Casey Deccio <casey@deccio.net> writes:
> Being unfamiliar with the internals, what's the risk here? If postgres
> thinks something is a bigint, but previously stored it as an int, does that
> mean it will try to extract data beyond the boundary of some of the (old)
> 32-bit values and potentially throw off offsets for other values?
Yes. This *will* break your table, spectacularly. The other person who
was opining that it would work has no understanding of the actual storage
layout.
regards, tom lane