Re: [BUG] Fix DETACH with FK pointing to a partitioned table fails

On 2024-Oct-25, Alexander Lakhin wrote:

> I've also discovered another anomaly with a similar setup, but it's not
> related to 53af9491a.

Hmm, it may well be a preexisting problem, but I do think it involves
the same code. As far as I can tell, the value "2" here

> This script ends up with:
> ERROR:  invalid attribute number 2
> ERROR:  cache lookup failed for attribute 2 of relation 16398

is coming from riinfo->confdelsetcols which was set up by
DetachPartitionFinalize during the last DETACH operation.

> (Perhaps it deserves a separate discussion.)

No need for that.

A backtrace from the errfinish gives me this (apologies for the
wide terminal):

#0 errfinish (filename=0x5610d93e4510 "../../../../../pgsql/source/master/src/backend/parser/parse_relation.c", lineno=3618,
funcname=0x5610d93e5518 <__func__.5> "attnumAttName") at ../../../../../../pgsql/source/master/src/backend/utils/error/elog.c:475
#1 0x00005610d8d68bcc in attnumAttName (rd=rd(at)entry=0x7f69dd0c1a90, attid=2) at ../../../../../pgsql/source/master/src/backend/parser/parse_relation.c:3618
#2 0x00005610d924a8c5 in ri_set (trigdata=0x7ffe38d924c0, is_set_null=<optimized out>, tgkind=<optimized out>)
at ../../../../../../pgsql/source/master/src/backend/utils/adt/ri_triggers.c:1219
#3 0x00005610d8f897ea in ExecCallTriggerFunc (trigdata=trigdata(at)entry=0x7ffe38d924c0, tgindx=tgindx(at)entry=2, finfo=0x5611153f5768, finfo(at)entry=0x5611153f5708,
instr=instr(at)entry=0x0, per_tuple_context=per_tuple_context(at)entry=0x561115471c30) at ../../../../../pgsql/source/master/src/backend/commands/trigger.c:2362
#4 0x00005610d8f8b646 in AfterTriggerExecute (trig_tuple_slot2=0x0, trig_tuple_slot1=0x0, per_tuple_context=0x561115471c30, instr=0x0, finfo=0x5611153f5708,
trigdesc=0x5611153f53e8, dst_relInfo=<optimized out>, src_relInfo=<optimized out>, relInfo=0x5611153f51d8, event=0x56111546fd4c, estate=0x5611153f4d50)
at ../../../../../pgsql/source/master/src/backend/commands/trigger.c:4498
#5 afterTriggerInvokeEvents (events=events(at)entry=0x5611154292a0, firing_id=1, estate=estate(at)entry=0x5611153f4d50, delete_ok=delete_ok(at)entry=false)
at ../../../../../pgsql/source/master/src/backend/commands/trigger.c:4735
#6 0x00005610d8f8ff10 in AfterTriggerEndQuery (estate=estate(at)entry=0x5611153f4d50) at ../../../../../pgsql/source/master/src/backend/commands/trigger.c:5090
#7 0x00005610d8fb2457 in standard_ExecutorFinish (queryDesc=0x561115460810) at ../../../../../pgsql/source/master/src/backend/executor/execMain.c:442
#8 0x00005610d917df48 in ProcessQuery (plan=0x56111545ef50, sourceText=0x56111539cfb0 "DELETE FROM t;", params=0x0, queryEnv=0x0, dest=0x56111545f0b0,
qc=0x7ffe38d92830) at ../../../../../pgsql/source/master/src/backend/tcop/pquery.c:193

and then

(gdb) frame 2
#2 0x00005610d924a8c5 in ri_set (trigdata=0x7ffe38d924c0, is_set_null=<optimized out>, tgkind=<optimized out>)
at ../../../../../../pgsql/source/master/src/backend/utils/adt/ri_triggers.c:1219
1219 quoteOneName(attname, RIAttName(fk_rel, set_cols[i]));
(gdb) print *riinfo
$5 = {constraint_id = 16400, valid = true, constraint_root_id = 16400, oidHashValue = 3001019032, rootHashValue = 3001019032, conname = {
data = "pt_a_fkey", '\000' <repeats 54 times>}, pk_relid = 16384, fk_relid = 16397, confupdtype = 97 'a', confdeltype = 100 'd', ndelsetcols = 1,
confdelsetcols = {2, 0 <repeats 31 times>}, confmatchtype = 115 's', hasperiod = false, nkeys = 1, pk_attnums = {1, 0 <repeats 31 times>}, fk_attnums = {1,
0 <repeats 31 times>}, pf_eq_oprs = {96, 0 <repeats 31 times>}, pp_eq_oprs = {96, 0 <repeats 31 times>}, ff_eq_oprs = {96, 0 <repeats 31 times>},
period_contained_by_oper = 0, agged_period_contained_by_oper = 0, valid_link = {prev = 0x5611153f4c30, next = 0x5610d96b51b0 <ri_constraint_cache_valid_list>}}

--
Álvaro Herrera 48°01'N 7°57'E — https://www.EnterpriseDB.com/
"Digital and video cameras have this adjustment and film cameras don't for the
same reason dogs and cats lick themselves: because they can." (Ken Rockwell)