Wrong Reverse result

Wrong Reverse result

[Marc Khouzam]

Hi,

I'm getting a weird behaviour in reverse, where GDB
jumps to far backwards.

I have GDB HEAD from two weeks ago with the three
missing PRrecord patches applied (that were pinged
by Hui this very week).

I'm running on Ubuntu.  
Is it me or can someone reproduce it?
(It happened during my presentation at the GCC summit,
so I have witnesses :-))

Thanks

Marc

GNU gdb (GDB) 6.8.50.20090522-cvs
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later
<http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show
copying"
and "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
(gdb) l
1	void foo() {
2	    return;
3	}
4	
5	int main() {
6	
7		foo(); // 1
8		foo(); // 2
9		foo(); // 3
10	
(gdb) l
11		return 0;
12	}
(gdb) start
Temporary breakpoint 1 at 0x80483f7: file r.c, line 7.
Starting program: /home/marc/testing/a.out 

Temporary breakpoint 1, main () at r.c:7
7		foo(); // 1
Current language:  auto; currently c++
(gdb) record
(gdb) n
8		foo(); // 2
(gdb) n
9		foo(); // 3
(gdb) n
11		return 0;
(gdb) rn
9		foo(); // 3
(gdb) s
foo () at r.c:3
3	}
(gdb) rn

No more reverse-execution history.
main () at r.c:7
7		foo(); // 1

======== This is weird.  Why does it go back to the top?

Re: Wrong Reverse result

[Hui Zhu]

I have reproduced this issue and find out the cause.

infrun: TARGET_WAITKIND_STOPPED
infrun: stop_pc = 0x8048361
infrun: stepped into subroutine
infrun: inserting step-resume breakpoint at 0x8048349

stop_stepping is better.

infrun: resume (step=0, signal=0), trap_expected=0
infrun: prepare_to_wait
infrun: target_wait (-1, status) =
infrun:   10798 [process 10798],
infrun:   status->kind = no-history
infrun: infwait_normal_state

No more reverse-execution history.
infrun: stop_stepping
main () at 1.c:7

On Thu, Jun 11, 2009 at 00:49, Marc Khouzam<marc.khouzam@ericsson.com> wrote:
> Hi,
>
> I'm getting a weird behaviour in reverse, where GDB
> jumps to far backwards.
>
> I have GDB HEAD from two weeks ago with the three
> missing PRrecord patches applied (that were pinged
> by Hui this very week).

All of patches are not for prec, they are for reverse.

>
> I'm running on Ubuntu.
> Is it me or can someone reproduce it?
> (It happened during my presentation at the GCC summit,
> so I have witnesses :-))

I am not sure Michael in GCC summit or not.  If so, you can let he fix
it at there directly.  :)

Thanks for your help.

Hui

>
> Thanks
>
> Marc
>
> GNU gdb (GDB) 6.8.50.20090522-cvs
> Copyright (C) 2009 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later
> <http://gnu.org/licenses/gpl.html>
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law.  Type "show
> copying"
> and "show warranty" for details.
> This GDB was configured as "i686-pc-linux-gnu".
> For bug reporting instructions, please see:
> <http://www.gnu.org/software/gdb/bugs/>...
> (gdb) l
> 1       void foo() {
> 2           return;
> 3       }
> 4
> 5       int main() {
> 6
> 7               foo(); // 1
> 8               foo(); // 2
> 9               foo(); // 3
> 10
> (gdb) l
> 11              return 0;
> 12      }
> (gdb) start
> Temporary breakpoint 1 at 0x80483f7: file r.c, line 7.
> Starting program: /home/marc/testing/a.out
>
> Temporary breakpoint 1, main () at r.c:7
> 7               foo(); // 1
> Current language:  auto; currently c++
> (gdb) record
> (gdb) n
> 8               foo(); // 2
> (gdb) n
> 9               foo(); // 3
> (gdb) n
> 11              return 0;
> (gdb) rn
> 9               foo(); // 3
> (gdb) s
> foo () at r.c:3
> 3       }
> (gdb) rn
>
> No more reverse-execution history.
> main () at r.c:7
> 7               foo(); // 1
>
> ======== This is weird.  Why does it go back to the top?
>

Re: Wrong Reverse result

[Hui Zhu]

