Re: [COMMITTERS] pgsql: Fix freezing of a dead HOT-updated tuple

On Thu, Sep 28, 2017 at 2:47 PM, Peter Geoghegan <pg(at)bowt(dot)ie> wrote:
> In the end, commit 6bfa88a fixed that old recovery bug by making sure
> the recovery routine heap_xlog_lock() did the right thing. In both
> cases (Feb 2014 and today), the index wasn't really corrupt -- it just
> pointed to the root of a HOT chain when it should point to some child
> tuple (or maybe a successor HOT chain).

Just to be clear, obviously I don't mean that the index should have
been magically updated to compensate for that 2014 heap_lock_tuple()
recovery bug (say, via an in-place IndexTuple update during recovery).
Certainly, the index AM was entitled to assume that the heap TID it
pointed to would continue to be the root of the same HOT chain, at
least until the next VACUUM. That was what I meant by "the index
wasn't actually corrupt".

All I meant here is that the index scan and sequential scan would have
been in agreement if something like that *had* happened artificially
(say, when the index tuple was edited with a hex editor). And, that
the actual high level problem was that we failed to treat a
HOT-updated tuple as a HOT-updated tuple, leading to consequences that
were mostly noticeable during index scans. Probably on both occasions
(back in 2014, and now).

--
Peter Geoghegan