| 3.3 hours ago |
vec1: over-optimization may cause floating point divide by zero exception | no replies | Open |
| 3.5 hours ago |
INSERT INTO d SELECT * FROM s INDEXED BY no_such_index can succeed instead of failing with "no such index" | 2 posts spanning 7.5 days | Resolved |
| 5.3 hours ago |
INSERT INTO parent SELECT * FROM an identically declared table makes COMMIT fail on a deferred foreign key | 2 posts spanning 7.4 days | Resolved |
| 5.8 hours ago |
17–125x slower queries to ALTER TABLE/ADD COLUMN on STRICT tables than on non-STRICT tables. | no replies | Open |
| 13.6 hours ago |
RENAME COLUMN silently drops the renamed column from a NATURAL JOIN in a view or trigger (with USING it is refused) | no replies | Open |
| 16.2 hours ago |
REPLACE that fails after resolving a UNIQUE conflict can leave the conflicting row deleted in a transaction | 2 posts spanning 7.0 days | Resolved |
| 16.2 hours ago |
Multi-row INSERT or UPDATE that fails with SQLITE_TOOBIG inside a transaction keeps earlier rows' changes | no replies | Open |
| 16.3 hours ago |
Failed multi-row INSERT or UPDATE on a STRICT table keeps earlier rows' changes inside a transaction | no replies | Open |
| 17.0 hours ago |
UPDATE that fails part-way inside json_each() can keep the rows it already changed inside a transaction | 2 posts spanning 6.9 days | Resolved |
| 17.0 hours ago |
UPDATE that fails with an integer overflow in sum() can keep the rows it already changed in a transaction | 2 posts spanning 6.9 days | Resolved |
| 21.2 hours ago |
Batch-atomic-write fallback overwrites the database before the rollback journal is synced | 2 posts spanning 31.2 hours | Resolved |
| 24.1 hours ago |
sqlite3's .recover output can smuggle shell commands into the next sqlite3 run | 3 posts spanning 5.3 days | Resolved |
| 35.1 hours ago |
Wrong result: in a UTF-16 database, CAST('1.5' || char(0x20AC) AS NUMERIC) gives the integer 1 | 2 posts spanning 6.2 days | Resolved |
| 35.4 hours ago |
Wrong result: two-argument substr(X, -9223372036854775808) drops the last character of X | 3 posts spanning 6.2 days | Resolved |
| 1.5 days ago |
Wrong result: the time shift modifiers '+24:00' and '-24:00' are accepted but leave the time-value unchanged | 3 posts spanning 5.1 days | Resolved |
| 1.5 days ago |
Wrong result: date('2024-02-30','start of day') returns the non-existent date '2024-02-30' | 3 posts spanning 7.0 days | Resolved |
| 1.5 days ago |
Wrong result: a second 'floor' modifier moves date('2024-01-31','+1 month','floor') back two more days | 3 posts spanning 7.0 days | Resolved |
| 1.6 days ago |
CLI: .mode --titlelimit 65540 acts as a limit of 4, and --linelimit 65537 as a limit of 1 | 2 posts spanning 5.0 days | Resolved |
| 1.6 days ago |
CLI: .mode box --escape off still shows a control character as a symbol and can split the value across lines | 2 posts spanning 4.9 days | Resolved |
| 1.6 days ago |
CLI: .mode html prints NULL as 'null' regardless of --null or .nullvalue | 2 posts spanning 4.9 days | Resolved |
| 1.7 days ago |
Failed DELETE inside a transaction is not rolled back when RAISE(IGNORE) blocks part of an ON DELETE CASCADE | 2 posts spanning 6.0 days | Resolved |
| 1.7 days ago |
A single-row INSERT ... VALUES that fails a foreign key check can leave the transaction unable to COMMIT | 2 posts spanning 6.0 days | Resolved |
| 1.7 days ago |
Planner regression: LEFT JOIN ... ON 0 plus GROUP BY makes an FTS5 MATCH run once per outer row | no replies | Open |
| 1.7 days ago |
Wrong result: EXISTS (SELECT DISTINCT ... LIMIT ... OFFSET ...) can return 1 when the subquery yields no rows | 2 posts spanning 6.7 days | Resolved |
| 1.8 days ago |
Wrong result: x=b AND x=1.0, with x a materialized CTE column and b TEXT, returns the row where x=b is false | 2 posts spanning 6.7 days | Resolved |
| ↓ Older... |