Name: cl74495 Date: 05/06/2000


The mapping files for Arabic iso8859-6 in sun.io convert numerals
incorrectly. Numerals in iso8859-6 are mapped into Hindu numerals
instead of Arabic (Western) numbers.
However the mapping for the windows Arabic encoding Cp1256 is correct!
This means that if you have an ASCII file containing only '0123456789'
numbers and then you read it in a files stream using the iso8859-6
encoding, these numbers will be converted into the Unicode Hindu
numbers: '\u0660\u0661\u0662\u0663\u0664\u0665\u0666\u0667\u0668\u0669'
The ASCII characters are not ASCII anymore eventhough ISO8859-6 is a
superset of ASCII.
.
This issue was inherited from the Unicode Group mapping files for Arabic
which have the same problem.
.
This issue affects all I/O operations in Java and cause all sorts of
errors if you are using iso8859-6.
.
The mapping between iso8859-6<---->UCS2 for numbers must be the same
as in Cp1256<---->UCS2.
.
The two affected files are:
sun.io.CharToByteISO8859_6.java & sun.io.ByteToCharISO8859_6.java
.
This bug reproduces for JDK1.1X JDK1.2.X and JDK1.3
(Review ID: 104483)
======================================================================