Prefix indexes are specified in characters while this error message
returns a limit in bytes. By adding the byte length in the message this
is makes it easier to understand.
This would deviate a little bit from the message in MySQL.
Before:
```
sql> CREATE TABLE t1(id int primary key, c1 text, key(c1(1000)));
ERROR: 1071 (42000): Specified key was too long; max key length is 3072 bytes
```
After:
```
sql> CREATE TABLE t1(id int primary key, c1 text, key(c1(1000)));
ERROR: 1071 (42000): Specified key was too long (4000 bytes); max key length is 3072 bytes
```
* Added test case
* ddl fix#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 #24427
* Disabled tests for CI testing
* Revert "Disabled tests for CI testing"
This reverts commit 17c28f30ba8802c578fed5653107aa6995e17607.
* Revert "Revert "Disabled tests for CI testing""
This reverts commit 65c84d94f7ab4440c739703312f5329619c6ccdc.
* 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 <ti-community-prow-bot@tidb.io>
