This is an automated comment for commit e9ab8e8 with description of existing statuses. It's updated for the latest CI running

❌ Click here to open a full report in a separate page

AST fuzzer (debug) — Assertion `left_columns_count == result_sample_block.columns() - required_right_keys.columns() - sample_block_with_columns_to_add.columns... Details

#58652

Stateless tests (release, analyzer) — Some queries hung, fail: 0, passed: 5988, skipped: 6 Details

Looks like the same as should be fixed by #58213 , but for some reason is reproduced again

CREATE table ttta (x Int32, t Tuple(t Tuple(t UInt32, s String), s String)) ENGINE = MergeTree ORDER BY x;
INSERT INTO ttta VALUES (1, ((1, 's'), 's'));

CREATE table tttb (x Int32, t Tuple(t Tuple(t Int32, s String), s String)) ENGINE = MergeTree ORDER BY x;
INSERT INTO tttb VALUES (1, ((1, 's'), 's'));

t.t has super-type, while type of t remains the same:

-- ((1,'s'),'s')   (1,'s') Tuple(\n    t Tuple(\n        t UInt32,\n        s String),\n    s String)      Tuple(Int64, String)
SELECT t, t.t, toTypeName(t), toTypeName(t.t) FROM ttta AS a FULL OUTER JOIN tttb AS b USING t.t;

It's happening because we select t and t.t as just different columns from table. Perhaps we can check such case somehow and forbid it too?

While with subquery we resolve t.t.t into function getSubcolumn(t, 't.t') and do not allow it:

SELECT t.t.t FROM (
    SELECT ((1, 's'), 's') :: Tuple(t Tuple(t UInt32, s String), s String) as t
) t1 FULL JOIN (
    SELECT ((1, 's'), 's') :: Tuple(t Tuple(t Int32, s String), s String) as t
) t2 USING (t);

Added comment to code, probably we can go with it, providing best-effort result while selecting sub/super-columns and having them in USING list. When an identifier is resolved into the function nested or getSubcolumn, and some column in its argument is in the USING list and its type has to be updated, we throw an error.

With current implementation, at least we do not have LOGICAL_ERRORS or type mismatches. (I added test with different subcolumn levels in select and using list, maybe it can be even more complex, e.g Tuples in Nested or several different columns in using list ? 😵‍💫)

@novikd you may check the PR

Stateful tests (debug) — fail: 1, passed: 128, skipped: 2 Details

timeout 00165_jit_aggregate_functions ? It was recently updated with SET max_bytes_before_external_group_by='200M';, maybe it's still not enough ?

Stress test (tsan) — Sanitizer assert (in stderr.log) Details

#58437