/ Timeline

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

200 most recent check-ins using file test/orderby1.test version e4501f54

Minor tweaks to query planning weights so that when STAT4 is enabled and functioning, a full table scan is more likely to be selected if that seems like the fastest solution. Only do this when STAT4 info is available because an error has a large potential downside. (Leaf check-in: 263293f1 user: drh tags: branch-3.28)
Import further corner-case fixes for the in-scan-vs-index or OP_SeekScan optimization from trunk. (check-in: c144d640 user: drh tags: branch-3.28)
For the OP_SeekScan optimization, the OP_IdxGT does not necessarily come right after the OP_SeekGE. So use the P2 operand of OP_SeekScan to point to the first instruction after OP_IdxGT. Problem found by dbsqlfuzz. (check-in: ec37744c user: drh tags: branch-3.28)
Disable the OP_SeekScan opcode of the in-scan-vs-index optimization when in PRAGMA reverse_unordered_selects mode, as the OP_SeekScan only works with forwards scans. Thanks to OSSFuzz for pointing out the problem to us. (check-in: 9a1cdf7e user: drh tags: branch-3.28)
Improved query optimization for multi-column indexes where the second or later columns are constrained by an IN operator and the earlier index columns limit the search to a small number of rows. Use the new OP_SeekScan opcode which does scanning of the relevant range of the index but gives up and falls back to doing a seek if the number of rows scanned grows to large, in order to guard against pathological cases where the estimated number of rows to be scanned is far too small. (check-in: f07ac3fb user: drh tags: branch-3.28)
Improvements to the IN-early-out optimization so that it works more efficiently when there are two or more indexed IN clauses on a single table. (check-in: 49b7631e user: drh tags: branch-3.28)
An attempt to improve the performance of the IN-early-out optimization (see check-in [09fffbdf9f2f6ce3]) by avoiding unnecessary calls to the b-tree search algorithm in OP_IfNoHope when the index key is at hand and the same answer can be obtained by doing a quick key comparison. Update: Experiment did not work out. (Closed-Leaf check-in: 8301da31 user: drh tags: branch-3.28-in-early-out-fail)
When doing an UPDATE or DELETE using a multi-column index where only a few of the earlier columns of the index are useful for the index lookup, postpone doing the main table seek until after all WHERE clause constraints have been evaluated, in case those constraints can be covered by unused later terms of the index, thus avoiding unnecessary main table seeks. (check-in: 0ecda433 user: dan tags: branch-3.28)
Provide the SQLITE_DEFAULT_LEGACY_ALTER_TABLE compile-time option. (check-in: b2325a6e user: drh tags: branch-3.28)
More restrictions on changes to shadow tables when in defensive mode. (check-in: b302b260 user: drh tags: branch-3.28)
Validate the type, name, and tbl_name fields of the sqlite_master table when loading the schema, unless writable_schema is engaged. (check-in: 3d13fbf2 user: drh tags: branch-3.28)
Do not allow shadow tables to be dropped in defensive mode. (check-in: ad1f760f user: drh tags: branch-3.28)
Import the 3.31.0-beta FTS3/4 code directly into the 3.28 branch, thus providing 3.28 with all the latest 3.31 fixes. FTS3 has not been enhanced to use any core functionality that was not already available in 3.23, so no modifications were made to FTS3 sources for this import. (check-in: be4269c6 user: drh tags: branch-3.28)
Fix the OP_Copy-coalesce optimization so that if the previous row happens to end with OP_Copy but is not a candidate for the optimization due to jumps, then the optimization is correctly bypassed. (check-in: b36126c1 user: drh tags: trunk)
Ensure that when a Select object is reset a new, empty, SrcList is allocated. (check-in: 4889cbf8 user: dan tags: trunk)
When generating the name of a view (or common table expression) because the SQL does not specify a name, avoid the names "true" and "false" which might be confused for the boolean literals of the same name, leading to an inconsistent abstract syntax tree. (check-in: ff9492d3 user: drh tags: trunk)
Fix a problem in sqlite3TreeViewBareExprList() in which the routine was not interpreting the new ExprList format correctly. (check-in: cd56872a user: drh tags: trunk)
Merge enhancements from trunk. (check-in: 091403a6 user: drh tags: new-security-options)
Add the two-size lookaside memory allocator. Also, reduce the per-entry size of the ExprList object. (check-in: 51665bf0 user: drh tags: trunk)
NEVER() and ALWAYS() macros tagging unreachable branches. (Closed-Leaf check-in: 34b87774 user: drh tags: two-size-lookaside)
Merge the latest enhancements from trunk. (check-in: bd57e6d9 user: drh tags: two-size-lookaside)
Do not allow the zipfile virtual table to start a transaction if no filename has been specified. (check-in: 7405e982 user: drh tags: trunk)
