It's very annoying that the DELIMITER command has a different syntax in the client and in mysqltest:

• Client:

  DELIMITER $$

  BEGIN NOT ATOMIC

  SELECT 1;

  END;

  $$

  DELIMITER ;

• mysqltest:

  DELIMITER $$;

  BEGIN NOT ATOMIC

  SELECT 1;

  END;

  $$

  DELIMITER ;$$

We want to unify "mysqltest" and "mariadb" so any script can be pasted to either of them without modifications.

Originally, there was a proposal to support both styles and to make it possible to choose between the two styles by introducing new switches:

• A new command to specify the delimiter style in a *.test file:

  --delimiter=client

  --delimiter=mysqltest

  with mysqltest being default (the current behavior).

• Add a command line option to mysqltest to specify the default delimiter behavior:

  mysqltest --delimiter=client

• Add a way to define delimiter style per the entire suite. This can probably be done by putting a my.cnf file into the suite directory:

  [mysqltest]

  delimiter=client

After discussion with the team (nikitamalyavin, serg, bar) we decided that having switches to choose between DELIMITER style would be over-engineering. Let's have the new behaviour hard-coded.

So non of these switches are needed.

We'll do the following:

1. Change mysqltest so instead of this style:

DELIMITER $$;

BEGIN NOT ATOMIC

  SELECT 1;

END;

$$

DELIMITER ;$$

it only understands just this style, compatible to the "mariadb" command line client:

DELIMITER $$

BEGIN NOT ATOMIC

  SELECT 1;

END;

$$

DELIMITER ;

2. Instead of adding the switches, we just fix all test result files *.result accordingly, to use the new DELIMITER style.

This change in the task means less programming on the step N1, but some time on the step N2.

A quick estimation in MariaDB-10.9 sources gives 10658 matches in 641 test files:

$ ag -i --stats-only delimiter mysql-test/

10658 matches

641 files contained matches

14818 files searched

But it seems most of them can most likely be changed using a simple script.