Message-ID: <bca3e7e6-ac1f-476d-9725-d6aec9100ed0@illuminatedcomputing.com>
Date: Thu, 23 Jan 2025 15:20:10 -0800
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: SQL:2011 application time
To: Peter Eisentraut <peter@eisentraut.org>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, jian he <jian.universality@gmail.com>,
 PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>,
 vignesh C <vignesh21@gmail.com>
References: <CA+renyUApHgSZF9-nd-a0+OPGharLQLO=mDHcY4_qQ0+noCUVg@mail.gmail.com>
 <57ea0668-5205-426e-b934-efc89f2186c2@illuminatedcomputing.com>
 <6dc196d4-8853-49d7-8dd3-e391058451c1@eisentraut.org>
 <bf6b3fc5-b1c2-4fdf-ad27-5de45770c58c@illuminatedcomputing.com>
 <66e47fc0-77b7-4763-9b70-5c0b7ff15f21@eisentraut.org>
 <a2ee0c1a-5ef0-43ab-897c-b56adb25b34f@illuminatedcomputing.com>
 <333d3886-b737-45c3-93f4-594c96bb405d@eisentraut.org>
 <e8e51e74-10d7-410d-a1d1-8ec681d8436a@eisentraut.org>
 <f4056005-0fb7-417a-bcb3-e704f9aeadc3@illuminatedcomputing.com>
 <6a2247d9-461d-4b0a-b692-f0f3c0fab27f@eisentraut.org>
 <8f995a4c-64e3-40cd-bfed-06b54a280d36@illuminatedcomputing.com>
 <0832f18d-523f-4c15-acb0-338a785150c5@illuminatedcomputing.com>
 <5f0c25de-2c13-4fb5-a4df-a1e03c43cd14@illuminatedcomputin! g.com>
 <1b45f81d-4e99-4813-a2c6-72bf5fd80980@eisentraut.org>
 <3611540.1737518430@sss.pgh.pa.us>
 <9bb3eea7-820f-4932-92cb-cb564fc44caf@eisentraut.org>
Content-Language: en-US
From: Paul Jungwirth <pj@illuminatedcomputing.com>
In-Reply-To: <9bb3eea7-820f-4932-92cb-cb564fc44caf@eisentraut.org>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Archived-At: <https://www.postgresql.org/message-id/bca3e7e6-ac1f-476d-9725-d6aec9100ed0%40illuminatedcomputing.com>
Precedence: bulk

On 1/23/25 07:31, Peter Eisentraut wrote:
> On 22.01.25 05:00, Tom Lane wrote:
>> Peter Eisentraut <peter@eisentraut.org> writes:
>>> I have committed the fix for foreign key NO ACTION (patch 0002, this did
>>> not require patch 0001).
>>
>> That commit seems to be causing occasional buildfarm failures:
>>
>> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=indri&dt=2025-01-22%2001%3A29%3A35
>> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=mylodon&dt=2025-01-22%2001%3A17%3A14
>>
>> Both of these look like
>>
>> --- /Users/buildfarm/bf-data/HEAD/pgsql.build/src/test/regress/expected/without_overlaps.out    
>> 2025-01-21 20:29:36
>> +++ /Users/buildfarm/bf-data/HEAD/pgsql.build/src/test/regress/results/without_overlaps.out    
>> 2025-01-21 20:43:08
>> @@ -1792,8 +1792,6 @@
>>     SET valid_at = CASE WHEN lower(valid_at) = '2018-01-01' THEN daterange('2018-01-01', 
>> '2018-01-05')
>>                         WHEN lower(valid_at) = '2018-02-01' THEN daterange('2018-01-05', 
>> '2018-03-01') END
>>     WHERE id = '[6,7)';
>> -ERROR:  update or delete on table "temporal_rng" violates RESTRICT setting of foreign key 
>> constraint "temporal_fk_rng2rng_fk" on table "temporal_fk_rng2rng"
>> -DETAIL:  Key (id, valid_at)=([6,7), [2018-01-01,2018-02-01)) is referenced from table 
>> "temporal_fk_rng2rng".
>>   -- a PK update that fails because both are referenced (even before commit):
>>   BEGIN;
>>     ALTER TABLE temporal_fk_rng2rng
>>
>> ie, an expected error did not get thrown.
> 
> I suspect the nested locking clauses in the new SQL query in the patch. I don't see anything else in 
> the patch that would possibly create this kind unstable behavior.

I can't find a regression.diffs in the second link. Is there one? I can't tell if it's the same 
failure as in the first link as not.

I ran installcheck-parallel on my own machine continuously over night and haven't been able to 
reproduce this yet. How many cases have appeared on the build farm? More than these two? And just to 
confirm: they are only since committing 1772d554b0?

The strange thing is that the omitted error message is for a RESTRICT foreign key, and 1772d554b0 
only changes behavior when is_no_action. That makes me think the bug is with the original temporal 
FK commit. But that has been running on the build farm for a long time, so probably not.

Likewise, I don't see how it can be the nested locking, when that SQL isn't used for RESTRICT 
constraints.

The infrequent failure made me suspect a memory error. It's hard to come up with explanations.

What about caching the FK's query plan? Could the RESTRICT test ever reuse the constraint oid from 
the NO ACTION tests just above it? I'm pretty sure that's not supposed to happen, but if it used a 
plan generated from the NO ACTION SQL, it would exhibit the behavior we're seeing. It also makes 
sense why it only appeared after 1772d554b0.

I'll dig into that hypothesis and let you know if I figure anything out.

Yours,

-- 
Paul              ~{:-)
pj@illuminatedcomputing.com