Fix some test logic in the OP_Delete opcode so that it works after a cursor-trip. (check-in: 28900e5c user: drh tags: trunk)
Ifdef-out code that is only possible with the sessions extension. (check-in: 4fcf07f7 user: drh tags: trunk)
Have the OP_ReleaseReg opcode also invalidate the registers if P5 is non-zero. (check-in: 937be221 user: drh tags: trunk)
Handle blobs that are the return values of functions being cast to text in utf16 databases in the same way as blobs read directly from the database. Fix for [771fe617]. (check-in: e782096a user: dan tags: trunk)
Fix the OP_Move opcode so that it correctly manages dependency tracking. This change impacts debugging builds only. (check-in: 5377add4 user: drh tags: trunk)
Enhancements to aid testing and debugging: In PRAGMA vdbe_trace=on output, show pScopyFrom dependencies on register values. Add the sqlite3VdbeRegisterDump() procedure, callable from a debugger, that shows the values of all registers. Pass the VDBE pointer into test_trace_breakpoint() so that sqlite3VdbeRegisterDump() is callable from the breakpoint. (check-in: 9886cb4b user: drh tags: trunk)
Add the test_trace_breakpoint() subroutine that is invoked after each instruction is printed while running PRAGMA vdbe_trace=on. Only works for SQLITE_DEBUG builds. Also add parameters "pc" and "pOp" to test_addop_breakpoint() to make it easier to set conditionals. (check-in: 49a6368c user: drh tags: trunk)
Use OP_Copy instead of OP_SCopy to move the results of a scalar subquery. (check-in: 435c272d user: drh tags: trunk)
The sqlite3WhereEnd() call now unwinds all Expr modifications made by the sqlite3WhereBegin(). (check-in: 7bfd42f1 user: drh tags: trunk)
Provide the -DSQLITE_ENABLE_INTERNAL_FUNCTIONS=1 compile-time option. Fix the ".testctrl internal_function" command in the CLI so that it does not signal an error on a valid input. (check-in: 8ee2ce92 user: drh tags: trunk)
When generating code for a subquery, make a copy of the Select object and generate the code out of the copy, in case the code generator makes modifications to expression and the Select object needs to be reused. (check-in: 4edddcc0 user: drh tags: trunk)
Ensure that when code for a scalar SELECT featuring window functions is generated more than once by the planner, separate ephemeral tables are opened for each instance. (check-in: ce141732 user: dan tags: trunk)
Fix the sqlite3ExprImpliesNonNullRow() routine so that it correctly handles a numeric comparison of two AND subexpressions. (check-in: 07e504d5 user: drh tags: trunk)
New test-only SQL functions: implies_nonnull_row(), expr_compare(), and expr_implies_expr(). The SQLITE_TESTCTRL_INTERNAL_FUNCTIONS test-control is modified to toggle internal function access on and off for a single database connection. (check-in: 473892a8 user: drh tags: trunk)
Factor out code generation for in-line SQL functions into a separate routine inside of expr.c. (check-in: 586a65a2 user: drh tags: trunk)
Issue OP_ReleaseReg opcodes against the array of registers used to compute index records after the index record has been computed. (check-in: e3330861 user: drh tags: trunk)
Remove an incorrect assert() statement. Ticket [46fcd700b855e6d1] (check-in: eca7ec9c user: drh tags: trunk)
Experimental branch with new sqlite3_db_config() options that could possible enhance security for applications reading potentially compromised database files. (check-in: 96a2db26 user: drh tags: new-security-options)
Also set the SQLITE_DIRECTONLY flag on the load_extension() function. (check-in: 3bd095a5 user: drh tags: trunk)
Set the SQLITE_DIRECTONLY flag on the fts3_tokenizer() function of FTS3, thus preventing that function from being called from within a trigger or view. (check-in: f3171dc2 user: drh tags: trunk)
Refactor names. Use "small" instead of "mini" to describe the smaller of the two lookaside memory allocation sizes. (check-in: 88d24498 user: drh tags: two-size-lookaside)
Merge recent enhancements from trunk. (check-in: 39d55579 user: drh tags: two-size-lookaside)
The OP_ResultRow opcode releases the SCopy dependences on all its registers, as the values in those registers will not be reused. (check-in: 1dc83c5d user: drh tags: trunk)
Back out the NEVER() from check-in [40d10e7aad5b8992] because it is reachable after all. Ticket [892575cdba4e1e36] (check-in: f481636f user: drh tags: trunk)
Add an ALWAYS() to a conditional which is apparently always true. (check-in: ea44c607 user: drh tags: trunk)
Put a NEVER() on a defensive branch. (check-in: 40d10e7a user: drh tags: trunk)
Fix a typo in a comment. No changes to code. (check-in: d4813a8d user: drh tags: trunk)
Do not use HIDDEN columns for NATURAL joins. Fix for [7c0e06b16]. (check-in: ab09ef42 user: dan tags: trunk)
In ALTER TABLE, rename columns and tables in expressions that are optimized out by the "AND 0" optimization. Doing this also fixes an otherwise harmless assert() failure. (check-in: a9e0354c user: dan tags: trunk)
