Always print "Detaching after fork from child..."
Commit Message
On 01/24/2018 08:56 PM, Sergio Durigan Junior wrote:
> On Wednesday, January 24 2018, Jan Kratochvil wrote:
>
>> On Wed, 24 Jan 2018 20:47:14 +0100, Sergio Durigan Junior wrote:
>>> https://bugzilla.redhat.com/show_bug.cgi?id=235197
>>
>> As a justification for this patch:
>>
>> ------------------------------------------------------------------------------
>> cat >fork2.c <<EOH
>> #include <unistd.h>
>> #include <stdio.h>
>> static void printit(void) {
>> puts("printed");
>> }
>> int main(void) {
>> if (!fork()) printit();
>> return 0;
>> }
>> EOH
>> gcc -o fork2 fork2.c -Wall -g
>> gdb -q ./fork2
>> (gdb) b printit
>> Breakpoint 1 at 0x40052b: file fork2.c, line 4.
>> (gdb) r
>> Starting program: /quad/home/jkratoch/t/fork2
>> printed
>> [Inferior 1 (process 15812) exited normally]
>> (gdb) q
>> ------------------------------------------------------------------------------
>>
>> As the GDB user does not expect the program could do any forks s/he is
>> confused the breakpoint did not get hit and assumes GDB is just broken.
>
> Thanks for the extra justification and the useful example.
>
>> But then I cannot say this patch is too great, it produces many uninteresting
>> Detaching after fork from child process 24905.
>> messages rather just annoying in most cases. So nowadays I feel the message
>> is more an excuse how to show it is user's fault s/he did not read it.
>> But I think nobody reads them as there are too many such messages.
>
> I understand where you're coming from, but I still think this is a good
> patch because I read the messages, and as I said, they even helped me in
> one occasion.
I'm of two minds here. On the one hand, "help, GDB doesn't stop at
my breakpoint!" is a FAQ on IRC. OTOH, I also worry a bit about
printing too many messages.
Maybe the middle ground is having a way to toggle this output on/off
other than the over-broad "set verbose". An idea would be to make it
conditional on the existing "set print inferior-events" instead, and
flip "set print inferior-events" on by default, like
"set print thread-events" is on by default. One trouble with the
"set print inferior-events" setting is that currently it prints
redundant info, like:
~~
[Inferior 1 (process 2629) exited normally]
[Inferior 2629 exited]
~~
But we can fix that, I think. See a quick straw-man patch below.
(Really just a a straw-man; there's more redundancy if you do "follow-fork child",
for example.)
>
>> I believe the right fix would be to make "set detach-on-fork off" the default.
>> But that is sure a new can of worms I do not want to speculate about.
>
> Yeah, I can see the rationale for this, and I think it's an idea worth
> discussing. Not sure what others think, but I also don't want deviate
> much from what the current patch is proposing.
Yeah, I'm wondered about doing that too, though I don't think we're ready
for that yet. For example, in all-stop mode, execution stops whenever one of
the children exit. The user then has to switch back to the parent, and
resume it. (Something Tromey was looking at a while back.) Once things
like that are sorted out, then we can reconsider. It may be a useful
exercise to add "set detach-on-fork off" in one's ~/.gdb_init. Until that
works neatly, we can't consider flipping the default, IMO.
Thanks,
Pedro Alves
From c222c0b2f1bce4b23245681222461df9e75e497e Mon Sep 17 00:00:00 2001
From: Pedro Alves <palves@redhat.com>
Date: Thu, 25 Jan 2018 15:42:37 +0000
Subject: [PATCH] inferior-events
---
gdb/inferior.c | 7 ++-----
gdb/infrun.c | 9 ++++++---
2 files changed, 8 insertions(+), 8 deletions(-)
Comments
On Thursday, January 25 2018, Pedro Alves wrote:
> On 01/24/2018 08:56 PM, Sergio Durigan Junior wrote:
>> On Wednesday, January 24 2018, Jan Kratochvil wrote:
>>
>>> On Wed, 24 Jan 2018 20:47:14 +0100, Sergio Durigan Junior wrote:
>>>> https://bugzilla.redhat.com/show_bug.cgi?id=235197
>>>
>>> As a justification for this patch:
>>>
>>> ------------------------------------------------------------------------------
>>> cat >fork2.c <<EOH
>>> #include <unistd.h>
>>> #include <stdio.h>
>>> static void printit(void) {
>>> puts("printed");
>>> }
>>> int main(void) {
>>> if (!fork()) printit();
>>> return 0;
>>> }
>>> EOH
>>> gcc -o fork2 fork2.c -Wall -g
>>> gdb -q ./fork2
>>> (gdb) b printit
>>> Breakpoint 1 at 0x40052b: file fork2.c, line 4.
