Prefix Index Scan Returns Wrong Results for Unicode String Range Comparison

[ZyanNo1]

1. Minimal reproduce step

DROP DATABASE IF EXISTS bug_repro;
CREATE DATABASE bug_repro;
USE bug_repro;
CREATE TABLE t (c2 VARCHAR(32) CHARACTER SET utf8mb4) ENGINE=InnoDB;
INSERT INTO t VALUES ('#а|'), ('ׄׄ׉');
CREATE INDEX pfx ON t (c2(1));
ANALYZE TABLE t;

-- [CORRECT] Full table scan (IGNORE INDEX)
SELECT '[CORRECT] IGNORE INDEX' AS plan, COUNT(*) FROM t IGNORE INDEX(pfx)
 WHERE c2 > '#а|';

-- [BUG] Prefix index scan (FORCE INDEX)
SELECT '[BUG]    FORCE INDEX' AS plan, COUNT(*) FROM t FORCE INDEX(pfx)
 WHERE c2 > '#а|';

2. What did you expect to see?

Both queries should return 1. The value 'ׄׄ׉' (starts with Hebrew character U+05C4, UTF-8: 0xD7 0x93) compares greater than '#а|' (starts with ASCII '#', UTF-8: 0x23) in both byte order and Unicode code point order. The prefix index on c2(1) stores the first character which is sufficient to determine 'ׄ' > '#':

plan                    COUNT(*)
[CORRECT] IGNORE INDEX  1
[BUG]    FORCE INDEX    1

3. What did you see instead

The FORCE INDEX query returns 0 — the prefix index scan incorrectly excludes the Hebrew row 'ׄׄ׉':

plan                    COUNT(*)
[CORRECT] IGNORE INDEX  1
[BUG]    FORCE INDEX    0

Root cause: When evaluating the range condition c2 > '#а|' via a prefix index on c2(1), the index scan path incorrectly determines that the Hebrew character 'ׄ' (U+05C4) does NOT compare greater than '#' (U+0023), even though it does. The prefix index should be able to use just the first character for this comparison, but the comparison logic for multi-byte UTF-8 characters in the prefix index scan path is flawed.

4. What is your PolarDB-X version?

VERSION(): 5.6.29-PXC-5.4.19-20250825

Docker image: polardbx/polardb-x:v2.4.2_5.4.19