Do not allow triggers that run as part of REPLACE conflict resolution during an UPDATE to modify the the table being updated. Otherwise, those triggers might delete content out from under the update operation, leading to all kinds of problems. Ticket [314cc133e5ada126] (check-in: db4b7e1d user: drh tags: trunk)
Add the OP_FinishSeek opcode which completes an OP_DeferredSeek if the seek has not already completed. Also add the sqlite3WhereUsesDeferredSeek() interface to the query planner. The UPDATE implementation adds an OP_FinishSeek before running the final OP_Insert if one is needed. Ticket [ec8abb025e78f40c] and also an assertion fault reported by Yongheng. (check-in: 21ef6e99 user: drh tags: trunk)
Do not attempt to flatten compound sub-queries in a FROM clause into the parent if any component of the sub-query uses a window function. (check-in: eeb76f62 user: dan tags: trunk)
Change an assert() in where.c to a testcase() macro, since the condition may be false. This was a problem with the assert() only, there is no bug in release builds that omit assert(). (check-in: 82be135d user: dan tags: trunk)
Disable early coding of transitive constraints at the end of each loop in the WHERE clause processing if the loop being coded is for a LEFT JOIN, even if the loop is part of an OR-clause optimization for virtual tables. Test cases in TH3. (check-in: 9421b442 user: drh tags: trunk)
Fix an instance where the planner might choose to use the OR-optimization when it adds no benefit. The same quirk causes an assert() to fail. This is not a bug in released versions - without the assert() the library still gets the right answer, it just does so less efficiently than it should. (check-in: f4bed1d7 user: dan tags: trunk)
Further improvements to .wheretrace during loop code generation. (check-in: c4d5b75c user: drh tags: trunk)
New enhancements to .wheretrace. The 0x20000 flag shows the WHERE clause before and after coding each loop. The 0x800 flag shows status at the start and at the end of each loop. An extra "C" tag is shown on coded terms. (check-in: 59cc46e5 user: drh tags: trunk)
Expose some of the Where data structure debug printing routines to the entire WHERE-clause processing module. (check-in: 85e76887 user: drh tags: trunk)
Add the --enable-all option to the main configure script as a short-hand to enable FTS4, FTS5, Geopoly/Rtree, JSON, and Sessions. In the amalgamation-autoconf, the --enable-rtree option (which is enabled by default) also now activates Geopoly. (check-in: 52ea0672 user: drh tags: trunk)
Simplifications to the initialization of the sqlite3_index_info structure that is used to communicate with virtual table modules. Avoid adding unused constraints to the sqlite3_index_info structure. Extra constraints are harmless, but might be confusing to people trying to understand the code. (check-in: 5e6357fc user: drh tags: trunk)
If the ".wheretrace 0x10000" bit is set, print WhereTerm.prereqAll and .prereqRight fields in the WhereTerm trace output. (check-in: a4f330b1 user: drh tags: trunk)
Two new test cases added to test/fuzzdata8.db. (check-in: 1be2c18f user: drh tags: trunk)
Second attempt to fix a problem with unwinding the WITH stack of the Parse object following an error. (check-in: 315d1f1a user: dan tags: trunk)
Merge latest trunk changes with this branch. (Closed-Leaf check-in: d693be37 user: dan tags: better-error-handling-1)
Fix an assert() failure in altertable3-22.4. (check-in: c566a91d user: dan tags: better-error-handling-1)
Fix a faulty assert() associated with query search limiting query plans. (check-in: 0d743585 user: drh tags: trunk)
Convert an assert() back into a conditional. The conditional was converted into an assert() by check-in [6ae4ad6ebee4db88] (2009-05-28) because we were unable to find a test case for it. Yongheng's fuzzer just now found that test case. (check-in: 4d0b9109 user: drh tags: trunk)
When an INSERT is receiving content from a SELECT, run an OP_ReleaseReg opcode at the top of each iteration of the loop in order to prevent spurious OP_SCopy misuse complaints. Ticket [de4b04149b9fdeae] (check-in: 6afadd3b user: drh tags: trunk)
Recompute the values for all generated columns after NOT NULL ON CONFLICT REPLACE constraints fire. Tickets [37823501c68a09f9] and [5fbc159eeb092130]. (check-in: 4cc12c18 user: drh tags: trunk)
Do not attempt to unwind the WITH stack in the Parse object following an error. This fixes a separate case to [de6e6d68], but also causes an assertion fault at select.c:4666 for test case altertab3-22.4. (check-in: d29edef9 user: dan tags: better-error-handling-1)
Remove a NEVER() that is no longer true. Fix for [36ffedcb9]. (check-in: 597896ed user: dan tags: trunk)
Fix a problem involving window function aliases being referenced from sub-selects. (check-in: e3b5fc05 user: dan tags: trunk)
