Source-Git-URL: https://git.postgresql.org/git/pgtranslation/messages.git
Source-Git-Hash: a2d024d57415123f7c9c6e7a71796c7cee8cabc6

"Triggers on partitioned tables cannot have transition tables." is
incorrect as we allow statement-level triggers on partitioned tables to
have transition tables.

This has been wrong since commit 86f57594; back-patch to v11 where that
commit came in.

Reviewed by Tom Lane.

Discussion: https://postgr.es/m/CAPmGK17gk4vXLzz2iG%2BG4LWRWCoVyam70nZ3OuGm1hMJwDrhcg%40mail.gmail.com

Commit f56f8f8d added some code in CloneFkReferencing that's way too
lax about a Constraint node it manufactures, not initializing enough
struct members -- initially_valid in particular was forgotten.  This
causes some FKs in partitions added by ALTER TABLE ATTACH PARTITION to
be marked as not validated.  Set initially_valid true, which fixes the
bug.

While at it, make the struct initialization more complete.  Very similar
code was added in two other places by the same commit; make them all
follow the same pattern for consistency, though no bugs are apparent
there.

This bug has never been reported: I only happened to notice while
working on commit 614a406b4ff1.  The test case that was added there with
the improper result is repaired.

Backpatch to 12.

Discussion: https://postgr.es/m/20221005105523.bhuhkdx4olajboof@alvherre.pgsql

If a syntax error occurred in a SQL-language or PL/pgSQL-language
CREATE FUNCTION or DO command executed in a logical replication worker,
we'd suffer a null pointer dereference or assertion failure.  That
seems like a rather contrived case, but nonetheless worth fixing.

The cause is that function_parse_error_transpose assumes it must be
executing within the context of a Portal, but logical/worker.c
doesn't create a Portal since it's not running the standard executor.
We can just back off the hard Assert check and make it fail gracefully
if there's not an ActivePortal.  (I have a feeling that the aggressive
check here was my fault originally, probably because I wasn't sure if
the case would always hold and wanted to find out.  Well, now we know.)

The hazard seems to exist in all branches that have logical replication,
so back-patch to v10.

Maxim Orlov, Anton Melnikov, Masahiko Sawada, Tom Lane

Discussion: https://postgr.es/m/b570c367-ba38-95f3-f62d-5f59b9808226@inbox.ru
Discussion: https://postgr.es/m/adf0452f-8c6b-7def-d35e-ab516c80088e@inbox.ru

Casting the result of palloc etc. to the intended type is more per
project style anyway.

(The fact that cpluspluscheck doesn't notice these problems is
because it doesn't expand any macros, which seems like a troubling
shortcoming.  Don't have a good idea about improving that.)

Back-patch to v13, which is as far as the patch applies cleanly;
doesn't seem worth working harder.

David Geier

Discussion: https://postgr.es/m/aa5d88a3-71f4-3455-11cf-82de0372c941@gmail.com

If we have no special-case code in s_lock.h for the current platform,
but the compiler has __sync_lock_test_and_set, use that instead of
failing.  It's unlikely that anybody's __sync_lock_test_and_set
would be so awful as to be worse than our semaphore-based fallback,
but if it is, they can (continue to) use --disable-spinlocks.

This allows removal of the RISC-V special case installed by commit
c32fcac56, which generated exactly the same code but only on that
platform.  Usefully, the RISC-V buildfarm animals should now test
at least the int variant of this patch.

I've manually tested both variants on ARM by dint of removing the
ARM-specific stanza.  We don't want to drop that, because it already
has some special knowledge and is likely to grow more over time.
Likewise, this is not meant to preclude installing special cases
for other arches if that proves worthwhile.

Per discussion of a request to install the same code for loongarch64.
Like the previous patch, we might as well back-patch to supported
branches.

Discussion: https://postgr.es/m/761ac43d44b84d679ba803c2bd947cc0@HSMAILSVR04.hs.handsome.com.cn

Since partitions can be foreign tables not only plain tables, but
logical replication only supports plain tables, we'd better check the
relkind of a partition after we find it.  (There was some discussion
of checking this when adding a partitioned table to a subscription;
but that would be inadequate since the troublesome partition could be
added later.)  Without this, the situation leads to a segfault or
assertion failure.

