Many hyperlinks are disabled.
Use anonymous login
to enable hyperlinks.
44 check-ins using file src/select.c version 64a4b48a
|
2023-03-24
| ||
| 19:17 | Fix a problem in cursor-hints for WITHOUT ROWID tables used in a RIGHT JOIN. Forum post 591006b1cc. (check-in: 221fdcec user: drh tags: trunk) | |
| 17:15 | Add a JS test which checks for the issue described in forum post 895425b49a. (check-in: 98d30400 user: stephan tags: trunk) | |
| 16:57 | Fix byte-code register allocation in ANALYZE for STAT4 when there multiple indexes with differing numbers of columns. forum post bc39e531e5. This problem arose when expression indexes were added by check-in [2131a5ca53f0e9b0]. (check-in: 2bf5413d user: drh tags: trunk) | |
|
2023-03-23
| ||
| 19:22 | Fix test cases so that they work when SQLITE_ENABLE_NAN_INF is defined. (check-in: 95c7af79 user: drh tags: nan-inf) | |
| 12:00 | Fix #ifdefs that use the wrong preprocessor macro. (check-in: 0aecf360 user: drh tags: nan-inf) | |
| 10:54 | The attempt to bring STAT4 up to 100% MC/DC at [55a26c67ed4a3a93] and at [168fa2fb22b8c1ad] are incorrect. Back them out and replace them with a simple NEVER() macro. Error reported by forum post dc4854437b. (check-in: 5992370a user: drh tags: trunk) | |
|
2023-03-22
| ||
| 20:21 | Add the SQLITE_ENABLE_NAN_INF compile-time option which makes the following behavior changes: (1) sqlite3_value_double(NULL) returns NaN, (2) SQLite preserves NaN values rather than converting them to NULL. (3) CAST statements understand "NaN" and "Inf" and make the right conversions. (4) Non-standard JSON is never generated by SQLite JSON routines, but those routines will accept floating point literals "NaN", "Inf", and "-Inf". (check-in: 96ec8306 user: drh tags: nan-inf) | |
| 19:57 | Internal cleanups in JS code. No functional changes. (check-in: 8fbdf7d1 user: stephan tags: trunk) | |
| 16:55 | The floating-point-to-text conversion with the zero-padding option now renders NaN as "null". (check-in: ad59fa17 user: drh tags: trunk) | |
| 16:37 | For consistency, the ".mode json" output from the CLI now renders infinity in the same format as the JSON functions. (check-in: abee339d user: drh tags: trunk) | |
| 16:24 | The double-to-text conversion renders infinity as 9e999, so that JSON output is compliant and so that values can be round-tripped. (check-in: b52081d0 user: drh tags: trunk) | |
| 16:01 | In the CLI, the magic parameter :inf and :nan bind floating point values Infinity and NaN, respectively, as an aid to testing SQLite's handling of those quantities. (check-in: c70a61d8 user: drh tags: trunk) | |
| 14:51 | Update the version number for the TEA tarball to 3.42.0, to match the core. (check-in: 03e6918e user: drh tags: trunk) | |
|
2023-03-21
| ||
| 14:20 | Add ALWAYS() on a branch this is always true now due to [84417bbd144b2197]. (check-in: badf7d0e user: drh tags: trunk) | |
| 12:29 | Add the fuzzcheck-asan.exe target to the MSVC makefile. (check-in: 0901bc02 user: drh tags: trunk) | |
| 11:56 | Add the fuzzcheck-asan target to the main posix makefile. (check-in: 55e94add user: drh tags: trunk) | |
| 11:13 | Fix a valgrind error and potential buffer overread when handling a corrupt database. (check-in: b1e0cd64 user: dan tags: trunk) | |
|
2023-03-20
| ||
| 20:22 | Reinsert two NEVER() macros for b-tree branches that were previously needed for [b6a82f3c3b9d89fd] but which are now obsolete due to [73f0036f045bf371]. Later: dbsqlfuzz quickly found new cases for which those two branches are needed. The new test cases have been added to TH3. (Closed-Leaf check-in: 3065dadb user: drh tags: backout) | |
| 18:35 | Minor change to btreeNext() to facilitate coverage testing. (check-in: 20b3ef04 user: drh tags: trunk) | |
| 14:59 | Fix a problem causing a cursor to retain an out-of-date cell-info cache when processing a DISTINCT query on values that are identical according to their collation sequence, but different on disk. Forum post e123e6cde4. (check-in: 1b3abc1d user: dan tags: trunk) | |
| 10:43 | Back out the extra margin added to the input buffer of the CLI, as it is not needed. (check-in: ac8d1e5d user: drh tags: trunk) | |
| 01:55 | A better fix for the sqlite3_error_offset() problem on generated columns. (check-in: 770b3e67 user: drh tags: trunk) | |
| 00:53 | Expression errors in generated columns should not generate non-negative sqlite3_error_offset() returns. Second of two defenses against [33aa4c0de8a62e33]. (check-in: 2adb4e0d user: drh tags: trunk) | |
| 00:48 | When reporting errors in the CLI, ignore the output of sqlite3_error_offset() if the value returned is clearly out-of-range. One of two lines of defense against [33aa4c0de8a62e33]. (check-in: 26adbb80 user: drh tags: trunk) | |
|
2023-03-19
