[jira] [Closed] (CAY-2944) MergerTokenFactoryProvider does not resolve adapter subclasses

["Andrus Adamchik (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CAY-2944?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Andrus Adamchik closed CAY-2944.
--------------------------------
    Resolution: Fixed

> MergerTokenFactoryProvider does not resolve adapter subclasses
> --------------------------------------------------------------
>
>                 Key: CAY-2944
>                 URL: https://issues.apache.org/jira/browse/CAY-2944
>             Project: Cayenne
>          Issue Type: Bug
>            Reporter: Andrus Adamchik
>            Assignee: Andrus Adamchik
>            Priority: Minor
>             Fix For: 5.0-M2
>
>
> We had this bug like forever. It was exposed by my recent attempt to unwind test suite customizations (namely let the test stack guess its DbAdapter). So... 
>  
> MergerTokenFactoryProvider.get(DbAdapter) looks up the per-adapter MergerTokenFactory by exact class name. When the underlying adapter is a subclass of a
>   mapped adapter, the lookup misses and falls back to DefaultMergerTokenFactory, producing SQL that is invalid for the actual database.
> Example: against HSQLDB 2.x, HSQLDBSniffer returns HSQLDBNoSchemaAdapter (extends HSQLDBAdapter). The factory map only has HSQLDBAdapter, so a column-type
>   change emits ALTER ... TYPE VARCHAR(20) instead of HSQL's ALTER COLUMN ... VARCHAR(20). HSQL rejects it, leaving the schema out of sync.
>   Fix: Walk up the adapter's class hierarchy when resolving the factory; fall back to the default only if no ancestor matches.
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)