50 most recent check-ins
|
2026-09-30
| ||
| 10:37 | Correct the xCheckReserveLock() impl of the OPFS SAHPool VFS to fix a corruption case involving multiple handles in the same JS thread to the same filename when one leaves a journal behind. Addresses bug report 2026-09-29T18:25:28Z. (leaf check-in: ea1d55e202 ... user: stephan tags: branch-3.53) | |
| 10:36 | Minor JS doc correction. (leaf check-in: 6b470917b2 ... user: stephan tags: trunk) | |
| 09:53 | Correct the xCheckReserveLock() impl of the OPFS SAHPool VFS to fix a corruption case involving multiple handles in the same JS thread to the same filename when one leaves a journal behind. Addresses bug report 2026-09-29T18:25:28Z. (check-in: 9168a6f1be ... user: stephan tags: trunk) | |
| 09:17 | Generic internal JS cleanups. No functional changes. Update the preprocessor for the ability to use C-style comments in directive arguments. (check-in: b2699eb967 ... user: stephan tags: trunk) | |
|
2026-09-29
| ||
| 19:42 | Change the way the session module applies changesets containing two or more UPDATE changes that form a dependency loop in order to avoid triggering ON DELETE CASCADE actions. (leaf check-in: 05acfd4dc9 ... user: dan tags: session-update-to-update) | |
| 18:49 | An early attempt to implement the tree-tracker optimization. (leaf check-in: ead82958b3 ... user: drh tags: tree-tracker) | |
| 14:25 | Add the preserveOnInitFailure option to the opfs-sahpool VFS. (check-in: f2d3432320 ... user: stephan tags: trunk) | |
| 12:02 | Fix a harmless compiler warning in fts3ChecksumIndex(). (check-in: 44e0d3e2ad ... user: drh tags: trunk) | |
| 11:52 | Reset the nullRow flag on the pseudo-cursor on each iteration of the loop when the RHS of a LEFT JOIN is a multi-row VALUES clause. Fix for Bug 2026-09-29T00:41:09Z. Problem introduced by check-in [a120c9235f125e05] with a missed opportunity to fix the problem at check-in [8c0f69e0e4ae0a44]. (check-in: 8e59d3c0a4 ... user: drh tags: trunk) | |
| 08:31 | OPFS SAHPool: more cleanly handle an init failure. Add an init option which tells it not to wipe the storage if initialization fails. Based on the discussion and patch in forum post 5664cd4b. Minor doc updates in unrelated code. (closed check-in: 0560c3cc20 ... user: stephan tags: opfs-sahpool-forum-5664cd4b) | |
|
2026-09-28
| ||
| 16:30 | Fix a duplicate test case identifier label in speedtest1.c. (check-in: 6f7db1d731 ... user: drh tags: trunk) | |
|
2026-09-27
| ||
| 15:54 | Enable the --column-metadata configuration flag in the 'autoconf' build. It was previously disabled with a comment that it changed how sqlite3.c was generated, but that's not the case. Addresses bug report 2026-09-27T13:25:48Z. (check-in: 03bb7acfe8 ... user: stephan tags: trunk) | |
|
2026-09-26
| ||
| 11:21 | Improvements to detection of invalid JSONB in the json_pretty() function. This is an enhancement, not a bug fix, as json_pretty() does not guarantee detection of invalid JSONB, as long as the invalid JSONB does not cause a memory error. Bug 2026-09-25T15:59:04Z. Test cases in TH3. (check-in: c1027e6aee ... user: drh tags: trunk) | |
| 11:01 | New assert()s added to demonstrate that calls to sqlite3_aggregate_context() within an xInverse() callback never return NULL. (check-in: d28a4b4167 ... user: drh tags: trunk) | |
| 09:51 | Never let the SQLITE_FP_PRECISION_LIMIT exceed 100 million. Bug 2026-09-26T07:06:12Z (check-in: ba6f5fb7b8 ... user: drh tags: trunk) | |