| ||
| 21:48 | Increase the size of ref-count values in the pager layer to 64-bits, to avoid any reasonable possiblity of overflowing the counters. There is a performance and memory penality for this. Forum post b741f15a35. (check-in: 6c5d99a8 user: drh tags: trunk) | |
|
2023-03-18
| ||
| 16:12 | Avoid a buffer overread in fts3 that could occur when processing a corrupt record. (check-in: 02ac2297 user: dan tags: trunk) | |
|
2023-03-17
| ||
| 20:31 | Fix json rendering so that it shows positive and negative infinity as 9.0e+999 and -9.0e+999 respectively. (Closed-Leaf check-in: efce4690 user: drh tags: numeric-only-json) | |
| 19:18 | Add the ability to name functions using one of the join keywords like CROSS FULL INNER LEFT NATURAL OUTER RIGHT. (check-in: 0910b192 user: drh tags: trunk) | |
| 14:18 | Fix a potential buffer overread in the recovery extension. (check-in: 0b3b5bf9 user: dan tags: trunk) | |
| 10:43 | Ensure that an error does not delete the Table object out from under the xConstruct method of a virtual table. dbsqlfuzz 7cc8804a1c6d4e3d554d79096e6ea75a7c1c7d2d (check-in: df4928c9 user: drh tags: trunk) | |
| 00:42 | Add safety margin on the CLI input buffer for tickets [33aa4c0de8a62e33], [b97e6c5e6a91d97f], [2971fbe3f993e95a], and [2971fbe3f993e95a]. (check-in: 741af08a user: drh tags: trunk) | |
| 00:01 | Fix assert() statements that would (incorrectly) fire if an IF NOT EXISTS trigger that already exists contained two or more RETURNING clauses. Tickets [89d259d45b855a0d] and [d15b3a4ea901ef0d]. (check-in: 648899e4 user: drh tags: trunk) | |
|
2023-03-16
| ||
| 20:54 | Correctly handle SELECT DISTINCT ... ORDER BY when all of the result set terms are constant and there are more result set terms than ORDER BY terms. Fix for these tickets: [c36cdb4afd504dc1], [4051a7f931d9ba24], [d6fd512f50513ab7]. (check-in: 12ad822d user: drh tags: trunk) | |
| 12:28 | Additional debug/test output from the query invariant checker showing the row-number that is being checked. (check-in: e4b6eb58 user: drh tags: trunk) | |
| 11:50 | Update the tracing output for the query-invariant checker such that it shows the SQL that is run to verify that a found query-invariant discrepency is valid. Changes to testing logic only. (check-in: 8f45ad27 user: drh tags: trunk) | |
| 10:17 | Do not use the one-pass optimization on an UPDATE if there is a subquery in the WHERE clause, since if the subquery is hidden behind a short-circuit operator, the subquery might not be evaluated until after one or more rows have been updated. Fix for the problem reported by forum post 0007d1fdb1. This is the same problem that was fixed by [73f0036f045bf371] only for UPDATE instead of DELETE. (check-in: 2c56b984 user: drh tags: trunk) | |
| 09:07 | Remove a NEVER() from btreeNext() that dbsqlfuzz 460aa158f9a2c41145831cc924296cde1f312b3f found could sometimes be reached. I will find a way to test that branch later. (check-in: 1dffeffe user: drh tags: trunk) | |
| 02:30 | Another approach at attempting to contain the damage caused by corruption that leaves MemPage.isInit clear. Works better than the previous but is still not perfect. (Closed-Leaf check-in: ba964eb0 user: drh tags: corruption-in-btree-init) | |
| 01:20 | When the btreeInitPage() routine detects database corruption, it should continue to the end and set MemPage.isInit before it returns SQLITE_CORRUPT, because if it leaves MemPage.isInit unset, then can cause difficulty later. dbsqlfuzz 460aa158f9a2c41145831cc924296cde1f312b3f (check-in: 44e83f8b user: drh tags: corruption-in-btree-init) | |
|
2023-03-15
| ||
| 17:58 | Disallow the one-pass optimization for DELETE if the WHERE clause contains a subquery. Fix for the problem reported by forum post e61252062c9d286d. This fix is more restrictive than necessary. It could be relaxed if the subquery does not involve the table that is the subject of the DELETE. (check-in: 73f0036f user: drh tags: trunk) | |
| 13:53 | Fix a broken assert() in the recovery extension. (check-in: 4c4e66f2 user: dan tags: trunk) | |
|
2023-03-14
| ||
| 20:08 | Fix Bloom filters on an expression index. forum post 2e427099d5 and forum post d47a0e8e3a. This problem goes back to the original introduction of Bloom filters (check-in [633bfeeea2bccdd4]) for SQLite version 3.38.0. (check-in: c028fb66 user: drh tags: trunk) | |
|
2023-03-13
| ||
| 16:08 | Include CLI's tip for -- in all builds. Better show optionality of its non-option arguments. (check-in: 9e2c771d user: larrybr tags: trunk) | |
|
2023-03-11
| ||
| 23:21 | The check-in at [198b3e33dcfd74c7] caused a performance regression for some queries, which is here fixed. Problem reported by forum post b405033490fa56d9. (check-in: dc9f025d user: drh tags: trunk) | |