On 2016-02-27 01:45:57 +0000, Simon Riggs wrote:
> Surely then the fix is to make heap_inplace_update() assign an xid? That
> way any catalog change will always generate a commit record containing the
> invalidation that goes with the change. No need to fix up the breakage
> later.
Well, we could, but it'd also break things where we rely on
heap_inplace_update not assigning an xid. I'm not seing why that's
better than my proposal of doing this
(http://archives.postgresql.org/message-id/20160227002958.peftvmcx4dxwe244%40alap3.anarazel.de),
by emitting a commit record in RecordTransactionCommmit() if nrels != 0
|| nmsgs != 0 || RelcacheInitFileInval
.
> The other heap_insert|update|delete functions (and similar) all assign xid,
> so it is consistent for us to do that for inplace_update also.
I don't think that follows. Inplace updates an absolute special case,
where are *not allowed* to include an xid in the tuple.
Andres