Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Any change here is *not* a bug fix, it is a change of clearly
> documented and not-obviously-unreasonable behavior. We have to take
> seriously the likelihood that it will break existing code.
Perhaps plpgsql could support tests of SQLSTATE, and recognize '02000'
(the standard value for "zero rows affected") to support the desired
new semantics?
-Kevin