[poe-commits] SF.net SVN: poe: [2319] trunk/poe/lib/POE/Kernel.pm

[<rc...@us...>]

Revision: 2319
          http://poe.svn.sourceforge.net/poe/?rev=2319&view=rev
Author:   rcaputo
Date:     2008-04-27 23:16:20 -0700 (Sun, 27 Apr 2008)

Log Message:
-----------
Per Marc Lehmann's feedback, I have documented how to avoid a
potential race condition when catching child processes with
sig_child().  Basically, if you want sig_child() to catch a process,
call the method in the same event handler that forked the process.
Otherwise POE::Kernel may reap the process before sig_child() is
called.

Modified Paths:
--------------
    trunk/poe/lib/POE/Kernel.pm

Modified: trunk/poe/lib/POE/Kernel.pm
===================================================================
--- trunk/poe/lib/POE/Kernel.pm	2008-04-26 09:13:50 UTC (rev 2318)
+++ trunk/poe/lib/POE/Kernel.pm	2008-04-28 06:16:20 UTC (rev 2319)
@@ -4525,7 +4525,7 @@
 the PROCESS_ID.
 
 A session may register as many sig_child() handlers as necessary, but
-there may only be one per PROCESS_ID.
+a session may only have one per PROCESS_ID.
 
 sig_child() watchers are one-shot.  They automatically unregister
 themselves once the EVENT_NAME has been delivered.
@@ -4533,6 +4533,11 @@
 sig_child() watchers keep a session alive for as long as they are
 active.  This is unique among signal watchers.
 
+Programs that wish to reliably reap child processes should be sure to
+call sig_child() before returning from the event handler that forked
+the process.  Otherwise POE::Kernel may have an opportunity to call
+waitpid() before an appropriate event watcher has been registered.
+
 sig_chid() does not return a meaningful value.
 
   sub forked_parent {


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.