On Wed, Mar 12, 2025 at 12:54 PM Artur Zakirov <zaartur@xxxxxxxxx> wrote:I can reproduce this with the table `t` on PG 15.10.I didn't mention I'm running 16.6, but I'm pretty sure it is reproducible on other versions too.In your case `base/357283/365810` file is a new index file. For some reason Postgres tries to read the new index. I suppose this is because during reading the table `t` within the function `f_t` it tries to access the new index.Yeah, even if it is not clear to me why it is trying to read the index that is under creation (i.e., not usable yet).According to the documentation, IMMUTABLE functions should not only modify the database, but also return the same results given the same arguments forever, which might not be true when you query a table within such a function. Such a function should be defined as STABLE or VOLATILE.As I stated, this example is controversial, and as the documentation states, the IMMUTABLE set of functions should not perform database lookups, as in my example. However, the error message is quite obscure to me, and reminds me a disk corruption rather a stability/function/lookup problem.
Test on 17.0:
It seems inconsistent to me :
creating a fresh ttt (no PK no constraint) :
amantzio@[local]/test=#
CREATE INDEX IF NOT EXISTS idx_t ON ttt( f_ttt( pk ) );
CREATE INDEX
amantzio@[local]/test=# drop index idx_t ;
DROP INDEX
amantzio@[local]/test=# ALTER TABLE ttt ALTER pk SET NOT NULL ;
ALTER TABLE
amantzio@[local]/test=# CREATE INDEX IF NOT EXISTS idx_t ON ttt(
f_ttt( pk ) );
ERROR: could not read blocks 0..0 in file
"base/17753/596558047": read only 0 of 8192 bytes
CONTEXT: SQL statement "SELECT pk
FROM public.ttt
WHERE pk = i"
PL/pgSQL function f_ttt(integer) line 5 at SQL statement
amantzio@[local]/test=# ALTER TABLE ttt ALTER pk DROP NOT NULL ;
ALTER TABLE
amantzio@[local]/test=# CREATE INDEX IF NOT EXISTS idx_t ON ttt(
f_ttt( pk ) );
ERROR: could not read blocks 0..0 in file
"base/17753/596558048": read only 0 of 8192 bytes
CONTEXT: SQL statement "SELECT pk
FROM public.ttt
WHERE pk = i"
PL/pgSQL function f_ttt(integer) line 5 at SQL statement
amantzio@[local]/test=#
\q
So yep you are definitely right as far as the error message is
concerned plus the inconsistent pattern shown above.
Luca