Post a patch for it.
http://sourceware.org/ml/gdb-patches/2009-06/msg00291.html

Hui

On Thu, Jun 11, 2009 at 13:33, Hui Zhu<teawater@gmail.com> wrote:
> I have reproduced this issue and find out the cause.
>
> infrun: TARGET_WAITKIND_STOPPED
> infrun: stop_pc = 0x8048361
> infrun: stepped into subroutine
> infrun: inserting step-resume breakpoint at 0x8048349
>
> stop_stepping is better.
>
> infrun: resume (step=0, signal=0), trap_expected=0
> infrun: prepare_to_wait
> infrun: target_wait (-1, status) =
> infrun:   10798 [process 10798],
> infrun:   status->kind = no-history
> infrun: infwait_normal_state
>
> No more reverse-execution history.
> infrun: stop_stepping
> main () at 1.c:7
>
> On Thu, Jun 11, 2009 at 00:49, Marc Khouzam<marc.khouzam@ericsson.com> wrote:
>> Hi,
>>
>> I'm getting a weird behaviour in reverse, where GDB
>> jumps to far backwards.
>>
>> I have GDB HEAD from two weeks ago with the three
>> missing PRrecord patches applied (that were pinged
>> by Hui this very week).
>
> All of patches are not for prec, they are for reverse.
>
>>
>> I'm running on Ubuntu.
>> Is it me or can someone reproduce it?
>> (It happened during my presentation at the GCC summit,
>> so I have witnesses :-))
>
> I am not sure Michael in GCC summit or not.  If so, you can let he fix
> it at there directly.  :)
>
> Thanks for your help.
>
> Hui
>
>>
>> Thanks
>>
>> Marc
>>
>> GNU gdb (GDB) 6.8.50.20090522-cvs
>> Copyright (C) 2009 Free Software Foundation, Inc.
>> License GPLv3+: GNU GPL version 3 or later
>> <http://gnu.org/licenses/gpl.html>
>> This is free software: you are free to change and redistribute it.
>> There is NO WARRANTY, to the extent permitted by law.  Type "show
>> copying"
>> and "show warranty" for details.
>> This GDB was configured as "i686-pc-linux-gnu".
>> For bug reporting instructions, please see:
>> <http://www.gnu.org/software/gdb/bugs/>...
>> (gdb) l
>> 1       void foo() {
>> 2           return;
>> 3       }
>> 4
>> 5       int main() {
>> 6
>> 7               foo(); // 1
>> 8               foo(); // 2
>> 9               foo(); // 3
>> 10
>> (gdb) l
>> 11              return 0;
>> 12      }
>> (gdb) start
>> Temporary breakpoint 1 at 0x80483f7: file r.c, line 7.
>> Starting program: /home/marc/testing/a.out
>>
>> Temporary breakpoint 1, main () at r.c:7
>> 7               foo(); // 1
>> Current language:  auto; currently c++
>> (gdb) record
>> (gdb) n
>> 8               foo(); // 2
>> (gdb) n
>> 9               foo(); // 3
>> (gdb) n
>> 11              return 0;
>> (gdb) rn
>> 9               foo(); // 3
>> (gdb) s
>> foo () at r.c:3
>> 3       }
>> (gdb) rn
>>
>> No more reverse-execution history.
>> main () at r.c:7
>> 7               foo(); // 1
>>
>> ======== This is weird.  Why does it go back to the top?
>>
>

RE: Wrong Reverse result

[Marc Khouzam]

> -----Original Message-----
> From: Hui Zhu [mailto:teawater@gmail.com] 
> Sent: June-11-09 1:34 AM
> To: Marc Khouzam
> Cc: gdb@sourceware.org; Michael Snyder
> Subject: Re: Wrong Reverse result
> 
> I have reproduced this issue and find out the cause.

Great.  Thanks for doing that so quick.
I forgot how to turn on those nice traces.  Can you 
remind me?
The traces don't mean much to me, but I can include them
in any future (if any :-)) issues.

