…reorg. (pingcap#38738)

* Added test case

* ddl fix pingcap#38669.

The issue was that mysql.tidb_ddl_reorg table was updated by an
inner transaction after the outer transaction started,
which then made a commit conflict in the outer transaction,
when it deleted the same row.

* Fixed typo in comment

* Added test case for pingcap#24427

* Disabled tests for CI testing

* Revert "Disabled tests for CI testing"

This reverts commit 17c28f3.

* Revert "Revert "Disabled tests for CI testing""

This reverts commit 65c84d9.

* removed test skips

* Clean up the tidb_ddl_reorg entry after DDL is completed

* Use a cleanup job afterwards instead.

* Fixed test

* Moved cleanup before asyncNotify

* More detailed test failure log

* Refined test error message

* Injecting timoeut to get stack traces from CI

* Updated Debug Dump on timeout

* Delete mulitple entries in tidb_ddl_reorg if needed

* Linting

* Linting

* Added CI debug logs

* Linting + CI debugs

* fixed CI debug

* Try to cleanup also if job.State == synced

* check for non-error of runErr instead of error...

* Use a new session, instead of reusing worker.sess

* Also handle case when job == nil

* Removed CI debug logs

* Misssed change session from w.sess to newly created sess

* Improved TestConcurrentDDLSwitch and added CI debug logs

* Always cleaning up all orphan mysql.tidb_ddl_reorg entries

* linting

* Also cleanup if job is nil

* Updated TestModifyColumnReorgInfo + CI debug logs

* more CI debug

* refactored the cleanupDDLReorgHandle code

* Added missing cleanup in handleDDLJobQueue

* Removed debug panic

* Code cleanup

* Test updates

* Debug cleanup

* Cleaned up test after removal of old non-concurrent DDL code merge

* Linting

* always wrap changes to tidb_ddl_reorg in an own transaction

+ fixed some typos

* Minimum fix

* Always update reorg meta, not only on error

* Issue is here :)

* Fixed newReorgHandler

* Wrapped more tidb_ddl_reorg changes into separate transactions

* linting

* Removed updateDDLReorgStartHandle

* cleanups

* Made runInTxn a method on *session, instead of normal function

* Update test

* Final touches

* Removed duplicate test

* CleanupDDLReorgHandles should only be called from HandleJobDone.

* Variable rename

* Renamed 'delete' variabel name

* Updated test

* small revert

* Removed timeout debugging code

* Simplified the cleanup to only start a new txn and not a new session

* Reverted the change of GetDDLInfo

Co-authored-by: Ti Chi Robot <[email protected]>

…8738-6.5

…8738-6.5

…8738-6.5

…8738-6.5