cgdb freezes on exit: Mavericks + gdb 7.7

[cgdb]

cgdb freezes on exit: Mavericks + gdb 7.7

[oblitum]

Sometimes cgdb freezes for me too on exit (on Ubuntu). I have to send a kill -9...

[brasko]

On Tue, Mar 04, 2014 at 06:44:39PM -0800, Francisco Lopes wrote:

Sometimes cgdb freezes for me too on exit (on Ubuntu). I have to send a kill -9...

When it's frozen, can a stack trace be sent to us?

I've never seen this behavior.

Thanks,
Bob Rossi

[oblitum]

@brasko anytime I get it I'll send it here.

[oblitum]

Hi @brasko, I just freezed now while debugging some Go code, backtrace is the following:

(gdb) bt
#0  0x00007f4478478d3b in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#1  0x00007f44783f5f1c in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#2  0x00007f44783f34c5 in malloc () from /lib/x86_64-linux-gnu/libc.so.6
#3  0x00007f4478ec5939 in xmalloc () from /lib/x86_64-linux-gnu/libreadline.so.6
#4  0x00007f4478ebefb3 in rl_copy_text () from /lib/x86_64-linux-gnu/libreadline.so.6
#5  0x00007f4478ec2b19 in rl_delete_text () from /lib/x86_64-linux-gnu/libreadline.so.6
#6  0x00000000004181a0 in rline_clear (rline=<optimized out>) at ../../../../cgdb/lib/rline/rline.c:207
#7  0x0000000000404d50 in rl_sigint_recved () at ../../../cgdb/cgdb/cgdb.c:638
#8  0x00000000004096d5 in signal_handler (signo=17) at ../../../cgdb/cgdb/interface.c:720
#9  <signal handler called>
#10 0x00007f44783f0750 in ?? () from /lib/x86_64-linux-gnu/libc.so.6
#11 0x0000000000418521 in ibuf_free (s=0xdba010) at ../../../../cgdb/lib/adt/ibuf.c:37
#12 0x0000000000404d81 in cleanup () at ../../../cgdb/cgdb/cgdb.c:1422
#13 0x000000000040510f in process_commands (tgdb=0xdbd350) at ../../../cgdb/cgdb/cgdb.c:1040
#14 gdb_input () at ../../../cgdb/cgdb/cgdb.c:1116
#15 0x0000000000403f8f in main_loop () at ../../../cgdb/cgdb/cgdb.c:1388
#16 main (argc=<optimized out>, argv=<optimized out>) at ../../../cgdb/cgdb/cgdb.c:1737

I got freezed in this situation:

(gdb) q
A debugging session is active.

        Inferior 1 [process 28310] will be killed.

Quit anyway? (y or n) y
A debugging session is active.

        Inferior 1 [process 28310] will be killed.

After that I have to kill it with kill -9.

Notice that I'm on tip, but also with some very minimal changes applied so to have some display of UTF-8 text.

GNU gdb (GDB) 7.6.50.20140101-cvs

[zhangyafeikimi]

I met the issue in Mac OS X 10.9.2.
GNU gdb (GDB) 7.7, installed from homebrew.
CGDB 0.6.7, installed from homebrew.

[brasko]

On Thu, Mar 20, 2014 at 11:43:33PM -0700, Francisco Lopes wrote:

Hi @brasko, I just freezed now while debugging some Go code, backtrace is the following:

(gdb) bt
#0  0x00007f4478478d3b in ?? () from /lib/x86_64-linux-gnu/libc.so.6
...
#16 main (argc=<optimized out>, argv=<optimized out>) at ../../../cgdb/cgdb/cgdb.c:1737

I got freezed in this situation:

(gdb) q
A debugging session is active.

        Inferior 1 [process 28310] will be killed.

Quit anyway? (y or n) y
A debugging session is active.

        Inferior 1 [process 28310] will be killed.

After that I have to kill it with kill -9.

Notice that I'm on tip, but also with some very minimal changes applied so to have some display of UTF-8 text.

Please don't ask why, but i just saw this email now.

For my sanity sake, do you know if you get the freezes with out your
patches?

Thanks,
Bob Rossi

[oblitum]

I used to get freezes before this patch, not sure whether they're more frequent now... Also, I expect to be the only one using this patch,

[oblitum]

I'm using mavericks now, no patches. It's also freezing as stated here.

[oblitum]

It has freezed on process termination, I'm unable to enter a command to quit cgdb.

