RE: [Dev-C++] memory violation....

[Ioannis V. <no...@ho...>]

> -----Original Message-----
> From: dev...@li... 
> [mailto:dev...@li...] On Behalf 
> Of Raghunath Marella.S
> Sent: Monday, June 10, 2002 10:25 AM
> To: Dev-C++ Mailing List
> Subject: RE: [Dev-C++] memory violation....
> 
> 
> Hi Ioannis/Jason/Others,
> 
> Please find the piece of code where the memory
> violation is seen. I have not deleted the pointer and
> accessed after that. No error has been done from my
> side.


It is 100% some logical error (bug). To avoid situations like this and
have bullet proof code use the standard library containers instead of
explicit memory allocation and deallocation. Examples can be seen below.

#include <vector>

using namespace std;


>  int *EleTopo=new int[seneles*sennds];

vector<int>EleTopo(seneles*sennds);


>  char *NodeSign=new char[this->NNodes+1];

list, or vector, or string, or something...



> Similarly few other variables have been dynamically
> allocated....
> 
>  for(int i=0;i<seneles;i++)
>    for(int j=0;j<ncn;j++)
>    {
>      int cuEl = (theSubEle+ise)->SEtopo[i];
> 
>      *(EleTopo+i*ncn+j)=(theElement+cuEl-1)->topo[j];
>      NodeSign[(theElement+cuEl-1)->topo[j]] =
> 'P';//All nodes are marked as positive initially.
>    }
>  After this piece of code, 'EleTopo'  values are
> printed and they confirmed to be correct.


It is too messy! *(EleTopo+i*ncn+j)? What about the equivallent
EleTopo[i*ncn+j] (which is somewhat messy too). Obviously you come from
a hardcore C background and carry all that low stuff bagage. :)



> 
>  After few lines of other code.....
> 
>  for(int iel=seneles-1;iel>=0 ; iel--)
>  {
> 	int iLocalForEle=0;
> 	int curElement = (theSubEle+ise)->SEtopo[iel];
> 	if(iel==seneles-1)
> 		coutf << "\n Final element in the SE is...:"<<iel;
> //newAdd
> 	int AnyNode;
> 
> 
> 	int  noOfDf = 0;
> 	NoOfNegNodes =0;
> 	for(int j=0;j<ncn;j++)
> 	{
> 	   AnyNode = *(EleTopo+iel*ncn+j);
> 	   int isThisNodeABN = 0;// is this a boundary node
> of SE.
> 	   for(int it=0;it<SeBNodes;it++)
> 	   {
> 	    	if(AnyNode ==
> ((theSubEle+ise)->SEBounNodes[it]))
> 		{
> 			isThisNodeABN = 1;
> 			break;
> 		}
> 		else
> 		isThisNodeABN = 0;
> 	   }
> 
> 	   if(isThisNodeABN==0)
> 	   {
> 
> ******	   	if(NodeSign[*(EleTopo+iel*ncn+j)]=='N')
> 	   	{
> 			//Do nothing...
> 	   	}
> 	   	else
> 	   	{
> 			NodeSign[*(EleTopo+iel*ncn+j)]='N';
> 			int nde = *(EleTopo+iel*ncn+j);
> ******			*(EleTopo+iel*ncn+j) = 
> -*(EleTopo+iel*ncn+j);
> 			NoOfNegNodes++;
> 
> 			//saving the the dfs that are going to 
> be removed
> after this element being assembled.
> 			for(int ir=0;ir<NDF;ir++)
> 			{
> 			  *(DfToRemove+iel*(ncn*NDF)+iLocalForEle)=
> (nde-1)*NDF+ir;
> 			  iLocalForEle++;
> 			}
> 			  noOfDf+=NDF;
> 	   	}
> 	   }
> 	   else
> 	      cout <<"\n this Node is a BoundaryNode of
> SE...:"<<AnyNode<<".."<<ise;
> 	}
> 	noOfDf = iLocalForEle;
> 	NoNegDf[iel]=NDF*NoOfNegNodes;
>  }//sign check ends here.....



If any exception is thrown from somewhere there, much of all those
allocated resources remain allocated.



> 
> 
>  The problem appears in the line marked with ******.
> Actually this problem is seen only in some of the
> problems. To give you more idea, I have tested this
> code  as standalone, it worked fine for 'EleTopo'
> sized upto 3200*6, then it worked  properly and no
> memory violation was observed.
> 
>  But, when the code used as a server implementation in
> CORBA environment, the  so called memory violation
> took place when 'EleTopo' was sized to 800*6. Briefly,
> 
>  'EleTopo' is like an two dimentional array of size
> [800][6]. I have to access  data in 'EleTopo' from the
> end, ie., from [800][].

[799][].



 The values are correct till
> [48][6] and started giving junk values for some of the
> indexes between [47][] to [0][].
> 
>  I have shown in array form just for illustration
> purpose.
> 
>  I am clueless how this memory violation is taking
> place in some cases only.
> 
>  Will you pls suggest me what is going wrong in this.


What is going wrong basically here is that you use C++ in the place of
assembly programming. Deal C++ as a high level programming language and
you can even gain some time/space! I suggest you read "The C++
Programming Language" 3rd Edition or Special Edition. And be prepared
for a life change. :)


Ioannis

* Ioannis Vranos
* Programming pages: http://www.noicys.cjb.net
* Alternative URL: http://run.to/noicys