This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] stepi/nexti: skip signal handler if "handle nostop" signal arrives
- From: Pedro Alves <palves at redhat dot com>
- To: Eli Zaretskii <eliz at gnu dot org>
- Cc: gdb-patches at sourceware dot org
- Date: Mon, 27 Oct 2014 20:34:43 +0000
- Subject: Re: [PATCH] stepi/nexti: skip signal handler if "handle nostop" signal arrives
- Authentication-results: sourceware.org; auth=none
- References: <1413308910-30423-1-git-send-email-palves at redhat dot com> <83ppdu5wx7 dot fsf at gnu dot org> <543D7044 dot 2000703 at redhat dot com> <83oate5uec dot fsf at gnu dot org> <543E7961 dot 5090002 at redhat dot com> <838ukh5rry dot fsf at gnu dot org> <544BAD08 dot 1050601 at redhat dot com> <83tx2p2z12 dot fsf at gnu dot org> <544EA096 dot 5050803 at redhat dot com> <83oasx2sjv dot fsf at gnu dot org>
On 10/27/2014 07:58 PM, Eli Zaretskii wrote:
>>>> +@cindex stepping into signal handlers
>>>> +@anchor{stepping into signal handlers}
>>>
>>> I would remove this @cindex entry: it doesn't add anything useful to
>>> the previous one, and will likely point to the same page.
>>
>> I'd prefer to keep it, if you don't mind. The queue-signal reference wants
>> to point here directly, and I can imagine the text above expanding in
>> directions not relevant for that cross reference.
>
> Both entries use the same words, so why do we need both? There's one
> paragraph of only a dozen of text lines between them.
I was thinking that "stepping into signal handler" was a concept
on its own. But I can certainly live without it; removed in the
final version.
>>>> +thread resumes (@pxref{Signaling, ,Giving your Program a Signal}),
>>>> +then a stepping command steps into the signal's handler.
>>>
>>> Not sure I understand the sequence here. First, I queue-signal, then
>>> the signal is delivered and the thread stops, and _then_ I issue si?
>>> I guess the "when execution of the thread resumes" confused me.
>>
>> Sounds like you're thinking of "queue-signal" like "kill" from the shell,
>> but that's not how "queue-signal" works. "queue-signal" instead
>> passes the signal to the program immediately as if the thread had
>> _already_ stopped for the signal. GDB doesn't intercept
>> the signal anymore.
>
> How is this different from what I wrote? The program behaves as if
> the signal was delivered to it, right?
Yes, but the "then the signal is delivered and the thread stops" part
doesn't exist. The thread is (and must be) already stopped.
>> I've added a queue-signal example. Hopefully that makes things clearer.
>
> It does, thanks.
>
>> Let me know how this version looks.
>
> LGTM.
Great, I've pushed it in then, as below.
Thanks!
>From e5f8a7cc2d376c81749b6e4a4efc034201cf683c Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Mon, 27 Oct 2014 20:24:59 +0000
Subject: [PATCH] stepi/nexti: skip signal handler if "handle nostop" signal
arrives
I noticed that "si" behaves differently when a "handle nostop" signal
arrives while the step is in progress, depending on whether the
program was stopped at a breakpoint when "si" was entered.
Specifically, in case GDB needs to step off a breakpoint, the handler
is skipped and the program stops in the next "mainline" instruction.
Otherwise, the "si" stops in the first instruction of the signal
handler.
I was surprised the testsuite doesn't catch this difference. Turns
out gdb.base/sigstep.exp covers a bunch of cases related to stepping
and signal handlers, but does not test stepi nor nexti, only
step/next/continue.
My first reaction was that stopping in the signal handler was the
correct thing to do, as it's where the next user-visible instruction
that is executed is. I considered then "nexti" -- a signal handler
could be reasonably considered a subroutine call to step over, it'd
seem intuitive to me that "nexti" would skip it.
But then, I realized that signals that arrive while a plain/line
"step" is in progress _also_ have their handler skipped. A user might
well be excused for being confused by this, given:
(gdb) help step
Step program until it reaches a different source line.
And the signal handler's sources will be in different source lines,
after all.
I think that having to explain that "stepi" steps into handlers, (and
that "nexti" wouldn't according to my reasoning above), while "step"
does not, is a sign of an awkward interface.
E.g., if a user truly is interested in stepping into signal handlers,
then it's odd that she has to either force the signal to "handle
stop", or recall to do "stepi" whenever such a signal might be
delivered. For that use case, it'd seem nicer to me if "step" also
stepped into handlers.
This suggests to me that we either need a global "step-into-handlers"
setting, or perhaps better, make "handle pass/nopass stop/nostop
print/noprint" have have an additional axis - "handle
stepinto/nostepinto", so that the user could configure whether
handlers for specific signals should be stepped into.
In any case, I think it's simpler (and thus better) for all step
commands to behave the same. This commit thus makes "si/ni" skip
handlers for "handle nostop" signals that arrive while the command was
already in progress, like step/next do.
To be clear, nothing changes if the program was stopped for a signal,
and the user enters a stepping command _then_ -- GDB still steps into
the handler. The change concerns signals that don't cause a stop and
that arrive while the step is in progress.
Tested on x86_64 Fedora 20, native and gdbserver.
gdb/
2014-10-27 Pedro Alves <palves@redhat.com>
* infrun.c (handle_signal_stop): Also skip handlers when a random
signal arrives while handling a "stepi" or a "nexti". Set the
thread's 'step_after_step_resume_breakpoint' flag.
gdb/doc/
2014-10-27 Pedro Alves <palves@redhat.com>
* gdb.texinfo (Continuing and Stepping): Add cross reference to
info on stepping and signal handlers.
(Signals): Explain stepping and signal handlers. Add context
index entry, and cross references.
gdb/testsuite/
2014-10-27 Pedro Alves <palves@redhat.com>
* gdb.base/sigstep.c (dummy): New global.
(main): Issue a couple writes to the new global.
* gdb.base/sigstep.exp (get_next_pc, test_skip_handler): New
procedures.
(skip_over_handler): Use test_skip_handler.
(top level): Call skip_over_handler for stepi and nexti too.
(breakpoint_over_handler): Use test_skip_handler.
(top level): Call breakpoint_over_handler for stepi and nexti too.
---
gdb/ChangeLog | 6 ++++
gdb/doc/ChangeLog | 7 ++++
gdb/testsuite/ChangeLog | 11 +++++++
gdb/doc/gdb.texinfo | 67 +++++++++++++++++++++++++++++++++++++-
gdb/infrun.c | 7 ++--
gdb/testsuite/gdb.base/sigstep.c | 7 ++--
gdb/testsuite/gdb.base/sigstep.exp | 55 +++++++++++++++++++++++--------
7 files changed, 141 insertions(+), 19 deletions(-)
diff --git a/gdb/ChangeLog b/gdb/ChangeLog
index 32f788c..8a34118 100644
--- a/gdb/ChangeLog
+++ b/gdb/ChangeLog
@@ -1,3 +1,9 @@
+2014-10-27 Pedro Alves <palves@redhat.com>
+
+ * infrun.c (handle_signal_stop): Also skip handlers when a random
+ signal arrives while handling a "stepi" or a "nexti". Set the
+ thread's 'step_after_step_resume_breakpoint' flag.
+
2014-10-27 Luis Machado <lgustavo@codesourcery.com>
* arm-tdep.c (INSN_S_L_BIT_NUM): Document.
diff --git a/gdb/doc/ChangeLog b/gdb/doc/ChangeLog
index f1b2329..f60fd8f 100644
--- a/gdb/doc/ChangeLog
+++ b/gdb/doc/ChangeLog
@@ -1,3 +1,10 @@
+2014-10-27 Pedro Alves <palves@redhat.com>
+
+ * gdb.texinfo (Continuing and Stepping): Add cross reference to
+ info on stepping and signal handlers.
+ (Signals): Explain stepping and signal handlers. Add context
+ index entry, and cross references.
+
2014-10-20 Simon Marchi <simon.marchi@ericsson.com>
* python.texi (Breakpoints In Python): Add parenthesis after
diff --git a/gdb/testsuite/ChangeLog b/gdb/testsuite/ChangeLog
index 2eea791..5cc7b0b 100644
--- a/gdb/testsuite/ChangeLog
+++ b/gdb/testsuite/ChangeLog
@@ -1,3 +1,14 @@
+2014-10-27 Pedro Alves <palves@redhat.com>
+
+ * gdb.base/sigstep.c (dummy): New global.
+ (main): Issue a couple writes to the new global.
+ * gdb.base/sigstep.exp (get_next_pc, test_skip_handler): New
+ procedures.
+ (skip_over_handler): Use test_skip_handler.
+ (top level): Call skip_over_handler for stepi and nexti too.
+ (breakpoint_over_handler): Use test_skip_handler.
+ (top level): Call breakpoint_over_handler for stepi and nexti too.
