JDK-4238709 : AWT Arabic Windows handling of Arabic and European number contexts
  • Type: Bug
  • Component: client-libs
  • Sub-Component: java.awt
  • Affected Version: 1.3.0
  • Priority: P4
  • Status: Open
  • Resolution: Unresolved
  • OS: generic
  • CPU: x86
  • Submitted: 1999-05-17
  • Updated: 2005-12-29
Related Reports
Relates :  
Description

Name: vr73486			Date: 05/17/99

This bug report supercedes 4212598 which was poorly written.

The bidi test applet numeralbuttonA.java version 1.2 was tested using
Cricket 1.2.2 "S" on three Arabic Windows platforms, 95, 98 and NT4.0.

The applet contains the following test cases:

1. A button with an Arabic number 
2. A button with Arabic text followed by an Arabic number
3. A button with a European number
4. A button with Arabic text followed by a European number
5. A button with a European number followed by Arabic text

By "Arabic number" we mean a run of Unicode characters from the range
0660 to 0669. By "European number" we mean a run of Unicode characters
in the range 0030 to 0039.

We noted the following results.

1. All platforms displayed the first button, Arabic number, as expected.

2. NT displayed the Arabic text to the right of the Arabic number, as expected.
   98 displayed the Arabic text to the left of the Arabic number. This is a bug.
   95 displayed the Arabic text without joining to the left of the Arabic
   number this is a bug.

3. All platforms displayed the European number correctly.

4. NT displayed the Arabic text followed by an Arabic number of the same
   value as the European number. That is, since the characters in the
   context were Arabic, it displayed the European number using glyphs that
   are normally used for Arabic numbers. This was unexpected but not
   necessarily incorrect, at least from the user view.

   98 displayed the same behavior as NT.

   95 displayed the Arabic text followed by the European number, which was
   the expected behavior.

5. All platforms displayed the European number to the left of the Arabic text,
   as expected.

Summary:

NT behaves acceptably in all cases.
Both NT and 98 use Arabic glyphs for European numbers when the context is
Arabic.
98 reverses the word order when Arabic numbers follow Arabic text, a bug.
95 looses context when Arabic numbers follow Arabic text.

I do not know how much of the above behavior is caused by JRE and how much
is caused by Windows.

WorkAround:
======================================================================

Comments
EVALUATION Commit to Merlin. ====================================================================== Should test this to see if it still exists after 4341196 is integrated. ###@###.### indicated that some problems with entering Arabic glyphs disappeared when we converted from common edit controls to rich edit controls (he tested with Arabic Windows98). eric.hawkes@eng 2000-10-31
31-10-2000