> infrun: TARGET_WAITKIND_STOPPED
> infrun: stop_pc = 0x8048361
> infrun: stepped into subroutine
> infrun: inserting step-resume breakpoint at 0x8048349
> 
> stop_stepping is better.
> 
> infrun: resume (step=0, signal=0), trap_expected=0
> infrun: prepare_to_wait
> infrun: target_wait (-1, status) =
> infrun:   10798 [process 10798],
> infrun:   status->kind = no-history
> infrun: infwait_normal_state
> 
> No more reverse-execution history.
> infrun: stop_stepping
> main () at 1.c:7
> 
> On Thu, Jun 11, 2009 at 00:49, Marc 
> Khouzam<marc.khouzam@ericsson.com> wrote:
> > Hi,
> >
> > I'm getting a weird behaviour in reverse, where GDB
> > jumps to far backwards.
> >
> > I have GDB HEAD from two weeks ago with the three
> > missing PRrecord patches applied (that were pinged
> > by Hui this very week).
> 
> All of patches are not for prec, they are for reverse.

I realized that just after pressing 'send' :-)

> > I'm running on Ubuntu.
> > Is it me or can someone reproduce it?
> > (It happened during my presentation at the GCC summit,
> > so I have witnesses :-))
> 
> I am not sure Michael in GCC summit or not.  If so, you can let he fix
> it at there directly.  :)

The summit is over now, but I did meet with Michael.  It's nice to see
people face to face.  I was able to put a bit of pressure on him
about reviewing the missing patches ;-).  I'm sure that as soon as he
has time, he'll have a look at these four things.

Thanks Hui

marc

> 
> Thanks for your help.
> 
> Hui
> 
> >
> > Thanks
> >
> > Marc
> >
> > GNU gdb (GDB) 6.8.50.20090522-cvs
> > Copyright (C) 2009 Free Software Foundation, Inc.
> > License GPLv3+: GNU GPL version 3 or later
> > <http://gnu.org/licenses/gpl.html>
> > This is free software: you are free to change and redistribute it.
> > There is NO WARRANTY, to the extent permitted by law.  Type "show
> > copying"
> > and "show warranty" for details.
> > This GDB was configured as "i686-pc-linux-gnu".
> > For bug reporting instructions, please see:
> > <http://www.gnu.org/software/gdb/bugs/>...
> > (gdb) l
> > 1       void foo() {
> > 2           return;
> > 3       }
> > 4
> > 5       int main() {
> > 6
> > 7               foo(); // 1
> > 8               foo(); // 2
> > 9               foo(); // 3
> > 10
> > (gdb) l
> > 11              return 0;
> > 12      }
> > (gdb) start
> > Temporary breakpoint 1 at 0x80483f7: file r.c, line 7.
> > Starting program: /home/marc/testing/a.out
> >
> > Temporary breakpoint 1, main () at r.c:7
> > 7               foo(); // 1
> > Current language:  auto; currently c++
> > (gdb) record
> > (gdb) n
> > 8               foo(); // 2
> > (gdb) n
> > 9               foo(); // 3
> > (gdb) n
> > 11              return 0;
> > (gdb) rn
> > 9               foo(); // 3
> > (gdb) s
> > foo () at r.c:3
> > 3       }
> > (gdb) rn
> >
> > No more reverse-execution history.
> > main () at r.c:7
> > 7               foo(); // 1
> >
> > ======== This is weird.  Why does it go back to the top?
> >
> 

Re: Wrong Reverse result

[Ramana Radhakrishnan]

> Great.  Thanks for doing that so quick.
> I forgot how to turn on those nice traces.  Can you
> remind me?
> The traces don't mean much to me, but I can include them
> in any future (if any :-)) issues.


help set debug on the gdb command line .

