"record save" failure

[Cedric Roux <sed at free dot fr>]

Hi GNU gdb people,

I'm trying to analyse an unfriendly program and I gave a try
to the "record" thing gdb provides.

Except it fails.

Below is an excerpt of the log.
Am I doing something wrong? forgetting to type something?
Does the "record" thing works with multi-threaded programs?
I used the "set scheduler-locking on" command to run only the
desired thread.

As you see, I record more than 5 millions instructions between
breakpoint 3 and breakpoint 2. The "record save" fails and a
following "info record" crashes gdb. My computer is a 64b intel p6000
thing with debian testing (last updated in october of last year or
something). I compiled gdb 7.6.2 locally with gcc 4.8.1.

I hacked a bit gdb (but just a bit) to dump the "record list" at
the beginning of "record_full_save" and I get a different dump
compared to "record_full_save" (which does more than dump the
list, it replays the instructions, which I don't do). So I'm a
bit lost. (I compared with a simple multi-threaded program I quickly
wrote, and for which "record" works.)

Any hint or whatever (maybe there is a bug somewhere in gdb...) is
very welcome.

Regards and thanks for gdb which is a great program,
Cédric.

momo@sed:~$ dbg/bin/gdb -p 23553
GNU gdb (GDB) 7.6.2
[...]
This GDB was configured as "x86_64-unknown-linux-gnu".
[...]
Attaching to process 23553
[...]
[New LWP 23554]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
[...]
0x00007ffff325824d in poll () from /lib/x86_64-linux-gnu/libc.so.6
(gdb) hb *0x00007fffe8c63f22
Hardware assisted breakpoint 1 at 0x7fffe8c63f22
(gdb) cont
Continuing.
[New Thread 0x7fffe4372700 (LWP 23570)]
[New Thread 0x7fffe4b73700 (LWP 23571)]
[New Thread 0x7fffdeaff700 (LWP 23572)]
[New Thread 0x7fffca553700 (LWP 23594)]
[New Thread 0x7fffc87ff700 (LWP 23595)]
[New Thread 0x7fffc7ffe700 (LWP 23596)]
[New Thread 0x7fffc77fd700 (LWP 23597)]
[New Thread 0x7fffc6ce9700 (LWP 23598)]
[New Thread 0x7fffc64e8700 (LWP 23599)]
[Thread 0x7fffc77fd700 (LWP 23597) exited]
[Thread 0x7fffc7ffe700 (LWP 23596) exited]
[Thread 0x7fffc87ff700 (LWP 23595) exited]
[Thread 0x7fffc6ce9700 (LWP 23598) exited]
[Thread 0x7fffc64e8700 (LWP 23599) exited]
[Thread 0x7fffca553700 (LWP 23594) exited]
[New Thread 0x7fffca553700 (LWP 23600)]
[New Thread 0x7fffc64e8700 (LWP 23601)]
[Thread 0x7fffc64e8700 (LWP 23601) exited]
[Switching to Thread 0x7fffca553700 (LWP 23600)]

Breakpoint 1, 0x00007fffe8c63f22 in ?? () from XXX.so
(gdb) cont
Continuing.

Breakpoint 1, 0x00007fffe8c63f22 in ?? () from XXX.so
(gdb) hb *0x00007fffe8ad361c
Hardware assisted breakpoint 2 at 0x7fffe8ad361c
(gdb) hb *0x00007fffe8c0f066
Hardware assisted breakpoint 3 at 0x7fffe8c0f066
(gdb) cont
Continuing.

Breakpoint 3, 0x00007fffe8c0f066 in ?? () from XXX.so
(gdb) set record full insn-number-max 10000000
(gdb) record full
(gdb) set scheduler-locking on
(gdb) cont
Continuing.

Breakpoint 2, 0x00007fffe8ad361c in ?? () from XXX.so
(gdb) info record
Active record target: record-full
Record mode:
Lowest recorded instruction number is 1.
Highest recorded instruction number is 5681093.
Log contains 5681093 instructions.
Max logged instructions is 10000000.
(gdb) record save TOTOTATA
warning: target file /proc/23553/cmdline contained unexpected null 
characters
Couldn't get registers: No such process.
(gdb) info record
Active record target: record-full
Segmentation fault