[3.0-dev] Backport prepared-statement / connector fixes (Fixes #24186)#24191
Open
jiangxinmeng1 wants to merge 10 commits intomatrixorigin:3.0-devfrom
Open
[3.0-dev] Backport prepared-statement / connector fixes (Fixes #24186)#24191jiangxinmeng1 wants to merge 10 commits intomatrixorigin:3.0-devfrom
jiangxinmeng1 wants to merge 10 commits intomatrixorigin:3.0-devfrom
Conversation
…23413) - Return BOOL as TINYINT(1) (0/1) instead of VARCHAR ("true/false") in MySQL protocol - Adjust column definition (type/length/charset) and result row encoding (text/binary) - Update related unit test and add jdbc-type-compatibility case - Refresh BVT expected results Approved by: @XuPeng-SH, @heni02
Adds the missing server_id system variable to MatrixOne so ODBC clients can read @@server_id during connection initialization without error. Also adds regression coverage in the distributed system variable tests. This prevents Tableau/ODBC connections from failing with “system variable does not exist.” Approved by: @fengttt, @heni02, @XuPeng-SH
…stmt null json panic (matrixorigin#23864) When a prepared statement inserts NULL into a JSON column multiple times, the second execution causes a panic during SELECT: ``` panic runtime error: index out of range [0] with length 0 ByteJson.Unmarshal (bytejson.go:96) DecodeJson (encoding.go:93) ColumnSlices.GetStringBased (output.go:538) ``` **Reproduce:** ```sql CREATE TABLE test_ps (id INT, meta JSON); PREPARE stmt FROM 'INSERT INTO test_ps (id, meta) VALUES (?, ?)'; SET @A=1, @b=NULL; EXECUTE stmt USING @A, @b; SET @A=2, @b=NULL; EXECUTE stmt USING @A, @b; DEALLOCATE PREPARE stmt; SELECT meta FROM test_ps where id = 1; -- ok SELECT meta FROM test_ps where id = 2; -- panic ``` **Root cause:** `FunctionExpressionExecutor.doFold()` calls `ToConst()` on the result vector after constant folding. On the second execution, `PreExtendAndReset()` calls `ResetWithSameType()` which clears nsp (null bitmap) but preserves the CONSTANT class. When `FunctionResult.AppendBytes(nil, true)` is called on a CONSTANT vector, it skips setting the null marker entirely, resulting in a non-null empty byte slice being stored instead of NULL. **Fix:** Reset vector class to FLAT in `ResetWithSameType()`, so the vector is in a clean state for re-population. `doFold()` will call `ToConst()` again after evaluation if needed. Approved by: @heni02, @aunjgr
This PR fixes a missing `MYSQL_TYPE_JSON` branch in the prepared-statement binary execute parser on `main`. Without this change, `COM_STMT_EXECUTE` falls through to `unsupport parameter type` when a client sends a JSON-typed parameter through the binary protocol. Focused validation: - `go test ./pkg/frontend -run 'TestParseExecuteDataWithJSONParam'` - `go test ./pkg/frontend -run 'TestParseExecuteDataWithJSONParam|Test_StripPassword'` Approved by: @LeftHandCold
This PR reclaims prepared-statement binary parameter state after `COM_STMT_EXECUTE` and on malformed execute cleanup. It fixes two closely related problems on `main`: 1. repeated binary executes with varlen parameters reused the same params vector and kept appending to its area, growing session memory until statement close; 2. malformed `COM_STMT_EXECUTE` packets could leave a partially populated params vector attached to the prepared statement because the parse-error path dropped the statement pointer before cleanup. Focused validation: - `go test ./pkg/frontend -run 'TestPrepareStmtClearBinaryParamStateReleasesParamArea|Test_ExecRequestStmtExecuteErrorClearsPreparedParamState'` - `go test ./pkg/frontend -run 'TestPrepareStmtClearBinaryParamStateReleasesParamArea|Test_ExecRequestStmtExecuteErrorClearsPreparedParamState|Test_parseStmtSendLongData'` Approved by: @LeftHandCold
