David E. Wheeler

Yeah, that sounds like the Oracle approach. If you use the Oracle username `sqitch` to do a deploy, the objects are created in the `sqitch` schema, which means you can...

> So we are stuck with the separate database approach, yes? Yeah, I think so. > Table-qualify-ing all references to the column is good, but not enough :) CUBRID wants...

So it is throwing the proper error, that's good. Maybe it's a bug in DBD::Cubrid that `HandleError` does not work? Or maybe it does, but it turns it into a...

Maybe stick some debugging code in that `HandleError` sub to see if it's reached. Something like ``` print STDERR "here I am!\n"; ```

> It's not called, looks like I bug to me. Okay, I suggest wrapping it in a `SKIP` block and only running the test if `$class ne App::Sqitch::Engine::cubrid`. You can...

BTW, I have now merged the `mysql` branch into `master`. You might want to `rebase` from there. Also, some of what I ended up having to do might be of...

Crap, the quoting of `changes` makes Oracle crap out. The quotes force case-sensitivity, and unlike MySQL, Postgres, and SQLite, the default case internal to Oracle is uppercase. So `"CHANGE"` will...

Yeah. Besides, some other engine will be added that reserves some other term we use. Might as well generalize the solution now. Commit coming shortly.

So you will probably need to add the use of that method to a few more places for the CUBRID support, @stefansbv.

MySQL has similar limitations, which is why its columns are not all `TEXT`. Setting things up to be UTF-8, I made them as big as could be indexed, which is...