Add a missing "ifcapable rtree {...}" line to the window1.test test module. (check-in: 4e6fbb1a user: drh tags: trunk)
Fix a problem with window functions in aggregate queries that do not have GROUP BY clauses. (check-in: 99609786 user: dan tags: trunk)
Do not mistake constant integers in a PARTITION BY expression for references to ORDER BY style references to values returned by the SELECT statement. (check-in: 45c64d39 user: dan tags: trunk)
Follow-up to check-in [c8c6dd0e6582ec91] - change the xAccess() method to return true if the file exists and it is anything other than a regular file, or if it is a regular file with a non-zero file size. (check-in: 8a39803e user: drh tags: trunk)
When determining if an aggregate within a sub-query should be processed as part of the sub-query or an outer query, consider any FILTER clause in the same way as the arguments to the aggregate. (check-in: 1ffc045d user: dan tags: trunk)
Ensure that the Pager.changeCountDone flag is cleared whenever dropping the write lock, even when transitioning from EXCLUSIVE locking mode into NORMAL locking mode while in WAL mode. Ticket [fb3b3024ea238d5c]. (check-in: 846b1de6 user: drh tags: trunk)
Fix buffer underflows in the zipfile extension associated with zero-length or NULL filename in the ZIP archive. But report on the mailing list by Yongheng and Rui. (check-in: 465a15c5 user: drh tags: trunk)
If an UPSERT can cause an Abort due to a constraint failure, make sure the query planner knows this. Ticket [7c13db5c3bf74001]. (check-in: f14ce948 user: drh tags: trunk)
An UPDATE of a table that is indexed by a constant virtual column that uses the one-pass optimization might cause the table seek to be omitted before reaching row DELETE/INSERT. Fix this by coding an extra OP_Column in that circumstance. Ticket [ec8abb025e78f40c] (check-in: e5456049 user: drh tags: trunk)
Fix an assert() in fts5 that could fail if an xSavepoint() call on another vtab fails. Fix for [167b2aac] . (check-in: a5d7f5d2 user: dan tags: trunk)
Makefile.in fix so that it works on systems that require a .EXE suffix on executables. (check-in: f482a4cd user: drh tags: trunk)
Add two new fuzzer test cases from dbsqlfuzz. (check-in: f8d5586a user: drh tags: trunk)
In the xAccess() method of the unix VFS, return true if the named object is a directory, regardless of what stat() reports as the st_size for the object. Different filesystems report st_size differently for directories. Problem reported on the mailing list by Stefan BrĂ¼ns. (check-in: c8c6dd0e user: drh tags: trunk)
Fix an FTS3 test case that depends on the ICU extension so that it only runs if SQLite is compiled with ICU. (check-in: 19c6240b user: drh tags: trunk)
Fix an incorrect assert() added by check-in [fa58aad48a788802]. Problem found by dbsqlfuzz. (check-in: a17b29f0 user: drh tags: trunk)
Disable the optimization that tries to pull the value of an expression from an index on that expression if the expression is a constant. (check-in: e5fd8b50 user: drh tags: trunk)
When the sqlite3WindowRewrite() routine detects and error, have it convert the SELECT statement into just "SELECT null" so that it does not leave the parse tree in a goofy state that can cause problems with subsequent code before the stack has a chance to unwind and report the error. Ticket [d87336c81c7d0873] (check-in: fa58aad4 user: drh tags: trunk)
Fix a minor performance regression from check-in [401c9d30e06191d9] (check-in: 76f54ee8 user: drh tags: trunk)
Backout the early VTable cursor close change from yesterday, as Yongheng and Rui have found a test case for which it fails. The new test case is added to fuzzcheck. (check-in: ddb10f03 user: drh tags: trunk)
Fix a problem causing fts5 integrity-check failures if numeric values are inserted into a table within a utf-16 database. Fix for [752fdbf6]. (check-in: bae060f3 user: dan tags: trunk)
Make the zipfile() extension function more robust against zero-length filenames. (check-in: b9c2005f user: drh tags: trunk)
Fix another case where malformed utf-8 was being mishandled in fts5. Fix for [df46a6f3]. (check-in: 1c0a05b0 user: dan tags: trunk)
Fix an assert() added as part of commit [a11b393dc] that can fail if fts5 database records are corrupt. (check-in: 4630c1ec user: dan tags: trunk)
Extra defense against problems following an OOM. dbsqlfuzz find. Also import the latest dbsqlfuzz test cases. (check-in: 0a70f5dd user: drh tags: trunk)
Fix a spurious report of corruption that could be made by the fts5 integrity-check in SQLITE_DEBUG builds if the fts5 index contains malformed utf text. Ticket [d62981b76de521e3] (check-in: a11b393d user: dan tags: trunk)
Convert an ALWAYS() into an assert() with an extra error term. Dbsqlfuzz find, with test case in TH3. (check-in: b473ad35 user: drh tags: trunk)
Allow comparison operators of a register against itself. Ticket [188f912b51cd802a], (check-in: 401c9d30 user: drh tags: trunk)
