[YSQL][SQLsmith] TRAP: FailedAssertion("!(!((allPgXact[proc->pgprocno].xid) != ((TransactionId) 0)))", File: "../../../../../../../src/postgres/src/backend/storage/ipc/procarray.c", Line: 440)

[def-]

Jira Link: DB-1168

Description

SQLsmith issue, sometimes ROLLBACK fails sporadically, can't reproduce. Postgres tserver log:

2022-01-26 18:44:03.159 CET [56123] ERROR:  operator does not exist: polygon = polygon at character 2013
2022-01-26 18:44:03.159 CET [56123] HINT:  No operator matches the given name and argument types. You might need to add explicit type casts.
2022-01-26 18:44:03.159 CET [56123] STATEMENT:  select  
	  subq_0.c0 as c0
	from 
	  (select  
	        ref_0.gid as c0, 
	        ref_3.pid as c1, 
	        ref_3.datid as c2, 
	        ref_2.typowner as c3
	      from 
	        pg_catalog.pg_prepared_xacts as ref_0
	            right join information_schema.usage_privileges as ref_1
	            on (false)
	          inner join pg_catalog.pg_type as ref_2
	            left join pg_catalog.pg_stat_progress_vacuum as ref_3
	              left join pg_catalog.pg_attribute as ref_4
	              on (ref_3.relid = ref_4.attrelid )
	            on (ref_2.typarray = ref_4.attrelid )
	          on (((true) 
	                or (cast(null as timetz) <= cast(null as timetz))) 
	              and (cast(null as circle) @> cast(null as circle)))
	      where pg_catalog.circle(
	          cast(case when (false) 
	              or (true) then cast(null as point) else cast(null as point) end
	             as point),
	          cast(cast(null as float8) as float8)) ~ case when cast(null as lseg) <@ (select b from public.mvtest_boxes limit 1 offset 2)
	               then pg_catalog.circle(
	            cast(cast(null as point) as point),
	            cast((select pg_catalog.stddev_samp(total_time) from pg_catalog.pg_stat_statements)
	               as float8)) else pg_catalog.circle(
	            cast(cast(null as point) as point),
	            cast((select pg_catalog.stddev_samp(total_time) from pg_catalog.pg_stat_statements)
	               as float8)) end
	          
