Alexander Korotkov <a.korotkov@postgrespro.ru> writes:
> [ aminterface-13.patch ]
I've committed this after some rather significant rework, not all of
it cosmetic in nature. For example, the patch fell over under
CLOBBER_CACHE_ALWAYS (planner failing to copy data out of relcache
entries that might not survive long) and on 32-bit machines (failure
to fix index_getbitmap API properly). I might've missed some other
portability issues; we'll see what the buildfarm says.
The amvalidate functions could do with quite a bit more work in the
long run. For now, they more or less replace the queries we had to
remove from opr_sanity, but I'd like to see almost all of what
opr_sanity does with opclasses get pushed down, so that these functions
can provide a test facility for extensions' opclasses. That does not
need to hold up adoption of the patch, though.
regards, tom lane