Re: _bt_split(), and the risk of OOM before its critical section

On Wed, May 8, 2019 at 3:37 PM Peter Geoghegan <pg(at)bowt(dot)ie> wrote:
> It makes perfect sense for _bt_split() to call _bt_findsplitloc()
> directly, since _bt_findsplitloc() is already aware of almost every
> _bt_split() implementation detail, whereas those same details are not
> of interest anywhere else.

I discovered that it even used to work like that until 1997, when
commit 71b3e93c505 added handling of duplicate index tuples. Tom
ripped the duplicate handling stuff out a couple of years later, for
what seemed to me to be very good reasons, but _bt_findsplitloc()
remained outside of _bt_split() until now.

I intend to push ahead with the fix for both v11 and HEAD on Monday.
--
Peter Geoghegan