> Obviously they are not identical, the question is why are they not identical after following the instructions that sqldiff itself gave? That's a key question with respect to the OP's problem. I would argue that if sqldiff gave the "instructions" that made one DB almost but not quite like the other, then something non-visible in the representation coming out of sqldiff **must** be at work. It's enough to make me wish SQLite had PostgreSQL's C-style string escape sequences. > It may be that the CRLFs are lost/converted when capturing or running the query -- OP what process are you using to apply the changes specified by sqldiff? That question is apropos, but this should be added: OP, how are you capturing the changes specified by sqldiff?