Test case for the zipfile-extension bug fix of the previous check-in. (check-in: bc8bfc7f user: drh tags: trunk)
Fix the zipfile() function in the zipfile extension so that it is able to deal with goofy filenames that contain embedded zeros. (check-in: cc0fb00a user: drh tags: trunk)
Thoroughly reset the rtree cursor at the start of each VFilter operation, including clearing its cache. This prevents left over pages in the cache which can cause problems on shutdown after a LEFT JOIN. Ticket [5eadca17c4dde90c] (check-in: 4c50afaf user: drh tags: trunk)
Remove an extra deflateInit2() call accidently left in check-in [f5ee30426e8876e7] (check-in: 953e6aa6 user: drh tags: trunk)
Do an early close of virtual table cursors to avoid unnecessary cursor contention in UPDATE for some virtual table implementations. Ticket [56a74875be799b85] (check-in: eb95dac7 user: drh tags: trunk)
Early detection of database corruption in balance_deeper(). (check-in: 61c22336 user: drh tags: trunk)
Fix a case in which SQLite could fail to identify "x BETWEEN ? AND ?" being true as implying that x is not null. Ticket [dfd66334]. (check-in: 2f179749 user: dan tags: trunk)
For expressions like (x, y) IN (SELECT ...) where the SELECT uses window-functions, require that all columns on the LHS be indexed before an index can be used. Fix for [d9ed4ebe]. (check-in: 0b1dbd60 user: dan tags: trunk)
Fix a shift-overflow problem in yesterday's check-in [36fdeb4f0a66970a] that OSSFuzz helpfully discovered overnight. Thanks Google. (check-in: bff38e2b user: drh tags: trunk)
Fix the OP_Cast operator so that when casting to TEXT, it always leaves the result in the encoding of the database. Ticket [0911b5d161b039c6]. Test cases in TH3. (check-in: f347744e user: drh tags: trunk)
Change the assert() back into a testcase(). See also check-ins [9ab985a9c8160b90] and [ddb17d92df194337] and other check-ins that those reference. Fix for ticket [9d708e474201c001] (check-in: 2c44c734 user: drh tags: trunk)
Enhance the sqlite3VdbeMemAboutToChange() shallow-copy validation mechanism by adding the new OP_ReleaseReg opcode to tell MemAboutToChange() that a range of registers is no longer needed so that the source register can be freely changed. This is a change to debugging and test builds only and does not impact release builds. Fix for ticket [c62c5e58524b204d] and [5ad2aa6921faa1ee]. The previous fix to ticket [5ad2aa6921faa1ee] is backed out by this change since this change is a better fix. (check-in: 36fdeb4f user: drh tags: trunk)
Change the code generator for the IN operator so that it avoids creating OP_Eq and OP_Ne opcode with the same P1 and P3 arguments. This enables us to back out check-in [ddb17d92df194337] and also fix ticket [188f912b51cd802]. (check-in: 9ab985a9 user: drh tags: trunk)
Make a hard copy of strings in constraint checks prior to applying OP_RealAffinity, to avoid problems with a pointer accounting assert. This change is not strictly necessary - the correct answer is obtained without it and no UB occurs - however the pointer accounting asserts are useful to prevent other problems so it is a simple matter to bring this piece into compliance. Ticket [5ad2aa6921faa1ee] (check-in: 89a9dad6 user: drh tags: trunk)
When constructing the virtual MATCH term of the WHERE clause for a virtual table that is in a LEFT JOIN, be sure to set the correct Expr.iRightJoinTable value. This value does not appear to ever be used, except inside of a single assert(). But it is good to set it correctly, nevertheless. This fixes ticket [7929c1efb2d67e98], which as far as I can tell is completely harmless. (check-in: ef604882 user: drh tags: trunk)
In the WHERE clause debugging output (the .wheretrace output) show the parent index of any WhereTerm that is a child. (check-in: 7fc73332 user: drh tags: trunk)
Fix to the optimization of check-in [a47efb7c8520a011] that reads the values of expressions used in an index-on-expression directly from the index rather than recomputing the value. If the expression has a top-level COLLATE or unlikely() operator, be sure to clear the corresponding flags prior to converting it into a TK_COLUMN expression. Failure to do this is most likely harmless in production, but might cause an assertion fault in debugging builds. Ticket [b0cb8aff561a6dcd]. Test cases in TH3. (check-in: 56539e1c user: drh tags: trunk)
Back out the asserts of check-ins [a500893b6f64aced] and [d9c9fe9f5ad3fc91] as ticket [1b06916e01c82b66] demonstrates a case that refutes them. (check-in: ddb17d92 user: drh tags: trunk)
When parsing a CREATE TABLE from the sqlite_master table, delete the CHECK constraints if there are any errors, since there might otherwise be attempts to use those CHECK constraints if PRAGMA writable_schema=ON is set. This undoes the fix in check-in [ea721b34477ab8b4] for a more general solution. (check-in: a982e643 user: drh tags: trunk)