In passing, add a separate variable for the target Relation of
a cross-partition UPDATE; reusing partrel seemed mighty confusing
and error-prone.

Shi Yu and Tom Lane, per report from Ilya Gladyshev.  Back-patch
to v13 where logical replication into partitioned tables became
a thing.

Discussion: https://postgr.es/m/6b93e3748ba43298694f376ca8797279d7945e29.camel@gmail.com

DST law changes in Chile, Fiji, Iran, Jordan, Mexico, Palestine,
and Syria.  Historical corrections for Chile, Crimea, Iran, and
Mexico.

Also, the Europe/Kiev zone has been renamed to Europe/Kyiv
(retaining the old name as a link).

The following zones have been merged into nearby, more-populous zones
whose clocks have agreed since 1970: Antarctica/Vostok, Asia/Brunei,
Asia/Kuala_Lumpur, Atlantic/Reykjavik, Europe/Amsterdam,
Europe/Copenhagen, Europe/Luxembourg, Europe/Monaco, Europe/Oslo,
Europe/Stockholm, Indian/Christmas, Indian/Cocos, Indian/Kerguelen,
Indian/Mahe, Indian/Reunion, Pacific/Chuuk, Pacific/Funafuti,
Pacific/Majuro, Pacific/Pohnpei, Pacific/Wake and Pacific/Wallis.
(This indirectly affects zones that were already links to one of
these: Arctic/Longyearbyen, Atlantic/Jan_Mayen, Iceland,
Pacific/Ponape, Pacific/Truk, and Pacific/Yap.)  America/Nipigon,
America/Rainy_River, America/Thunder_Bay, Europe/Uzhgorod, and
Europe/Zaporozhye were also merged into nearby zones after discovering
that their claimed post-1970 differences from those zones seem to have
been errors.

While the IANA crew have been working on merging zones that have no
post-1970 differences for some time, this batch of changes affects
some zones that are significantly more populous than those merged
in the past, notably parts of Europe.  The loss of pre-1970 timezone
history for those zones may be troublesome for applications
expecting consistency of timestamptz display.  As an example, the
stored value '1944-06-01 12:00 UTC' would previously display as
'1944-06-01 13:00:00+01' if the Europe/Stockholm zone is selected,
but now it will read out as '1944-06-01 14:00:00+02'.

There exists a "packrat" option that will build the timezone data
files with this old data preserved, but the problem is that it also
resurrects a bunch of other, far less well-attested data; so much so
that actually more zones' contents change from 2022a with that option
than without it.  I have chosen not to do that here, for that reason
and because it appears that no major OS distributions are using the
"packrat" option, so that doing so would cause Postgres' behavior
to diverge significantly depending on whether it was built with
--with-system-tzdata.  However, for anyone for whom these changes pose
significant problems, there is a solution: build a set of timezone
files with the "packrat" option and use those with Postgres.

When executing a utility statement, we must fetch everything
we need out of the PlannedStmt data structure before calling
standard_ProcessUtility.  In certain cases (possibly only ROLLBACK
in extended query protocol), that data structure will get freed
during command execution.  The situation is probably often harmless
in production builds, but in debug builds we intentionally overwrite
the freed memory with garbage, leading to picking up garbage values
of statement location and length, typically causing an assertion
failure later in pg_stat_statements.  In non-debug builds, if
something did go wrong it would likely lead to storing garbage
for the query string.

Report and fix by zhaoqigui (with cosmetic adjustments by me).
It's an old problem, so back-patch to all supported versions.

Discussion: https://postgr.es/m/17663-a344fd0675f92128@postgresql.org
Discussion: https://postgr.es/m/1667307420050.56657@hundsun.com

Contrary to what is documented in src/backend/access/transam/README,
ginHeapTupleFastInsert() had a few ordering issues with the way it does
its WAL operations when inserting items in its fast path.

First, when using a separate list, XLogBeginInsert() was being always
called before START_CRIT_SECTION(), and in this case a second thing was
wrong when merging lists, as an exclusive lock was taken on the tail
page *before* calling XLogBeginInsert().  Finally, when inserting items
into a tail page, the order of XLogBeginInsert() and
START_CRIT_SECTION() was reversed.  This commit addresses all these
issues by moving the calls of XLogBeginInsert() after all the pages
logged are locked and pinned, within a critical section.

