Re: SimpleLruTruncate() mutual exclusion

On Fri, Jan 31, 2020 at 12:42:13AM -0800, Noah Misch wrote:
> On Thu, Jan 30, 2020 at 04:34:33PM +0100, Dmitry Dolgov wrote:
> > Just to clarify, since the CF item for this patch was withdrawn
> > recently. Does it mean that eventually the thread [1] covers this one
> > too?
> > 
> > [1]: https://www.postgresql.org/message-id/flat/20190202083822.GC32531%40gust.leadboat.com
> 
> I withdrew $SUBJECT because, if someone reviews one of my patches, I want it
> to be the one you cite in [1].  I plan not to commit [1] without a Ready for
> Committer, and I plan not to commit $SUBJECT before committing [1].  I would
> be willing to commit $SUBJECT without getting a review, however.

After further reflection, I plan to push $SUBJECT shortly after 2020-08-13,
not waiting for [1] to change status.  Reasons:

- While I put significant weight on the fact that I couldn't reproduce
  $SUBJECT problems without first fixing [1], I'm no longer confident of that
  representing real-world experiences.  Reproducing [1] absolutely requires a
  close approach to a wrap limit, but $SUBJECT problems might not.

- Adding locks won't change correct functional behavior to incorrect
  functional behavior.

- By pushing at that particular time, the fix ordinarily will appear in v13.0
  before appearing in a back branch release.  If problematic contention arises
  quickly in the field, that's a more-comfortable way to discover it.