On 20 April 2015 at 18:33, Bruce Momjian <bruce@momjian.us> wrote:
Also, I am also not sure we should be designing features at this stage in our release process.
I see this more as a process of gaining approval. I don't think patches at the back of the queue should get the "its too late treatment" just because they are at the back of the queue. They are logically all at the same stage. There should be a way to allow people that show patience and respect for the process to get the same timeshare as those that push their patches daily.
Anyway, in this case, the patch conflicts with other things going in now, so changing things isn't really sensible for this release, in terms of my time.
We clearly need to do a better job of piggybacking actions on a dirty block. The discussion has been about whether to do early cleanup, but we should also consider post-access cleanup if we set hint bits or dirtied the block in other ways.
Since we have many votes in favour of change in this area I'll post a new version and look for an early review/commit for next release.
--
Simon Riggs http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services