This has been applied first only on HEAD as of 56b6625, but as per
discussion with Tom Lane and Álvaro Herrera, a backpatch is preferred to
keep all the branches consistent and to respect the transam's README
where we can.

Author:  Matthias van de Meent, Zhang Mingli
Discussion: https://postgr.es/m/CAEze2WhL8uLMqynnnCu1LAPwxD5RKEo0nHV+eXGg_N6ELU88HQ@mail.gmail.com
Backpatch-through: 10

Specifically, when pg_basebackup is invoked with -Tx=y, don't error
out if x could plausibly be an absolute path either on Windows or on
non-Windows systems. We don't know whether the remote system is
running the same OS as the local system, so it's not appropriate to
assume that our local rule about absolute pathnames is the same as
the rule on the remote system.

Patch by me, reviewed by Tom Lane, Andrew Dunstan, and
Davinder Singh.

Discussion: http://postgr.es/m/CA+TgmoY+jC3YiskomvYKDPK3FbrmsDU7_8+wMHt02HOdJeRb0g@mail.gmail.com

Previously in commit 42681dffaf, we added CFI during decoding changes but
missed another similar case that can happen while restoring changes
spilled to disk back into memory in a loop.

Reported-by: Robert Haas
Author: Amit Kapila
Backpatch-through: 10
Discussion: https://postgr.es/m/CA+TgmoaLObg0QbstbC8ykDwOdD1bDkr4AbPpB=0DPgA2JW0mFg@mail.gmail.com

This problem has been introduced by commit 272248a0c1 where we started
assigning the subtransactions to the top-level transaction when we mark
both the top-level transaction and its subtransactions as containing
catalog changes. After we assign subtransactions to the top-level
transaction, we were not allowed to execute any invalidations associated
with it when we decide to skip the transaction.

The reason to assign the subtransactions to the top-level transaction was
to avoid the assertion failure in AssertTXNLsnOrder() as they have the
same LSN when we sometimes start accumulating transaction changes for
partial transactions after the restart. Now that with commit 64ff0fe4e8,
we skip this assertion check until we reach the LSN at which we start
decoding the contents of the transaction, so, there is no reason for such
an assignment anymore.

The assignment change was introduced in 15 and prior versions but this bug
doesn't exist in branches prior to 14 since we don't add invalidation
messages to subtransactions. We decided to backpatch through 11 for
consistency but not for 10 since its final release is near.

Reported-by: Kuroda Hayato
Author: Masahiko Sawada
Reviewed-by: Amit Kapila
Backpatch-through: 11
Discussion: https://postgr.es/m/TYAPR01MB58660803BCAA7849C8584AA4F57E9%40TYAPR01MB5866.jpnprd01.prod.outlook.com
Discussion: https://postgr.es/m/a89b46b6-0239-2fd5-71a9-b19b1f7a7145%40enterprisedb.com

56788d21 adjusted the parallel seq scan code so that instead of handing
out a single block at a time to parallel workers, it now hands out ranges
of blocks.

Here we update the documentation which still claimed that workers received
just 1 block at a time.

Reported-by: Zhang Mingli
Discussion: https://postgr.es/m/17c99615-2c3b-4e4e-9d0b-424a66a7bccd@Spark
Backpatch-through: 14, where 56788d21 was added.

When the logical decoding restarts from NEW_CID, since there is no
association between the top transaction and its subtransaction, both are
created as top transactions and have the same LSN. This caused the
assertion failure in AssertTXNLsnOrder().

This patch skips the assertion check until we reach the LSN at which we
start decoding the contents of the transaction, specifically
start_decoding_at LSN in SnapBuild. This is okay because we don't
guarantee to make the association between top transaction and
subtransaction until we try to decode the actual contents of transaction.
The ordering of the records prior to the start_decoding_at LSN should have
been checked before the restart.

The other assertion failure is due to the reason that we forgot to track
that we have considered top-level transaction id in the list of catalog
changing transactions that were committed when one of its subtransactions
is marked as containing catalog change.

