访问thread_data_base的偏移量时，在pthread_join上增强线程连接段错误

Question

Crash is seen when pthread code tries to access an offset from thread_handle
#0  0x00007f6c7459b7d8 in pthread_join () from /lib64/libpthread.so.0
#1  0x00007f6c7438c310 in boost::thread::join (this=0x4069ef0)

Here is the thread info from object (this=0x4069ef0):

$14 = (boost::detail::thread_data_base *) 0x3c2fd90
(gdb) p  *$9.px
$15 = {
  <boost::enable_shared_from_this<boost::detail::thread_data_base>> = {
    weak_this_ = {
      px = 0x3c2fd90,
      pn = {
        pi_ = 0x3c2ff20
      }
    }
  },
  members of boost::detail::thread_data_base:
  _vptr.thread_data_base = 0x102c3f0,
  self = {
    px = 0x0,
    pn = {
      pi_ = 0x0
    }
  },
  thread_handle = 140103459141968,

Accessing contents from thread_handle causing segfault. On trying to view contents from *thread_handle

Cannot access memory at address 0x7f6c60e9f950

Boost version is boost-1.41.0

Here is the code which creates the thread

t1 = new boost::thread(boost::bind(&test::Th, this));

Later while cleanup

t1->join();

cause seg fault

Answer 1

We had some code which was intercepting pthreads and releasing the stack allocated by using pthread api. Later when boost was accessing the it as the stack was gone we were seeing issue.