This PR fixes three categories of issues: - Repeated COM_STMT_SEND_LONG_DATA chunks for the same parameter now **append** instead of overwriting, matching MySQL protocol semantics. - Truncated COM_STMT_EXECUTE packets now return a malformed-packet error instead of risking a panic while reading the new parameter bound flag. - PreInsert.refreshAutoIncrementTableID() wrote directly to the shared *plan.TableDef.TblId during Prepare(). - MultiUpdate.Prepare() read the same TblId concurrently in a different ants worker pool goroutine. - **Fix**: Store the refreshed TblId in a local container field (ctr.tblId) instead of mutating the shared plan object. - PartitionInsert and PartitionDelete temporarily mutated the shared ObjRef.ObjName during partition iteration (save/restore via defer) -- replaced with stack-local copies. - PartitionMultiUpdate.writeTable() directly mutated shared MultiUpdateCtx entries -- now uses clone() to match the writeS3 code path. - Also fixed a pre-existing copy-paste bug in MultiUpdateCtx.clone() where PartitionCols was incorrectly copying DeleteCols. Approved by: @ouyuanning
This PR hardens prepared-statement binary protocol cleanup on `main`. It fixes three closely related problems in the frontend prepare path: 1. `PrepareStmt.Close()` now keeps a valid process context for freeing binary parameter vectors after binary parsing. 2. `COM_STMT_RESET` / `reset prepare` now clears accumulated binary prepared-statement state instead of acting like a no-op. 3. `COM_STMT_CLOSE`, `COM_STMT_RESET`, and malformed `COM_STMT_EXECUTE` error paths now cleanly return and clean up instead of dereferencing nil state or leaving dirty binary state behind. Focused validation: - `go test ./pkg/frontend -run 'Test_doResetClearsPreparedBinaryState|Test_ExecRequestPrepareCommandMissingStmt|Test_ExecRequestStmtExecuteErrorClearsPreparedBinaryState|TestPrepareStmtCloseAfterBinaryParamParsing|TestMysqlProtocolImpl_Close|Test_parseStmtSendLongData'` Approved by: @LeftHandCold
jiangxinmeng1
requested review from
XuPeng-SH,
aunjgr,
heni02 and
ouyuanning
as code owners
|
|
heni02
approved these changes
Apr 24, 2026
…ayofweek, create_sequence
XuPeng-SH
approved these changes
Apr 24, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
8 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What type of PR is this?
Which issue(s) this PR fixes:
fixes #24186
What this PR does / why we need it:
Backports the prepared-statement / connector compatibility fixes from
mainto3.0-devlisted in #24186:server_idsystem variable for ODBCResetWithSameTypeto fix prepared stmt null json panicAdditional changes required to make the cherry-picks compile and pass tests on 3.0-dev:
clearBinaryParamStateruns;doResetnow clears binary state; COM_STMT_RESET / COM_STMT_CLOSE error paths return early instead of dereferencing a nil prepareStmt;parseStmtExecutereturnspreStmton error).MultiUpdateCtx.clone()helper (withPartitionColsinitialization) on 3.0-dev that fix: harden prepare long data parsing #23980 expects.deletion_partition_test.go,insert_partition_test.go) to use the 2-argumenttestutil.MakeInt64Vector/MakeRowIdVectorsignatures present on 3.0-dev, and renamed the partitionExprWithRowIDfield reference toExpr.TestResetWithSameTypeResetsClass/TestFunctionResultAppendNullAfterToConst(already present on 3.0-dev).MYSQL_TYPE_DATETIME/MYSQL_TYPE_TIMESTAMPinappendResultSetBinaryRow2and brought scale/timezone-aware DATE/DATETIME/TIME/TIMESTAMP formatting intoappendResultSetTextRow(matchesmain).Special notes for your reviewer:
CI on this branch will validate the BVT delta. Tested locally with
go test -tags matrixone_test ./pkg/frontend/... ./pkg/sql/colexec/{insert,deletion,preinsert,multi_update}/... ./pkg/container/vector/...— all green (withTZ=UTCfor the pre-existing TZ-sensitiveTestRowToString).