On 11/30/14 12:54 PM, Tom Lane wrote:
> "Not usable" is hyperbole. The actual effect would be that the particular
> function would fail to restore, but the rest of it would be fine.
>
> I should note that it's not like we've never seen any other cases of
> pg_dump failing to deal with circular dependencies. Issuing a warning
> and pushing on is *way* better than refusing the dump entirely.
Maybe there should be an option to fail or not fail in these cases.
Both cases are useful, but I tend to agree that it should fail hard by
default.