Re: delta relations in AFTER triggers

On Fri, May 5, 2017 at 12:39 AM, Neha Sharma
<neha(dot)sharma(at)enterprisedb(dot)com> wrote:
> While testing the feature we encountered one more crash,below is the
> scenario to reproduce.
>
> create table t1 ( a int);
> create table t2 ( a int);
> insert into t1 values (11),(12),(13);
>
> create or replace function my_trig() returns trigger
> language plpgsql as $my_trig$
> begin
> insert into t2(select a from new_table);
> RETURN NEW;
> end;
> $my_trig$;
>
> create trigger my_trigger
> after truncate or update on t1
> referencing new table as new_table old table as oldtab
> for each statement
> execute procedure my_trig();
>
> truncate t1;
> server closed the connection unexpectedly
> This probably means the server terminated abnormally
> before or while processing the request.
> The connection to the server was lost. Attempting reset: Failed.

Thanks. Reproduced here. The stack looks like this:

frame #3: 0x0000000103e5e8b0
postgres`ExceptionalCondition(conditionName="!((((((trigdata->tg_event)
& 0x00000003) == 0x00000000) || (((trigdata->tg_event) & 0x00000003)
== 0x00000002) || (((trigdata->tg_event) & 0x00000003) == 0x00000001))
&& (((trigdata->tg_event) & 0x00000018) == 0x00000000) &&
!(trigdata->tg_event & 0x00000020) && !(trigdata->tg_event &
0x00000040)) || (trigdata->tg_oldtable == ((void*)0) &&
trigdata->tg_newtable == ((void*)0)))", errorType="FailedAssertion",
fileName="trigger.c", lineNumber=2045) + 128 at assert.c:54
frame #4: 0x0000000103a6f542
postgres`ExecCallTriggerFunc(trigdata=0x00007fff5c40bad0, tgindx=0,
finfo=0x00007fd8ba0817b8, instr=0x0000000000000000,
per_tuple_context=0x00007fd8b906f928) + 258 at trigger.c:2039
frame #5: 0x0000000103a754ed
postgres`AfterTriggerExecute(event=0x00007fd8ba092460,
rel=0x00000001043fd9c0, trigdesc=0x00007fd8ba068758,
finfo=0x00007fd8ba0817b8, instr=0x0000000000000000,
per_tuple_context=0x00007fd8b906f928,
trig_tuple_slot1=0x0000000000000000,
trig_tuple_slot2=0x0000000000000000) + 1469 at trigger.c:3860
frame #6: 0x0000000103a73080
postgres`afterTriggerInvokeEvents(events=0x00007fd8ba07fb00,
firing_id=1, estate=0x00007fd8ba090440, delete_ok='\x01') + 592 at
trigger.c:4051
frame #7: 0x0000000103a72b7b
postgres`AfterTriggerEndQuery(estate=0x00007fd8ba090440) + 203 at
trigger.c:4227
frame #8: 0x0000000103a498aa
postgres`ExecuteTruncate(stmt=0x00007fd8ba059f40) + 2026 at
tablecmds.c:1485

There's an assertion that it's (one of INSERT, UPDATE, DELETE, an
AFTER trigger, not deferred) *or* there are no transition tables.
Here it's TRUNCATE and there are transition tables, so it fails:

/*
* Protect against code paths that may fail to initialize transition table
* info.
*/
Assert(((TRIGGER_FIRED_BY_INSERT(trigdata->tg_event) ||
TRIGGER_FIRED_BY_UPDATE(trigdata->tg_event) ||
TRIGGER_FIRED_BY_DELETE(trigdata->tg_event)) &&
TRIGGER_FIRED_AFTER(trigdata->tg_event) &&
!(trigdata->tg_event & AFTER_TRIGGER_DEFERRABLE) &&
!(trigdata->tg_event & AFTER_TRIGGER_INITDEFERRED)) ||
(trigdata->tg_oldtable == NULL && trigdata->tg_newtable == NULL));

We can't possibly support transition tables on TRUNCATE (the whole
point of TRUNCATE is not to inspect all the rows so we can't collect
them), and we already reject ROW triggers on TRUNCATE, so we should
reject transition tables on STATEMENT triggers for TRUNCATE at
creation time too. See attached. Thoughts?

> Log file and core dump attached for reference.

Thanks! Just by the way, it'd be better to post just an interesting
stack trace fragment rather than a core file, because core files can't
really be used without the exact executable that you built.

--
Thomas Munro
http://www.enterprisedb.com