Review correctness of integer division in floating point context

[dekobon]

See:

AesCbcCipherDetails.java:124

[indianwhocodes]

This was an extremely fun bug to verify since it's nature was mathematical, I also got a chance to use the Wolfram-Alpha-Calculator which is a fun-tool like MATLAB to perform complex mathematical operations. To surmise, Math.floor() works perfectly fine since its returns the lesser than closest integer towards +infinity which is essentially required to calculate the blocks for contentLength in AESCBCCiphers while running AesCbcCipherDetailsTest#calculateContentLengthWorks().
I understand why you reported this, since the aforementioned function is a bit generic nevertheless we can safely close #479.

[indianwhocodes]

Although, I suggest I should come up with a pull-request to define the test-group:unlimited-crypto in the testng.xml file.

[dekobon]

Ok, I just took a look at this. This is interesting:

    public static void main(String[] argv) {
        int a = -902342342;
        int b = 3233337;
        int intDivision = a / b;
        double floor = Math.floor(a / b);
        double floorDiv = Math.floorDiv(a, b);
        double casted = (double)a / (double)b;
        double floorCasted = Math.floor(casted);

        System.out.printf("a / b as int = %d\n", intDivision);
        System.out.printf("floor(a / b) as double = %f\n", floor);
        System.out.printf("floorDiv(a / b) as double = %f\n", floorDiv);
        System.out.printf("floor((double)a / (double)b) as double = %f\n", floorCasted);
    }

Output:

a / b as int = -279
floor(a / b) as double = -279.000000
floorDiv(a / b) as double = -280.000000
floor((double)a / (double)b) as double = -280.000000

All three values are the same if a and b are positive. In our case, Math.floorDiv() might make the most sense because it converts the ints to a double and does the floor operation all at once.

What are your thoughts?

[indianwhocodes]

I agree, Also within this code context (AesCbcCipherDetails.java#L124) the return type for Math.floorDiv() would be long which is what we need here.
The aforementioned approach would also reduce the number of type-casting we perform supplemented by the fact that it's a relatively more accurate method of performing this mathematical operation and checks for Overflow-Error.

[indianwhocodes]

Fixed by Pull-Request:#500 .