On Tue, Jun 2, 2015 at 10:10 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> What it seems like we should do, if we want to back-patch this, is apply
> it without the add_path_precheck changes. Then as an independent
> HEAD-only patch, change add_path_precheck so that it's behaving as
> designed. It looks to me like that will save some planning time in any
> case --- changing add_path_precheck to disregard startup cost when
> appropriate seems to let it reject a lot more paths than it used to.
I'd just like to mention that I really appreciate the time and thought
that went into keeping the back-patched portion of this fix narrow.
Thanks!
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company