Andres Freund <andres@2ndquadrant.com> writes:
> On 2013-11-27 19:24:35 -0300, Alvaro Herrera wrote:
>> The other idea is to just not backpatch this.
> I think backpatching is a good idea, I have seen GetMultiXactIdMembers()
> + slru code take up 80% cpu in strange workloads. But it possibly might
> be a good idea to wait till after the next point release to give people
> at least a minimal chance of catching problems.
Agreed on both counts. We're close enough now to Monday's wrap that we
should avoid anything that risks destabilizing 9.3.x, unless it's to fix
a serious bug. AIUI this is just a performance issue, so let's wait till
after 9.3.2 is done to push in the fix. But since it is a performance
regression from pre-9.3, never back-patching the fix at all isn't very
attractive.
regards, tom lane