From: Antonin Houska <ah@cybertec.at>
To: Bruce Momjian <bruce@momjian.us>
cc: Alvaro Herrera <alvherre@2ndquadrant.com>,
    pgsql-hackers@postgresql.org
Subject: Re: Incorrect visibility test function assigned to snapshot
In-reply-to: <20180623151856.GG21575@momjian.us>
References: <23215.1527665193@localhost>
 <20180530132854.tkv37qiamepu3slh@alvherre.pgsql>
 <20180623151856.GG21575@momjian.us>
Comments: In-reply-to Bruce Momjian <bruce@momjian.us>
   message dated "Sat, 23 Jun 2018 11:18:56 -0400."
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="=-=-="
Date: Fri, 08 Feb 2019 11:59:05 +0100
Message-ID: <23440.1549623545@localhost>
Precedence: bulk

--=-=-=
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable

Bruce Momjian <bruce@momjian.us> wrote:

> On Wed, May 30, 2018 at 09:28:54AM -0400, Alvaro Herrera wrote:
> > On 2018-May-30, Antonin Houska wrote:
> >=20
> > > In the header comment, SnapBuildInitialSnapshot() claims to set
> > > snapshot->satisfies to the HeapTupleSatisfiesMVCC test function, and =
indeed it
> > > converts the "xid" array to match its semantics (i.e. the xid items e=
ventually
> > > represent running transactions as opposed to the committed ones). How=
ever the
> > > test function remains HeapTupleSatisfiesHistoricMVCC as set by
> > > SnapBuildBuildSnapshot().
> >=20
> > Interesting.  While this sounds like an oversight that should have
> > horrible consequences, it's seems not to because the current callers
> > don't seem to care about the ->satisfies function.  Are you able to come
> > up with some scenario in which it causes an actual problem?
>=20
> Uh, are we going to fix this anyway?  Seems we should.

Sorry, I forgot. Patch is below and I'm going to add an entry to the next C=
F.

--=20
Antonin Houska
Cybertec Sch=C3=B6nig & Sch=C3=B6nig GmbH
Gr=C3=B6hrm=C3=BChlgasse 26, A-2700 Wiener Neustadt
Web: https://www.cybertec-postgresql.com


--=-=-=
Content-Type: text/x-diff
Content-Disposition: inline; filename=snapshot_type.patch

diff --git a/src/backend/replication/logical/snapbuild.c b/src/backend/replication/logical/snapbuild.c
index 2f185f7823..3394abb602 100644
--- a/src/backend/replication/logical/snapbuild.c
+++ b/src/backend/replication/logical/snapbuild.c
@@ -615,6 +615,8 @@ SnapBuildInitialSnapshot(SnapBuild *builder)
 
 		TransactionIdAdvance(xid);
 	}
+	/* And of course, adjust snapshot type accordingly. */
+	snap->snapshot_type = SNAPSHOT_MVCC;
 
 	snap->xcnt = newxcnt;
 	snap->xip = newxip;
diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index 6e02585e10..bc0166cc6f 100644
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -1502,6 +1502,13 @@ ImportSnapshot(const char *idstr)
 	 */
 	memset(&snapshot, 0, sizeof(snapshot));
 
+	/*
+	 * Do not rely on the fact that SNAPSHOT_MVCC is zero. (The core code
+	 * currently does not use this field of imported snapshot, but let's keep
+	 * things consistent.)
+	 */
+	snapshot.snapshot_type = SNAPSHOT_MVCC;
+
 	parseVxidFromText("vxid:", &filebuf, path, &src_vxid);
 	src_pid = parseIntFromText("pid:", &filebuf, path);
 	/* we abuse parseXidFromText a bit here ... */

--=-=-=--