Re: BUG in cbq.cc?

[Sally Floyd <floyd@ee.lbl.gov>]

>It seems that there is a BUG in cbq.cc. 

Actually, I think it works fine as it is now.  I have just added a
validation test, "../../ns test-suite-cbq.tcl TwoDynamic", that
illustrates the correct operation of "$cbqclass newallot $bw".

...
>If the current allotment for 
>the pri_ is 0.9, then new allotment for pri_ is 1.3 which is more than 1. 

That is right, there is nothing in the code that prevents the total
allocation for a priority level to be more than 100% of the link 
bandwidth.  

>On the other hand, if the current allotment is 0.4 and new allotment is 1, 
>then it will execute addallot(pri_,-0.6). 
...

No, it will execute "addallot(pri_,0.6)"

I have added a validation test "../../ns test-suite-cbq.tcl TwoDynamic1" 
to show that "newallot" works correctly even for two
classes that are at the same priority level.  This test also
shows a case when the total allotment for classes in a priority
level is greater than 100%.  These tests should be
available in the tonight's nightly snapshot of the ns code.

- Sally
--------------------------------
http://www-nrg.ee.lbl.gov/floyd/
--------------------------------