By using this site, you agree to our updated Privacy Policy and our Terms of Use. Manage your Cookies Settings.
443,923 Members | 1,279 Online
Bytes IT Community
+ Ask a Question
Need help? Post your question and get tips & solutions from a community of 443,923 IT Pros & Developers. It's quick & easy.

Threads and signals

P: n/a
I've run into an "opportunity" in a Python application using threads and
signals.
Basically, there is a main process that spawns off a child thread that loops
forever.
In between iterations, the child thread sleeps for X seconds. All the while, the
main
thread loops forever doing its thing and also sleeps in between iterations (see
the code
at the end this message). Normally the application would be daemonized before
the main
process starts up. Hence, during a system shutdown or stoppage of the daemon,
it's
desired that the daemon catch a few signals (TERM/INT) and perform a few cleanup
routines. According to the Python docs, only the main thread will receive
signals. The
problem I have, on FreeBSD systems, is that the sleep function in the child gets
interrupted
and the signal never gets handled until the main thread's sleep concludes. It
works as
expected on a Linux box (main thrd's sleep is interrupted). Sample output from
multiple
systems is directly below.

Just looking for insight from others in the know.

Thanks,
Chad
Linux 2.6.17
------------------------------------------------------------------------
test_bed$ python2.5 ./thrd_test.py
Starting up
MainThread.run(): 14332
MainThread before sleep(15)
ChildThread.run(): 14332
ChildThread before sleep(10)
Received signal 2 <-- Interrupted here (Ctrl-C), correctly
flag: True interrupts sleep in main thread
flag: False
MainThread after sleep(15)
Shutting down

test_bed$ python2.4 ./thrd_test.py
Starting up
MainThread.run(): 14338
MainThread before sleep(15)
ChildThread.run(): 14338
ChildThread before sleep(10)
Received signal 2 <-- Interrupted here (Ctrl-C), correctly
flag: True interrupts sleep in main thread
flag: False
MainThread after sleep(15)
Shutting down
test_bed$

FreeBSD 4.11, 6.1, and 6.2
------------------------------------------------------------------------
bash-2.05b# python2.5 ./thrd_test.py
Starting up
MainThread.run(): 65930
ChildThread.run(): 65930
ChildThread before sleep(10)
MainThread before sleep(15)
^CChildThread after sleep(10) <-- Interrupted here (Ctrl-C), but
ChildThread before sleep(10) interrupts the child thrd's sleep
ChildThread after sleep(10)
ChildThread before sleep(10)
Received signal 2 <-- Main sleep concludes and then
the
flag: True signal gets handled
flag: False
MainThread after sleep(15)
Shutting down
bash-2.05b#
#--- CODE BEGIN ---#

#!/usr/bin/env python

import os
import sys
import time
import signal
import threading

def sigHandle(signo, stkframe):
print "Received signal ", signo
print "flag: ", mthrd.flag.isSet()
mthrd.flag.clear()
print "flag: ", mthrd.flag.isSet()

class ChildThread(threading.Thread):

def __init__(self):
threading.Thread.__init__(self)

def run(self):
print "ChildThread.run(): ", os.getpid()
while (True):
print "ChildThread before sleep(10)"
time.sleep(10)
print "ChildThread after sleep(10)"

class MainThread(object):

def __init__(self):
self.flag = threading.Event()
self.cthrd = ChildThread()
self.cthrd.setDaemon(True)

def run(self):
print "MainThread.run(): ", os.getpid()
self.flag.wait()
self.cthrd.start()
while (self.flag.isSet()):
print "MainThread before sleep(15)"
time.sleep(15)
print "MainThread after sleep(15)"
return # or cleanup routine

if __name__ == "__main__":

# Normally, the process is daemonized first. That's been
# left out for testing purposes.

signal.signal(signal.SIGINT, sigHandle)
signal.signal(signal.SIGQUIT, sigHandle)
signal.signal(signal.SIGTERM, sigHandle)

mthrd = MainThread()
print "Starting up"
mthrd.flag.set()
mthrd.run()
print "Shutting down"

sys.exit(0)

#--- CODE END ---#

Nov 15 '06 #1
Share this Question
Share on Google+
1 Reply


P: n/a
By writing a little C module I fixed the problem. By simply calling
the function in the child thrd's run method, things work as expected.

