"Merlin Moncure" <mmoncure@gmail.com> writes:
> imo, the most likely scenario would be a begin/exception/end block in
> pg/sql. i would venture to guess that very little true savepointing
> happens in practice. maybe add a little note of caution pg/sql error
> handling documentation?
I mentioned exception blocks as a risk factor, but I think the right
place to document it is under FOR UPDATE/SHARE:
http://developer.postgresql.org/cvsweb.cgi/pgsql/doc/src/sgml/ref/select.sgml.diff?r1=1.93;r2=1.94
regards, tom lane