-remote target.
- the current inferior is attached to any process.
- some other inferior than than the current one is live.
In current master, we get:
(gdb) tar extended-remote :9999
A program is being debugged already. Kill it? (y or n)
While after multi-target, since each inferior may have its own target
connection, we'll get:
(gdb) tar extended-remote :9999
Already connected to a remote target. Disconnect? (y or n)
That change made gdb.server/extended-remote-restart.exp expose a gdb
bug, because it made "target remote", via gdb_reconnect, just
disconnect from the previous connection, while in current master that
command would kill the inferior before disconnecting. In turn, that
would make a multi-target gdb find processes already running under
control of gdbserver as soon as it reconnects, while in current master
there is never any process around when gdb reconnects, since they'd
all been killed prior to disconnection.
The bug this exposed is that remote_target::remote_add_inferior was
always reusing current_inferior() for the new process, even if the
current inferior was already bound to a process. In the testcase's
case, when we reconnect, the remote is debugging two processes. So
we'd bind the first remote process to the empty current inferior the
first time, and then bind the second remote process to the same
inferior again, essencially losing track of the first process. That
resulted in failed assertions when we look up the inferior for the
first process by PID. The fix is to still prefer binding to the
current inferior (so that plain "target remote" keeps doing what you'd
expect), but not reuse the current inferior if it is already bound to
a process.
This patch tweaks the test to explicitly disconnect before
reconnecting, to avoid GDB killing processes, thus making current GDB
behave the same as it will behave when the multi-target work lands.
That change alone without the GDB fix exposes the bug like so:
(gdb) PASS: gdb.server/extended-remote-restart.exp: kill: 0, follow-child 0: disconnect
target extended-remote localhost:2350
Remote debugging using localhost:2350
src/gdb/thread.c:93: internal-error: thread_info* inferior_thread(): Assertion `tp' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
The original bug that the testcase was written for was related to
killing, (git 9d4a934ce604 ("gdb: Fix assert for extended-remote
target (PR gdb/18050)")), but since the testcase tries reconnecting
with both explicitly killing and not explicitly killing, I think we're
covering the original bug with this testcase change.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* remote.c (remote_target::remote_add_inferior): Don't bind a
process to the current inferior if the current inferior is already
bound to a process.
gdb/testsuite/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* gdb.server/extended-remote-restart.exp (test_reload): Explicitly
disconnect before reconnecting.
Avoid another inferior_ptid reference in gdb/remote.c
The multi-target patch makes inferior_ptid point to null_ptid before
calling into target_wait, which catches bad uses of inferior_ptid,
since the current selected thread in gdb shouldn't have much relation
to the thread that reports an event.
One such bad use is found in remote_target::remote_parse_stop_reply,
where we handle the 'W' or 'X' packets (process exit), and the remote
target does not support the multi-process extensions, i.e., it does
not report the PID of the process that exited.
With the multi-target patch, that would result in a failed assertion,
trying to find the inferior for process pid 0.
gdb/ChangeLog:
2020-01-10 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Pedro Alves <palves@redhat.com>
* remote.c (remote_target::remote_parse_stop_reply) <W/X packets>:
If no process is specified, return null_ptid instead of
inferior_ptid.
(remote_target::wait_as): Handle TARGET_WAITKIND_EXITED /
TARGET_WAITKIND_SIGNALLED with no pid.
gdb/testsuite/ChangeLog:
2020-01-10 Tankut Baris Aktemur <tankut.baris.aktemur@intel.com>
Pedro Alves <palves@redhat.com>
* gdb.server/connect-without-multi-process.exp: Also test
continuing to end.
Pedro Alves [Fri, 10 Jan 2020 20:05:52 +0000 (20:05 +0000)]
Tweak handling of remote errors in response to resumption packet
With current master, on a Fedora 27 machine with a kernel with buggy
watchpoint support, I see:
(gdb) PASS: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: hardware breakpoints work
continue
Continuing.
warning: Remote failure reply: E01
Remote communication error. Target disconnected.: Connection reset by peer.
(gdb) FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: watchpoints work
continue
The program is not being run.
(gdb) FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: breakpoint after the first fork (the program is no longer running)
The FAILs themselves aren't what's interesting here. What is
interesting is that with the main multi-target patch applied, I was getting this:
(gdb) PASS: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: hardware breakpoints work
continue
Continuing.
warning: Remote failure reply: E01
/home/pedro/brno/pedro/gdb/binutils-gdb-2/build/../src/gdb/inferior.c:285: internal-error: inferior* find_inferior_pid(process_stratum_target*, int): Assertion `pid != 0' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n) FAIL: gdb.threads/watchpoint-fork.exp: parent: singlethreaded: watchpoints work (GDB internal error)
The problem is that in remote_target::wait_as, we're hitting this:
switch (buf[0])
{
case 'E': /* Error of some sort. */
/* We're out of sync with the target now. Did it continue or
not? Not is more likely, so report a stop. */
rs->waiting_for_stop_reply = 0;
which leaves event_ptid as null_ptid. At the end of the function, we then reach:
else if (status->kind != TARGET_WAITKIND_EXITED
&& status->kind != TARGET_WAITKIND_SIGNALLED)
{
if (event_ptid != null_ptid)
record_currthread (rs, event_ptid);
else
event_ptid = inferior_ptid; <<<<< here
}
and the trouble is that with the multi-target patch, we'll get here
with inferior_ptid as null_ptid too. That is done exactly to find
these implicit assumptions that inferior_ptid is a good choice for
default thread, which isn't generaly true.
I first thought of fixing this in the "case 'E'" path, but, given that
this "event_ptid = inferior_ptid" path is also taken when the remote
target does not support threads at all, no thread-related packets or
extensions, it's better to fix it in latter path, to handle all
scenarios that miss reporting a thread.
That's what this patch does.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* remote.c (first_remote_resumed_thread): New.
(remote_target::wait_as): Use it as default event_ptid instead of
inferior_ptid.