| 06:23 | Update some long-dormant JS test code for refactoring-/renaming breakage induced since it was last used. Remove a separate long-dormant JS demo which cannot work with the current code. (check-in: fc4e246dc2 ... user: stephan tags: trunk) | |
|
2026-09-25
| ||
| 17:42 | Avoid dropping an error code in fts5 contentless-delete mode after parsing an %_docize record. Report 2026-09-25T15:52:10Z. (check-in: 52f84cfcdb ... user: dan tags: trunk) | |
| 16:43 | Have RBU refuse to update any table with GENERATED columns. (check-in: 2a2ed4dc5a ... user: dan tags: branch-3.53) | |
| 16:35 | Have RBU refuse to update any table with GENERATED columns. Bug report 2026-09-25T10:43:02Z. (check-in: 2cde8fd0a2 ... user: dan tags: trunk) | |
| 15:16 | Fix 8 byte buffer overwrite that could occur when applying RBU updates to databases containing indexes with more than 16 columns. (check-in: 15e5ba7e54 ... user: dan tags: branch-3.53) | |
| 15:12 | Fix 8 byte buffer overwrite that could occur when applying RBU updates to databases containing indexes with more than 16 columns. Report 2026-09-25T11:18:49Z. (check-in: 7b4fa1d8cd ... user: dan tags: trunk) | |
| 14:55 | Fix some cases where fts3 integrity-check was not detecting corrupt databases. Report 2026-09-25T11:18:11Z. (check-in: 9b23f99212 ... user: dan tags: trunk) | |
|
2026-09-24
| ||
| 20:18 | Avoid a use-after-free problem in rtree triggered by an attacker manipulating the shadow table schema to ensure sqlite3_last_insert_rowid() returns 0 after certain INSERTs. Problem reported by Claude and Ada Logics. (check-in: 11ce330557 ... user: dan tags: branch-3.53) | |
| 20:14 | Avoid a use-after-free problem in rtree triggered by an attacker manipulating the shadow table schema to ensure sqlite3_last_insert_rowid() returns 0 after certain INSERTs. Problem reported by Claude and Ada Logics. (check-in: 6a9e141565 ... user: dan tags: trunk) | |
| 17:09 | Avoid marking little-known extension function next_char() as SQLITE_INNOCUOUS by default. (check-in: 481a1df638 ... user: dan tags: trunk) | |
| 16:32 | Avoid potential buffer overwrite in debugging tool showwal when dealing with a wal file containing records with a very large number of fields. (check-in: bc7641d1f1 ... user: dan tags: trunk) | |
| 15:53 | Avoid leaving an fts3 porter tokenizer cursor in an inconsistent state following an OOM error. At the moment, this does not cause problems because fts3 always frees a tokenizer cursor following an OOM. But it might if some other code used the porter tokenizer implementation. (check-in: 75c1ee9de6 ... user: dan tags: trunk) | |
| 15:34 | Fix an off-by-one error in sqlite3GetUInt32(). The only user-visible effect of fixing this problem is that the error message emitted for a corrupt database has changed in one obscure case. (check-in: 7f0149c969 ... user: dan tags: trunk) | |
| 10:55 | Work around quirks in MSVC associated with [d046d5f4c6855]. (check-in: 18b6a43eb8 ... user: drh tags: trunk) | |
| 10:37 | Fix test case inaccuracy introduced by [4f4ca57309c82d9e]. (check-in: 1b097b30ac ... user: drh tags: trunk) | |
| 10:04 | In accordance with Unicode Corrigendum #9, allow characters U+fffe and U+fffd to pass through unchanged. Bug 2026-09-23T02:39:21Z. (check-in: 0f5ddd0041 ... user: drh tags: trunk) | |
| 09:34 | Avoid truncating the argument of flag pragmas to 8 bits. Bug 2026-09-23T02:42:23Z. (check-in: 9400baf192 ... user: drh tags: trunk) | |
| 00:18 | In date/time functions, the hour can be 24 as long as the minutes and seconds are zero. Bug 2026-09-23T02:44:48Z (check-in: 52eec02592 ... user: drh tags: trunk) | |
