Configuration.GenerateSchemaCreationScript() Exception with Oracle8iDialect and above

[greenlight2k]

Issue since NH5:

When I use:

NHibernate.Cfg.Configuration.GenerateSchemaCreationScript(new Oracle8iDialect());

an exception is thrown:

Message:
Dialect does not support DbType.String
Parametername: typecode

StackTrace:
at NHibernate.Dialect.TypeNames.Get(DbType typecode)
at NHibernate.Dialect.Dialect.GetTypeName(SqlType sqlType)
at NHibernate.Mapping.Column.GetDialectTypeName(Dialect dialect, IMapping mapping)
at NHibernate.Mapping.Column.GetSqlType(Dialect dialect, IMapping mapping)
at NHibernate.Mapping.Table.SqlCreateString(Dialect dialect, IMapping p, String defaultCatalog, String defaultSchema)
at NHibernate.Cfg.Configuration.GenerateSchemaCreationScript(Dialect dialect)

I checked the sourcecode and found the problem:
Inside the constructor of Oracle8iDialect there was a line removed:
RegisterCharacterTypeMappings(); (line 84)

I inherited a new class from Oracle8iDialect and added that line in the constructor.
This should fix the issue. But it needs to be confirmed and checked further.

Question is: Why was that mapping removed?

public sealed class Oracle10gFixed : Oracle10gDialect
{
    public Oracle10gFixed()
    {
        RegisterCharacterTypeMappings();
    }
}

UPDATE:

This issue does not occur with SchemaUpdate or SchemaExport,
only with the GenerateSchemaCreationScript method in Configuration.

[fredericDelaporte]

RegisterCharacterTypeMappings has been moved to this dialect's Configure method, because it now depends on configuration settings. So the actual bug is the use of an un-configured dialect.

For obtaining a configured dialect, better use NHibernate.Dialect.Dialect.GetDialect(cfg.Properties) rather than a direct call to its constructor. (Assuming that cfg is a Configuration object and that the dialect to use has been specified in the configuration.)

Otherwise call the Configure method on the dialect before using it.

The fact that dialects need to be configured before being used is a breaking change from 5.0.0 that we have forgotten to document.

[hazzik]

Closing as not an issue as per @fredericDelaporte explanation