[brasko]

On Tue, Mar 04, 2014 at 01:17:57AM -0800, Gregory Pakosz wrote:

Hello, I seem to face the same problem as reported in #35 (which as been closed by the reporter without further comment).

When using gdb 7.7 (installed with homebrew and self signed), quitting gdb while debugging works. When using cgdb, quitting while debugging hangs. I then need to hit CTRL-C.

Hi,

I finally reproduced this behavior and understand why it hangs on exit.

The basic issue is, when gdb dies, the OS sends a SIGCHLD to CGDB. CGDB
is in the middle of a malloc (or some other operation) when that signal
is received. In the signal handler it again calls malloc, which locks a
mutex causing CGDB to hang.

I'll be fixing the issue over the next few days and probably doing a
release after that. Thanks for continuing to follow up with this issue.

Below is the stack trace that shows the issue.

Thanks,
Bob Rossi

(gdb) bt
#0 __lll_lock_wait_private () at
../nptl/sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:95
#1 in _L_lock_12779 () at malloc.c:5206
#2 in __GI___libc_malloc (bytes=1) at malloc.c:2887
#3 in xmalloc () from /lib/x86_64-linux-gnu/libreadline.so.6
#4 in rl_copy_text () from /lib/x86_64-linux-gnu/libreadline.so.6
#5 in rl_delete_text () from /lib/x86_64-linux-gnu/libreadline.so.6
#6 in rline_clear (rline=) at
../../../cgdb/lib/rline/rline.c:207
#7 in rl_sigint_recved () at ../../cgdb/cgdb/cgdb.c:638
#8 in signal_handler (signo=17) at ../../cgdb/cgdb/interface.c:720
#9
#10 malloc_consolidate (av=av@entry=0x7f02b3a64760 <main_arena>) at
malloc.c:4143
#11 in _int_malloc (av=0x7f02b3a64760 <main_arena>, bytes=4097) at
malloc.c:3423
#12 in __GI___libc_malloc (bytes=4097) at malloc.c:2891
#13 in gdb_input () at ../../cgdb/cgdb/cgdb.c:1100
#14 in main_loop () at ../../cgdb/cgdb/cgdb.c:1388
#15 main (argc=, argv=) at
../../cgdb/cgdb/cgdb.c:1737

[oblitum]

looks similar to the previous :-)

[brasko]

On Fri, Jul 11, 2014 at 06:48:20AM -0700, Francisco Lopes wrote:

looks similar to the previous :-)

You are right. I'm sorry I didn't recognize this earlier.

The good news here is that we can confirm the behavior is the same issue
and this means this should probably fix the issue for good.

I'll be working on a formal fix shortly.

Thanks,
Bob Rossi

[brasko]

On Fri, Jul 11, 2014 at 01:17:55PM -0400, Bob Rossi wrote:

On Fri, Jul 11, 2014 at 06:48:20AM -0700, Francisco Lopes wrote:

looks similar to the previous :-)

You are right. I'm sorry I didn't recognize this earlier.

The good news here is that we can confirm the behavior is the same issue
and this means this should probably fix the issue for good.

I'll be working on a formal fix shortly.

Finally fixed upstream.

Bob Rossi

[oblitum]

@brasko hi, sadly, it's still hanging consistently for me. For the following simple testcase it has hanged the two times I've tried:

I had to kill it with kill -9 and also, when hanging cgdb gets 100% CPU usage.

I'm on OS X Mavericks, configure line was

../../cgdb/configure --prefix=/opt/local --with-readline=/usr/local/opt/readline

where readline is provided by homebrew.

[oblitum]

by the way, it seems usage of cgdb coupled with tmux is not very good, it's the first time I tried it.

[oblitum]

Also, I didn't quit. It hanged on continue command, and the process terminated but I didn't get the control back.

[brasko]

On Fri, Jul 18, 2014 at 08:50:11PM -0700, Francisco Lopes wrote:

Also, I didn't quit. It hanged on continue command, and the process terminated but I didn't get the control back.

You need to rebuild cgdb from source and then do 'make && make install'
and run the newly installed cgdb.

If you do that, you should be at hash
da718d2
and the hangs should no longer occur. I'm positive I've fixed the
original issue.

If you still see an issue, please provide me with another backtrace.

Thanks!
Bob Rossi

[oblitum]

ok @brasko, I think I will open another issue as this seems unrelated and it's the first time I'm using it on Mavericks.