This is all the function (blockall) does:

sigset_t omask, nmask;
sigfillset(&nmask);
sigprocmask(SIG_SETMASK, &nmask, NULL);

bash-2.05b# python2.5 ./thrd_test.py
Starting up
MainThread.run(): 71799
ChildThread.run(): 71799
ChildThread before sleep(10)
MainThread before sleep(15)
^CReceived signal 2
flag: True
flag: False
MainThread after sleep(15)
Shutting down
bash-2.05b#

Chad
I've run into an "opportunity" in a Python application using threads
and signals. Basically, there is a main process that spawns off a child
thread that loops forever. In between iterations, the child
thread sleeps for X seconds. All the while, the main thread loops
forever doing its thing and also sleeps in between iterations (see the code at
the end this message). Normally the application would
be daemonized before the main process starts up. Hence, during a
system shutdown or stoppage of the daemon, it's desired that the
daemon catch a few signals (TERM/INT) and perform a few cleanup
routines. According to the Python docs, only the main thread will receive
signals. The problem I have, on FreeBSD systems, is that the sleep
function in the child gets interrupted and the signal never gets handled
until the main thread's sleep concludes. It works as expected on a Linux
box (main thrd's sleep is interrupted). Sample output from multiple systems
is directly below.

Just looking for insight from others in the know.

Thanks,
Chad
Linux 2.6.17
------------------------------------------------------------------------
test_bed$ python2.5 ./thrd_test.py
Starting up
MainThread.run(): 14332
MainThread before sleep(15)
ChildThread.run(): 14332
ChildThread before sleep(10)
Received signal 2 <-- Interrupted here (Ctrl-C), correctly
flag: True interrupts sleep in main thread
flag: False
MainThread after sleep(15)
Shutting down

test_bed$ python2.4 ./thrd_test.py
Starting up
MainThread.run(): 14338
MainThread before sleep(15)
ChildThread.run(): 14338
ChildThread before sleep(10)
Received signal 2 <-- Interrupted here (Ctrl-C), correctly
flag: True interrupts sleep in main thread
flag: False
MainThread after sleep(15)
Shutting down
test_bed$

FreeBSD 4.11, 6.1, and 6.2
------------------------------------------------------------------------
bash-2.05b# python2.5 ./thrd_test.py
Starting up
MainThread.run(): 65930
ChildThread.run(): 65930
ChildThread before sleep(10)
MainThread before sleep(15)
^CChildThread after sleep(10) <-- Interrupted here (Ctrl-C), but
ChildThread before sleep(10) interrupts the child thrd's sleep
ChildThread after sleep(10)
ChildThread before sleep(10)
Received signal 2 <-- Main sleep concludes and then
flag: True the signal gets handled
flag: False
MainThread after sleep(15)
Shutting down
bash-2.05b#
#--- CODE BEGIN ---#

#!/usr/bin/env python

import os
import sys
import time
import signal
import threading

def sigHandle(signo, stkframe):
print "Received signal ", signo
print "flag: ", mthrd.flag.isSet()
mthrd.flag.clear()
print "flag: ", mthrd.flag.isSet()

class ChildThread(threading.Thread):

def __init__(self):
threading.Thread.__init__(self)

def run(self):
print "ChildThread.run(): ", os.getpid()
mymod.blockall()
while (True):
print "ChildThread before sleep(10)"
time.sleep(10)
print "ChildThread after sleep(10)"

class MainThread(object):

def __init__(self):
self.flag = threading.Event()
self.cthrd = ChildThread()
self.cthrd.setDaemon(True)

def run(self):
print "MainThread.run(): ", os.getpid()
self.flag.wait()
self.cthrd.start()
while (self.flag.isSet()):
print "MainThread before sleep(15)"
time.sleep(15)
print "MainThread after sleep(15)"
return # or cleanup routine

if __name__ == "__main__":

# Normally, the process is daemonized first. That's been
# left out for testing purposes.

signal.signal(signal.SIGINT, sigHandle)
signal.signal(signal.SIGQUIT, sigHandle)
signal.signal(signal.SIGTERM, sigHandle)

mthrd = MainThread()
print "Starting up"
mthrd.flag.set()
mthrd.run()
print "Shutting down"

sys.exit(0)

#--- CODE END ---#
Nov 16 '06 #2

This discussion thread is closed

Replies have been disabled for this discussion.