Fixed in master, leading to several follow up bugs that all had to be fixed:
1. generate identifier allowed first char to be a number
2. postgresql composite foreign key detection left a space in the second (and more) column names
3. Index column names were not sanitized to lower-case, leading to comparison bugs.
There has been a major refactoring now such that, for each foreign key there is always an explicit index being created. On SQLite, Postgres and Oracle this can lead to quite some additional indexes being created now using SchemaTool --update. MySQL already did this implicitly.
There are now heuristics that detect duplicate indexes (based on columns indexed) and override rules (adding primary on columns foo, bar will delete index on columns foo bar).