Name: dfC67450 Date: 02/13/98



No specifications are available for locale specific data. Because of
total absence of any specifications for locale date pattern, number pattern,
month names etc. it is impossible to write JCK tests for classes
java.txt.***Format.


======================================================================

Name: mgC56079 Date: 02/23/98



The JCK tests are not (and are not going to be) based on the internal
representation of locale-specific data (java.text.resources package).
But the tests have to be based on the data itself because
many of the i18n-related methods (NumberFormat.getInstance(),
DateFormatSymbols.getMonthNames(), DateFormat.getDateInstance() etc)
depend on this data.

The intent of the JCK tests is to verify that these methods return the same
values in all implementations. The tests should be based on specification.

The problem is that different JDK builds produce different results and
it is not clear from specification which output is correct.
It should be specified what output to expect in each supported locale
for each particular method call (this means the locale-specific data should
be documented in some way).

It should also be specified which locales are required to be supported by any
implementation and which are optional (but should be compatible with specification
and reference implementation if supported).

The following are examples of the differences between jdk1.1.4 and jdk1.2:
--- ErasTest.java ---
import java.text.*;
import java.util.*;

class ErasTest {
    public static void main(String[] args) {
        Locale locale = Locale.CANADA_FRENCH;
        DateFormatSymbols dfs = new DateFormatSymbols(locale);
        String eras[] = dfs.getEras();
        System.out.println("locale: " + locale.getLanguage() + "_"
                                    + locale.getCountry());
        System.out.println("Eras: {\"" + eras[0] + "\", \"" + eras[1] + "\"}");
    }
}
---------output for jdk1.1.4 -----------
locale: fr_CA
Eras: {"BC", "AD"}
----------------------------------------
---------output for jdk1.1.6 , jdk1.2beta3G -----------
locale: fr_CA
Eras: {"BC", "ap. J.-C."}
----------------------------------------

Yet another test demonstrating the changes:

--- CurrencyInstanceTest.java ---
  
import java.text.*;
import java.util.*;

class CurrencyInstanceTest {
    public static void main(String[] args) {
        Locale locale;
        double num;

          // Italian locale
        locale = Locale.ITALIAN;
        num = 123.567;
        System.out.println("locale: " + locale.getLanguage() + "_"
                                    + locale.getCountry());
        System.out.println(num + " is formatted as: \"" +
          NumberFormat.getCurrencyInstance(locale).format(num) + "\"");

          // Italian locale
        locale = Locale.JAPANESE;
        num = 333.777;
        System.out.println("locale: " + locale.getLanguage() + "_"
                                    + locale.getCountry());
        System.out.println(num + " is formatted as: \"" +
          NumberFormat.getCurrencyInstance(locale).format(num) + "\"");
    }
}
---------output for jdk1.1.4 -----------
locale: it_
123.567 is formatted as: "L. 123,57"
locale: ja_
333.777 is formatted as: "¥334"
----------------------------------------
---------output for jdk1.1.6, jdk1.2beta3G -----------
locale: it_
123.567 is formatted as: "L. 124"
locale: ja_
333.777 is formatted as: "¥333.78"
----------------------------------------


======================================================================