[sqlsmith] parallel worker errors "subplan ... was not initialized" - Mailing list pgsql-hackers

From Andreas Seltenreich
Subject [sqlsmith] parallel worker errors "subplan ... was not initialized"
Date
Msg-id 87v9doorz7.fsf@aurora.ydns.eu
Whole thread Raw
Responses Re: [sqlsmith] parallel worker errors "subplan ... was not initialized"  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-hackers
Hi,

yesterday's sqlsmith run on master at 3df51ca8b3 also yielded 4855
queries failing with ERRORs like the following:

,----
| ERROR:  subplan "InitPlan 3 (returns $3)" was not initialized
| CONTEXT:  parallel worker
`----

They all run fine with parallel scans off.  Below is a recipe that
reproduces one for me on a fresh regression database.

regards,
Andreas

set min_parallel_table_scan_size = '8kB';
set parallel_setup_cost = 1;
select
  sample_0.a as c0,
  sample_0.a as c1,
  pg_catalog.pg_stat_get_bgwriter_requested_checkpoints() as c2,
  sample_0.b as c3
from
  public.itest6 as sample_0
where EXISTS (
  select
      cast(coalesce(cast(nullif(sample_1.z,
          sample_1.x) as int4),
        sample_1.x) as int4) as c0,
      pg_catalog.macaddr_cmp(
        cast(case when false then cast(null as macaddr) else cast(null as macaddr) end
           as macaddr),
        cast(cast(null as macaddr) as macaddr)) as c1
    from
      public.insert_tbl as sample_1
    where (select timestamptzcol from public.brintest limit 1 offset 29)
         < pg_catalog.date_larger(
        cast((select d from public.nv_child_2011 limit 1 offset 6)
           as date),
        cast((select pg_catalog.min(filler3) from public.mcv_lists)
           as date)));



pgsql-hackers by date:

Previous
From: Alexander Korotkov
Date:
Subject: Re: range_agg
Next
From: Justin Pryzby
Date:
Subject: Re: CREATE INDEX CONCURRENTLY on partitioned index