Re: SIREAD lock versus ACCESS EXCLUSIVE lock

On 1 May 2011 I wrote:

> Consider this a WIP patch

Just so people know where this stands...

By 8 May 2011 I had the attached.  I didn't post it because I was
not confident I had placed the calls to the SSI functions for DROP
TABLE and TRUNCATE TABLE correctly.  Then came PGCon and also the
row versus tuple discussion -- the patch for which had conflicts
with this one.

I will be reviewing this on the weekend and making any final
adjustments, but the patch is very likely to look just like this
with the possible exception of placement of the DROP TABLE and
TRUNCATE TABLE calls.  If anyone wants to start reviewing this now,
it would leave more time for me to make changes if needed.

Also, if anyone has comments or hints about the placement of those
calls, I'd be happy to receive them.

This patch compiles with `make world`, passes `make check-world`,
and passes both `make installcheck-world` and `make -C
src/test/isolation installcheck` against a database with
default_transaction_isolation = 'serializable'.  It also passed a
few ad hoc tests of the implemented functionality.

-Kevin