Ensure sqlite3WindowRewrite() is called on a SELECT statement before any terms aremoved from it as part of IN() clause processing. Fix for [f00d096ca]. (check-in: 8c856404 user: dan tags: trunk)
Fix "PRAGMA data_version" so that it works the same way with locking_mode=PERSIST and journal_mode=PERSIST configured. Fix for 7a458c2a5f. (check-in: 45748e2d user: dan tags: trunk)
Fix a spurious report of corruption from the fts3/4 integrity-check triggered by using the languageid option. (check-in: 70815e27 user: dan tags: trunk)
When creating a new virtual table, ensure that the OP_ParseSchema opcode processes the correct entry in the sqlite_master table even if there is a second entry with the same name and table values due to database corruption and the use of writable_schema=ON. Dbsqlfuzz find. (check-in: 4dbb6e1c user: drh tags: trunk)
When a corrupt schema is loaded using writable_schema=ON, the CHECK constraints (or other expressions in the table definition) might not be fully resolved. Ensure that the code generator can deal with this if the table is subsequently used in a DML statement. dbsqlfuzz find. (check-in: ea721b34 user: drh tags: trunk)
Do not try to access a generated column through an index if the collating sequence for the generated column is non-standard. Part 2 of ticket [e0a8120553f4b082] (check-in: 056bb8dc user: drh tags: trunk)
Apply real affinity to generated columns of type REAL that are extract from an index. Ticket [e0a8120553f4b082] (check-in: 728ad39e user: drh tags: trunk)
Debugging enhancment: Show the Expr.y.pTab pointer on TK_COLUMN nodes of an expression tree in the treeview. (check-in: 64154ac4 user: drh tags: trunk)
Fix a bad interaction between RBU and [df51ae19]. (check-in: 0b9d8a12 user: dan tags: trunk)
Fix two more cases in fts5 where sqlite3_value_bytes() was being called before sqlite3_value_text(). Fix for e431c355. (check-in: a1ba9a37 user: dan tags: trunk)
Export the public RBU entry points from the Win32 DLL. (check-in: e62d1791 user: mistachkin tags: trunk)
Debugging improvement: when tracing the VDBE, output the register value for OP_Cast after the cast is completed. (check-in: dc5f1d28 user: drh tags: trunk)
When computing dependencies on WHERE-clause terms, be sure to take into account the FILTER clause of aggregate functions. Problem reported by Manuel Rigger. (check-in: 3cc2b570 user: drh tags: trunk)
Improved assert() statements on the OP_Column opcode. (check-in: c538601e user: drh tags: trunk)
Load new dbsqlfuzz test cases into test/fuzzdata8.db. (check-in: 01aeeb5c user: drh tags: trunk)
Fix a potential problem with ALTER TABLE commands on schemas that contains a triggers or view featuring a sub-query that uses an ON clause with a join that is not a LEFT JOIN. (check-in: 365dd8b6 user: dan tags: trunk)
Fix harmless compiler warnings. (check-in: 2bbd014c user: drh tags: trunk)
Ensure fts5 consistently calls sqlite3_value_text() before sqlite3_value_bytes(). This matters when storing blobs in fts5 tables within utf-16 databases. Fix for [cf36cb4e02]. (check-in: c16305eb user: dan tags: trunk)
Do not continue an ALTER TABLE tree walk of a subquery if a prior error is seen in the tree. Report the problem immediately, without continuing. (check-in: 01ca865f user: drh tags: trunk)
Fix an assert() in btree.c that can fail if an INSERT is attempted on a corrupted database while in PRAGMA writable_schema=ON. (check-in: 07beb362 user: drh tags: trunk)
Continuation of [e2bddcd4c55ba3cb]: Add another spot where it is necessary to abort early due to prior errors in sqlite3WindowRewrite(). (check-in: cba2a2a4 user: drh tags: trunk)
Fix the regexp extension so that it correctly translates all over-length 3-byte UTF8 sequences into 0xfffd. (check-in: 3d4c0bf8 user: drh tags: trunk)
When an error occurs while rewriting the parser tree for window functions in the sqlite3WindowRewrite() routine, make sure that pParse->nErr is set, and make sure that this shuts down any subsequent code generation that might depend on the transformations that were implemented. This fixes a problem discovered by the Yongheng and Rui fuzzer. (check-in: e2bddcd4 user: drh tags: trunk)
Detect the invalid use of NULLS LAST on an INTEGER PRIMARY KEY definition of a WITHOUT ROWID table. (check-in: 4417c5bf user: drh tags: trunk)
Fix the zipfile extension so that INSERT works even if the pathname of the file being inserted is a NULL. Bug discovered by the Yongheng and Rui fuzzer. (check-in: a80f84b5 user: drh tags: trunk)
Fix incorrect SQL generated by the Lemon change of the previous check-in. (check-in: fccfb8a9 user: drh tags: trunk)
Modify the parse.sql output file from Lemon so that the RULE table contains a column with the complete text of the rule. (check-in: 329cbb37 user: drh tags: trunk)
More restrictions on changes to shadow tables when in defensive mode. (Leaf check-in: 4146c629 user: drh tags: branch-3.30)
