From: Pedro Alves <pedro@codesourcery.com>
To: Vladimir Prus <vladimir@codesourcery.com>
Cc: gdb-patches@sourceware.org
Subject: Re: [1/3] broken -thread-info output in non-stop mode, suppress_stop_observer
Date: Tue, 17 Mar 2009 17:27:00 -0000 [thread overview]
Message-ID: <200903171724.54512.pedro@codesourcery.com> (raw)
In-Reply-To: <200903171026.31197.vladimir@codesourcery.com>
On Tuesday 17 March 2009 07:26:29, Vladimir Prus wrote:
> On Tuesday 17 March 2009 09:21:02 Pedro Alves wrote:
> > This patch removes the suppress_stop_observer global. As indicated
> > in the comments I'm removing, this was problematic for non-stop
> > mode.
> >
> > To fix the broken -thread-info output mentioned here:
> >
> > http://sourceware.org/ml/gdb-patches/2009-03/msg00236.html
> >
> > ... I wanted to add a thread_info->in_infcall flag to use in
> > the 3rd patch in the series. That meant I needed to get rid of
> > the suppress_resume_observer global, which then brings
> > us to this.
> >
> > Vladimir, et al, how does this look?
>
> I did not look very carefully at details of conditions, but the
> general idea looks solid. One detail is the new command you
> add in infrun.c -- can it explicitly write down *why* we need to
> suppress *stopped at this point, when doing finish?
Certainly, here it is. Thank you.
Hmmm, I do wonder if we could rewrite "finish" to not rely on
bpstat_find_breakpoint after normal_stop, but to instead add a new
BPSTAT_WHAT_FINISH and handle it from within handle_inferior_event's
internal breakpoint handling switch (where we handle
BPSTAT_WHAT_STEP_RESUME, etc.). Since that runs before normal_stop,
we wouldn't have to suppress the normal_stop observer.
?
--
Pedro Alves
2009-03-17 Pedro Alves <pedro@codesourcery.com>
* infcall.c (run_inferior_call): Remove references to
suppress_stop_observer.
* infcmd.c (suppress_stop_observer): Delete.
(finish_command_continuation): Remove NOTE. Don't clear
suppress_stop_observer anymore.
(finish_command_continuation_free_arg): Likewise.
(finish_forward): Remove references to suppress_stop_observer.
Call normal_stop observer if we haven't already.
* inferior.h (suppress_stop_observer): Delete.
* infrun.c (normal_stop): When deciding to suppress the
normal_stop observer, check for proceed_to_finish instead of
suppress_stop_observer.
---
gdb/infcall.c | 3 ---
gdb/infcmd.c | 24 ++++++++----------------
gdb/inferior.h | 4 ----
gdb/infrun.c | 24 +++++++++++++++++++-----
4 files changed, 27 insertions(+), 28 deletions(-)
Index: src/gdb/infcall.c
===================================================================
--- src.orig/gdb/infcall.c 2009-03-17 05:25:17.000000000 +0000
+++ src/gdb/infcall.c 2009-03-17 16:14:06.000000000 +0000
@@ -331,7 +331,6 @@ run_inferior_call (struct thread_info *c
volatile struct gdb_exception e;
int saved_async = 0;
int saved_suppress_resume_observer = suppress_resume_observer;
- int saved_suppress_stop_observer = suppress_stop_observer;
ptid_t call_thread_ptid = call_thread->ptid;
char *saved_target_shortname = xstrdup (target_shortname);
@@ -344,7 +343,6 @@ run_inferior_call (struct thread_info *c
saved_async = target_async_mask (0);
suppress_resume_observer = 1;
- suppress_stop_observer = 1;
TRY_CATCH (e, RETURN_MASK_ALL)
proceed (real_pc, TARGET_SIGNAL_0, 0);
@@ -355,7 +353,6 @@ run_inferior_call (struct thread_info *c
call_thread = NULL;
suppress_resume_observer = saved_suppress_resume_observer;
- suppress_stop_observer = saved_suppress_stop_observer;
/* Don't restore the async mask if the target has changed,
saved_async is for the original target. */
Index: src/gdb/infcmd.c
===================================================================
--- src.orig/gdb/infcmd.c 2009-03-17 05:25:19.000000000 +0000
+++ src/gdb/infcmd.c 2009-03-17 16:14:06.000000000 +0000
@@ -169,8 +169,6 @@ struct gdb_environ *inferior_environ;
/* When set, no calls to target_resumed observer will be made. */
int suppress_resume_observer = 0;
-/* When set, normal_stop will not call the normal_stop observer. */
-int suppress_stop_observer = 0;
\f
/* Accessor routines. */
@@ -1346,13 +1344,16 @@ static void
