[Alexander Lakhin <exclusion@gmail.com>]

Hi Bertrand,

25.04.2024 10:20, Bertrand Drouvot wrote:
> postgres=# CREATE FUNCTION f() RETURNS int LANGUAGE SQL RETURN f() + 1;
> ERROR:  cache lookup failed for function 16400
>
> This stuff does appear before we get a chance to call the new depLockAndCheckObject()
> function.
>
> I think this is what Tom was referring to in [1]:
>
> "
> So the only real fix for this would be to make every object lookup in the entire
> system do the sort of dance that's done in RangeVarGetRelidExtended.
> "
>
> The fact that those kind of errors appear also somehow ensure that no orphaned
> dependencies can be created.

I agree; the only thing that I'd change here, is the error code.

But I've discovered yet another possibility to get a broken dependency.
Please try this script:
res=0
numclients=20
for ((i=1;i<=100;i++)); do
for ((c=1;c<=numclients;c++)); do
   echo "
CREATE SCHEMA s_$c;
CREATE CONVERSION myconv_$c FOR 'LATIN1' TO 'UTF8' FROM iso8859_1_to_utf8;
ALTER CONVERSION myconv_$c SET SCHEMA s_$c;
   " | psql >psql1-$c.log 2>&1 &
   echo "DROP SCHEMA s_$c RESTRICT;" | psql >psql2-$c.log 2>&1 &
done
wait
pg_dump -f db.dump || { echo "on iteration $i"; res=1; break; }
for ((c=1;c<=numclients;c++)); do
   echo "DROP SCHEMA s_$c CASCADE;" | psql >psql3-$c.log 2>&1
done
done
psql -c "SELECT * FROM pg_conversion WHERE connamespace NOT IN (SELECT oid FROM pg_namespace);"

It fails for me (with the v4 patch applied) as follows:
pg_dump: error: schema with OID 16392 does not exist
on iteration 1
   oid  | conname  | connamespace | conowner | conforencoding | contoencoding |      conproc      | condefault
-------+----------+--------------+----------+----------------+---------------+-------------------+------------
  16396 | myconv_6 |        16392 |       10 |              8 |             6 | iso8859_1_to_utf8 | f

Best regards,
Alexander