Pedro Alves [Fri, 10 Jan 2020 20:05:49 +0000 (20:05 +0000)]
Some get_last_target_status tweaks
- Make get_last_target_status arguments optional. A following patch
will add another argument to get_last_target_status (the event's
target), and passing nullptr when we don't care for some piece of
info is handier than creating dummy local variables.
- Declare nullify_last_target_wait_ptid in a header, and remove the
local extern declaration from linux-fork.c.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* break-catch-sig.c (signal_catchpoint_print_it): Don't pass a
ptid to get_last_target_status.
* break-catch-syscall.c (print_it_catch_syscall): Don't pass a
ptid to get_last_target_status.
* infcmd.c (continue_command): Don't pass a target_waitstatus to
get_last_target_status.
(info_program_command): Don't pass a target_waitstatus to
get_last_target_status.
* infrun.c (init_wait_for_inferior): Use
nullify_last_target_wait_ptid.
(get_last_target_status): Handle nullptr arguments.
(nullify_last_target_wait_ptid): Clear target_last_waitstatus.
(print_stop_event): Don't pass a ptid to get_last_target_status.
(normal_stop): Don't pass a ptid to get_last_target_status.
* infrun.h (get_last_target_status, set_last_target_status): Move
comments here and update.
(nullify_last_target_wait_ptid): Declare.
* linux-fork.c (fork_load_infrun_state): Remove local extern
declaration of nullify_last_target_wait_ptid.
* linux-nat.c (get_detach_signal): Don't pass a target_waitstatus
to get_last_target_status.
Pedro Alves [Fri, 10 Jan 2020 20:05:48 +0000 (20:05 +0000)]
switch inferior/thread before calling target methods
Once each inferior has its own target stack, we'll need to make sure
that the right inferior is selected before we call into target
methods.
It kind of sounds worse than it is in practice. Not that many places
need to be concerned.
In thread.c, we add a new switch_to_thread_if_alive function that
centralizes the switching before calls to target_thread_alive. Other
cases are handled with explicit switching.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* gdbthread.h (scoped_restore_current_thread)
<dont_restore, restore, m_dont_restore>: Declare.
* thread.c (thread_alive): Add assertion. Return bool.
(switch_to_thread_if_alive): New.
(prune_threads): Switch inferior/thread.
(print_thread_info_1): Switch thread before calling target methods.
(scoped_restore_current_thread::restore): New, factored out from
...
(scoped_restore_current_thread::~scoped_restore_current_thread):
... this.
(scoped_restore_current_thread::scoped_restore_current_thread):
Add assertion.
(thread_apply_all_command, thread_select): Use
switch_to_thread_if_alive.
* infrun.c (proceed, restart_threads, handle_signal_stop)
(switch_back_to_stepped_thread): Switch current thread before
calling target methods.
Pedro Alves [Fri, 10 Jan 2020 20:05:47 +0000 (20:05 +0000)]
Introduce switch_to_inferior_no_thread
Several places want to switch context to an inferior and its pspace,
while at the same time switch to "no thread selected". This commit
adds a function that does that, and uses it in a few places.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* inferior.c (switch_to_inferior_no_thread): New function,
factored out from ...
(inferior_command): ... here.
* inferior.h (switch_to_inferior_no_thread): Declare.
* mi/mi-main.c (run_one_inferior): Use
switch_to_inferior_no_thread.
Pedro Alves [Fri, 10 Jan 2020 20:05:46 +0000 (20:05 +0000)]
Delete unnecessary code from kill_command
I believe this comment:
/* Killing off the inferior can leave us with a core file. If
so, print the state we are left in. */
Referred to the fact that a decade ago, by design, GDB would let you
type "run" when debugging a core dump, keeping the core open. That
"run" would push a process_stratum target on the target stack for the
live process, and, the core would remain open -- we used to have a
core_stratum. When the live process was killed/detached or exited,
GDB would go back to debugging the core, since the core_stratum target
was now at the top of the stack. That design had a number of
problems, see here for example:
In 2010, core_stratum was finaly eliminated and cores now have
process_stratum too, with commit c0edd9edadfe ("Make core files the
process_stratum."). Pushing a live process on the stack while you're
debugging a core discards the core completely.
I also thought that this might be in use with checkpoints, but it does
not -- "kill" when you have multiple checkpoints kills all the
checkpoints.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
Pedro Alves [Fri, 10 Jan 2020 20:05:45 +0000 (20:05 +0000)]
Don't check target is running in remote_target::mourn_inferior
I believe the tail end of remote_target::mourn_inferior is broken, and
it's been broken for too long to even bother trying to fix. Most
probably nobody needs it. If the code is reached and we find the
target is running, we'd need to resync the thread list, at least,
since generic_mourn_inferior got rid of all the threads in the
inferior, otherwise, we'd hit an assertion on the next call to
inferior_thread(), for example. A "correct" fix would probably
involve restarting the whole remote_target::start_remote requence,
exactly as if we had completely disconnected and reconnected from
scratch.
Note that regular stub debugging usually uses plain target remote, but
this code is only reachable in target extended-mode:
- The !remote_multi_process_p check means that it's only reacheable if
the stub does not support multi-process. I.e., there can only ever
be one live process.
- remote_target::mourn_inferior has this at the top:
/* In 'target remote' mode with one inferior, we close the connection. */
if (!rs->extended && number_of_live_inferiors () <= 1)
{
unpush_target (this);
/* remote_close takes care of doing most of the clean up. */
generic_mourn_inferior ();
return;
}
Which means that if we only had one live inferior (which for our
case, must be true), we'll have closed the connection already,
unless we're in extended-remote mode.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* remote.c (remote_target::mourn_inferior): No longer check
whether the target is running.