finish_command_continuation (void *arg)
{
struct finish_command_continuation_args *a = arg;
-
+ struct thread_info *tp = NULL;
bpstat bs = NULL;
if (!ptid_equal (inferior_ptid, null_ptid)
&& target_has_execution
&& is_stopped (inferior_ptid))
- bs = inferior_thread ()->stop_bpstat;
+ {
+ tp = inferior_thread ();
+ bs = tp->stop_bpstat;
+ }
if (bpstat_find_breakpoint (bs, a->breakpoint) != NULL
&& a->function != NULL)
@@ -1369,22 +1370,15 @@ finish_command_continuation (void *arg)
}
/* We suppress normal call of normal_stop observer and do it here so
- that that *stopped notification includes the return value. */
- /* NOTE: This is broken in non-stop mode. There is no guarantee the
- next stop will be in the same thread that we started doing a
- finish on. This suppressing (or some other replacement means)
- should be a thread property. */
- observer_notify_normal_stop (bs, 1 /* print frame */);
- suppress_stop_observer = 0;
+ that the *stopped notification includes the return value. */
+ if (bs != NULL && tp->proceed_to_finish)
+ observer_notify_normal_stop (bs, 1 /* print frame */);
delete_breakpoint (a->breakpoint);
}
static void
finish_command_continuation_free_arg (void *arg)
{
- /* NOTE: See finish_command_continuation. This would go away, if
- this suppressing is made a thread property. */
- suppress_stop_observer = 0;
xfree (arg);
}
@@ -1469,8 +1463,6 @@ finish_forward (struct symbol *function,
old_chain = make_cleanup_delete_breakpoint (breakpoint);
tp->proceed_to_finish = 1; /* We want stop_registers, please... */
- make_cleanup_restore_integer (&suppress_stop_observer);
- suppress_stop_observer = 1;
proceed ((CORE_ADDR) -1, TARGET_SIGNAL_DEFAULT, 0);
cargs = xmalloc (sizeof (*cargs));
Index: src/gdb/inferior.h
===================================================================
--- src.orig/gdb/inferior.h 2009-03-17 05:25:19.000000000 +0000
+++ src/gdb/inferior.h 2009-03-17 16:14:06.000000000 +0000
@@ -368,10 +368,6 @@ extern int debug_displaced;
void displaced_step_dump_bytes (struct ui_file *file,
const gdb_byte *buf, size_t len);
-
-/* When set, normal_stop will not call the normal_stop observer. */
-extern int suppress_stop_observer;
-
/* When set, no calls to target_resumed observer will be made. */
extern int suppress_resume_observer;
Index: src/gdb/infrun.c
===================================================================
--- src.orig/gdb/infrun.c 2009-03-17 05:25:21.000000000 +0000
+++ src/gdb/infrun.c 2009-03-17 16:24:17.000000000 +0000
@@ -4433,11 +4433,25 @@ Further execution is probably impossible
done:
annotate_stopped ();
- if (!suppress_stop_observer
- && !(target_has_execution
- && last.kind != TARGET_WAITKIND_SIGNALLED
- && last.kind != TARGET_WAITKIND_EXITED
- && inferior_thread ()->step_multi))
+
+ /* Suppress the stop observer if we're in the middle of:
+
+ - a step n (n > 1), as there still more steps to be done.
+
+ - a "finish" command, as the observer will be called in
+ finish_command_continuation, so it can include the inferior
+ function's return value.
+
+ - calling an inferior function, as we pretend we inferior didn't
+ run at all. The return value of the call is handled by the
+ expression evaluator, through call_function_by_hand. */
+
+ if (!target_has_execution
+ || last.kind == TARGET_WAITKIND_SIGNALLED
+ || last.kind == TARGET_WAITKIND_EXITED
+ || (!inferior_thread ()->step_multi
+ && !(inferior_thread ()->stop_bpstat
+ && inferior_thread ()->proceed_to_finish)))
{
if (!ptid_equal (inferior_ptid, null_ptid))
observer_notify_normal_stop (inferior_thread ()->stop_bpstat,
next prev parent reply other threads:[~2009-03-17 17:25 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-03-17 6:23 Pedro Alves
2009-03-17 7:29 ` Vladimir Prus
2009-03-17 17:27 ` Pedro Alves [this message]
2009-03-18 15:16 ` Joel Brobecker
2009-03-22 21:24 ` Pedro Alves
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200903171724.54512.pedro@codesourcery.com \
--to=pedro@codesourcery.com \
--cc=gdb-patches@sourceware.org \
--cc=vladimir@codesourcery.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox