Skip to content
  • Jakub Łopuszański's avatar
    3cdd6309
    BUG#27646322 RECLOCK::ADD_TO_WAITQ CALLS LOCK_UPDATE_AGE ON DB_DEADLOCK INFLATING TRX->AGE · 3cdd6309
    Jakub Łopuszański authored
    
    
    - two new MTR tests (appended to already existing suite/innodb/innodb_cats.test)
    - added DBUG_EXECUTE_IF required to verify that trx->age is now well behaving in above tests
    - change in RecLock::add_to_waitq, such that the lock_update_age is not called in case of DB_DEADLOCK
    - removal of the redundant wait parameter of lock_update_age
    - added C++11 [[ noreturn ]] attribute to ut_error,
    - removed unreachable code after ut_error
    - in particular removed the only use of DB_QUE_THR_SUSPENDED and thus simplified handling of lock_table
    - adjusted @return doxygen comments to reflect that DB_QUE_THR_SUSPENDED is now an impossible value
    - added some ut_ad asserts which verify that return value indeed matches the values advertised by @return documentation
    - optimized lock_rec_add_to_queue by avoiding a search for similar lock, unless the type_mode permits reuse of a lock object
    - fixed a comment for lock_prdt_lock to reflect that it can return DB_SUCCESS_LOCKED_REC
    
    Reviewed-by: default avatarJimmy Yang <jimmy.yang@oracle.com>
    Reviewed-by: default avatarMarcin Babij <marcin.babij@oracle.com>
    3cdd6309
    BUG#27646322 RECLOCK::ADD_TO_WAITQ CALLS LOCK_UPDATE_AGE ON DB_DEADLOCK INFLATING TRX->AGE
    Jakub Łopuszański authored
    
    
    - two new MTR tests (appended to already existing suite/innodb/innodb_cats.test)
    - added DBUG_EXECUTE_IF required to verify that trx->age is now well behaving in above tests
    - change in RecLock::add_to_waitq, such that the lock_update_age is not called in case of DB_DEADLOCK
    - removal of the redundant wait parameter of lock_update_age
    - added C++11 [[ noreturn ]] attribute to ut_error,
    - removed unreachable code after ut_error
    - in particular removed the only use of DB_QUE_THR_SUSPENDED and thus simplified handling of lock_table
    - adjusted @return doxygen comments to reflect that DB_QUE_THR_SUSPENDED is now an impossible value
    - added some ut_ad asserts which verify that return value indeed matches the values advertised by @return documentation
    - optimized lock_rec_add_to_queue by avoiding a search for similar lock, unless the type_mode permits reuse of a lock object
    - fixed a comment for lock_prdt_lock to reflect that it can return DB_SUCCESS_LOCKED_REC
    
    Reviewed-by: default avatarJimmy Yang <jimmy.yang@oracle.com>
    Reviewed-by: default avatarMarcin Babij <marcin.babij@oracle.com>
Loading