Date: Sun, 11 Apr 2021 16:23:00 -0500
From: Justin Pryzby <pryzby@telsasoft.com>
To: Alvaro Herrera <alvherre@alvh.no-ip.org>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Amit Langote <amitlangote09@gmail.com>,
	pgsql-hackers@lists.postgresql.org
Subject: Re: ALTER TABLE .. DETACH PARTITION CONCURRENTLY
Message-ID: <20210411212259.GB20530@telsasoft.com>
References: <3269784.1617215412@sss.pgh.pa.us>
 <20210411212035.GA15048@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210411212035.GA15048@alvherre.pgsql>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20210411212259.GB20530%40telsasoft.com>
Precedence: bulk

On Sun, Apr 11, 2021 at 05:20:35PM -0400, Alvaro Herrera wrote:
> On 2021-Mar-31, Tom Lane wrote:
> 
> > diff -U3 /home/buildfarm/trilobite/buildroot/HEAD/pgsql.build/src/test/isolation/expected/detach-partition-concurrently-4.out /home/buildfarm/trilobite/buildroot/HEAD/pgsql.build/src/test/isolation/output_iso/results/detach-partition-concurrently-4.out
> > --- /home/buildfarm/trilobite/buildroot/HEAD/pgsql.build/src/test/isolation/expected/detach-partition-concurrently-4.out	2021-03-29 20:14:21.258199311 +0200
> > +++ /home/buildfarm/trilobite/buildroot/HEAD/pgsql.build/src/test/isolation/output_iso/results/detach-partition-concurrently-4.out	2021-03-30 18:54:34.272839428 +0200
> > @@ -324,6 +324,7 @@
> >  1              
> >  2              
> >  step s1insert: insert into d4_fk values (1);
> > +ERROR:  insert or update on table "d4_fk" violates foreign key constraint "d4_fk_a_fkey"
> >  step s1c: commit;
> >  
> >  starting permutation: s2snitch s1b s1s s2detach s1cancel s3vacfreeze s1s s1insert s1c
> 
> Hmm, actually, looking at this closely, I think the expected output is
> bogus and trilobite is doing the right thing by throwing this error
> here.  The real question is why isn't this case behaving in that way in
> every *other* animal.

I was looking/thinking at it, and wondered whether it could be a race condition
involving pg_cancel_backend()

-- 
Justin