|
2019-05-14
| ||
| 20:25 | Fix a problem with the fix for [9cf6c9bb51] (commit [658b84d7]) that could cause a cursor to be left in an invalid state following a (rowid < text-value) search. (check-in: bc7d2c1656 user: dan tags: trunk) | |
|
2019-05-08
| ||
| 14:03 | • Ticket [9cf6c9bb51] "<" or "<=" comparison of rowid and non-numeric text value sometimes gets the wrong answer. status still Closed with 4 other changes (artifact: bbfde8a2e2 user: drh) | |
| 11:54 | • Ticket [9cf6c9bb51]: 3 changes (artifact: 8cbc48333c user: dan) | |
| 11:53 | • Closed ticket [9cf6c9bb51]. (artifact: 4ec82a65a6 user: dan) | |
| 11:52 | Fix VDBE opcodes OP_SeekLT and OP_SeekLE so that they work on intkey tables with non-numeric text values. Fix for [9cf6c9bb]. (check-in: 658b84d7f4 user: dan tags: trunk) | |
| 11:40 | • Ticket [9cf6c9bb51] "<" or "<=" comparison of rowid and non-numeric text value sometimes gets the wrong answer. status still Open with 6 other changes (artifact: 4622d12dc3 user: dan) | |
| 11:37 | • New ticket [9cf6c9bb51]. (artifact: c18df1c460 user: dan) | |
| Ticket Hash: | 9cf6c9bb51e53bded445bedb200de5cb516ce867 | ||
| Title: | "<" or "<=" comparison of rowid and non-numeric text value sometimes gets the wrong answer. | ||
| Status: | Closed | Type: | Code_Defect |
| Severity: | Severe | Priority: | Immediate |
| Subsystem: | Unknown | Resolution: | Fixed |
| Last Modified: |
2019-05-08 14:03:50 6.17 years ago |
Created: |
2019-05-08 11:37:44 6.17 years ago |
| Version Found In: | |||
| User Comments: | ||||
|
dan added on 2019-05-08 11:37:44:
In the following, the second SELECT statement returns 0 rows: CREATE TABLE t14(x INTEGER PRIMARY KEY); INSERT INTO t14(x) VALUES (100); SELECT * FROM t14 WHERE x < 'a' ORDER BY rowid ASC; SELECT * FROM t14 WHERE x < 'a' ORDER BY rowid DESC; dan added on 2019-05-08 11:40:33: This has been a problem since before version 3.8.0. dan added on 2019-05-08 11:54:06: Fixed by [658b84d7]. drh added on 2019-05-08 14:03:50: This problem was originally reported on the SQLite mailing list by Manuel Rigger. The original problem report used PRAGMA reverse_unordered_selects. However, as the script above shows, the problem has nothing to do with reverse_unordered_selects. The problem was a defect in the implementation of some of the (rarely used) comparison opcodes in the bytecode engine. | ||||