Pedro Alves [Fri, 10 Jan 2020 20:05:44 +0000 (20:05 +0000)]
Make target_ops::has_execution take an 'inferior *' instead of a ptid_t
With the multi-target work, each inferior will have its own target
stack, so to call a target method, we'll need to make sure that the
inferior in question is the current one, otherwise target->beneath()
calls will find the target beneath in the wrong inferior.
In some places, it's much more convenient to be able to check whether
an inferior has execution without having to switch to it in order to
call target_has_execution on the right inferior/target stack, to avoid
side effects with switching inferior/thread/program space.
The current target_ops::has_execution method takes a ptid_t as
parameter, which, in a multi-target world, isn't sufficient to
identify the target. This patch prepares to address that, by changing
the parameter to an inferior pointer instead. From the inferior,
we'll be able to query its target stack to tell which target is
beneath.
Also adds a new inferior::has_execution() method to make callers a bit
more natural to read.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* corelow.c (core_target::has_execution): Change parameter type to
inferior pointer.
* inferior.c (number_of_live_inferiors): Use
inferior::has_execution instead of target_has_execution_1.
* inferior.h (inferior::has_execution): New.
* linux-thread-db.c (thread_db_target::update_thread_list): Use
inferior::has_execution instead of target_has_execution_1.
* process-stratum-target.c
(process_stratum_target::has_execution): Change parameter type to
inferior pointer. Check the inferior's PID instead of
inferior_ptid.
* process-stratum-target.h
(process_stratum_target::has_execution): Change parameter type to
inferior pointer.
* record-full.c (record_full_core_target::has_execution): Change
parameter type to inferior pointer.
* target.c (target_has_execution_1): Change parameter type to
inferior pointer.
(target_has_execution_current): Adjust.
* target.h (target_ops::has_execution): Change parameter type to
inferior pointer.
(target_has_execution_1): Change parameter type to inferior
pointer. Change return type to bool.
* tracefile.h (tracefile_target::has_execution): Change parameter
type to inferior pointer.
Pedro Alves [Fri, 10 Jan 2020 20:05:43 +0000 (20:05 +0000)]
exceptions.c:print_flush: Remove obsolete check
Commit 20f0d60db4fb ("Avoid crash when calling warning too early"),
added a "current_top_target () != NULL" check to
target_supports_terminal_ours, so this check in exceptions.c is now
obsolete.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
Pedro Alves [Fri, 10 Jan 2020 20:05:42 +0000 (20:05 +0000)]
Make "show remote exec-file" inferior-aware
The "set remote exec-file" setting is per-inferior, but the "show
remote exec-file" command always shows the last set exec-file,
irrespective of the current inferior. E.g.:
# Set inferior 1's exec-file:
(gdb) set remote exec-file prog1
# Add inferior 2, switch to it, and set its exec-file:
(gdb) add-inferior
Added inferior 2
(gdb) inferior 2
(gdb) set remote exec-file prog2
# Switch back to inferior 1, and show its exec-file:
(gdb) inferior 1
(gdb) show remote exec-file
prog2
^^^^^ should show "prog1" instead here.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* remote.c (show_remote_exec_file): Show the current inferior's
exec-file instead of the command variable's value.
gdb/testsuite/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
Pedro Alves [Fri, 10 Jan 2020 20:05:41 +0000 (20:05 +0000)]
Don't rely on inferior_ptid in record_full_wait
The multi-target patch sets inferior_ptid to null_ptid before handling
a target event, and thus before calling target_wait, in order to catch
places in target_ops::wait implementations that are incorrectly
relying on inferior_ptid (which could otherwise be a ptid of a
different target, for example). That caught this instance in
record-full.c.
Fix it by saving the last resumed ptid, and then using it in
record_full_wait_1, just like how the last "step" argument passed to
record_full_target::resume is handled too.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
* record-full.c (record_full_resume_ptid): New global.
(record_full_target::resume): Set it.
(record_full_wait_1): Use record_full_resume_ptid instead of
inferior_ptid.
Pedro Alves [Fri, 10 Jan 2020 20:05:41 +0000 (20:05 +0000)]
Preserve selected thread in all-stop w/ background execution
In non-stop mode, if you resume the program in the background (with
"continue&", for example), then gdb makes sure to not switch the
current thread behind your back. That means that you can be sure that
the commands you type apply to the thread you selected, even if some
other thread that was running in the background hits some event just
while you're typing.
In all-stop mode, however, if you resume the program in the
background, gdb let's the current thread switch behind your back.
This is bogus, of course. All-stop and non-stop background
resumptions should behave the same.
This patch fixes that, and adds a testcase that exposes the bad
behavior in current master.
The fork-running-state.exp changes are necessary because that
preexisting testcase was expecting the old behavior:
Before:
continue &
Continuing.
(gdb)
[Attaching after process 8199 fork to child process 8203]
[New inferior 2 (process 8203)]
info threads
Id Target Id Frame
1.1 process 8199 "fork-running-st" (running)
* 2.1 process 8203 "fork-running-st" (running)
(gdb)
After:
continue &
Continuing.
(gdb)
[Attaching after process 24660 fork to child process 24664]
[New inferior 2 (process 24664)]
info threads
Id Target Id Frame
* 1.1 process 24660 "fork-running-st" (running)
2.1 process 24664 "fork-running-st" (running)
(gdb)
Here we see that before this patch GDB switches current inferior to
the new inferior behind the user's back, as a side effect of handling
the fork.
The delete_exited_threads call in inferior_appeared is there to fix an
issue that Baris found in a previous version of this patch. The
fetch_inferior_event change increases the refcount of the current
thread, and in case the fetched inferior event denotes a thread exit,
the thread will not be deleted right away. A non-deleted but exited
thread stays in the inferior's thread list. This, in turn, causes the
"init_thread_list" call in inferior.c to be skipped. A consequence is
that the global thread ID counter is not restarted if the current
thread exits, and then the inferior is restarted:
(gdb) start
Temporary breakpoint 1 at 0x4004d6: file main.c, line 21.