+
2014-10-27 Yao Qi <yao@codesourcery.com>
* gdb.trace/tfile.c (adjust_function_address)
diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
index a1b8ac7..15c2908 100644
--- a/gdb/doc/gdb.texinfo
+++ b/gdb/doc/gdb.texinfo
@@ -5079,7 +5079,9 @@ line of source code, or one machine instruction (depending on what
particular command you use). Either when continuing or when stepping,
your program may stop even sooner, due to a breakpoint or a signal. (If
it stops due to a signal, you may want to use @code{handle}, or use
-@samp{signal 0} to resume execution. @xref{Signals, ,Signals}.)
+@samp{signal 0} to resume execution (@pxref{Signals, ,Signals}),
+or you may step into the signal's handler (@pxref{stepping and signal
+handlers}).)
@table @code
@kindex continue
@@ -5573,6 +5575,66 @@ a result of the fatal signal once it saw the signal. To prevent this,
you can continue with @samp{signal 0}. @xref{Signaling, ,Giving your
Program a Signal}.
+@cindex stepping and signal handlers
+@anchor{stepping and signal handlers}
+
+@value{GDBN} optimizes for stepping the mainline code. If a signal
+that has @code{handle nostop} and @code{handle pass} set arrives while
+a stepping command (e.g., @code{stepi}, @code{step}, @code{next}) is
+in progress, @value{GDBN} lets the signal handler run and then resumes
+stepping the mainline code once the signal handler returns. In other
+words, @value{GDBN} steps over the signal handler. This prevents
+signals that you've specified as not interesting (with @code{handle
+nostop}) from changing the focus of debugging unexpectedly. Note that
+the signal handler itself may still hit a breakpoint, stop for another
+signal that has @code{handle stop} in effect, or for any other event
+that normally results in stopping the stepping command sooner. Also
+note that @value{GDBN} still informs you that the program received a
+signal if @code{handle print} is set.
+
+@anchor{stepping into signal handlers}
+
+If you set @code{handle pass} for a signal, and your program sets up a
+handler for it, then issuing a stepping command, such as @code{step}
+or @code{stepi}, when your program is stopped due to the signal will
+step @emph{into} the signal handler (if the target supports that).
+
+Likewise, if you use the @code{queue-signal} command to queue a signal
+to be delivered to the current thread when execution of the thread
+resumes (@pxref{Signaling, ,Giving your Program a Signal}), then a
+stepping command will step into the signal handler.
+
+Here's an example, using @code{stepi} to step to the first instruction
+of @code{SIGUSR1}'s handler:
+
+@smallexample
+(@value{GDBP}) handle SIGUSR1
+Signal Stop Print Pass to program Description
+SIGUSR1 Yes Yes Yes User defined signal 1
+(@value{GDBP}) c
+Continuing.
+
+Program received signal SIGUSR1, User defined signal 1.
+main () sigusr1.c:28
+28 p = 0;
+(@value{GDBP}) si
+sigusr1_handler () at sigusr1.c:9
+9 @{
+@end smallexample
+
+The same, but using @code{queue-signal} instead of waiting for the
+program to receive the signal first:
+
+@smallexample
+(@value{GDBP}) n
+28 p = 0;
+(@value{GDBP}) queue-signal SIGUSR1
+(@value{GDBP}) si
+sigusr1_handler () at sigusr1.c:9
+9 @{
+(@value{GDBP})
+@end smallexample
+
@cindex extra signal information
@anchor{extra signal information}
@@ -16654,6 +16716,9 @@ be used to pass a signal whose handling state has been set to @code{nopass}
@end table
@c @end group
+@xref{stepping into signal handlers}, for information on how stepping
+commands behave when the thread has a signal queued.
