Re: SCRAM authentication, take three

On Fri, Mar 3, 2017 at 2:43 PM, Michael Paquier
<michael.paquier@gmail.com> wrote:
> I am attaching 0009 and 0010 that address those problems (pushed on
> github as well) that can be applied on top of the latest set.

While doing more tests with my module able to do SASLprep, I have
noticed that calculations related to Hangul characters were incorrect:
 /* Constants for calculations wih Hangul characters */
-#define SBASE      0xAC00
-#define LBASE      0x1100
-#define VBASE      0x1161
-#define TBASE      0x11A7
+#define SBASE      0xEAB080    /* U+AC00 */
+#define LBASE      0xE18480    /* U+1100 */
+#define VBASE      0xE185A1    /* U+1161 */
+#define TBASE      0xE186A7    /* U+11A7 */

Once the following is applied things get better:
-- Test for U+FB01, Latin Small Ligature Fi
=# select array_to_utf8(pg_sasl_prepare(utf8_to_array('fi')));
 array_to_utf8
---------------
 fi
(1 row)
-- Test for U+1E9B, Latin Small Letter Long S with Dot Above
-- This one was failing previously.
=# select array_to_utf8(pg_sasl_prepare(utf8_to_array('ẛ')));
 array_to_utf8
---------------
 ṡ
(1 row)
-- Test for U+2075, superscript 5
=# select array_to_utf8(pg_sasl_prepare(utf8_to_array('⁵')));
 array_to_utf8
---------------
 5
(1 row)
-- 
Michael