Starting program: /tmp/main
Temporary breakpoint 1, main () at main.c:21
21 foo ();
(gdb) info threads -gid
Id GId Target Id Frame
* 1 1 process 16106 "main" main () at main.c:21
(gdb) c
Continuing.
[Inferior 1 (process 16106) exited normally]
(gdb) start
Temporary breakpoint 2 at 0x4004d6: file main.c, line 21.
Starting program: /tmp/main
Temporary breakpoint 2, main () at main.c:21
21 foo ();
(gdb) info threads -gid
Id GId Target Id Frame
* 1 2 process 16138 "main" main () at main.c:21
^^^
Notice that GId == 2 above. It should have been "1" instead.
The new tids-git-reset.exp testcase exercises the problem above.
gdb/ChangeLog:
2020-01-10 Pedro Alves <palves@redhat.com>
George Barrett [Fri, 10 Jan 2020 19:30:28 +0000 (06:30 +1100)]
Fix handling of null stap semaphores
According to the SystemTap documentation on user-space probes[0], stap
probe points without semaphores are denoted by setting the semaphore
address in the probe's note to zero. At present the code does do a
comparison of the semaphore address against zero, but only after it's
been relocated; as such it will (almost?) always fail, commonly
resulting in GDB trying to overwrite the ELF magic located at the
image's base address.
This commit tests the address as specified in the SDT note rather than
the relocated value in order to correctly detect absent probe
semaphores.
George Barrett [Fri, 10 Jan 2020 19:30:01 +0000 (06:30 +1100)]
gdb/testsuite/gdb.base/stap-probe: Minor clean-up
This patch resolves a couple of issues with the test case for SystemTap
user-space probe points:
1. The preprocessor macro guarding the semaphore variables in the C
file is (rather confusingly) named USE_PROBES. This has been
renamed to USE_SEMAPHORES, to better reflect its function.
2. The test procedures in the expect file improperly pass the flag
defining USE_PROBES to prepare_for_testing; as such, the test
binary that's supposed to have probes with semaphores is the same
as the one without. This has also been fixed.
3. No test is performed to check that `info probes' returns
information about probe semaphores. Such a test is included in this
patch.
gdb/testsuite/ChangeLog
2020-01-10 George Barrett <bob@bob131.so>
* gdb.base/stap-probe.c: Rename USE_PROBES to USE_SEMAPHORES.
* gdb.base/stap-probe.exp: Likewise.
(stap_test): Pass argument as an additional flag.
(stap_test_no_debuginfo): Likewise.
(stap_test): Check `info probes stap' output for semaphore
addresses if the test binary is supposed to have them.
Szabolcs Nagy [Thu, 9 Jan 2020 17:20:56 +0000 (17:20 +0000)]
[PR ld/22269] arm: Avoid dynamic relocs for undefweak symbols in static PIE
With static PIE linking undefined weak symbols are resolved to 0, so no
dynamic relocation is needed for them. The UNDEFWEAK_NO_DYNAMIC_RELOC
macro was introduced so this case can be handled easily, but it was not
applied consistently in the first attempt to fix ld/22269 for arm:
Tamar Christina [Fri, 10 Jan 2020 13:48:57 +0000 (13:48 +0000)]
AArch64: Revert setting of elf class in linker stub.
This changes the fix to PR 25210 by removing the ELF class change.
As it turns out the correct change was only the change in compress.c.
Everything else is unneeded and setting the elf class is making the linker
behave very oddly under LTO. The first stub is correctly written out but for
the rest the suddenly don't have a pointer to the stub section anymore.
This caused SPEC to fail as the program would branch to the stub and it wouldn't
be filled in.
Committed to master under the trivial rule as this is partially reverting a previous commit.
Alan Modra [Wed, 8 Jan 2020 20:29:42 +0000 (06:59 +1030)]
ubsan: m10300: shift exponent -4
* m10300-dis.c (disassemble): Move extraction of DREG, AREG, RREG,
and XRREG value earlier to avoid a shift with negative exponent.
* m10200-dis.c (disassemble): Similarly.
Alan Modra [Wed, 8 Jan 2020 20:14:16 +0000 (06:44 +1030)]
ubsan: spu: left shift of negative value
Also fixes a real bug. The DECODE_INSN_I9a and DECODE_INSN_I9b both
use UNSIGNED_EXTRACT for 7 low bits of the result, but this was an
unsigned value due to "insn" being unsigned. DECODE_INSN_I9* is
therefore unsigned too, leading to a zero extension in an expression
using a bfd_vma if bfd_vma is 64 bits.
* opcode/spu.h: Formatting.
(UNSIGNED_EXTRACT): Use 1u.
(SIGNED_EXTRACT): Don't sign extend with shifts.
(DECODE_INSN_I9a, DECODE_INSN_I9b): Avoid left shift of signed value.
Keep result signed.
(DECODE_INSN_U9a, DECODE_INSN_U9b): Delete.
This commit restores the synchronised scrolling and also maintains the
horizontal scroll within the source view when it is vertically
scrolled, something that was broken before.
This commit does not mean that scrolling the assembler view scrolls
the source view. The connection this way never existed, though maybe
it should, but I'll leave adding this feature for a separate commit.
gdb/ChangeLog:
* tui/tui-source.c (tui_source_window::do_scroll_vertical): Update
all source windows, and maintain horizontal scroll status while
doing so.
Tom Tromey [Sun, 22 Dec 2019 23:52:56 +0000 (16:52 -0700)]
gdb: Fix scrolling in TUI
Hannes Domani pointed out that my previous patch to fix the "list"
command in the TUI instead broke vertical scrolling. While looking at
this, I found that do_scroll_vertical calls print_source_lines, which
seems like a very roundabout way to change the source window. This
patch removes this oddity and fixes the bug at the same time.
