No errors and a crash ! Yes I could use some help..

[codestain]

Don't know what to really put here.. You'd need to know some theory first..

Generally I'm trying to hide (watermark) information (text) into an image. I do that with ANDing the blue pixel with 254->"11111110". So the LSB of the pixel is 0 and then I add the bits of the text the user puts (after making the text to a binary table).

Variables aside, which you don't know what they represent without the UI, do you see any problems with the code below? Qt turns out no errors.. But when I run it, it crashes (windows 7).

So there goes:

void MainWindow::on_InsertWatermarkButton_clicked()
{
    if (imageloaded==false)
    {
        ui->ImagePreviewLabel->setText("No image is loaded, you idiot !");
    }
    else
    {
    QString watermarkdata = ui->WatermarkTextEdit->toPlainText();
    int length = watermarkdata.size();

    while (length%8!=0)							// for the Blowfish encryption
    {
    length++;
    }

    image->convertToFormat(QImage::Format_RGB888);                  // convert to 24-bit image, 8 bits per pixel

    numx = image->width();                                          //width in pixels probably
    numy = image->height();                                         //height

    if (3*7*(length*8)>3*numx*numy)                               //if more, it doesn't fit according to the key
    {
    QMessageBox msgBox;
    msgBox.setText("Watermark text is\ntoo big for this image.");
    msgBox.exec();
    }
    else
    {
        QString pass = ui->BlowfishLineEdit->text();
        int passlength = pass.size();

        char *key = new char[passlength];                       //table holding the bytes of password

        key = pass.toAscii().data();

        char *bytes = new char[length];                         //table holding the bytes of the watermark message

        bytes = watermarkdata.toAscii().data();

        CBlowFish blow( reinterpret_cast<unsigned char*>(key), passlength);

        blow.Encrypt(reinterpret_cast<unsigned char*>(bytes),length);       // Blowfish encryption


        int *binary=new int[length*8];			//this will hold the binary table

        QString *bin = new QString[256];						//look up table

                bin[0]="00000000";
                bin[1]="00000001";
                bin[2]="00000010";
                  ...........
                  ...........
                  ...........
                bin[253]="11111101";
                bin[254]="11111110";
                bin[255]="11111111";

        QString temp;				//temporary string
        int a=0;                                //counter
        for (int i=0;i<length;i++)
        {
            temp=bin[bytes[i]];

            binary[a]  = temp[0].digitValue();		//MSB
            binary[a+1]= temp[1].digitValue();		// .
            binary[a+2]= temp[2].digitValue();		// .
            binary[a+3]= temp[3].digitValue();		// .
            binary[a+4]= temp[4].digitValue();		// .
            binary[a+5]= temp[5].digitValue();		// .
            binary[a+6]= temp[6].digitValue();		// .
            binary[a+7]= temp[7].digitValue();		//LSB
            a=a+8;
        }


        for (int h=0;h<a;h++)		//converting from ascii to binary
        binary[h]-=48;


        int x,y;
        int g=3;

        int cnt=0;
        for (int s=0;s<length*8;s++)
        {
        x=3*7*g;						//random number generator

            if (x>numx)
            y=(int) (x/numx);

        QRgb value = image->pixel(x,y);
        int red = qRed(value);
        int green = qGreen(value);
        int blue = qGreen (value);
        blue = blue & 254;                                  //ANDing pixels with 254->"11111110"
        blue+=binary[cnt];                                  //adding information
        value = qRgb( red, green, blue);

        image->setPixel(x, y, value);

        cnt++;
        g++;
        }
        delete [] key;
        delete [] bin;
        delete [] bytes;
        delete [] binary;

        QString temp2= "Watermarked ";
        QString savefilename = temp2 + filename;
        image->save(savefilename,"BMP",100);

        ui->ImagePreviewLabel->setPixmap(QPixmap::fromImage(*image));
        ui->WatermarkTextEdit->setPlainText("Watermarked:\nImage is saved in a Bitmap format");
        delete image;
        }
    }

}

[danhicksbyron]

How does it crash? Have you tried putting a breakpoint on qFatal?

[danhicksbyron]

I'll admit that my C++ was learned in the gutters, but this sequence seems flaky to me:

        char *key = new char[passlength];                       //table holding the bytes of password

        key = pass.toAscii().data();

        char *bytes = new char[length];                         //table holding the bytes of the watermark message

        bytes = watermarkdata.toAscii().data();

The first line allocates a string "passlength" long and stores it in "key", but the very next line then stores whatever "pass.toAscii().data()" returns over the top of the string pointer you just stored. Same thing for "bytes". If nothing else, it's a storage leak (unless the "data" function returns a type which has an assignment operator to char* that uses the existing string -- pretty obscure).

That binary array conversion stuff is kind going around behind Robin Hood's barn to get things done, but can't seen anything to it that should cause a crash. (Why not just code the "bin" array as octal or hex constants, and skip the conversion entirely, or maybe code them like "0x10110101" and then just compress the hex digits to bits?)

Most of the rest seems unlikely to cause a crash -- are you sure it's not due to one of the routines you're calling from your code? (Getting a stack traceback would definitely help here.)

[codestain]

I'll admit that my C++ was learned in the gutters, but this sequence seems flaky to me:

        char *key = new char[passlength];                       //table holding the bytes of password

        key = pass.toAscii().data();

        char *bytes = new char[length];                         //table holding the bytes of the watermark message

        bytes = watermarkdata.toAscii().data();

The first line allocates a string "passlength" long and stores it in "key", but the very next line then stores whatever "pass.toAscii().data()" returns over the top of the string pointer you just stored. Same thing for "bytes". If nothing else, it's a storage leak (unless the "data" function returns a type which has an assignment operator to char* that uses the existing string -- pretty obscure).

