Re: snapshot too old, configured by time

On Mon, Apr 18, 2016 at 8:41 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Kevin Grittner <kgrittn(at)gmail(dot)com> writes:
>> On Sun, Apr 17, 2016 at 10:38 PM, Alvaro Herrera
>> <alvherre(at)2ndquadrant(dot)com> wrote:
>>> I understand the backpatching pain argument, but my opinion was the
>>> contrary of yours even so.
>
>> The other possibility would be to backpatch the no-op patch which
>> just uses the new syntax without any change in semantics.
>
> That would break 3rd-party extensions in a minor release, wouldn't it?
> Or do I misunderstand your suggestion?

With a little bit of a change to the headers I think we could avoid
that breakage.

The original no-op patch didn't change the executable code, but it
would have interfered with 3rd-party compiles; but with a minor
adjustment (using a modified name for the BufferGetPage with the
extra parameters), we could avoid that problem. That would seem to
address Álvaro's concern while avoiding five years of backpatch
nightmares.

I don't claim it's an *elegant* solution, but it might be a workable compromise.

--
Kevin Grittner
EDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company