|
2026-09-23
| ||
| 21:04 | The --escape mode for CSV should default to "off". Bug 2026-09-23T16:25:07Z (check-in: c356641fad ... user: drh tags: trunk) | |
| 14:15 | Cherrypick JS kvvfs fixes from [ce09bfe406728312] (superfluous writes) and [c43287d8855b3f9a] (a potential buffer ovewwrite). (check-in: eefb455f1d ... user: stephan tags: branch-3.53) | |
| 14:00 | Fix a small potential buffer overwrite in the JS impl of kvvfs to resolve bug report 2026-09-23T10:50:46Z. Remove some adjacent dead code. (check-in: c43287d885 ... user: stephan tags: trunk) | |
| 12:27 | Minor adjustment to [987ba5a3dc5f2a22] to deal with a corner case involving the maximum integer. Bug 2026-09-23T03:07:00Z. (check-in: 4a66e328a0 ... user: drh tags: trunk) | |
| 12:15 | Improved comment on the previous check-in. (leaf check-in: 4ee6d72487 ... user: drh tags: large-realint-keys) | |
| 11:37 | An attempt to improve the handling of rowid lookups using very large floating point values. (check-in: 881da860d6 ... user: drh tags: large-realint-keys) | |
| 11:31 | Avoid a potential crash by checking the arguments to sqlite_expert_rem() (part of the "expert" extension). This function was normally only called internally, but a sufficiently determined attacker could also find a way. Report 2026-09-23T10:50:15Z. (check-in: 3529f81f53 ... user: dan tags: trunk) | |
| 11:12 | Do not allow the arguments of table-valued-functions to the left of a LEFT JOIN to refer to tables to the right of the LEFT JOIN. Fix for 2026-09-20T09:57:48Z. (check-in: e9e341013f ... user: dan tags: trunk) | |
| 11:06 | Handle correctly very large (or small) out-of-range column indexes passed to various fts5 auxiliary functions. Report 2026-09-23T02:40:15Z. (check-in: 55c634d96d ... user: dan tags: trunk) | |
| 10:06 | Preprocessor: remove more not-relevant-to-this-copy pieces and add support for C-style comments in directive arguments. (check-in: 5c321483e0 ... user: stephan tags: trunk) | |
| 09:34 | Change the behavior of rtree_i32 such that coordinate inputs that are outside the range of signed 32-bit integers are converted into the nearest representable signed 32-bit integer rather than just being truncated to 32-bits. The prior behavior of truncating is what programmers expected years ago, but apparently the new generation of programmers expects the values to be capped at the nearest representable value. We could argue about which is more "correct", but it seems easier to just do what th... (check-in: d046d5f4c6 ... user: drh tags: trunk) | |
| 01:29 | Fix a possible double-free in the (unused) sqlite3_intck_test_sql() test interface to the intck extension. Bug 2026-09-22T22:05:25Z (check-in: 3b1e5538da ... user: drh tags: trunk) | |
|
2026-09-22
| ||
| 19:39 | Improved output when errors occur during ".dump" in the CLI. (check-in: c49c3890bb ... user: drh tags: trunk) | |
| 19:11 | Fix goofy condition in the previous check-in (check-in: 03e14c5002 ... user: drh tags: trunk) | |
| 17:12 | Improved capture and reporting of errors by the CLI. Bug 2026-09-21T02:55:55Z (check-in: 6698a7e187 ... user: drh tags: trunk) | |
| 12:29 | Add more precision limitations on width and precision in the printf() logic. Changes are not actually needed though, so they are parked on this branch. (leaf check-in: a77ebe4f98 ... user: drh tags: limit-printf-precision) | |
| 10:30 | Fix calculation of the virtual table dbstat.pgoffset field for overflow pages. Report 2026-09-22T01:43:42Z. (check-in: 4f4ca57309 ... user: dan tags: trunk) | |
