Andrew Dunstan <andrew@dunslane.net> writes:
> The pg_upgrade part of this seems seriously awry (see buildfarm animal
> crake). We can't assume that the upgrade source is using the new wording.
Ugh. I'll look into a fix tomorrow --- but my first instinct is
that pg_upgrade shouldn't be so dependent on the exact wording of
pg_controldata output.
regards, tom lane