[svn] w32pth - r19 - trunk

[svn author marcus]

Author: marcus
Date: 2008-02-13 13:53:28 +0100 (Wed, 13 Feb 2008)
New Revision: 19

Modified:
   trunk/TODO
Log:
Update.


Modified: trunk/TODO
===================================================================
--- trunk/TODO	2008-02-12 12:50:10 UTC (rev 18)
+++ trunk/TODO	2008-02-13 12:53:28 UTC (rev 19)
@@ -1,19 +1,6 @@
-* pth_waitpid should be implemented for dirmngr, at least for the 
+* pth_waitpid should be implemented for dirmngr, at least for the
   non-blocking case.
 
-* dirmngr will create one new thread for each pth_accept_ev until a
-  connection is accepted, when all these threads will wake up and 
-  terminate.  This goes up to 3600 threads (the current max timeout).
-  We could limit the timeout by the minimum of all timeout events that
-  are passed along with the pth_wait call.  That's one step in the right 
-  direction.  Alternatively, we could cancel the threads (but we need to 
-  figure out if this is safe---it probably isn't) or wake them up in 
-  some other way.  For example, we can have one shared event, a shutdown
-  notification, by all select threads in one invocation, and have the 
-  threads listen on both the socket and the shutdown event.
-
-* I think the pth_wait on socket implementation leaks event objects.
-
 * The whole wait stuff needs to be cleaned up.  In particular, it is 
   vital that it is clarified which events get reset before pth_wait 
   returns and which are not.  Tentatively I think that none should be 
@@ -23,5 +10,7 @@
 * Speaking of signals, that needs to be properly implemented and 
   tested.
 
+* Dirmngr still leaks some events under load.  Find out if this is a
+  bug in dirmngr or in w32pth or a combination of both.