Looked around the source. I think org.codehaus.groovy.runtime.TimeCategory (in src/main) is our class. And the test will go in the same package of src/test folder?

Things you should note:

• the fixme_testDurationArithmetic() method won't be run as part of the normal test suite (just remove the 'fixme_' part of the name) to enable it. This was because it was breaking the build whenever daylight saving was starting or stopping.
• you can find out coverage information about what is tested from the coverage report on the CI server: http://build.canoo.com/groovy/artifacts/20070617013352/reports/cobertura/index.html

Here is the test method that is failing (just sample code to demonstrate the issue):

/**

• Months evaluates relative to the current date, so if this is
• July then months has 31 days. If this is November then months
• has 30 days. The test uses java.util.Calendar to determine if
• groovy works correctly.
• WARNING: This test will sometimes fail if the month changes during
• the execution of the test method. Incredibly rare event.
  */
  void testMonths()
  Unknown macro: { use(TimeCategory) { def daysInMonth = Calendar.getInstance().getActualMaximum(Calendar.DAY_OF_MONTH) def actual = new BigInteger(1.month.toMilliseconds()); def expected = new BigInteger(daysInMonth); expected = expected.multiply(new BigInteger(1000)) //milliseconds expected = expected.multiply(new BigInteger(60)) //seconds expected = expected.multiply(new BigInteger(60)) //minutes expected = expected.multiply(new BigInteger(24)) //hours println("actual = " + actual); println("expected = " + expected); assert actual == expected } }

The assertion fails. The output is:
actual = 2682000000
expected = 2678400000

It is as if the TimeCategory is calculating the number of days in October as 31.041 and two thirds.

Any suggestions?

It is out by an hour. This is to be expected if your system is set to a timezone which has daylight saving starting/ending.

A user's regional settings, timezone, and DST status should not effect the outcome of these tests.

The test case contains 4 todo notes, which correspond to the places in the test effected by the defect discussed on the mailing list (http://www.nabble.com/TimeCategory-subtraction-not-working--p13246074.html)

Attached file is output of "svn diff" command.

Please let me know if anything is out of order or needs changing.