On Wed, Dec 6, 2017 at 5:01 AM, Amit Khandekar <amitdkhan...@gmail.com> wrote: > This looks good. > > In attached revised patch, just added some comments in the changes that you > did.
Committed, thanks. It's rather embarrassing that I didn't notice this problem, because I did compare that logic with the preceding loop. I concluded it was OK on the theory the previous loop would have already given up if there were no partial plans. But that's wrong, of course: the previous loop will not have given up if it grabbed the last plan in a list of only non-partial plans. Oops. -- Robert Haas EnterpriseDB: http://www.enterprisedb.com The Enterprise PostgreSQL Company
