Discussion:
Seg faults when using boost::regex_replace in multiple threads
m***@kamba.net.nz
2006-07-10 22:16:34 UTC
Permalink
Hello,



I'm having a few problems with boost::regex_replace seg faulting when used
in a multithreaded program. The problem seems to be identical to the one
described in http://lists.boost.org/boost-users/2006/06/20235.php, with a
seg fault on line 133 of regex_format.hpp, though I have linked against the
multithreaded version of libboost_regex.so and the problem is still
happening. Unfortunately, I never saw any definitive answer for what was
causing the original poster's problems.



===========================================

Stack trace:

===========================================



Program received signal SIGSEGV, Segmentation fault.

[Switching to Thread -584664144 (LWP 28385)] 0x080520e0 in
boost::re_detail::basic_regex_formatter<boost::re_detail::string_out_ite

rator<std::string>, boost::match_results<char const*,
std::allocator<boost::sub_match<char const*> > >,
boost::regex_traits_wrapper<boost::regex_traits<char,

boost::cpp_regex_traits<char> > > >::format_all (this=0xdd26b8d0) at

regex_format.hpp:133

133 switch(*m_position)

(gdb) where

#0 0x080520e0 in

boost::re_detail::basic_regex_formatter<boost::re_detail::string_out_ite

rator<std::string>, boost::match_results<char const*,
std::allocator<boost::sub_match<char const*> > >,
boost::regex_traits_wrapper<boost::regex_traits<char,

boost::cpp_regex_traits<char> > > >::format_all (this=0xdd26b8d0) at

regex_format.hpp:133

#1 0x080518be in

boost::re_detail::basic_regex_formatter<boost::re_detail::string_out_ite

rator<std::string>, boost::match_results<char const*,
std::allocator<boost::sub_match<char const*> > >,
boost::regex_traits_wrapper<boost::regex_traits<char,

boost::cpp_regex_traits<char> > > >::format (this=0xdd26b8d0, p1=0xd55005dc
"$1", p2=0xd55005de "",

f=format_all) at regex_format.hpp:123

#2 0x08051073 in

boost::re_detail::regex_format_imp<boost::re_detail::string_out_iterator

<std::string>, char const*, std::allocator<boost::sub_match<char const*>
, char, boost::regex_traits_wrapper<boost::regex_traits<char,
boost::cpp_regex_traits<char> > > > (out=

{<std::iterator<std::output_iterator_tag, char, int, char*, char&>> =
{<No data fields>}, out = 0xdd26ba70}, m=@0xd5500518, p1=0xd55005dc "$1",
p2=0xd55005de "",

flags=format_all, t=@0x8064140) at regex_format.hpp:569

#3 0x08050710 in

format<boost::re_detail::string_out_iterator<std::basic_string<char,

std::char_traits<char>, std::allocator<char> > >, boost::basic_regex<char,
boost::regex_traits<char, boost::cpp_regex_traits<char> > > >
(this=0xd5500518, out={<std::iterator<std::output_iterator_tag, char, int,
char*, char&>> = {<No data fields>}, out = 0xdd26ba70},

fmt=@0xdd26b990, flags=format_all, re=@0x805f6ec) at
match_results.hpp:170

#4 0x0804fe35 in

boost::regex_replace<boost::re_detail::string_out_iterator<std::string>,

char const*, boost::regex_traits<char, boost::cpp_regex_traits<char> >,

char> (out=

{<std::iterator<std::output_iterator_tag, char, int, char*, char&>> =
{<No data fields>}, out = 0xdd26ba70}, first=0xd550048c "abcdefgh",
last=0xd5500494 "", e=@0x805f6ec,

fmt=0x805a2a9 "$1", flags=format_all) at regex_replace.hpp:53

#5 0x0804f7e1 in boost::regex_replace<boost::regex_traits<char,

boost::cpp_regex_traits<char> >, char> (s=@0xdd26ba80, e=@0x805f6ec,

fmt=0x805a2a9 "$1", flags=format_all)