I've added a new test case. This is somewhat tricky, because the
obvious approach of sending a dummy command after the scroll did not
work -- due to how the TUI works, sennding a command causes the scroll
to take effect.
gdb/ChangeLog
2019-12-22 Tom Tromey <tom@tromey.com>
PR tui/18932:
* tui/tui-source.c (tui_source_window::do_scroll_vertical): Call
update_source_window, not print_source_lines.
gdb/testsuite/ChangeLog
2019-12-22 Tom Tromey <tom@tromey.com>
PR tui/18932:
* lib/tuiterm.exp (Term::wait_for): Rename from _accept. Return a
meangingful value.
(Term::command, Term::resize): Update.
* gdb.tui/basic.exp: Add scrolling test.
Andrew Burgess [Tue, 7 Jan 2020 00:41:08 +0000 (00:41 +0000)]
gdb/tui: Fix 'layout asm' before the inferior has started
Currently if a user starts the tui with 'layout asm' then they will be
presented with the 'src' layout.
What happens is:
1. Layout command enables TUI, selecting the SRC layout by default.
2. As part of tui_enable we call tui_display_main, which calls
tui_get_begin_asm_address, which calls
set_default_source_symtab_and_line. This changes core GDBs
current symtab and line, which triggers a call to the symtab
changed hook tui_symtab_changed, which sets the flag
from_source_symtab.
3. Back in the layout command, the layout is changed from SRC to
ASM. After this the layout command completes and we return to
core GDB which prints the prompt, however...
4. The before prompt hook is called which sees the
from_source_symtab flag is set and forces the SRC window to be
displayed. This switches us back to SRC view.
The solution I propose here is to delay installing the hooks into core
GDB until after we have finished setting up the tui and selecting the
default frame to view. In this way we effectively ignore the first
symtab changed event triggered when making main the default symtab.
gdb/ChangeLog:
* tui/tui.c (tui_enable): Register tui hooks after calling
tui_display_main.
Andrew Burgess [Tue, 7 Jan 2020 00:35:02 +0000 (00:35 +0000)]
gdb/testsuite/tui: Introduce check_box_contents
A new test procedure for matching the contents of one screen box
against a regexp. This can be used to match the contents of one TUI
window against a regexp without any of the borders, or other windows
being included in the matched output (as is currently the case with
check_contents).
This will be used in a later commit.
gdb/testsuite/ChangeLog:
* lib/tuiterm.exp (Term::check_box_contents): New proc.
Andrew Burgess [Tue, 7 Jan 2020 00:30:35 +0000 (00:30 +0000)]
gdb/testsuite/tui: Split enter_tui into two procs
Split Term::enter_tui into two procedures, a core which does the
setup, but doesn't actually enable tui mode, and the old enter_tui
that calls the new core, and then enables tui mode.
This is going to be useful in a later commit.
gdb/testsuite/ChangeLog:
* lib/tuiterm.exp (Term::prepare_for_tui): New proc.
(Term::enter_tui): Use Term::prepare_for_tui.
Andrew Burgess [Tue, 7 Jan 2020 00:26:22 +0000 (00:26 +0000)]
gdb/testsuite/tui: Always dump_screen when asked
The Term::dump_screen routine currently dumps the screen using calls
to 'verbose', this means it will only dump the screen when the
testsuite is running in verbose mode.
However, the Term::dump_screen is most often called when a test fails,
in this case I think it is useful to have the screen dumped even when
we're not in verbose mode.
This commit changes the calls to 'verbose' to be 'verbose -log' so we
always get the screen dump.
gdb/testsuite/ChangeLog:
* lib/tuiterm.exp (Term::dump_screen): Always dump the screen when
called.
Before the above change we expected GDB to behave like this:
(gdb) step
foo () at /path/to/gdb/testsuite/gdb.base/skip.c:42
42 return 0;
(gdb)
However, when the test is compiled with GCC 9.2.0 we get a different
behaviour, and so we need a second 'step', like this:
(gdb) step
main () at /path/to/gdb.base/skip.c:32
32 x = baz ((bar (), foo ()));
(gdb) step
foo () at /path/to/gdb/testsuite/gdb.base/skip.c:42
42 return 0;
(gdb)
Now the change to the test matches against 'main () at .*', however if
GDB or expect is being slow then we might only get to see output like
this:
(gdb) step
main () at /path/to/g
This will happily match the question pattern, so we send 'step' to GDB
again. Now GDB continues to produce output which expect accepts, we
now see this:
b.base/skip.c:32
32 x = baz ((bar (), foo ()));
(gdb)
This has carried on from where the previous block of output left off.
This doesn't match the final pattern 'foo \\(\\) at.*', but it does
match the prompt pattern that gdb_test_multiple adds, and so we report
the test as failing.
The solution is to simply ensure that the question consumes everything
up to, and including the prompt. This ensures that the prompt can't
then match the failure case. The new test line becomes:
Recent MinGW versions require -lssp when using _FORTIFY_SOURCE, which
gdb does (in common-defs.h)
https://github.com/msys2/MINGW-packages/issues/5868#issuecomment-544107564
To avoid all the complications with checking for -lssp and making sure it's
linked statically, just don't define it.
gdb/ChangeLog:
2020-01-09 Christian Biesinger <cbiesinger@google.com>
* gdbsupport/common-defs.h: Don't define _FORTIFY_SOURCE on MinGW.
compute_and_set_names would only free the name if we did not find the name
in the hashtable, but it needs to always free it. Solve this by moving the
smart pointer outside the if.
Thanks to PhilippeW for finding this.
gdb/ChangeLog:
2020-01-09 Christian Biesinger <cbiesinger@google.com>
* symtab.c (general_symbol_info::compute_and_set_names): Move the
unique_xmalloc_ptr outside the if to always free the demangled name.
Nick Clifton [Thu, 9 Jan 2020 16:51:04 +0000 (16:51 +0000)]
Fix an attempt to free a static pointer when using objcopy's symbol addition feature.
PR 25220
* objcopy.c (empty_name): New variable.
(need_sym_before): Prevent an attempt to free a static variable.
(filter_symbols): Avoid strcmp test by checking for pointer
equality.
Aaron Merey [Thu, 9 Jan 2020 13:19:20 +0000 (13:19 +0000)]
Add support for debuginfod to the binutils (disable by default, enabled via a configure time option).
debuginfod is a lightweight web service that indexes ELF/DWARF
debugging resources by build-id and serves them over HTTP. This patch
enables objdump and readelf to query debuginfod servers when they are
otherwise not able to find separate debug files. Binutils can be built
with debuginfod using the --with-debuginfod configure option. This
requires that libdebuginfod be installed and found at configure time.
debuginfod is packaged with elfutils, starting with version 0.178. For
more information see https://sourceware.org/elfutils/.
toplevel* config/debuginfod.m4: New file. Add macro AC_DEBUGINFOD. Adds
new configure option --with-debuginfod.
* configure: Regenerate.
* configure.ac: Call AC_DEBUGINFOD.
binutils* Makefile.am (readelf_LDADD, objdump_LDADD): Add libdebuginfod.
* Makefile.in: Regenerate.
* NEWS: Update.
* config.in: Regenerate.
* configure: Regenerate.
* configure.ac: Call AC_DEBUGINFOD.
* doc/Makefile.in: Regenerate.
* doc/binutils.texi: Add section on using binutils
with debuginfod.
* dwarf.c (debuginfod_fetch_separate_debug_info): New function.
Query debuginfod servers for the target debug file.
(load_separate_debug_info): Call
debuginfod_fetch_separate_debug_info if configured with
debuginfod.
(load_separate_debug_files): Add file argument to
load_separate_debug_info calls.
* dwarf.h (get_build_id): Add declaration.
* objdump.c (get_build_id): New function. Get build-id of file.
* readelf.c (get_build_id): Likewise.
* testsuite/binutils-all/debuginfod.exp: New tests.
* testsuite/binutils-all/linkdebug.s: Add .note.gnu.build-id
section.
Jan Beulich [Thu, 9 Jan 2020 10:40:04 +0000 (11:40 +0100)]
x86: refine when to trigger optimizations
Checking just the base opcode without also checking this isn't a VEX
encoding, and without there being other insn properties avoiding a match
once respective VEX/XOP/EXEX-encoded insns would appear, is at least
dangerous. Add respective checks. At the same time there's no real need
to check the extension opcode to be None for the 0xA8 form - there's
nothing it can be confused with, and non-VEX-and-alike forms also can't
appear.
Jan Beulich [Thu, 9 Jan 2020 10:38:59 +0000 (11:38 +0100)]
x86: consistently convert to byte registers for TEST w/ imm optimization
Commit ac0ab1842d ("i386: Also check R12-R15 registers when optimizing
testq to testb") didn't go quite far enough: In order to avoid confusing
other code registers would better be converted to byte ones uniformly.
Jan Beulich [Thu, 9 Jan 2020 10:38:01 +0000 (11:38 +0100)]
x86: SYSENTER/SYSEXIT are unavailable in 64-bit mode on AMD
The disassembler change is such that in default mode we'd disassemble
the insns (for there not ebing any conflicts), but when AMD64 mode was
explicitly requested, we'd show them as "(bad)".
Tom Tromey [Mon, 6 Jan 2020 21:34:52 +0000 (14:34 -0700)]
Change section_offsets to a std::vector
This changes section_offsets to be specialization of a std::vector and
updates all the users. It also removes the ANOFFSET and
SIZEOF_N_SECTION_OFFSETS macros.
Most of this is just a generic sort of cleanup, that reduces the
number of lines of code. However, a couple spots were doing weird
things.
Nick Clifton [Wed, 8 Jan 2020 16:30:20 +0000 (16:30 +0000)]
Make the assembler generate an error if there is an attempt to define a section with the same name as an already defined symbol.
PR 14891
* config/obj-elf.c (obj_elf_section): Fail if the section name is
already defined as a different symbol type.
* testsuite/gas/elf/pr14891.s: New test source file.
* testsuite/gas/elf/pr14891.d: New test driver.
* testsuite/gas/elf/pr14891.s: New test expected error output.
* testsuite/gas/elf/elf.exp: Run the new test.
Alan Modra [Wed, 8 Jan 2020 01:12:36 +0000 (11:42 +1030)]
ubsan: z8k: index 10 out of bounds for type 'unsigned int const[10]'
The fix is the additional ARRAY_SIZE test, the rest just tidies
variable types rather than adding a cast to avoid warnings.
opcodes/
* z8k-dis.c: Include libiberty.h
(instr_data_s): Make max_fetched unsigned.
(z8k_lookup_instr): Make nibl_index and tabl_index unsigned.
Don't exceed byte_info bounds.
(output_instr): Make num_bytes unsigned.
(unpack_instr): Likewise for nibl_count and loop.
* z8kgen.c (gas <opcode_entry_type>): Make noperands, length and
idx unsigned.
* z8k-opc.h: Regenerate.
gas/
* config/tc-z8k.c (md_begin): Make idx unsigned.
(get_specific): Likewise for this_index.
We use gas' expression function to parse the operands of an
instruction in a generic way. There are situations when we have labels
and registers having the same name as well as the substraction sign
doesn't always stands for the arithmetical operation but for the
register range (e.g. enter instruction). This patch improves parsing
symbols found in a instruction operand, cleans up code and avoids
using default or undefined variables.
* config/tc-arc.c (parse_reloc_symbol): New function.
(tokenize_arguments): Clean up, use parse_reloc_symbol function.
(md_operand): Set X_md to absent.
(arc_parse_name): Check for X_md.
gdb: Don't allow annotations to influence what else GDB prints
A change was accidentally made that moved a call to do_gdb_disassembly
out of an if block guarded by 'if (source_print && sal.symtab)'. The
result was that if a user has 'set disassemble-next-line on' then the
backtrace would now include some disassembly of a few instructions in
each frame.
This change was not intentional, but was not spotted by any tests.
This commit restores the old behaviour and adds a test to ensure this
doesn't break again in the future.
Shahab Vahedi [Mon, 6 Jan 2020 14:27:32 +0000 (15:27 +0100)]
GDB: Fix the overflow in addr/line_is_displayed()
In tui_disasm_window::addr_is_displayed(), there can be situations
where "content" is empty. For instance, it can happen when the
"content" was not filled in tui_disasm_window::set_contents(),
because tui_disassemble() threw an exception. Usually this exception
is the result of fetching invalid PC addresses like the ones beyond
the end of the program.
Having "content.size ()" zero leads to an overflow in this condition
check inside tui_disasm_window::addr_is_displayed():
int i = 0;
while (i < content.size () - threshold ...) {
... content[i] ...
}
"threshold" is 2 and there are times that "content.size ()" is 0.
This results into an overflow and the loop is entered whereas it
should have been skipped. Finally, "content[i]" access leads to
a segmentation fault.
Same problem applies to tui_source_window::line_is_displayed().
The issue has been discussed at length in bug 25345:
https://sourceware.org/bugzilla/show_bug.cgi?id=25345
This commit avoids the segmentation faults with an early check:
if (content.size () < SCROLL_THRESHOLD)
return false;
Moreover, those functions have been overhauled to a leaner code.
* tui/tui-disasm.c (tui_disasm_window::addr_is_displayed): Avoid
overflow by an early check of content vs threshold.
* tui/tui-source.c (tui_source_window::line_is_displayed):
Likewise.
Jon Turney [Thu, 2 Jan 2020 00:25:56 +0000 (00:25 +0000)]
Fix a crash with a malformed PE header
Don't try to read the PE export table when no section contains the RVA
for it.
(I have a PE executable [1] packed with UPX, where the export table data
directory entry contains a RVA which doesn't correspond to any section.
Mistakenly trying to debug this with gdb makes it crash.)
Nick Clifton [Mon, 6 Jan 2020 16:24:23 +0000 (16:24 +0000)]
Add -fcommon compiler command line option to linker tests that need common symbols.
PR 25327
* testsuite/ld-elf/shared.exp: Add -fcommon option to compiler
command line when building libcomm1.o and pr13250 tests.
* testsuite/ld-plugin/lto.exp: Likewise for pr20267 tests.
Hannes Domani [Sun, 22 Dec 2019 13:58:22 +0000 (14:58 +0100)]
Fix search in TUI
The variable last_line_listed is never set when print_source_lines_base is
called in TUI mode, so the search always started from the last line printed
outside of TUI mode.
gdb/ChangeLog:
2020-01-06 Hannes Domani <ssbssa@yahoo.de>
* source.c (print_source_lines_base): Set last_line_listed.
Eli Zaretskii [Mon, 6 Jan 2020 11:51:54 +0000 (11:51 +0000)]
Improve process exit status macros on MinGW
When a Windows program is terminated by a fatal exception, its exit
code is the value of that exception, as defined by the various
EXCEPTION_* symbols in the Windows API headers. This commit emulates
WTERMSIG etc. by translating the fatal exception codes to more-or-less
equivalent Posix signals.
gdb/ChangeLog:
2020-01-06 Eli Zaretskii <eliz@gnu.org>
Pedro Alves <palves@redhat.com>
* Makefile.in (COMMON_SFILES): Add gdbsupport/gdb_wait.c.
* windows-tdep.c: New enumeration of WINDOWS_SIG* signals.
(windows_gdb_signal_to_target): New function, uses the above
enumeration to convert GDB internal signal codes to equivalent
Windows codes.
(windows_init_abi): Call set_gdbarch_gdb_signal_to_target.
* windows-nat.c: Include "gdb_wait.h".
(get_windows_debug_event): Extract the fatal exception from the
exit status and convert to the equivalent Posix signal number.
* cli/cli-cmds.c (exit_status_set_internal_vars): Account for the
possibility that WTERMSIG returns GDB_SIGNAL_UNKNOWN.
* gdbsupport/gdb_wait.c: New file, implements
windows_status_to_termsig.
* gdbsupport/gdb_wait.h (WIFEXITED, WIFSIGNALED, WEXITSTATUS)
(WTERMSIG) [__MINGW32__]: Separate definitions for MinGW.
gdb/gdbserver/ChangeLog:
2020-01-06 Eli Zaretskii <eliz@gnu.org>
Pedro Alves <palves@redhat.com>
* win32-low.c (get_child_debug_event): Extract the fatal exception
from the exit status and convert to the equivalent Posix signal
number.
(win32_wait): Allow TARGET_WAITKIND_SIGNALLED status as well.
* Makefile.in (OBS, SFILES): Add gdb_wait.[co].
Alan Modra [Mon, 6 Jan 2020 09:21:22 +0000 (19:51 +1030)]
bfd_check_format: ignore errors from coff_real_object_p
Since 1993-11-05 git commit c188b0bec3b, bfd_check_format has failed
if any of the target object_p functions returns false with any error
but bfd_error_wrong_format. That's just weird. There is really no
reason why coff_real_object_p should be fixed to only return that
error instead of numerous other possible errors. Even an out of
memory condition for one target doesn't necessarily mean other targets
can't match, assuming the failing target nicely returns all memory it
might have used.
* format.c (bfd_check_format_matches): Ignore bfd_error on target
match failures. Don't init to bfd_error_wrong_format before
calling _bfd_check_format.
Alan Modra [Mon, 6 Jan 2020 07:59:14 +0000 (18:29 +1030)]
alpha-vms: don't exit on stack underflow/overflow
BFD is not supposed to exit or abort on anything the user can do.
* vms-alpha.c (_bfd_vms_push, _bfd_vms_pop): Return pass/fail
status rather than exiting on stack overflow or underflow.
(_bfd_vms_slurp_etir): Adjust to suit.
