On 28 January 2018 at 12:00, Tomas Vondra <tomas.vondra@2ndquadrant.com> wrote: > On 01/27/2018 10:45 PM, Tom Lane wrote: >> David Rowley <david.rowley@2ndquadrant.com> writes: >>> I'd offer to put it back to the order of the enum, but I want to >>> minimise the invasiveness of the patch. I'm not sure yet if it should >>> be classed as a bug fix or a new feature. >> >> FWIW, I'd call it a new feature. >> > > I'm not sure what exactly the feature would be? I mean "keep statistics > even if you only ask for indexes" does not seem particularly helpful to > me. So I see it more like a bug - I certainly think it should have been > handled differently in 10.
Now I'll ask; On me doing so, would anyone have pushed back on that request and said that what I'm asking is a separate feature?
If the answer to that is "no", then this is a bug that should be fixed and backpacked to v10.
Its a bug of omission (I'm going with no one saying no to your proposition) - though that still doesn't automatically allow us to back-patch it.
This bug has an obvious if annoying work-around and fixing the bug will likely cause people's code, that uses said work-around, to fail. Breaking people's working code in stable release branches is generally a no-no.
However, given that this was discovered 4 months after the feature was released suggests to me that we are justified, and community-serving, to back-patch this. Put more bluntly, we can ask for more leeway in the first few patch releases of a new feature since more people will benefit from 5 years of a fully-baked feature than may be harmed by said change. We shouldn't abuse that but an obvious new feature bug/oversight like this seems reasonable.
