Thread: BUG: PLPGSQL function causes PgSQL process to die when inserting into a view

Hi all;

I have two tables which have different security considerations wrapped into
a view.  Inserting into the view is done using a simple rule which inserts
into both tables (tables have a 1:1 relationship and reflect the
user-defined and admin-defined portions of a user profile).  From psql,
inserts into the table work as expected.  However from a PLPGSQL function,
any attempt to insert into the table causes the following error:

FATAL:  SPI: improper call to spi_dest_setup

The query is a simple insert and the rule merely splits it into two inserts.

The workaround is simply to insert into both tables in the view
independently within the plpgsql function.  But this seems like odd behavior
that should not be occurring.

Best Wishes,
Chris Travers

_________________________________________________________________
Never get a busy signal because you are always connected  with high-speed
Internet access. Click here to comparison-shop providers.
https://broadband.msn.com

On Thu, 23 Oct 2003, Christopher Travers wrote:

> Hi all;
>
> I have two tables which have different security considerations wrapped into
> a view.  Inserting into the view is done using a simple rule which inserts
> into both tables (tables have a 1:1 relationship and reflect the
> user-defined and admin-defined portions of a user profile).  From psql,
> inserts into the table work as expected.  However from a PLPGSQL function,
> any attempt to insert into the table causes the following error:
>
> FATAL:  SPI: improper call to spi_dest_setup
>
> The query is a simple insert and the rule merely splits it into two inserts.
>
> The workaround is simply to insert into both tables in the view
> independently within the plpgsql function.  But this seems like odd behavior
> that should not be occurring.

Can you give version information and a standalone example?

"Christopher Travers" <einhverfr@hotmail.com> writes:
> FATAL:  SPI: improper call to spi_dest_setup

I think this is a known and already-fixed bug.  What version are you
running?

            regards, tom lane