On Tue, Dec 18, 2018 at 10:26 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Yeah, I've been wondering about that as well. The original intention
> for dependency traversal was that it'd work independently of the ordering
> of entries in pg_depend. If it's not doing so, I'd call that a bug in
> dependency traversal rather than something the index code needs to be
> responsible for.
It's clearly not doing so. Is somebody else actually going to fix it,
though? I'm quite happy to stand aside to make way for a better
solution, but I don't consider myself particularly qualified to come
up with that better solution. The findDependentObjects() code is
pretty subtle.
> (Note that this statement doesn't disagree with our issues about needing
> to suppress dependency reports in the regression tests; that's because
> the order of reports about independent objects can legitimately depend on
> the index order. But there shouldn't be any semantic differences.)
The advantage of my admittedly kludgy approach is that it will almost
completely eliminate any further need to suppress acceptable
regression test differences -- those non-semantic output differences
that we're already suppressing semi-regularly. In practice, adding a
trailing attribute to each of the two pg_depend indexes almost
entirely eliminates the need to play whack-a-mole. That has to be an
advantage for the kind of approach that I've suggested.
--
Peter Geoghegan