"Kevin Grittner" wrote:
> A patch is attached which just covers the predicate lock
> acquisition
This patch rolls that up with snapshot checking in the conflict
detection function called on read. The only other two functions
which use that macro check for conflicts on write, and I can't see
why snapshot checking would make sense there. The write isn't into a
particular snapshot, it's in the context of the transaction; so
unless someone can make a good case for why snapshot checking makes
sense there, I think this is far as it makes sense to take this.
Except of course for those two functions in the btree AM which didn't
have snapshot available. I'm not sure what needs to be done there.
Thoughts?
-Kevin