Heikki Linnakangas wrote:
> While playing around, I bumped into another related bug, and after
> googling around I found out that it was already reported by Robert
> Haas earlier, but still not fixed:
> Kevin, the last message on that thread says you'll write a patch
> for that. Ping?
OK, attached is a first cut to confirm that the approach looks sane
to everyone; I *think* it is along the lines that we reached
consensus. After moving the check to the point where we get a
serializable snapshot, it was still behaving badly. It took a bit,
but forcing the snapshot acquisition in InitPostgres to use 'read
committed' instead of the default isolation level got reasonable
behavior in my initial tests. It sure looks a lot better to *me*
than what was happening before.
Besides confirming that this is the solution people want, this has
not been tested well enough yet to be ready for commit. It's getting
late, though, and I'm fading. If the overall approach looks good
I'll beat up on it some more tomorrow.
Since the existing behavior is so bad, I'm inclined to think this
merits backpatching to 9.1. Thoughts on that?
-Kevin