Feedback-ID: ia2694551:Fastmail
Date: Sat, 9 Nov 2024 13:29:12 +0100
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: jian he <jian.universality@gmail.com>,
	Pg Hackers <pgsql-hackers@lists.postgresql.org>,
 Tender Wang <tndrwang@gmail.com>
Subject: Re: not null constraints, again
Message-ID: <202411091229.k6wy32lley63@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <3649828.1731083171@sss.pgh.pa.us>
Archived-At: 
 <https://www.postgresql.org/message-id/202411091229.k6wy32lley63%40alvherre.pgsql>
Precedence: bulk

On 2024-Nov-08, Tom Lane wrote:

> Alvaro Herrera <alvherre@alvh.no-ip.org> writes:
> > But we'll see what else the buildfarm has to say now that I pushed it ...
> 
> A lot of the buildfarm is saying
> 
>  adder         | 2024-11-08 13:04:39 | ../pgsql/src/backend/catalog/pg_constraint.c:708:37: warning: comparison is always true due to limited range of data type [-Wtype-limits]
> 
> which evidently is about this:
> 
> 	Assert(colnum > 0 && colnum <= MaxAttrNumber);

Hah.

> The memcpy right before that doesn't seem like project style either.
> Most other places that are doing similar things just cast the
> ARR_DATA_PTR to the right pointer type and dereference it.

Hmm, yeah, that's easily removed,


diff --git a/src/backend/catalog/pg_constraint.c b/src/backend/catalog/pg_constraint.c
index e953000c01d..043bf7c24dd 100644
--- a/src/backend/catalog/pg_constraint.c
+++ b/src/backend/catalog/pg_constraint.c
@@ -704,11 +704,7 @@ extractNotNullColumn(HeapTuple constrTup)
 		ARR_DIMS(arr)[0] != 1)
 		elog(ERROR, "conkey is not a 1-D smallint array");
 
-	memcpy(&colnum, ARR_DATA_PTR(arr), sizeof(AttrNumber));
-	Assert(colnum > 0 && colnum <= MaxAttrNumber);
-
-	if ((Pointer) arr != DatumGetPointer(adatum))
-		pfree(arr);				/* free de-toasted copy, if any */
+	colnum = ((AttrNumber *) ARR_DATA_PTR(arr))[0];
 
 	return colnum;
 }


I notice I cargo-culted a "free de-toasted copy", but I think it's
impossible to end up with a toasted datum here, because the column is
guaranteed to have only one element, so not a candidate for toasting.
But also, if we don't free it (in case somebody does an UPDATE to the
catalog with a large array), nothing happens, because memory is going to
be released soon anyway, by the error that results by conkey not being
one element long.

-- 
Álvaro Herrera               48°01'N 7°57'E  —  https://www.EnterpriseDB.com/
"Si quieres ser creativo, aprende el arte de perder el tiempo"