After struggling with os.spawnxxx to get a detached process I tried using Pyhton2.4's new subprocess module.
I struggled with that as well even when trying to use the creation flags for DETACHED_PROCESS 0x8 and CREATE_NEW_PROCESS_GROUP = 0x200 I am using the following cgi script parent.cgi #!c:/python24/python.exe -u print 'content-type: text/plain' print print 'DONE' import subprocess pid = subprocess.Popen(['c:/python24/pythonw.exe', 'c:\\tmp\\child.py', 'arg1', 'arg2'], creationflags=0x208).pid print pid #end of parent script child is a simple time delay #child.py import time, sys, os time.sleep(10) print 'stdout IN THE CHILD' print >>sys.stderr, 'stderr IN THE CHILD' #end of child When I run the parent.cgi script through apache I see no output in the browser for 10 seconds; when the child dies (as observed using procexp) I see the expected DONE 3256 In other words it seems impossible to get standard subprocess.py to detach the child process properly. However, if I hack subprocess.py to alter the bInheritHandles flag passed into CreateProcess (line 718) from the constant 1 to not (creationflags & 0x8) and 1 or 0 Is this a buglet or a feature request? It seems subprocess punts on closefds for mswindows, but setting bInheritHandles to 0 seems to work fine ie when it is 0 my test seems to indicate that the parent has finished and gone to heaven long before the child ends its sleep. Of course it may be that it's just the handles that are being held. -- Robin Becker -- http://mail.python.org/mailman/listinfo/python-list