Minor enhancement to Win32 locking semantics. (Leaf check-in: ed66ffce user: mistachkin tags: win32LockFlags)
Change an unreachable testcase() into an assert() (check-in: 062e00c4 user: drh tags: trunk)
Continue to back away from the LEFT JOIN optimization of check-in [41c27bc0ff1d3135] by disallowing query flattening if the outer query is DISTINCT. Without this fix, if an index scan is run on the table within the view on the right-hand side of the LEFT JOIN, stale result registers might be accessed yielding incorrect results, and/or an OP_IfNullRow opcode might be invoked on the un-opened table, resulting in a NULL-pointer dereference. This problem was found by the Yongheng and Rui fuzzer. (check-in: 86297431 user: drh tags: trunk)
Add the --vdbe-debug option to fuzzcheck. (check-in: 289158aa user: drh tags: trunk)
Enhancements to fuzzcheck on unix so that it captures signals and prints the current test file and test number prior to terminating. (check-in: ae7cbb24 user: drh tags: trunk)
Add a test case for fix in check-in [1ca0bd982ab1183b] (check-in: 519864da user: drh tags: trunk)
Correctly unwind the savepoint stack if the database goes read-only in the middle of a savepoint. (check-in: a96d02b4 user: drh tags: trunk)
Remove an assert() in the code generator that can fail due to an invalid INSERT statement. The same assert() appears again in the VDBE and will fire there if the problem persists in an SQL statement that does not contain errors. Gramfuzz find. (check-in: 9bdd57cb user: drh tags: trunk)
When processing constant integer values in ORDER BY clauses of window definitions (see check-in [7e4809eadfe99ebf]) be sure to fully disable the constant value to avoid an invalid pointer dereference if the expression is ever duplicated. This fixes a crash report from Yongheng and Rui. (check-in: 1ca0bd98 user: drh tags: trunk)
Clean up the ExprList that holds the names of columns in a CTE before checking for unused references in the ALTER TABLE implementation. (check-in: 8223e79f user: drh tags: trunk)
When a table is part of a LEFT JOIN and should be a completely NULL row due to the semantics of a LEFT JOIN, make sure any generated columns on that row evaluate to NULL. Ticket [3b84b42943644d6f] (check-in: 02714914 user: drh tags: trunk)
Ensure that all ON CONFLICT REPLACE indexes are sorted to the end of the list of indexes for a table, even for weird cases where the same UNIQUE constraint occurs twice with the ON CONFLICT REPLACE clause only on the second one. This avoids an out-of-order contraint processing problem that can arise due to the optimization of check-in [469a62ca33081854]. (check-in: 1e3918ca user: drh tags: trunk)
Abort the OP_Savepoint opcode early if a virtual table xSync method fails, perhaps due to an interrupt. (check-in: 672e749a user: drh tags: trunk)
When a generated column expression is initially parsed, if it contains errors convert the expression to a simple NULL. This prevents additional pieces of the expression tree that are allocated from lookaside from leaking into the schema, where lookaside allocations are not allowed. (check-in: a89b3860 user: drh tags: trunk)
Due to the previous change, the p5 parameter to OP_VColumn no longer ever contains extraneous bits, so change a testcase() into an assert() to show as much. (check-in: 5b4a88cd user: drh tags: trunk)
Do not set OP_Column flags on the instructions generated by sqlite3ExprCodeGetColumn() if the opcode generated is not really an OP_Column, which might happen if the column is virtual. Fix for ticket [b439bfcfb7deedc6] (check-in: 2401e047 user: drh tags: trunk)
Make the sqlite3ExprCodeTarget() expression code generator routine robust in the face of unknown opcodes - it simply generates a NULL value. (check-in: f8e876c8 user: drh tags: trunk)
Fix problems in fts3 with processing corrupt records and undefined integer overflows. (check-in: 3b873029 user: dan tags: trunk)
Fix an utterly harmless "applying zero offset to null pointer" UB warning in sqlite3AtoF(): (check-in: 052fdf5e user: drh tags: trunk)
Initial exploration of blocky APIs, starting with a variant of sqlite3_exec. (Leaf check-in: fd849d5b user: numist tags: blocky-exec)
Patch to the page cache to avoid harmless pointer arithmetic that due to bugs in the STD-C spec is technically UB. This is to fix a harmless UBSAN complaint that OSSFuzz is hitting. (check-in: c29fc212 user: drh tags: trunk)
Change the default lookaside configuration to 40 slots of 1200-bytes each. This actually works out to 30 big slots and 93 small slots using the mini-lookaside allocator. We get the same (or better) lookaside coverage but with 72KB less memory per connection. (check-in: 47b71a84 user: drh tags: two-size-lookaside)
Cleanup and performance enhancements for mini-lookaside. (check-in: 74805668 user: drh tags: two-size-lookaside)
Merge fixes from trunk. (check-in: 9c471195 user: drh tags: two-size-lookaside)
