Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> How should it work?
> 1. Remove the BEGIN and COMMIT around blobs?
> 2. Use SAVEPOINT ?
> Presumably (1).
Yeah, there is no need for the per-blob begin/commit if we've got one
around the whole restore.
One thing to be careful of is not to suppress BEGINs that are sent on
the dumping side --- it's sometimes hard to tell which parts of pg_dump
execute when.
regards, tom lane