at regex_replace.hpp:84

#6 0x0804f3c4 in test_fn () at regexTest.C:32

#7 0xf63e5dec in start_thread () from /lib/tls/libpthread.so.0

#8 0xf6385a2a in clone () from /lib/tls/libc.so.6





(gdb) print *this

$1 = {m_traits = @0x8064140, m_results = @0xd5500518, m_out =
{<std::iterator<std::output_iterator_tag, char, int, char*, char&>> = {<No
data fields>}, out = 0xdd26ba70},

m_position = 0xd44ff5d6 <Address 0xd44ff5d6 out of bounds>, m_end =
0xd55005de "", m_flags = format_all, m_state = output_copy,
m_have_conditional = false}





==========================================



GCC version is 3.2.3 20030502 (Red Hat Linux 3.2.3-42), with Boost 1.33.1,
on a 4 CPU machine.



We're using the STL from RogueWave, though since the post from a couple of
weeks ago had the same problem with both GCC and STLport libraries I don't
think that's the problem. This problem also occurs with version 1.33.0, but
not with 1.32 or earlier.



Does anyone know if there is a known problem with using these new versions
of boost::regex_replace in multiple threads? If it's helpful, I have a very
simple test program that demonstrates the problem. Any help on this would be
greatly appreciated.



Kind regards,

Alistair Potts
John Maddock
2006-07-11 09:01:54 UTC
Permalink
Post by m***@kamba.net.nz
Does anyone know if there is a known problem with using these new
versions of boost::regex_replace in multiple threads? If it's
helpful, I have a very simple test program that demonstrates the
problem. Any help on this would be greatly appreciated.
If you can let me have the test program I can at least try and reproduce it.

I must admit to being at a loss to explain this otherwise: there really is
no thread sensitive code inside regex_replace.

John.
Alistair Potts
2006-07-12 18:20:48 UTC
Permalink
Sure. Here's a very simple program that does nothing useful, but which
crashes consistently on a multi-processor machine.

Kind regards,
Alistair

#include <boost/regex.hpp>
#include <iostream>
#include <string>
#include <pthread.h>

const char* const reExpr("abcd(.*)");
const char* const replaceExpr("$1");
boost::regex re(reExpr);

size_t noThreads(50);
pthread_barrier_t barrier;

extern "C"
{
void* test_fn(void*)
{
pthread_barrier_wait(&barrier);
while(1)
{
std::string expr("abcdefgh");
std::cout << boost::regex_replace(expr, re, replaceExpr,
boost::match_default|boost::format_all) << std::endl;
}
return 0;
}
}

int main(int argc,char* argv[])
{
pthread_t* threads = new pthread_t[noThreads];
pthread_barrier_init(&barrier, NULL, noThreads);

size_t i = 0;
for (; i < noThreads; ++i)
{
if(pthread_create(&(threads[i]), NULL, test_fn, NULL))
{
std::cerr << "Error creating thread" << std::endl;
return 1;
}
}
for(i = 0; i < noThreads; ++i)
pthread_join(threads[i], NULL);

exit(0);
}

-----Original Message-----
From: boost-users-***@lists.boost.org
[mailto:boost-users-***@lists.boost.org] On Behalf Of John Maddock
Sent: 11 July 2006 10:02
To: boost-***@lists.boost.org
Subject: Re: [Boost-users] Seg faults when using
boost::regex_replaceinmultiple threads
Post by m***@kamba.net.nz
Does anyone know if there is a known problem with using these new
versions of boost::regex_replace in multiple threads? If it's
helpful, I have a very simple test program that demonstrates the
problem. Any help on this would be greatly appreciated.
If you can let me have the test program I can at least try and reproduce it.

I must admit to being at a loss to explain this otherwise: there really is
no thread sensitive code inside regex_replace.

John.

Loading...