fork() and anydbm question

Ralf Hildebrandt hildeb at www.stahl.bau.tu-bs.de
Wed Feb 16 16:00:40 CET 2000


I've written a little program which mails me the daily Dilbert, Robotman etc.

It mainly consists of a function "fetch_and_mail()" which tries to get the
GIF, checks whether that particular GIF was already fetched yesterday (yes,
updates on those servers are not reliable!). This is done by writing the
GIF's md5 hash into a persistent dictionary using anydbm.

fetch_and_mail() calls fork() in order to parallelize the process of
fetching the GIF's:

def fetch_and_mail():
 
  parent = os.fork()   # return 0 in the child
  if (parent > 0):     # not 0, meaning parent process
     time.sleep(5)     # give the process 5s to fork
     return            # parent returns
  if (parent == -1):
     print "fork() failed"
     sys.exit(1)
		   
  # we're in the child now             
  ... lots of ugly code ...
  
  try:
     page = urllib.urlopen(mainurl)
  except:
     print "urlopen(" + mainurl + ") failed"
     os._exit(1)                        

  ...
  
Questions:

  * What about the database created by anydbm.open() ?
    Does it become inconsistent since multiple processes write to it?
    
  * Is the use of os._exit(1) in the child process correct? Or should I use
    sys.exit(1) instead?
    






More information about the Python-list mailing list