>>> (gdb) r
>>> Starting program: /quad/home/jkratoch/t/fork2
>>> printed
>>> [Inferior 1 (process 15812) exited normally]
>>> (gdb) q
>>> ------------------------------------------------------------------------------
>>>
>>> As the GDB user does not expect the program could do any forks s/he is
>>> confused the breakpoint did not get hit and assumes GDB is just broken.
>>
>> Thanks for the extra justification and the useful example.
>>
>>> But then I cannot say this patch is too great, it produces many uninteresting
>>> Detaching after fork from child process 24905.
>>> messages rather just annoying in most cases. So nowadays I feel the message
>>> is more an excuse how to show it is user's fault s/he did not read it.
>>> But I think nobody reads them as there are too many such messages.
>>
>> I understand where you're coming from, but I still think this is a good
>> patch because I read the messages, and as I said, they even helped me in
>> one occasion.
>
> I'm of two minds here. On the one hand, "help, GDB doesn't stop at
> my breakpoint!" is a FAQ on IRC. OTOH, I also worry a bit about
> printing too many messages.
>
> Maybe the middle ground is having a way to toggle this output on/off
> other than the over-broad "set verbose". An idea would be to make it
> conditional on the existing "set print inferior-events" instead, and
> flip "set print inferior-events" on by default, like
> "set print thread-events" is on by default. One trouble with the
> "set print inferior-events" setting is that currently it prints
> redundant info, like:
>
> ~~
> [Inferior 1 (process 2629) exited normally]
> [Inferior 2629 exited]
> ~~
>
> But we can fix that, I think. See a quick straw-man patch below.
> (Really just a a straw-man; there's more redundancy if you do "follow-fork child",
> for example.)
Thanks for the reply and the patch. It is my understanding that you
would like it to be extended in order to avoid redundancy in other
cases. I will take a look at it.
On 01/25/2018 08:20 PM, Sergio Durigan Junior wrote:
> On Thursday, January 25 2018, Pedro Alves wrote:
>> But we can fix that, I think. See a quick straw-man patch below.
>> (Really just a a straw-man; there's more redundancy if you do "follow-fork child",
>> for example.)
>
> Thanks for the reply and the patch. It is my understanding that you
> would like it to be extended in order to avoid redundancy in other
> cases. I will take a look at it.
Yes, if this is the direction to go then we'll need to make sure
that the output of both "set print inferior-events off/on" looks
reasonable in the multiple use cases that spawn / attach /
fork / detach / kill etc. inferiors, with both native and
remote debugging.
Thanks,
Pedro Alves
@@ -47,7 +47,7 @@ static int highest_inferior_num;
/* Print notices on inferior events (attach, detach, etc.), set with
`set print inferior-events'. */
-static int print_inferior_events = 0;
+int print_inferior_events = 1;
/* The Current Inferior. This is a strong reference. I.e., whenever
an inferior is the current inferior, its refcount is
@@ -232,9 +232,6 @@ exit_inferior (int pid)
struct inferior *inf = find_inferior_pid (pid);
exit_inferior_1 (inf, 0);
-
- if (print_inferior_events)
- printf_unfiltered (_("[Inferior %d exited]\n"), pid);
}
void
@@ -986,7 +983,7 @@ initialize_inferiors (void)
can only allocate an inferior when all those modules have done
that. Do this after initialize_progspace, due to the
current_program_space reference. */
- current_inferior_ = add_inferior (0);
+ current_inferior_ = add_inferior_silent (0);
current_inferior_->incref ();
current_inferior_->pspace = current_program_space;
current_inferior_->aspace = current_program_space->aspace;
@@ -408,6 +408,9 @@ show_follow_fork_mode_string (struct ui_file *file, int from_tty,
}
+/* Putme in header. */
+extern int print_inferior_events;
+
/* Handle changes to the inferior list based on the type of fork,
which process is being followed, and whether the other process
should be detached. On entry inferior_ptid must be the ptid of
@@ -461,7 +464,7 @@ holding the child stopped. Try \"set detach-on-fork\" or \
remove_breakpoints_pid (ptid_get_pid (inferior_ptid));
}
- if (info_verbose || debug_infrun)
+ if (print_inferior_events)
{
/* Ensure that we have a process ptid. */
ptid_t process_ptid = pid_to_ptid (ptid_get_pid (child_ptid));
@@ -549,7 +552,7 @@ holding the child stopped. Try \"set detach-on-fork\" or \
struct inferior *parent_inf, *child_inf;
struct program_space *parent_pspace;
- if (info_verbose || debug_infrun)
+ if (print_inferior_events)
{
target_terminal::ours_for_output ();
fprintf_filtered (gdb_stdlog,
@@ -594,7 +597,7 @@ holding the child stopped. Try \"set detach-on-fork\" or \
}
else if (detach_fork)
{
- if (info_verbose || debug_infrun)
+ if (print_inferior_events)
{
/* Ensure that we have a process ptid. */
ptid_t process_ptid = pid_to_ptid (ptid_get_pid (child_ptid));