+
@node Returning
@section Returning from a Function
diff --git a/gdb/infrun.c b/gdb/infrun.c
index 90a3123..df053e2 100644
--- a/gdb/infrun.c
+++ b/gdb/infrun.c
@@ -4463,9 +4463,9 @@ handle_signal_stop (struct execution_control_state *ecs)
return;
}
- if (ecs->event_thread->control.step_range_end != 0
- && ecs->event_thread->suspend.stop_signal != GDB_SIGNAL_0
- && pc_in_thread_step_range (stop_pc, ecs->event_thread)
+ if (ecs->event_thread->suspend.stop_signal != GDB_SIGNAL_0
+ && (pc_in_thread_step_range (stop_pc, ecs->event_thread)
+ || ecs->event_thread->control.step_range_end == 1)
&& frame_id_eq (get_stack_frame_id (frame),
ecs->event_thread->control.step_stack_frame_id)
&& ecs->event_thread->control.step_resume_breakpoint == NULL)
@@ -4485,6 +4485,7 @@ handle_signal_stop (struct execution_control_state *ecs)
"single-step range\n");
insert_hp_step_resume_breakpoint_at_frame (frame);
+ ecs->event_thread->step_after_step_resume_breakpoint = 1;
/* Reset trap_expected to ensure breakpoints are re-inserted. */
ecs->event_thread->control.trap_expected = 0;
keep_going (ecs);
diff --git a/gdb/testsuite/gdb.base/sigstep.c b/gdb/testsuite/gdb.base/sigstep.c
index aa2384a..cc69184 100644
--- a/gdb/testsuite/gdb.base/sigstep.c
+++ b/gdb/testsuite/gdb.base/sigstep.c
@@ -24,6 +24,7 @@
#include <errno.h>
static volatile int done;
+static volatile int dummy;
static void
handler (int sig)
@@ -74,8 +75,10 @@ main ()
return 1;
}
}
- /* Wait. */
- while (!done);
+ /* Wait. Issue a couple writes to a dummy volatile var to be
+ reasonably sure our simple "get-next-pc" logic doesn't
+ stumble on branches. */
+ dummy = 0; dummy = 0; while (!done);
done = 0;
}
return 0;
diff --git a/gdb/testsuite/gdb.base/sigstep.exp b/gdb/testsuite/gdb.base/sigstep.exp
index 184d46e..53152b8 100644
--- a/gdb/testsuite/gdb.base/sigstep.exp
+++ b/gdb/testsuite/gdb.base/sigstep.exp
@@ -269,9 +269,38 @@ proc skip_to_handler_entry { i } {
gdb_test "clear *handler" ".*" "$prefix; clear handler"
}
-skip_to_handler_entry step
-skip_to_handler_entry next
-skip_to_handler_entry continue
+foreach cmd {"stepi" "nexti" "step" "next" "continue"} {
+ skip_to_handler_entry $cmd
+}
+
+# Get the address of where a single-step should land.
+
+proc get_next_pc {test} {
+ global gdb_prompt
+ global hex
+
+ set next ""
+ gdb_test_multiple "x/2i \$pc" $test {
+ -re "$hex .*:\[^\r\n\]+\r\n\[ \]+($hex).*\.\r\n$gdb_prompt $" {
+ set next $expect_out(1,string)
+ pass $test
+ }
+ }
+
+ return $next
+}
+
+# Test that the command skipped over the handler.
+
+proc test_skip_handler {prefix i} {
+ if {$i == "stepi" || $i == "nexti"} {
+ set next_pc [get_next_pc "$prefix; get next PC"]
+ gdb_test "$i" "dummy = 0.*" "$prefix; performing $i"
+ gdb_test "p /x \$pc" " = $next_pc" "$prefix; advanced"
+ } else {
+ gdb_test "$i" "done = 0.*" "$prefix; performing $i"
+ }
+}
# Try stepping when there's a signal pending but no breakpoints.
# Should skip the handler advancing to the next line.
@@ -295,13 +324,13 @@ proc skip_over_handler { i } {
# Make the signal pending
sleep 1
-
- gdb_test "$i" "done = 0.*" "$prefix; performing $i"
+
+ test_skip_handler $prefix $i
}
-skip_over_handler step
-skip_over_handler next
-skip_over_handler continue
+foreach cmd {"stepi" "nexti" "step" "next" "continue"} {
+ skip_over_handler $cmd
+}
# Try stepping when there's a signal pending, a pre-existing
# breakpoint at the current instruction, and a breakpoint in the
@@ -385,7 +414,7 @@ breakpoint_to_handler_entry continue
# Try stepping when there's a signal pending, and a pre-existing
# breakpoint at the current instruction, and no breakpoint in the
-# handler. Should advance to the next line.
+# handler. Should advance to the next line/instruction.
proc breakpoint_over_handler { i } {
global gdb_prompt
@@ -409,10 +438,10 @@ proc breakpoint_over_handler { i } {
# Make the signal pending
sleep 1
- gdb_test "$i" "done = 0.*" "$prefix; performing $i"
+ test_skip_handler $prefix $i
gdb_test "clear $infinite_loop" ".*" "$prefix; clear infinite loop"
}
-breakpoint_over_handler step
-breakpoint_over_handler next
-breakpoint_over_handler continue
+foreach cmd {"stepi" "nexti" "step" "next" "continue"} {
+ breakpoint_over_handler $cmd
+}
--
1.9.3