cheers
Ramana

>
>> infrun: TARGET_WAITKIND_STOPPED
>> infrun: stop_pc = 0x8048361
>> infrun: stepped into subroutine
>> infrun: inserting step-resume breakpoint at 0x8048349
>>
>> stop_stepping is better.
>>
>> infrun: resume (step=0, signal=0), trap_expected=0
>> infrun: prepare_to_wait
>> infrun: target_wait (-1, status) =
>> infrun:   10798 [process 10798],
>> infrun:   status->kind = no-history
>> infrun: infwait_normal_state
>>
>> No more reverse-execution history.
>> infrun: stop_stepping
>> main () at 1.c:7
>>
>> On Thu, Jun 11, 2009 at 00:49, Marc
>> Khouzam<marc.khouzam@ericsson.com> wrote:
>> > Hi,
>> >
>> > I'm getting a weird behaviour in reverse, where GDB
>> > jumps to far backwards.
>> >
>> > I have GDB HEAD from two weeks ago with the three
>> > missing PRrecord patches applied (that were pinged
>> > by Hui this very week).
>>
>> All of patches are not for prec, they are for reverse.
>
> I realized that just after pressing 'send' :-)
>
>> > I'm running on Ubuntu.
>> > Is it me or can someone reproduce it?
>> > (It happened during my presentation at the GCC summit,
>> > so I have witnesses :-))
>>
>> I am not sure Michael in GCC summit or not.  If so, you can let he fix
>> it at there directly.  :)
>
> The summit is over now, but I did meet with Michael.  It's nice to see
> people face to face.  I was able to put a bit of pressure on him
> about reviewing the missing patches ;-).  I'm sure that as soon as he
> has time, he'll have a look at these four things.
>
> Thanks Hui
>
> marc
>
>>
>> Thanks for your help.
>>
>> Hui
>>
>> >
>> > Thanks
>> >
>> > Marc
>> >
>> > GNU gdb (GDB) 6.8.50.20090522-cvs
>> > Copyright (C) 2009 Free Software Foundation, Inc.
>> > License GPLv3+: GNU GPL version 3 or later
>> > <http://gnu.org/licenses/gpl.html>
>> > This is free software: you are free to change and redistribute it.
>> > There is NO WARRANTY, to the extent permitted by law.  Type "show
>> > copying"
>> > and "show warranty" for details.
>> > This GDB was configured as "i686-pc-linux-gnu".
>> > For bug reporting instructions, please see:
>> > <http://www.gnu.org/software/gdb/bugs/>...
>> > (gdb) l
>> > 1       void foo() {
>> > 2           return;
>> > 3       }
>> > 4
>> > 5       int main() {
>> > 6
>> > 7               foo(); // 1
>> > 8               foo(); // 2
>> > 9               foo(); // 3
>> > 10
>> > (gdb) l
>> > 11              return 0;
>> > 12      }
>> > (gdb) start
>> > Temporary breakpoint 1 at 0x80483f7: file r.c, line 7.
>> > Starting program: /home/marc/testing/a.out
>> >
>> > Temporary breakpoint 1, main () at r.c:7
>> > 7               foo(); // 1
>> > Current language:  auto; currently c++
>> > (gdb) record
>> > (gdb) n
>> > 8               foo(); // 2
>> > (gdb) n
>> > 9               foo(); // 3
>> > (gdb) n
>> > 11              return 0;
>> > (gdb) rn
>> > 9               foo(); // 3
>> > (gdb) s
>> > foo () at r.c:3
>> > 3       }
>> > (gdb) rn
>> >
>> > No more reverse-execution history.
>> > main () at r.c:7
>> > 7               foo(); // 1
>> >
>> > ======== This is weird.  Why does it go back to the top?
>> >
>>
>

Re: Wrong Reverse result

[Joel Brobecker]

> I forgot how to turn on those nice traces.  Can you 
> remind me?
> The traces don't mean much to me, but I can include them
> in any future (if any :-)) issues.

It's "set debug infrun 1". I don't know how I was ever able to
investigate infrun issues before they got introduced...

-- 
Joel

RE: Wrong Reverse result

[Marc Khouzam]

> -----Original Message-----
> From: Joel Brobecker [mailto:brobecker@adacore.com] 
> Sent: June-11-09 10:01 AM
> To: Marc Khouzam
> Cc: Hui Zhu; gdb@sourceware.org; Michael Snyder
> Subject: Re: Wrong Reverse result
> 
> > I forgot how to turn on those nice traces.  Can you 
> > remind me?
> > The traces don't mean much to me, but I can include them
> > in any future (if any :-)) issues.
> 
> It's "set debug infrun 1". I don't know how I was ever able to
> investigate infrun issues before they got introduced...

But now, you can use Eclipse, GDB and Reverse to debug 
those problems :-) 

Re: Wrong Reverse result

[Hui Zhu]

On Thu, Jun 11, 2009 at 22:00, Joel Brobecker<brobecker@adacore.com> wrote:
>> I forgot how to turn on those nice traces.  Can you
>> remind me?
>> The traces don't mean much to me, but I can include them
>> in any future (if any :-)) issues.
>
> It's "set debug infrun 1". I don't know how I was ever able to
> investigate infrun issues before they got introduced...

You can "set debug infrun 1".
When you are interseting with some part for example "infrun: stepped
into subroutine", you can search "infrun: stepped into subroutine" in
infrun.c.  You can find the code.  Debug this part with Eclipse is
more better than command mode.  :)

Thanks,
Hui