That binary array conversion stuff is kind going around behind Robin Hood's barn to get things done, but can't seen anything to it that should cause a crash. (Why not just code the "bin" array as octal or hex constants, and skip the conversion entirely, or maybe code them like "0x10110101" and then just compress the hex digits to bits?)

Most of the rest seems unlikely to cause a crash -- are you sure it's not due to one of the routines you're calling from your code? (Getting a stack traceback would definitely help here.)

Thanks for the reply,
I'm not a programmer too, and I'm also new to Qt :]

pass is a QString.
And
int passlength = pass.size(); (how many bytes is it)

also

QByteArray QString::toAscii () const (http://qt.nokia.com/doc/4.6/qstring.html#toAscii)

Returns an 8-bit ASCII representation of the string as a QByteArray.

and

char * QByteArray::data () (http://qt.nokia.com/doc/4.6/qbytearray.html#data)

Returns a pointer to the data stored in the byte array. The pointer can be used to access and modify the bytes that compose the array. The data is '\0'-terminated, i.e. the number of bytes in the returned character string is size() + 1 for the '\0' terminator

Do you see anything wrong with it?
Thanks

[divanov]

char * QByteArray::data () (http://qt.nokia.com/doc/4.6/qbytearray.html#data)


Do you see anything wrong with it?
Thanks

QByteArray::data () returns pointer to internal data stored in QByteArray and then you call delete on it:

key = pass.toAscii().data();
bytes = watermarkdata.toAscii().data();

delete [] key;
delete [] bytes;

Also note, this forum is devoted to Qt questions, not to general C++ programming.

[codestain]

I call delete when the entire process is done..
Why not delete, it will eat your memory for no use.. But I tried removing those lines, still crashes..

Also note, this forum is devoted to Qt questions, not to general C++ programming

Don't you consider questions based on the Qt Class reference, Qt questions?

Still, I appreciate your help!

[divanov]

Why not delete, it will eat your memory for no use.. But I tried removing those lines, still crashes..

Byte array owns that data, you cannot free it. Consider this test application:

#include <QCoreApplication>
#include <QByteArray>
#include <QDebug>

int
main (int argc, char **argv)
{
    QCoreApplication app(argc, argv);
    QByteArray array("test array");

    /* Very stupid! */
    delete [] array.data();

    return app.exec();
}

Don't you consider questions based on the Qt Class reference, Qt questions?

If you misuse Qt code do to lack of basic C++ knowledge it doesn't really make it Qt question.

You can ask on Windows developers' forum how to produce crash backtrace on Windows platform. It's basic knowledge and it will allow you to find every crash reason in 1 minute.

[danhicksbyron]

pass is a QString.

I missed that detail. Definitely no need to allocate the new arrays.

I agree with divanov that deleting those strings is at least partly the problem.

You'd have a MUCH better idea if you could get a traceback -- have you tried it with that breakpoint on qFatal??

[codestain]

Debugger shows me that it crashes inside this loop, when it goes for the second round:

        for (int i=0;i<length;i++)
        {
            temp=bin[bytes[i]];

            binary[a]  = temp[0].digitValue();		//MSB
            binary[a+1]= temp[1].digitValue();		// .
            binary[a+2]= temp[2].digitValue();		// .
            binary[a+3]= temp[3].digitValue();		// .
            binary[a+4]= temp[4].digitValue();		// .
            binary[a+5]= temp[5].digitValue();		// .
            binary[a+6]= temp[6].digitValue();		// .
            binary[a+7]= temp[7].digitValue();		//LSB
            a=a+8;
        }

The inferior stopped because it received a signal from the Operating System.

Signal name:SIGSEGV
Signal meaning: Segmentation fault

[divanov]

Debugger shows me that it crashes inside this loop, when it goes for the second round:

        for (int i=0;i<length;i++)
        {
            temp=bin[bytes[i]];

            binary[a]  = temp[0].digitValue();		//MSB
            binary[a+1]= temp[1].digitValue();		// .
            binary[a+2]= temp[2].digitValue();		// .
            binary[a+3]= temp[3].digitValue();		// .
            binary[a+4]= temp[4].digitValue();		// .
            binary[a+5]= temp[5].digitValue();		// .
            binary[a+6]= temp[6].digitValue();		// .
            binary[a+7]= temp[7].digitValue();		//LSB
            a=a+8;
        }

"bytes" is actually internal array of bytes belonging to watermark. "length" is length of "watermark" string ceiled to closest number proportional to 8. Thus, "i" will go out of "bytes" every time your string length is not proportional to 8.

[codestain]

"bytes" is actually internal array of bytes belonging to watermark. "length" is length of "watermark" string ceiled to closest number proportional to 8. Thus, "i" will go out of "bytes" every time your string length is not proportional to 8.

bytes has size of length, and length is always multiple of 8. bytes will have some empty slots if watermarkdata is not multiple of 8, since i increase length till it's multiple of 8.

[divanov]

bytes has size of length, and length is always multiple of 8. bytes will have some empty slots if watermarkdata is not multiple of 8, since i increase length till it's multiple of 8.

As I said already, bytes belongs to QByteArray and any manipulations with length doesn't have any effect to it.

[codestain]

This code-logic actually works with MFC classes. I have no idea why it doesn't here..

[divanov]

This code-logic actually works with MFC classes. I have no idea why it doesn't here..

Actually not:
http://www.augustcouncil.com/~tgibson/tutorial/arr.html

[codestain]

Actually not.

Trust me it works..

If you have an idea just post.

Yes. I'm not a programmer.