Alan Modra [Sat, 4 Jan 2020 09:23:19 +0000 (19:53 +1030)]
ubsan: m32c: left shift of negative value
There are probably a lot more of these still here.
cpu/
* m32c.cpu (f-dsp-8-u16, f-dsp-8-s16): Rearrange to mask any sign
bits before shifting rather than masking after shifting.
(f-dsp-16-u16, f-dsp-16-s16, f-dsp-32-u16, f-dsp-32-s16): Likewise.
(f-dsp-40-u16, f-dsp-40-s16, f-dsp-48-u16, f-dsp-48-s16): Likewise.
(f-dsp-64-u16, f-dsp-8-s24): Likewise.
(f-bitbase32-16-s19-unprefixed): Avoid signed left shift.
opcodes/
* m32c-ibld.c: Regenerate.
Alan Modra [Sun, 5 Jan 2020 22:22:39 +0000 (08:52 +1030)]
PR25344, z80 disassembler recursion
PR 25344
* z80-dis.c (suffix): Don't use a local struct buffer copy.
Peek at next byte to prevent recursion on repeated prefix bytes.
Ensure uninitialised "mybuf" is not accessed.
(print_insn_z80): Don't zero n_fetch and n_used here,..
(print_insn_z80_buf): ..do it here instead.
Andrew Burgess [Sun, 22 Dec 2019 23:45:31 +0000 (23:45 +0000)]
gdb: use tui_set_layout not show_layout to fix window focus
When calling tui_add_win_to_layout, use tui_set_layout not show_layout
so that window focus is correctly updated. If the focus is not
correctly maintained then GDB can be crashed like this:
start
tui enable
layout asm
list SOME_FUNCTION
At this point GDB will have "popped up" the source window to
display SOME_FUNCTION. Previously no window would have focus at this
point, and so if the user now does 'focus next' or 'focus prev', then
GDB would crash.
Calling tui_set_layout ensures that focus is correctly calculated as
the source window is "popped up", and this fixes the issue.
gdb/ChangeLog:
* tui/tui-layout.c (tui_add_win_to_layout): Use tui_set_layout not
show_layout.
gdb/testsuite/ChangeLog:
* gdb.tui/list.exp: Test 'focus next' after 'list main'.
Luis Machado [Fri, 3 Jan 2020 19:08:16 +0000 (16:08 -0300)]
[AArch64] Fix erroneous use of spu architecture bfd
While investigating some SVE code, i noticed the use of two spu bfd variables.
This looks like an oversight, as the "id" field is available for non-spu
architectures as well, even though its primary use was the Cell BE
architecture.
gdb/ChangeLog:
2020-01-05 Luis Machado <luis.machado@linaro.org>
* aarch64-linux-nat.c
(aarch64_linux_nat_target::thread_architecture): Use bfd_arch_aarch64
and bfd_mach_aarch64.
Eli Zaretskii [Sun, 5 Jan 2020 05:50:27 +0000 (09:50 +0400)]
libctf: Add configure check for asprintf (for MinGW)
This commit fixes a compilation warning when compiling libctf
on MinGW:
libctf/ctf-dump.c:118:8: warning: implicit declaration of function
'asprintf'; did you mean 'vasprintf'? [-Wimplicit-function-declaration]
if (asprintf (&bit, " %lx: [slice 0x%x:0x%x]",
^~~~~~~~
vasprintf
MinGW doesn't provide that function, so we depend on the one provided
by libiberty. However, the declaration is guarded by HAVE_DECL_ASPRINTF,
which we do not have in libctf's config.h.
Alan Modra [Thu, 2 Jan 2020 07:09:32 +0000 (17:39 +1030)]
Release bfd_alloc memory in bfd_check_format_matches
It's a little tricky. We can release any memory back when we have a
match failure, but after a match success which we might want to
preserve for later use the high water mark must change to that of the
matched bfd.
* format.c (bfd_check_format_matches): Add preserve_match.
Save initial bfd state in "preserve", matched bfd state in
"preserve_match". Save just the first match. Release
bfd_alloc memory. Restore and finish preserved state as
appropriate on all function exit paths.
Alan Modra [Tue, 31 Dec 2019 23:58:42 +0000 (10:28 +1030)]
coff: free malloc'd memory on successful target match too
object_p functions cannot allocate memory by malloc and not free it
before returning. Even a successful target match may not be the best
match. If a match isn't used then those malloc'd blocks won't be
freed.
* coffgen.c (coff_real_object_p): Free malloc'd memory on target
match too.
While handling the comments of Tom related to
[RFC] Have an option to tell GDB to detect and possibly handle mismatched exec-files.
https://sourceware.org/ml/gdb-patches/2019-12/msg00621.html
I saw that GDB warnings are produced ignoring the given styles.
This patch:
* ensures that style markups are properly handled by "warning".
* changes 'set/show data-directory' so that file style is used
in warnings and in 'show message'
* changes all other messages in top.c to use file style when appropriate.
* Uses the above data-directory changes in gdb.base/style.exp
2020-01-03 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* ui-file.c (stdio_file::can_emit_style_escape)
(tee_file::can_emit_style_escape): Ensure style is used also on
gdb_stderr when gdb_stderr is a tty supporting styling, similarly
to gdb_stdout.
* main.c (set_gdb_data_directory): Use file style to output the
warning that the given pathname is not a directory.
* top.c (show_history_filename, gdb_safe_append_history)
(show_gdb_datadir): Use file style.
2020-01-03 Philippe Waroquiers <philippe.waroquiers@skynet.be>
* gdb.base/style.exp: Test that warnings are styled.
Nick Clifton [Fri, 3 Jan 2020 16:17:53 +0000 (16:17 +0000)]
Fix potential illegal memory access when parsing a corrupt PEF format file.
PR 25307
(bfd_pef_parse_function_stubs): Correct the test that ensures that
there is enough data remaining in the code buffer before
attempting to read a function stub.
projects / deliverable / binutils-gdb.git / log