	      limit 61) as subq_0
	where ((subq_0.c3 is NULL) 
	    or ((case when true then cast(null as record) else cast(null as record) end
	           < cast(null as record)) 
	      or (case when (false) 
	            and (subq_0.c3 is not NULL) then cast(null as tsvector) else cast(null as tsvector) end
	           <> pg_catalog.jsonb_to_tsvector(
	          cast(cast(null as jsonb) as jsonb),
	          cast((select pg_catalog.jsonb_agg(implementation_info_id) from information_schema.sql_implementation_info)
	             as jsonb))))) 
	  and ((cast(nullif(cast(nullif(cast(null as polygon),
	          pg_catalog.polygon(
	            cast(subq_0.c1 as int4),
	            cast(cast(null as circle) as circle))) as polygon),
	        pg_catalog.poly_in(
	          cast(cast(coalesce(cast(null as cstring),
	            cast(null as cstring)) as cstring) as cstring))) as polygon) ~ pg_catalog.point(
	        cast((select b from public.mvtest_boxmv limit 1 offset 6)
	           as box))) 
	    or (case when pg_catalog.circle_in(
	            cast(cast(null as cstring) as cstring)) >> cast(null as circle) then cast(nullif(case when cast(null as macaddr8) > cast(null as macaddr8) then cast(null as date) else cast(null as date) end
	            ,
	          pg_catalog.to_date(
	            cast((select pg_catalog.max(md5) from public.mvtest_foo_data)
	               as text),
	            cast((select pg_catalog.max(locktype) from pg_catalog.pg_locks)
	               as text))) as date) else cast(nullif(case when cast(null as macaddr8) > cast(null as macaddr8) then cast(null as date) else cast(null as date) end
	            ,
	          pg_catalog.to_date(
	            cast((select pg_catalog.max(md5) from public.mvtest_foo_data)
	               as text),
	            cast((select pg_catalog.max(locktype) from pg_catalog.pg_locks)
	               as text))) as date) end
	         = case when ((cast(null as circle) < cast(null as circle)) 
	            and (cast(null as uuid) = cast(null as uuid))) 
	          and (((((subq_0.c3 is NULL) 
	                  and (subq_0.c1 is NULL)) 
	                and (cast(null as timestamptz) < cast(null as "timestamp"))) 
	              or ((subq_0.c2 is NULL) 
	                or (false))) 
	            and (cast(null as tsquery) <> cast(null as tsquery))) then cast(null as date) else cast(null as date) end
	        ))
	limit 83
2022-01-26 18:44:03.173 CET [56125] WARNING:  there is no transaction in progress
2022-01-26 18:44:03.181 CET [56124] WARNING:  there is no transaction in progress
2022-01-26 18:44:03.195 CET [56122] WARNING:  there is no transaction in progress
TRAP: FailedAssertion("!(!((allPgXact[proc->pgprocno].xid) != ((TransactionId) 0)))", File: "../../../../../../../src/postgres/src/backend/storage/ipc/procarray.c", Line: 440)
2022-01-26 18:44:03.196 CET [51547] LOG:  server process (PID 56122) was terminated by signal 6: Abort trap
2022-01-26 18:44:03.196 CET [51547] DETAIL:  Failed process was running: ROLLBACK;
2022-01-26 18:44:03.196 CET [51547] LOG:  terminating any other active server processes
2022-01-26 18:44:03.196 CET [56125] WARNING:  terminating connection because of crash of another server process
2022-01-26 18:44:03.196 CET [56125] DETAIL:  The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.
2022-01-26 18:44:03.196 CET [56125] HINT:  In a moment you should be able to reconnect to the database and repeat your command.
2022-01-26 18:44:03.196 CET [56123] WARNING:  terminating connection because of crash of another server process
2022-01-26 18:44:03.196 CET [56123] DETAIL:  The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.
2022-01-26 18:44:03.196 CET [56123] HINT:  In a moment you should be able to reconnect to the database and repeat your command.
2022-01-26 18:44:03.196 CET [56124] WARNING:  terminating connection because of crash of another server process
2022-01-26 18:44:03.196 CET [56124] DETAIL:  The postmaster has commanded this server process to roll back the current transaction and exit, because another server process exited abnormally and possibly corrupted shared memory.
2022-01-26 18:44:03.196 CET [56124] HINT:  In a moment you should be able to reconnect to the database and repeat your command.
I0126 18:44:03.196766 1830531072 poller.cc:66] Poll stopped: Service unavailable (yb/rpc/scheduler.cc:80): Scheduler is shutting down (system error 58)
W0126 18:44:03.196738 1832824832 outbound_call.cc:131] Failed to schedule invoking callback on response for request yb.tserver.TabletServerService.Read to 127.0.0.1: Aborted (yb/rpc/thread_pool.cc:260): Service is shutting down
I0126 18:44:03.196907 1831104512 poller.cc:66] Poll stopped: Service unavailable (yb/rpc/scheduler.cc:80): Scheduler is shutting down (system error 58)
W0126 18:44:03.197257 1832824832 rpc.cc:179] Aborted (yb/rpc/rpc.cc:178): Failed to schedule: 0x0000000109db1d08 -> Read(tablet: 00000000000000000000000000000000, num_ops: 1, num_attempts: 2, txn: 00000000-0000-0000-0000-000000000000, subtxn: [none])
I0126 18:44:03.197327 1831104512 poller.cc:66] Poll stopped: Service unavailable (yb/rpc/scheduler.cc:80): Scheduler is shutting down (system error 58)
2022-01-26 18:44:03.202 CET [51547] LOG:  all server processes terminated; reinitializing

[def-]

Was trying to find materialized view bugs, but the same error also happens without matviews.

[tedyu]

ERROR:  relation "public.mvtest_boxes" does not exist
LINE 25: ... ~ case when cast(null as lseg) <@ (select b from public.mvt...
                                                              ^

What's the DDL for mvtest_boxes ?

[def-]

DDLs are:

CREATE TABLE mvtest_foo_data AS SELECT i, md5(random()::text)
  FROM generate_series(1, 10) i;
CREATE TABLE mvtest_boxes (id serial primary key, b box);
INSERT INTO mvtest_boxes (b) VALUES
  ('(32,32),(31,31)'),
  ('(2.0000004,2.0000004),(1,1)'),
  ('(1.9999996,1.9999996),(1,1)');
CREATE MATERIALIZED VIEW mvtest_boxmv AS SELECT * FROM mvtest_boxes;

But can't reproduce here.

[def-]

Can sometimes reproduce with:

select  
  subq_0.c4 as c0, 
  ref_1.datallowconn as c1, 
  74 as c2, 
  subq_1.c0 as c3
from 
  (select  
          ref_0.indexrelname as c0, 
          ref_0.relid as c1, 
          ref_0.indexrelname as c2, 
          37 as c3, 
          ref_0.relid as c4
        from 
          pg_catalog.pg_statio_sys_indexes as ref_0
        where true
        limit 93) as subq_0
    inner join pg_catalog.pg_database as ref_1
    on (subq_0.c3 = ref_1.encoding ),
  lateral (select  
        subq_0.c2 as c0
      from 
        information_schema.role_column_grants as ref_2
          right join pg_catalog.pg_file_settings as ref_3
          on (ref_1.datistemplate is NULL)
      where cast(null as tinterval) #<> cast(null as reltime)) as subq_1
where ((select sent_lsn from pg_catalog.pg_stat_replication limit 1 offset 3)
       < pg_catalog.pg_current_wal_flush_lsn()) 
  and ((select type from pg_catalog.pg_hba_file_rules limit 1 offset 1)
       >= cast(nullif(cast(null as text),
      cast(null as text)) as text))
limit 28

But mostly #11233 happens, which leads to me to believe these two bugs are related.