On 08/21/2018 01:46 PM, Tom Lane wrote: > Andres Freund <and...@anarazel.de> writes: >> On 2018-08-21 13:29:20 -0400, Tom Lane wrote: >>> We've got a buildfarm handy that could answer the question. >>> Let's just stick a test function in there for a day and see >>> which animals fail. > >> I think we pretty much know the answer already, anything before 2013 >> will fail. > > Do we know that for sure? I thought it was theoretical.
I thought I remembered a message where it had been looked up in docs, but I think the one I was remembering was Peter's "According to my research (completely untested in practice), you need 2010 for mixed code and declarations and 2013 for named initialization of structs." [1] which didn't quite actually say it was documented. -Chap [1] https://www.postgresql.org/message-id/ef986aa7-c7ca-ec34-19d9-fef38716b109%402ndquadrant.com