pt-online-schema-change fails with "Duplicate entry" on MariaDB

Is this a known MariaDB bug with SHOW WARNINGS or does the query really only produce a warning on MariaDB? Does this happen consistently? for any table and data?

I think that it isn't a bug of MariaDB. I create simple table with primary key in MariaDB, then i try INSERT INTO ... with duplicated id, SHOW WARNINGS returns row with level "Error". And when i try INSERT IGNORE INTO ..., SHOW WARNINGS returns row with level "Warning".
MySQL returns same result in first case, and nothing in second case.
MariaDB seems to return just more verbose data.
It happens consistently for any table and data i've tried.

That's good to know. I'll set up a MariaDB sandbox and try to reproduce this then. Thanks for the info.

Indeed, MariaDB fixed this "bug" in 5.5.28:

https://kb.askmonty.org/en/mariadb-5528-changelog/ : "Added warnings for duplicate key errors when using INSERT IGNORE"

I say "bug" because although http://dev.mysql.com/doc/refman/5.5/en/insert.html says "If you use the IGNORE keyword, errors that occur while executing the http://dev.mysql.com/doc/refman/5.5/en/insert.html statement are treated as warnings instead.", even in MySQL 5.5.29 that is not actually the case.

So is it a MySQL or documentation bug? I checked 5.0.96 and it does not generate a warning. Yet, http://dev.mysql.com/doc/refman/4.1/en/insert.html says it's treated as a warning.

So I think we have a clear case here were historical precedence of how things actually working has become the rule, despite equally historical documentation to the contrary.

In any case, Brian is going to try making the tool simply ignore "Duplicate entry" warnings altogether because that's what we intend to do by using INSERT IGNORE. So far, I can't think of why ignoring this warning would cause any problems.

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PT-345