I apologize for the monologue, but I still have some issues with
*object instantiation of C++ classes in ns*
I read up on inherited classes and according to the book the constructor
of a base class is called when an object of an inherited class is
instantiated. Example:
class animal
{
public:
aninmal(){cout<<"making an animal"<<endl;};
~animal(){cout<<"unmaking an animal"<<endl;};
};
class dog: public animal
{
public:
dog(){cout<<"making a dog"<<endl};
~dog(){cout<<"making a dog"<<endl};
};
int main()
{
dog * bello = new dog;
cout<< " blah ..." <<endl;
delete bello; bello = 0;
return 0;
}
(should) lead to the following output:
making an animal
making a dog
blah ...
unmaking a dog
unmaking an animal
So the questions remain:
a) What happens in the mw-node code with the WirelessRoutingModule()
constructor? Why doesn't it seem to be called?
b) The constructor only calls init(), declared virtual function in
WirelessRoutingModule. Since BaseWirelessRoutingModule (derived from the
former) also has a init() function, the WirelessRoutingModule()
constructor should call the init() of BaseWirelessRoutingModule,
shouldn't it?
c) Resulting from that I followed Pedro's guide to get oTcl debugging
capabilities, but I couldn't get it to run with the mw-node patch. The
MASH inspector always kills ns with a segmentation fault when tryin gto
inspect a WirelessRoutingModule oTcl object. Anybody having some input
on that?
I apologize for this series of long emails, but somehow it seems that
there is some basic issues of C++/oTcl/ns2 which I do not understand.
Thanks in advance for any comment.
Claus
Claus Christmann wrote:
> OK, my mistake...
>
> BaseWirelessRouting module is derived from WirelessRoutingModule, but
> constructors and destructors are not inherited. (Which is a basic C++
> behavior...)
>
> Hence there is no constructor or destructor explicitly given and thus
> (an empyt) one is created. This auto generated constructor obviously
> does not call BaseWirelessRoutingModule::init() .
>
> Sorry for all the fuzz...
>
> Claus
>
>
> Claus Christmann wrote:
>> Hi list,
>>
>> I am working with Laurent's mw-node patch
>> (http://www.q2s.ntnu.no/~paquerea/ns.html) and am trying to get my own
>> routing protocol to work.
>>
>> Since my routing protocol utilizes several wirelss interfaces it should
>> be implemented as a WirelessRoutingModule akording to Laurent's
>> documentation. Unfortunately the constructor for this class seems to be
>> never called. I assume that is due to the oTcl/C++ duality, which I seem
>> to be unable to debug.
>>
>> The code under consideration is in wireless/wireless-rtmodule.cc, lines
>> 84pp:
>>
>> //! Constructor
>> /*! Create a new WirelessRoutingModule with an empty routing information
>> base.*/
>> WirelessRoutingModule::WirelessRoutingModule()
>> : RoutingModule(),lastUid_(-1),lastRibEntry_(0)
>> {
>> rib_ = new Rib();
>> init();
>> }
>>
>> This C++ class is used as a parent later on
>> (wireless/wirelss-rtmodule.h, lines 173pp):
>>
>> class BaseWirelessRoutingModule : public WirelessRoutingModule {
>> ...
>> }
>>
>> and BaseWirelssRoutingModule is used in the example provided with the
>> patch. Unfortunately BaseWirelessRoutingModule does not do anything in
>> init() (wireless/wireless-rtmodule.cc, lines 488pp)
>>
>> void
>> BaseWirelessRoutingModule::init()
>> {
>> // nothing to do.
>> }
>>
>> Thus it seemes not to matter, that at least using the provided tcl
>> script it seems never to be called.
>>
>> Am I doing somthing wrong in gdb?
>> *How is it possible that the constructor of a used class is never called?*
>>
>> Has anybody any experience in coding a routing protocol for the MW-node
>> patch or could point me to a source to look at?
>>
>> Thanks in advance,
>>
>> Claus
>>
>>
>
--
Claus Christmann
Graduate Research Assistant
Georgia Institute of Technology
270 Ferst Drive
Atlanta, GA 30332-0150
http://uav.ae.gatech.edu