Reported-by: Tomas Vondra, Osumi Takamichi
Author: Masahiko Sawada, Kuroda Hayato
Reviewed-by: Amit Kapila, Dilip Kumar, Kuroda Hayato, Kyotaro Horiguchi, Masahiko Sawada
Backpatch-through: 10
Discussion: https://postgr.es/m/a89b46b6-0239-2fd5-71a9-b19b1f7a7145%40enterprisedb.com
Discussion: https://postgr.es/m/TYCPR01MB83733C6CEAE47D0280814D5AED7A9%40TYCPR01MB8373.jpnprd01.prod.outlook.com

Per https://llvm.org/docs/OpaquePointers.html, support for non-opaque
pointers still exists and we can request that on our context.  We have
until LLVM 16 to move to opaque pointers, a much larger change.

Back-patch to 11, where LLVM support arrived.

Author: Thomas Munro <thomas.munro@gmail.com>
Author: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/CAMHz58Sf_xncdyqsekoVsNeKcruKootLtVH6cYXVhhUR1oKPCg%40mail.gmail.com

It was previously easily overlooked at the end of several tables.

Reported-by: Alex Denman

Discussion: https://postgr.es/m/166335888474.659.16897487975376230364@wrigleys.postgresql.org

Backpatch-through: 10

The fix is to run ANALYZE.

Discussion: https://postgr.es/m/YzRr+ys98UzVQJvK@momjian.us,
   https://postgr.es/m/flat/CAKJS1f8DTbCHf9gedU0He6ARsd58E6qOhEHM1caomqj_r9MOiQ%40mail.gmail.com,
   https://postgr.es/m/CAKJS1f80o98hcfSk8j%3DfdN09S7Sjz%2BvuzhEwbyQqvHJb_sZw0g%40mail.gmail.com

Backpatch-through: 10

DefineQueryRewrite() has long required that ON SELECT rules be named
"_RETURN".  But we overlooked the converse case: we should forbid
non-ON-SELECT rules that are named "_RETURN".  In particular this
prevents using CREATE OR REPLACE RULE to overwrite a view's _RETURN
rule with some other kind of rule, thereby breaking the view.

Per bug #17646 from Kui Liu.  Back-patch to all supported branches.

Discussion: https://postgr.es/m/17646-70c93cfa40365776@postgresql.org

The executor will dump core if it's asked to execute a seqscan on
a relation having no table AM, such as a view.  While that shouldn't
really happen, it's possible to get there via catalog corruption,
such as a missing ON SELECT rule.  It seems worth installing a defense
against that.  There are multiple plausible places for such a defense,
but I picked the planner's get_relation_info().

Per discussion of bug #17646 from Kui Liu.  Back-patch to v12 where
the tableam APIs were introduced; in older versions you won't get a
SIGSEGV, so it seems less pressing.

Discussion: https://postgr.es/m/17646-70c93cfa40365776@postgresql.org

If the non-recursive term of a SEARCH BREADTH FIRST recursive
query has only constants in its target list, the planner will
fold the starting RowExpr added by rewrite into a simple Const
of type RECORD.  The executor doesn't have any problem with
that --- but EXPLAIN VERBOSE will encounter the Const as the
ultimate source of truth about what the field names of the
SET column are, and it didn't know what to do with that.
Fortunately, we can pull the identifying typmod out of the
Const, in much the same way that record_out would.

For reasons that remain a bit obscure to me, this only fails
with SEARCH BREADTH FIRST, not SEARCH DEPTH FIRST or CYCLE.
But I added regression test cases for both of those options
too, just to make sure we don't break it in future.

Per bug #17644 from Matthijs van der Vleuten.  Back-patch
to v14 where these constructs were added.

Discussion: https://postgr.es/m/17644-3bd1f3036d6d7a16@postgresql.org

In the latest version of Apple's macOS SDK, <sys/socket.h>
fails to compile if "REF" is #define'd as something.
Apple may or may not agree that this is a bug, and even if
they do accept the bug report I filed, they probably won't
fix it very quickly.  In the meantime, our back branches will all
fail to compile gram.y.  v15 and HEAD currently escape the problem
thanks to the refactoring done in 98e93a1fc, but that's purely
accidental.  Moreover, since that patch removed a widely-visible
inclusion of <netdb.h>, back-patching it seems too likely to break
third-party code.

