[Quickfix-developers] Re: ThreadedSocketAcceptor Segmentation Faults

[Alberto B. <bellido@3.14financial.com>]

Hi

We're getting the same error on some testings.

#0  0x4048318c in pthread_detach () from /lib/tls/libpthread.so.0
#1  0x400d5edc in FIX::thread_detach (thread=0) at Utility.cpp:344
#2  0x4009323f in FIX::ThreadedSocketAcceptor::removeThread
(this=0xbffff500, s=57) at stl_map.h:332
#3  0x40093335 in FIX::ThreadedSocketAcceptor::socketThread (p=0x3) at
ThreadedSocketConnection.h:51
#4  0x40482b3c in start_thread () from /lib/tls/libpthread.so.0
#5  0x4032092a in clone () from /lib/tls/libc.so.6

We're running Mandrake Linux 10.1, Quickfix 1.10.2. As you said, there's
no way to find a pattern to reproduce it.

Have you find any clue?

Thanks 

Alberto





Since I upgraded to Quickfix 1.10.2 (prior we were using 1.9.4) i'm
getting
a random segmentation fault when sessions log off. I couldn't find a pattern
to reproduce it, or even a "more probable" scenario, it just happens from
time to time.

Before entering a bug report, i'd like to know if someone has faced this (or
similar) problem, or if someone thinks the problem is on my side.

All the dumped cores from this segmentation fault read the same:

#0  0x00cb8b60 in pthread_detach () from /lib/tls/libpthread.so.0
#1  0x00870b97 in FIX::thread_detach (thread=0) at Utility.cpp:344
#2  0x0081d4c8 in FIX::ThreadedSocketAcceptor::removeThread (this=0x8f596c8,
    s=19) at stl_map.h:221
#3  0x0081d5e7 in FIX::ThreadedSocketAcceptor::socketThread (p=0xf6a8a10)
    at ThreadedSocketConnection.h:51
#4  0x00cb7dec in start_thread () from /lib/tls/libpthread.so.0
#5  0x006f0a2a in clone () from /lib/tls/libc.so.6

At Level #2, the Acceptor member m_threads shows a suspicious "_M_node_count
= 0" which can lead to the thread=0 passed to thread_detach and causes the
segfault

I'm not sure if faulty code on my side can lead to that errors, but i've
found that problem in all my developments using a ThreadAcceptor, ranging
from a single FIX logger to a monstrous routing system.

I'm compiling QFix with mysql support (MySQL ver 4.14) on RedHat Enterprise
3 servers.
Some system libs:
glibc-devel-2.3.2.-95.3x (pthreads)
libstdc++-devel-3.2.3-52 (stl)

Thanks in advance and cheers (again) for your great work

Guillermo