MIME-Version: 1.0
References: <7B0EC0EC-5461-41EF-9B31-F9BBE608DEA5@gmail.com>
In-Reply-To: <7B0EC0EC-5461-41EF-9B31-F9BBE608DEA5@gmail.com>
From: Kirill Reshke <reshkekirill@gmail.com>
Date: Fri, 8 May 2026 22:47:39 +0500
Message-ID: 
 <CALdSSPh9TMhgiAai4i7RtAUXRFtD-Lzhxm-YAY7-dzUAu=9yzQ@mail.gmail.com>
Subject: Re: Fix REPACK with WITHOUT OVERLAPS replica identity indexes
To: Chao Li <li.evan.chao@gmail.com>
Cc: PostgreSQL-development <pgsql-hackers@postgresql.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Archived-At: 
 <https://www.postgresql.org/message-id/CALdSSPh9TMhgiAai4i7RtAUXRFtD-Lzhxm-YAY7-dzUAu%3D9yzQ%40mail.gmail.com>
Precedence: bulk

On Fri, 8 May 2026 at 09:22, Chao Li <li.evan.chao@gmail.com> wrote:
>
> Hi,
>
> While testing UPDATE FOR PORTION OF, I started wondering whether REPACK s=
upports temporal tables. In theory, it should, because temporal WITHOUT OVE=
RLAPS indexes can be used as replica identity indexes. So I created a test =
script, repack_temporal.spec, which is included in the attached patch, and =
it failed.
>
> I found that REPACK hard-codes BTEqualStrategyNumber when calling get_opf=
amily_member(). That seems wrong, because build_replindex_scan_key() uses I=
ndexAmTranslateCompareType() to get the equality strategy for COMPARE_EQ.
>
> After fixing the hard-coded BTEqualStrategyNumber, the temporal test pass=
ed. Then I added another test for multirange, repack_temporal_multirange.sp=
ec, which also failed. The reason is that find_target_tuple() uses the iden=
tity index to find the first tuple and returns it directly, but a lossy ind=
ex scan may return false positives and require recheck.
>
> Please see the attached patch for the fix details and test scripts.
>
> Best regards,
> --
> Chao Li (Evan)
> HighGo Software Co., Ltd.
> https://www.highgo.com/
>

your analysis appears correct to me

>  + while (index_getnext_slot(scan, ForwardScanDirection, retrieved))
> + {
> + if (scan->xs_recheck && !identity_key_equal(chgcxt, locator, retrieved)=
)
> + continue;
> +
> + retval =3D true;
> + break;
> + }

Should we add CFI() ?


Also, do we really need isolation tests and inj points here? Doesn't a
simple regression test for REPACK execute the same code?


--=20
Best regards,
Kirill Reshke