On Tue, Jan 25, 2005 at 12:32:57PM -0500, Tom Lane wrote:
> So the right fix might involve putting the portal into PORTAL_FAILED
> state rather than just zapping it completely.
Strangely, the code comes up simpler after the fix. Patch attached.
Regression test pass. Additionaly I tried both cases mentioned in this
thread; maybe it's worthy to add tests for them too.
alvherre=# begin;
BEGIN
alvherre=# savepoint x;
SAVEPOINT
alvherre=# create table abc (a int);
CREATE TABLE
alvherre=# insert into abc values (5);
INSERT 33616 1
alvherre=# declare foo cursor for select * from abc;
DECLARE CURSOR
alvherre=# rollback to x;
ROLLBACK
alvherre=# fetch from foo; -- hits an Assert()
ERROR: no existe el cursor «foo»
alvherre=# commit;
ROLLBACK
alvherre=# begin;
BEGIN
alvherre=# declare c cursor for select 1 union all select 2;
DECLARE CURSOR
alvherre=# savepoint x;
SAVEPOINT
alvherre=# fetch from c;
?column?
----------
1
(1 fila)
alvherre=# rollback to x;
ROLLBACK
alvherre=# fetch from c;
?column?
----------
2
(1 fila)
alvherre=# commit;
COMMIT
--
Alvaro Herrera (<alvherre[@]dcc.uchile.cl>)
"The ability to monopolize a planet is insignificant
next to the power of the source"