Instead, change the token's code name to REF_P, following our usual
convention for naming parser tokens that are likely to have symbol
conflicts.  The effects of that should be localized to the grammar
and immediately surrounding files, so it seems like a safer answer.

Per project policy that we want to keep recently-out-of-support
branches buildable on modern systems, back-patch all the way to 9.2.

Discussion: https://postgr.es/m/1803927.1665938411@sss.pgh.pa.us

snprintf.c has always fallen back on libc's *printf implementation
when printing pointers (%p) and floats.  When this code originated,
we were still supporting some platforms that lacked native snprintf,
so we used sprintf for that.  That's not actually unsafe in our usage,
but nonetheless builds on macOS are starting to complain about sprintf
being unconditionally deprecated; and I wouldn't be surprised if other
platforms follow suit.  There seems little reason to believe that any
platform supporting C99 wouldn't have standards-compliant snprintf,
so let's just use that instead to suppress such warnings.

Back-patch to v12, which is where we started to require C99.  It's
also where we started to use our snprintf.c everywhere, so this
wouldn't be enough to suppress the warning in older branches anyway
--- that is, in older branches these aren't necessarily all our
usages of libc's sprintf.  It is enough in v12+ because any
deprecation annotation attached to libc's sprintf won't apply to
pg_sprintf.  (Whether all our usages of pg_sprintf are adequately
safe is not a matter I intend to address here, but perhaps it could
do with some review.)

Per report from Andres Freund and local testing.

Discussion: https://postgr.es/m/20221015211955.q4cwbsfkyk3c4ty3@awork3.anarazel.de

When a query whose results were requested in single-row mode is the last
in the queue by the time those results are being read, the single-row
flag was not being reset, because we were returning early from
pqPipelineProcessQueue.  Move that stanza up so that the flag is always
reset at the end of sending that query's results.

Add a test for the situation.

Backpatch to 14.

Author: Denis Laxalde <denis.laxalde@dalibo.com>
Discussion: https://postgr.es/m/01af18c5-dacc-a8c8-07ee-aecc7650c3e8@dalibo.com

While at it, simplify wording a bit.

Author: Takamichi Osumi <osumi.takamichi@fujitsu.com>
Reviewed-by: Peter Smith <smithpb2250@gmail.com>
Discussion: https://postgr.es/m/TYCPR01MB8373F93F5D094A2BE648990DED259@TYCPR01MB8373.jpnprd01.prod.outlook.com

Add
    After=network-online.target
    Wants=network-online.target
to the suggested unit file for starting a Postgres server.
This delays startup until the network interfaces have been
configured; without that, any attempt to bind to a specific
IP address will fail.

If listen_addresses is set to "localhost" or "*", it might be
possible to get away with the less restrictive "network.target",
but I don't think we need to get into such detail here.

Per suggestion from Pablo Federico.

Discussion: https://postgr.es/m/166552157407.591805.10036014441784710940@wrigleys.postgresql.org

The postmaster is not supposed to do anything that depends
fundamentally on shared memory contents, because that creates
the risk that a backend crash that trashes shared memory will
take the postmaster down with it, preventing automatic recovery.
In commit 969d7cd4 I lost sight of this principle and coded
AssignPostmasterChildSlot() in such a way that it could fail
or even crash if the shared PMSignalState structure became
corrupted.  Remarkably, we've not seen field reports of such
crashes; but I managed to induce one while testing the recent
changes around palloc chunk headers.

To fix, make a semi-duplicative state array inside the postmaster
so that we need consult only local state while choosing a "child
slot" for a new backend.  Ensure that other postmaster-executed
routines in pmsignal.c don't have critical dependencies on the
shared state, either.  Corruption of PMSignalState might now
lead ReleasePostmasterChildSlot() to conclude that backend X
failed, when actually backend Y was the one that trashed things.
But that doesn't matter, because we'll force a cluster-wide reset
regardless.

Back-patch to all supported branches, since this is an old bug.

Discussion: https://postgr.es/m/3436789.1665187055@sss.pgh.pa.us

DEFAULT markers appearing in an INSERT on an updatable view
could be mis-processed if they were in a multi-row VALUES clause.
This would lead to strange errors such as "cache lookup failed
for type NNNN", or in older branches even to crashes.

