Re: On-disk compatibility for nbtree-unique-key enhancement

On 21/09/2018 01:18, Peter Geoghegan wrote:
> * This means that there is a compatibility issue for anyone that is
> already right on the threshold -- we *really* don't want to see a
> REINDEX fail, but that seems like a possibility that we need to talk
> about now.

When would the REINDEX need to happen?  Will the code still be able to
read and write v3 btrees?  Could there perhaps be an amcheck or
pageinspect feature that tells you ahead of time if there are too large
items in an old index?

-- 
Peter Eisentraut              http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services