During recent review of the INCLUDE covering index patch, I pushed to
formalize the slightly delicate assumptions that we make around how
index_truncate_tuple() is called. It's natural to call
index_truncate_tuple() during a page split, when a buffer lock is
held. This is what we actually do in most cases.
It occurred to me that it would be nice to be able to
Assert(!AnyBufferLockHeldByMe()) at a certain point within
index_form_tuple(), to make sure that our assumptions hold. If
index_truncate_tuple() (or any other function) ever called
index_form_tuple(), and ended up actually performing table access with
an exclusive buffer lock held, we'd at least be able to catch the bug
when assertions are enabled. A function that lets code assert that no
buffer locks are held, for the rare cases where external table access
is required seems like good general infrastructure.
Does this seem like a good idea? This could get pretty expensive if it
was overused, even by the standards of what we expect from
assertion-enabled builds, but we could make it optional if the
overhead got out of hand.
--
Peter Geoghegan