Subj : threaded program and wait() call
To   : comp.programming.threads
From : tadek
Date : Fri Aug 05 2005 04:28 pm

Hello,
I posted similar request to comp.unix.programmer, but since I don't see
it there and this group deals specifically with threads, please let me
present my problem and ask for help.

OS: RHEL4 Linux, i386 platform and i686 hardware (glibc-2.3.4-2.9)

Process creates two threads.  In the 'main' thread module, I install
signal handler for SIGTERM and SIGHUP.

The 'main' thread creates 'worker' thread and and then loops checking
wheather 'worker' loop is alive (pseudo code):
  ....
  while (bWorkerDead == FALSE)
    sleep(5);
    status = pthread_kill (thread_parms.sWorkerThread.tid, 0);
    if (status != EOK)
       ...
       if(g_shutdown)
            pthread_kill(thread_parms.sWorkerThread.tid, SIGHUP);
    ...

When SIGTERM is delivered to the signal handler, global flag
(g_shutdown) is set and in the loop above SIGHUP is sent to worker
thread. (I do make sure that SIGHUP is sent only once).

The 'worker' thread spawns n number of independent processes using
posix_spawn() function and when done, calls wait() function, waiting
for any child process to exit or for signal (SIGHUP). Unfortunately
SIGHUP is not delivered to the 'worker' thread (pseudo code)
   ...
   loop
     posix_spawn();          // start child processes
   while()                   // monitor them
      if( SIGHUP received)
         loop
            kill spawned processes
      ....
      pid=wait();            // wait for child process to die/exit
      if (pid == (pid_t)-1)
         if (EINTR == errno) // signal received (SIGHUP expected)
            continue;
      else
         ...

      if (WIFEXITED(iStatVal))
         .....
      etc;


I ported above code from QNX to Linux and I was expecting to behave
identically: when child process exits - it gets restarted, when SIGHUP
is deliver - 'worker' thread should kill all spawned child processes.
The call to wait() never returns as the result of signal delivery
(pid==-1). My signal handler receives SIGHUP.

Is anything missing in the logic above?

Tad Dragowski
PS I am also confused whether I am using LinuxThreads or NPTL when
linking with  glibc and making calls to pthread_* functions.

.