Re: Possible bugs in ns-2.1b4

[Yuri Pryadkin <yuri@who.isi.edu>]

Urs Thuermann <thuerman@ibr.cs.tu-bs.de> writes:

> Reading some source of ns-2.1b4 I've found two possible bugs:
> 
> 1. The function
> 
> 	int Queue::length()
> 
>    is not virtual, so that when I have a pointer Queue *p that in fact
>    points to a DRR Queue, when doing
> 
> 	p->length();
> 
>    this unexpectedly calls Queue::length() instead of DRR::length().
> 
>    Or is this behavior intended?

This must be a bug.  Will be changed to virtual in the next release.
Thanks.

> 
> 2. In DropTail and RED that command "packetqueue-attach" can leave the
>    queue in an incosistent state.  The code is (in {drop-tail,red}.cc):
> 
> 	if (!strcmp(argv[1], "packetqueue-attach")) {
> 		delete q_;
> 		if (!(q_ = (PacketQueue*) TclObject::lookup(argv[2])))
> 			return (TCL_ERROR);
> 		else {
> 			pq_ = q_;
> 			return (TCL_OK);
> 		}
> 	}
> 
>    When the TclObject::lookup() fails the pointer q_ is NULL but pq_,
>    inherited from the Queue class, still points to the old PacketQueue
>    which has already been deleted though.  When calling the length()
>    member function on the DropTail or RED object, Queue::length() will
>    access *pq_ which is invalid.

What you are saying is right, of course, but the command returns
TCL_ERROR, which means that the error is reported back to tcl.

 -Yuri