The cause is that commit 41531e42 tried to re-use rewriteValuesRTE()
to remove any SetToDefault nodes (that hadn't previously been replaced
by the view's own default values) appearing in "product" queries,
that is DO ALSO queries.  That's fundamentally wrong because the
DO ALSO queries might not even be INSERTs; and even if they are,
their targetlists don't necessarily match the view's column list,
so that almost all the logic in rewriteValuesRTE() is inapplicable.

What we want is a narrow focus on replacing any such nodes with NULL
constants.  (That is, in this context we are interpreting the defaults
as being strictly those of the view itself; and we already replaced
any that aren't NULL.)  We could add still more !force_nulls tests
to further lobotomize rewriteValuesRTE(); but it seems cleaner to
split out this case to a new function, restoring rewriteValuesRTE()
to the charter it had before.

Per bug #17633 from jiye_sw.  Patch by me, but thanks to
Richard Guo and Japin Li for initial investigation.
Back-patch to all supported branches, as the previous fix was.

Discussion: https://postgr.es/m/17633-98cc85e1fa91e905@postgresql.org

PostgreSQL::Test::Cluster and ::Utils were not being installed.  This is
very hard to notice, as it only seems to affect external modules that
want to run tests from 15 back in earlier versions.  Oversight in
b235d41d.

This applies only to branches 14 and back, because 15 had already been
made correct in commit b3b4d8e68ae8.

Discussion: https://postgr.es/m/20221010093415.poplkyn7pjeiv2y7@alvherre.pgsql

There are a number of bugs in this area.  Two of them are fixed here,
namely:
1. get_relation_idx_constraint_oid does not restrict the type of
   constraint that's returned, so with sufficient bad luck it can
   return the OID of a foreign key constraint.  This has the effect that
   a primary key in a partition can end up as a child of a foreign key,
   which makes no sense (it needs to be the child of the equivalent
   primary key.)
   Change the API contract so that only index-backed constraints are
   returned, mimicking get_constraint_index().

2. Both CloneFkReferenced and CloneFkReferencing clone a
   self-referencing foreign key, so the partition ends up with
   a duplicate foreign key.  Change the former function to ignore such
   constraints.

Add some tests to verify that things are better now.  (However, these
new tests show some additional misbehavior that will be fixed later --
namely that there's a constraint marked NOT VALID.)

Backpatch to 12, where these constraints are possible at all.

Author: Jehan-Guillaume de Rorthais <jgdr@dalibo.com>
Discussion: https://postgr.es/m/20220603154232.1715b14c@karst

Commit 34f581c3 intended to ensure that RelationGetBufferForTuple
would acquire a visibility-map page pin in case the otherBuffer's
all-visible bit had become set since we last had lock on that page.
But I missed a case: when we're extending the relation, VM concerns
were dealt with only in the relatively-less-likely case that we
fail to conditionally lock the otherBuffer.  I think I'd believed
that we couldn't need to worry about it if the conditional lock
succeeds, which is true for the target buffer; but the otherBuffer
was unlocked for awhile so its bit might be set anyway.  So we need
to do the GetVisibilityMapPins dance, and then also recheck the
page's free space, in both cases.

Per report from Jaime Casanova.  Back-patch to v12 as the previous
patch was (although there's still no evidence that the bug is
reachable pre-v14).

Discussion: https://postgr.es/m/E1lWLjP-00006Y-Ml@gemulon.postgresql.org

The compression parameter to PQsslAttribute has never returned the
compression method used, it has always returned "on" or "off since
it was added in commit 91fa7b47. Backpatch through v10.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/B9EC60EC-F665-47E8-A221-398C76E382C9@yesql.se
Backpatch-through: v10

Remove the stale text that is a leftover from an earlier version of the
patch to add support for batch insertion, and adjust the wording in the
remaining text.

Back-patch to v14 where batch insertion came in.

Review and wording adjustment by Tom Lane.

Discussion: https://postgr.es/m/CAPmGK14goatHPHQv2Aeu_UTKqZ%2BBO%2BP%2Bzd3HKv5D%2BdyyfWKDSw%40mail.gmail.com

Reported-by: Tom Lane

Discussion: https://postgr.es/m/3976627.1662651004@sss.pgh.pa.us

Backpatch-through: 10