Input problems with DJGPP?

Hi,

I´ve written this program to input some data, and make some calculation on this data. Problem is that I cannot get the input working. I´ve copied the code from another program of mine where there are no problems. I´m using DJGPP, and I´ve tried to compile it on Solaris too. On Solaris it works fine - THE SAME CODE!!1

Can anyone help me explain this??

Code is as goes:

Code:

#include <math.h>
#include <stdlib.h>
#include <fstream>
#include <iostream>
using namespace std;

void calcProbs(unsigned char* data ,int* probs,int size);
double getEntropy(int* probs,int size);
void makeDiffs(unsigned char* data,int size);

#define MAXFILESIZE 1000000

int main(int argc,char** argv){
	unsigned char* data;					// Data buffer
	unsigned char* data2;				// Data buffer 
	int i,j;									// Loop vars
	int probs[256];						// Probability vars
	double entropy;						// Calculated Entropy
	ifstream in;							// Input Filestream
	char* temp;								// Temporary var


	if(argc!=2){ cerr << "Wrong number of arguments" << endl; return 0;}

	data=(unsigned char*) malloc(MAXFILESIZE);		// Alloc some memory
	if(data==NULL){ return 0; } 	

	data2=(unsigned char*) malloc(MAXFILESIZE);		// Alloc some memory
	if(data2==NULL){ free(data); return 0; } 	

	i=0;
	in.open(argv[1],ifstream::in);						// open file
	while(in.good()){
		data[i++]=in.get();
	}
	in.close();													// close file
																	// Read input!
	
	cout << i << endl;										// Output numbers of bytes read

	calcProbs(data,probs,i);
	entropy=getEntropy(probs,i);	
	cout << "original entropy: " <<  entropy << endl;

	makeDiffs(data,i);
	calcProbs(data,probs,i);
	entropy=getEntropy(probs,i);	
	cout << "difference entropy: " << entropy << endl;

	free(data);
	free(data2);
}

void calcProbs(unsigned char* data ,int* probs,int size){
	int i;
	unsigned char c;
	for(i=0;i<256;i++){
		probs[i]=0;
	}	
	for(i=0;i<size;i++){
		c=data[i];
		probs[c]++;
	}
}
double getEntropy(int* probs,int size){
	int i;
	double size2,temp,temp2;
	temp2=0.0;
	size2=(double) size;
	
	for(i=0;i<256;i++){
		if(probs[i]){
			temp=(double) probs[i];
			temp/=size2;
			temp2+=temp*log(temp);
		}
	}
	temp2/=log(2.0);	
	temp2=-temp2;
	return temp2;
}



void makeDiffs(unsigned char* data,int size){
	unsigned char c,d;
	int i;
	c=0;
	for(i=0;i<size;i++){
		d=data[i];
		data[i]=d-c;
		c=d;
	}	
}

>>Problem is that I cannot get the input working
Care to be a bit explanatory than "it doesn't work" ?

>>while(in.good()){
>>data[i++]=in.get(); }
This is a buffer overflow just waiting to happen.

[edit]
>>in.open(argv[1],ifstream::in);
Check the file open worked OK.

>>>while(in.good()){
>>>data[i++]=in.get(); }
>This is a buffer overflow just waiting to happen.

I know that, but right now it´s only in test phase, I´ll adjust that later.

>>>in.open(argv[1],ifstream::in);
>Check the file open worked OK.

Ok, I should do that, but thats not what´s causing my problem.

Any suggestions??

>>Any suggestions??
Yes, for the second time:
>>Care to be a bit more explanatory than "it doesn't work" ?

I compiled/run it, and it said:

Code:

4764
original entropy: 4.26253
difference entropy: 5.07116

I've got no idea what it is supposed to do, or what it is doing wrong. So why don't you tell us, then we can help you better.

Sure! It´s a program which calculates the entropy of a source. The entropy is a statistic measure, and the calculations are working like they should. The problem is, that when run on window - compiled with DJGPP, it doesn´t load all the file I specify. You´ll see, first it outputs how many bytes read, which should equal the size of the file specified when runned. Under Windows - compiled with DJGPP, it just don´t load the hole file. More strange is, that when runned on Solaris - compiled with GCC - no problems. I´d like to here some suggestions as it appear strange to me...

>> it just don´t load the hole file
Do you mean it's a few bytes short? If so, it's because the file is being opened in text mode, and the CR LF bytes that mark and end of line are combined into one \n in your program. On *nix, an end of line marker is LF only, so there is a one to one relationship.

If you want to get the "exact" data into your program, open with binary mode (ios::binary). This will then work on both *nix and Windows.

Should that be like:

Code:

in.open(argv[1],ifstream::in|ios::binary);

Or how should I specify it?

Another things is - it´s not just a few bytes short, but it depends much on the file specified. You can have a look yourself!

That should do.

>>You can have a look yourself!
I did There's still a problem though, you might find it's one byte out. If you can't work that one out, ask again...

That helped! But - yeah - I´m getting a byte too much... Do you know why, cause I can´t understand this one as well....

Make your read loop like this:

Code:

int ch;
while((ch = in.get()) != EOF){
  data[i++]=ch;
}

This explains the problem, but in C terms.

Thanks! That solved all my problems.