Ensure that there is a containing SELECT statement when processing a normal aggregate function as if it were a window function. (check-in: c1014e80 user: drh tags: trunk)
*Remove* the macros, not just comment them out. (check-in: e5dc2939 user: drh tags: trunk)
Remove unused macros formerly used to define built-in aggregate functions. (check-in: 3ef0d44e user: drh tags: trunk)
Reduce the size of ExprList to 24-bytes per entry, down from 32-bytes per entry. This helps keep lookaside allocations in the mini-size. (check-in: bda92d92 user: drh tags: two-size-lookaside)
Tie up the loose ends in the ExprList size reduction. (Leaf check-in: 59d0f3af user: drh tags: two-size-lookaside, exprlist-size-reduction)
Work toward reducing the incremental size of an ExprList object to 24-byte per entry, from 32-bytes (on a 64-bit machine). This helps the new mini-lookaside allocator to run better by avoiding excessive reallocs. The current change mostly works, but still has a few loose ends to tie up. This check-in is merely a snapshot to save my work. (check-in: fdda76cf user: drh tags: two-size-lookaside, exprlist-size-reduction)
Reduce the number of call sites to dbMallocRawFinish in the case where n > lookaside.sz (check-in: e392e7f2 user: numist tags: two-size-lookaside)
Fix an issue where malloc could be used to fulfill a small allocation when a large lookaside slot could have beeen used instead. (check-in: 611020e3 user: numist tags: two-size-lookaside)
Change the name of the Expr.a.zName field to zEName, so that it has a name that is distinct from other fields and variables and is hence easier to grep for. (check-in: d3783357 user: drh tags: two-size-lookaside)
Change the size of a mini-lookaside allocation to a macro (MINI_SZ) rather than a magic number (128). (check-in: 5e1949bc user: drh tags: two-size-lookaside)
Merge enhancements from trunk. (check-in: 6cb053f9 user: drh tags: two-size-lookaside)
Factor out the conditional (which is only true for rare errors) from the notValid() function in resolve.c, for a performance improvement and size reduction. Also cause failures to set the Expr node to a NULL operator so that it does not cause problems later in case PRAGMA writable_schema=ON has been set. Test cases in TH3. (check-in: c6af9f65 user: drh tags: trunk)
More efficient implementation of a lookaside allocator that supports mini (in this case, harcoded to 128B) slots. (check-in: b02fdc09 user: numist tags: two-size-lookaside)
Fix goofy string formatting in lemon.c that dates from the K&R-C days. (check-in: 48ba5e5a user: drh tags: trunk)
Improved tracing output from the LEMON-generated parser. (check-in: 4d6d2fc0 user: drh tags: trunk)
Strengthen check-in [83da4d4104ee1870] by ignoring *all* WHERE-clause constraints for a virtual table that is the right table of a LEFT JOIN, as such constraints are never useful. This fixes an issue discovered by Manuel Rigger. (check-in: 840de36d user: drh tags: trunk)
When trying to drop a virtual table that has no xDestroy method, invoke the xDisconnect method rather than doing nothing, to avoid a memory leak. (check-in: 1fa29a5f user: drh tags: trunk)
The xDestroy method of a module might be NULL if the schema is corrupt. (check-in: 0457e7d1 user: drh tags: trunk)
Enhance LEMON to provide the ability to mark a rules as one that will never reduce due to intervention of actions. Use this new capability to designate the EXPLAIN rule as NEVER-REDUCES. (check-in: 136cdefb user: drh tags: trunk)
Use a statement journal on a CREATE VIRTUAL TABLE statement in case the VCreate opcode fails. (check-in: aa3b0eb8 user: drh tags: trunk)
Fix parser bugs: require a semicolon after an EXPLAIN command. (check-in: 707a058a user: drh tags: trunk)
Avoid a buffer overread in fts5 that could occur when parsing corrupt configuration records. (check-in: 355afd77 user: dan tags: trunk)
Set the affinity of regular columns prior to computing the values of generated columns. Ticket [d7c3f125c925c522] (check-in: d47d66e3 user: drh tags: trunk)
Rename a local variable to avoid masking a function parameter and thus causing a harmless compiler warning. (check-in: f065cf00 user: drh tags: trunk)
Add a NEVER() to the ALTER TABLE fix in check-in [1d2e53a39b87e364] (check-in: c7309ed3 user: drh tags: trunk)
The previous check-in was not quite correct, and introduced a new problem with the USING clause. Use this version instead. (check-in: ed28aaa4 user: drh tags: trunk)
Ensure that the SrcList_item.colUsed field is set correctly (set to have a 1 for all columns of the table) when a generated column appears in the USING clause of a join. (check-in: 1923efb2 user: drh tags: trunk)
Fix the NOT NULL verification logic in PRAGMA integrity_check so that it works for generated columns whose value is the result of a comparison operator. Ticket [bd8c280671ba44a7] (check-in: f3b39c71 user: drh tags: trunk)