- Old SQLite setups have the last IP column as NOT NULL but without a default value. With the new concept (where we don't set a last IP on player registration) it fails.
- Create an /authme debug child that allows to migrate SQLite (tricky because SQLite does not support dropping or modifying columns)
- Allow last IP column to be NOT NULL in MySQL as well (extend MySQL /authme debug child)
- Add TODO comments with follow-up issue to extend our commands with new registration IP field
* Introduce hasSession field in datasource
That makes isLogged more consistent as it will be '1' only when the player is online.
* Fixes
* Fix unit testing
* Update config doc
* Create SessionService
* Create test for SessionService, avoid DB operations if sessions are disabled
* Cleanup: remove outdated warning for session timeout = 0
- Remove outdated warning
- Encapsulate session enabled check in SessionService
* Fix failing SessionServiceTest, add data source integration tests for session methods
- Take MAX(reg_date, login_date) as timestamp to compare against
- Remove the second "all" parameter to include entries with 0 registration date -> we expect registration date to always be set to the current date, so the parameter becomes obsolete
- Create command under /authme debug that allows to change the 'nullable' status of MySQL columns (currently last date and email only)
- We need to offer a default value for forum integrations that have a NOT NULL email column. Offering a command avoids us from force-migrating existing databases while still offering migrations in both directions
- Change in default value handling: lack of values are not handled by setting default values to the PlayerAuth anymore, and reading a default value from the database into a PlayerAuth will be translated into null by the PlayerAuth builder
- When a new database is created, email and lastlogin are now nullable and lack a default a value
Open points:
- Finish MySqlDefaultChangerTest
- Revise purging logic (#792)
- Allow to have more columns nullable (#814)
As noticed by @Gnat008
- We need two different implementations for MySQL and SQLite because SQLite uses an older version where #isClosed is not implemented
- Extract logic used in SQLite and MySQL for logging and closing SQL objects
- Decided to leave buildAuthFromResultSet methods individually as this might be more implementation-specific
- Rename DataSource#close to DataSource#closeConnection to fix conflict with static import
- FlatFile doesn't store the "realname" - all names are always in all-lowercase. Converting from flatfile to other data source should therefore not take over an auth's realname
- Adjust sample flatfile file to only have all-lowercase usernames
- authme.cache to authme.data
- Rename PlayerData to LimboPlayer to match with LimboCache
- Move authme.converter to authme.datasource.converter
- Split output package into output and message
- /email recovery generates recovery code and resets password only if recovery code is also given
- Change data source method to return email and recovery code
- Change SendMailSSL to be injected into classes and created regardless of settings
- Various minor cleanups (remove accidentally committed test, add more precise logging statement)
- Drop initialization of all columns on table create in favor of checking each column individually. This is slower but guarantees that each column is only defined once in the code. Columns are only created once so having clean code outweighs performance.
- Write more datasource integration tests
- Rename autoPurging to isPurging: we should always register if a purge task is in progress (regardless if autopurge or not) and deny any new requests
- Reduce the same logic being coded multiple times by calling through the methods
- DataSource: remove purgeBanned in favor of purgeRecords, both do exactly the same thing
- Move console initialization for tests into TestHelper
- Remove unused properties in legacy Settings
- Add issue number to TODO comments where applicable
- Remove redundant uses of WrapperMock
- Use assertThat() from JUnit, not hamcrest
- Use hamcrest Matchers everywhere (not BaseMatchers etc.)
- Favor Mockito's argThat() over using ArgumentCaptor (more succinct)
- Delete useless test classes
