Bug 364216

Comment 1 Jing Qian 2011-11-21 20:11:16 EST

deferred

Comment 2 Huang Ji Yong 2011-12-29 00:50:41 EST

Created attachment 208837 [details]
fix for problem2

Comment 3 Huang Ji Yong 2011-12-29 00:55:03 EST

There are 2 problems here
1. modf::modfTest1: FAILED - modf1 - Failed: Expected value =
'0.03899999998975545' Actual value = '0.039' 
2. modf::modfTest2: FAILED - modf2 - Failed: Expected value = '771896' Actual
value = '0'

The fix for problem 2 is committed.

Regarding problem 1, according to the comment in bug 360863, the value should be the default of the platform.
There is no modf function in js, according to the definition of modf, I think the expected value should be '0.039'. Thus we may need to change test case for javascript.

Brian, what do you think?

Comment 4 Matt Heitz 2012-01-13 17:27:32 EST

I also think problem 1 is not a defect.

Comment 5 Huang Ji Yong 2012-01-15 20:34:54 EST

Route to Kathy to change the testcase.
If any problems, reopen it for me, thanks.

Comment 6 Brian Svihovec 2012-01-16 15:31:02 EST

I talked with Tim about this, and we agree that test 1 should print 0.039.

Comment 7 Kathy Carroll 2012-02-24 11:00:21 EST

I've modified the test variations based on comment 6.  Now Java gen 0.8 has new failures that aren't regressions from 0.70.  

Here are the failing variations:

package mini;

// basic library
library modfTester
	
    decimalX decimal(18, 6);
    decimalY decimal(18, 6);
    resultFloat float;
    bigIntY bigInt;
    
    y int; 
    x, result smallFloat;  
    
    aFloat float;
    answer float;
   	ir int;
   	aDecimal decimal(18, 6);

    function modfTest1(){@Test}
        decimalX = 000000771896.039000;
        decimalY = 000000771896.000000;
        bigIntY = decimalY as bigInt;
        resultFloat = mathLib.modf(decimalX, bigIntY);
        logresult.logStdOut(resultFloat);
        LogResult.assertFloatEqual("modf1 BUG 364216 reference comment 6", 0.039, resultFloat);
    end
    
    function modfTest2(){@Test}
        LogResult.assertBigIntEqual("modf2 BUG 364216", 771896, bigIntY);
    end
   
    function modfTest3(){@Test}
        x = 23.5678;
        try
            result = mathLib.modf(x, y);
        onException(oops AnyException)
            LogResult.failed("Exception");
            exit;
        end
        LogResult.assertFloatEqual("modf BUG 36217/364216 reference comment 6", 0.5678, result);
    end

    function modfTest4(){@Test}
        LogResult.assertFloatEqual("modf", 23, y);
    end

    function testModf6(){@Test}
	afloat = 3245.435;
      answer = MathLib.modf(afloat, ir);
     	LogResult.assertBigIntEqual("modf( float, smallint ) #2 integet", 3245, ir);
	LogResult.assertFloatEqual("modf( float, smallint ) #2 fractional", 0.435, answer);
    end

    function testModf7(){@Test}
      afloat = -987654.32109;
      answer = MathLib.modf(afloat, ir);    
      LogResult.assertBigIntEqual("modf( float, smallint ) #3 integet", -987654, ir);
	LogResult.assertFloatEqual("modf( float, smallint ) #3 fractional", -0.32109, answer);        
    end

    function testModf9(){@Test}
      adecimal = 3245.435;
      answer = MathLib.modf(adecimal, ir);
      LogResult.assertBigIntEqual("modf( decimal, smallint ) #2 integet", 3245, ir);
	LogResult.assertFloatEqual("modf( decimal, smallint ) #2 fractional", 0.435, answer);
    end

    function testModf10(){@Test}
      adecimal = -987654.32109;
      answer = MathLib.modf(adecimal, ir);
      LogResult.assertBigIntEqual("modf( decimal, bigint ) #3 integet", -987654, ir);
	LogResult.assertFloatEqual("modf( decimal, bigint ) #3 fractional", -0.32109, answer);  
    end
    
end

Comment 8 Jeff Douglas 2012-02-24 13:34:45 EST

The modf javagen runtime function gives an approximation for the decimals side of things, because the the new Decimal(numericField1) call. We are dealing with floating point values, so it is possible to have approximations made.

We discussed this in the scrum, and we are assigning this to the language defect list, to see what we want to do about it.

Comment 9 Matt Heitz 2012-02-24 13:54:41 EST

Part of the problem may be that the Java and JavaScript code give different results when converting from decimal to float.  

    function modfTest1(){@Test}
        x float = 771896.039e0;   // changed
        decimalY = 000000771896.000000;
        bigIntY = decimalY as bigInt;
        resultFloat = mathLib.modf(x, bigIntY);  // changed
        logresult.logStdOut(resultFloat);
        LogResult.assertFloatEqual("modf1 BUG 364216 reference comment 6",
0.039, resultFloat);
    end


Also, the little C program below says that the correct answer is 0.039.

#include <stdio.h>
#include <math.h>

main()
{
	double blah;
	printf( "%f\n", modf( 771896.039, &blah ) );
}

Comment 10 Matt Heitz 2012-11-18 22:35:00 EST

There's nothing more to do with this.  The code is working correctly.