Subj : Re: Problem with system() calls in a multithreaded program on HPUX
To   : comp.programming.threads,comp.sys.hp.hpux
From : Stefaan A Eeckels
Date : Mon Feb 14 2005 09:53 am

On 13 Feb 2005 23:09:37 -0800
maheshkumarjha@gmail.com (Mahesh Kumar) wrote:

> Can anyone suggest a possible reason for still hanging? 

Remember that the advice was to use fork()/exec() 
_AND_ to go single-threaded. 

The fork() functions suspend all threads in the process before
they run.  Threads that are already running and are in an
uninterruptible wait inside the kernel cause the fork() to pause
until they exit the uninterruptible wait. As a result, fork(),
and your whole process, appears to be hung. 

Using fork() In a multithreaded process can result in
interrupted blocking system calls - you should be prepared
to handle EINTR errors.

fork() per se isn't thread-safe, and requires TLC to mesh
well with threads. If you really have to continue with
your multi-threaded approach, read the popen(3) man
page carefully. Notice where it says:

| popen() and pclose() are thread-safe.  These interfaces are not
| async-cancel-safe. A cancellation point may occur when a thread is
| executing popen() or pclose().

and consider whether your objectives cannot achieved through
a less complex approach.

Take care,

-- 
Stefaan
-- 
As complexity rises, precise statements lose meaning,
and meaningful statements lose precision. -- Lotfi Zadeh

.