On Fri, Feb 3, 2017 at 7:41 AM, Ashutosh Sharma <ashu.coek88@gmail.com> wrote:
> I think UInt32GetDatum(metad->hashm_procid) looks fine, the reason
> being 'hashm_procid' is defined as 32-bit unsigned integer but then we
> may have to define procid as int8 in SQL function.
No, you're wrong. The GetDatum you choose macro has to match the SQL
type, not the type of the variable that you're passing to it. For
example, if you've got an "int" in the code and the SQL type is
"int8", you have to use Int64GetDatum, not Int32GetDatum. Otherwise,
stuff breaks, because on some systems 64-bit integers are passed by
reference, not by value, so the representation that Int32GetDatum
produces isn't valid when interpreted by DatumGetInt64 later on. The
latter is expecting a pointer, but the former didn't produce one.
> Note: I am extremely sorry for wrongly choosing some of the SQL types
> in the patch for pageinspect. I think there were few platform specific
> things that too should have been addressed by me. Moreover, I feel
> being the owner of this project I should have participated in this
> discussion a bit earlier but as I was not subscribed to
> pgsql-committers list I could not be on time.
It might be a good idea to subscribe to pgsql-committers; that way you
can follow what's getting committed, whether it is your patch or
otherwise. But we also should perhaps have migrated this discussion
to pgsql-hackers. Adjusting recipient list.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company