Liquibase has its share of gotchas, especially in the SQL syntax. My sense is that most of the problems I experienced had to do with weird edge cases introduced by trying to handle some complex enterprisey scenario that I would rather deal with by just not letting things get that complicated in the first place. That exclusive lock mechanism, for example, isn't foolproof, and I suspect that it has bitten me more times than it's helped me.

-but-

Roll-your-own is not necessarily the easy option. I'd say that, if your goal is to get something robust enough to use in production, a day to implement is a very optimistic estimate. I'd say that it's one to steer clear of unless you're sure you know what you're doing.

Which one is better depends, as usual, on your situation. I think my rubric would probably be: If the XML schema definition language is at all attractive to you, use Liquibase (or similar). That's far and away its most unique contribution. If, on the other hand, the knowledge that anyone would even consider defining their SQL schema in a language other than SQL fills you with holy rage, maybe choose an option that sits closer to the metal. This doesn't necessarily need to be DIY project, but, thanks to Zawinski's Law, that might be the only way to ensure that your solution also stays close to the metal.