Andrey Borodin <x4mmm@yandex-team.ru> writes:
> There is bunch of code in current GiST implementation checking for GistTupleIsInvalid(). PostgreSQL since 9.1 do not
createinvalid tuples. Should we support this tuples forever?
The question is not whether we still create such tuples. The reason
the code is still there is that an index that's been pg_upgraded from
before 9.1 might still contain such tuples. We can't drop the support
unless either we provide logic to clean up invalid entries, or we're
willing to force users to REINDEX old GiST indexes to get rid of them
that way. The latter seems like a pretty high price just to get rid of
some crufty old code.
regards, tom lane