Deadlock with concurrent SELECT / DROP TABLE

[bgranvea]

I get a random deadlock when performing concurrently a SELECT and a DROP TABLE on the same view. I've tested with Clickhouse 18.14.19 and 19.1.6.

To reproduce the problem, just launch these 2 scripts in different shells:

for i in {1..1000}; do echo $i; echo "drop table if exists test_view" | clickhouse-client; echo "create view test_view as select count(*),database,table from system.columns group by database,table" | clickhouse-client; done

for i in {1..1000}; do echo $i; echo "select * from test_view order by table" | clickhouse-client >/dev/null; done

I can see in system.processes my 2 queries which are still running:

select elapsed,query from system.processes limit 2 format Vertical

Row 1:
──────
elapsed: 502.919283347
query:   select * from test_view order by table

Row 2:
──────
elapsed: 502.918433233
query:   drop table if exists test_view

[alexey-milovidov]

There is fundamental issue with current implementation of table locks.
Nevertheless, we are going to fix it.