If the client sent a query cancel request with backend PID 0, it
tripped an assertion. With assertions disabled, you got this in the
log instead:
LOG: invalid cancel request with PID 0
LOG: wrong key in cancel request for process 0
Query cancellations don't even require authentication, so we better
tolerate bogus requests. Fix by turning the assertion into a regular
runtime check.
Spotted while testing libpq behavior with a modified server that
didn't send BackendKeyData to the client.
Tom Lane [Tue, 29 Jul 2025 19:17:40 +0000 (15:17 -0400)]
Don't put library-supplied -L/-I switches before user-supplied ones.
For many optional libraries, we extract the -L and -l switches needed
to link the library from a helper program such as llvm-config. In
some cases we put the resulting -L switches into LDFLAGS ahead of
-L switches specified via --with-libraries. That risks breaking
the user's intention for --with-libraries.
It's not such a problem if the library's -L switch points to a
directory containing only that library, but on some platforms a
library helper may "helpfully" offer a switch such as -L/usr/lib
that points to a directory holding all standard libraries. If the
user specified --with-libraries in hopes of overriding the standard
build of some library, the -L/usr/lib switch prevents that from
happening since it will come before the user-specified directory.
To fix, avoid inserting these switches directly into LDFLAGS during
configure, instead adding them to LIBDIRS or SHLIB_LINK. They will
still eventually get added to LDFLAGS, but only after the switches
coming from --with-libraries.
The same problem exists for -I switches: those coming from
--with-includes should appear before any coming from helper programs
such as llvm-config. We have not heard field complaints about this
case, but it seems certain that a user attempting to override a
standard library could have issues.
The changes for this go well beyond configure itself, however,
because many Makefiles have occasion to manipulate CPPFLAGS to
insert locally-desirable -I switches, and some of them got it wrong.
The correct ordering is any -I switches pointing at within-the-
source-tree-or-build-tree directories, then those from the tree-wide
CPPFLAGS, then those from helper programs. There were several places
that risked pulling in a system-supplied copy of libpq headers, for
example, instead of the in-tree files. (Commit cb36f8ec2 fixed one
instance of that a few months ago, but this exercise found more.)
The Meson build scripts may or may not have any comparable problems,
but I'll leave it to someone else to investigate that.
Reported-by: Charles Samborski <demurgos@demurgos.net>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/70f2155f-27ca-4534-b33d-7750e20633d7@demurgos.net
Backpatch-through: 13
Tom Lane [Tue, 29 Jul 2025 16:47:19 +0000 (12:47 -0400)]
Remove unnecessary complication around xmlParseBalancedChunkMemory.
When I prepared 71c0921b6 et al yesterday, I was thinking that the
logic involving explicitly freeing the node_list output was still
needed to dodge leakage bugs in libxml2. But I was misremembering:
we introduced that only because with early 2.13.x releases we could
not trust xmlParseBalancedChunkMemory's result code, so we had to
look to see if a node list was returned or not. There's no reason
to believe that xmlParseBalancedChunkMemory will fail to clean up
the node list when required, so simplify. (This essentially
completes reverting all the non-cosmetic changes in 6082b3d5d.)
Reported-by: Jim Jones <jim.jones@uni-muenster.de>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/997668.1753802857@sss.pgh.pa.us
Backpatch-through: 13
Tom Lane [Tue, 29 Jul 2025 14:35:01 +0000 (10:35 -0400)]
Split up pgfdw_report_error so that we can mark it pg_noreturn.
pgfdw_report_error has the same design fault as elog/ereport
do, namely that it might or might not return depending on elevel.
While those functions are too widely used to redesign, there are
only about 30 call sites for pgfdw_report_error, and it's not
exposed for extension use. So let's rethink it. Split it into
pgfdw_report_error() which hard-wires ERROR elevel and is marked
pg_noreturn, and pgfdw_report() which allows only elevels less
than ERROR. (Thanks to Álvaro Herrera for suggesting this naming.)
The motivation for doing this now is that in the wake of commit 80aa9848b, which removed a bunch of PG_TRYs from postgres_fdw,
we're seeing more thorough flow analysis there from C compilers
and Coverity. Marking pgfdw_report_error as noreturn where
appropriate should help prevent false-positive complaints.
We could alternatively have invented a macro wrapper similar
to what we use for elog/ereport, but that code is sufficiently
fragile that I didn't find it appetizing to make another copy.
Since 80aa9848b already changed pgfdw_report_error's signature,
this won't make back-patching any harder than it was already.
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/420221.1753714491@sss.pgh.pa.us
Tom Lane [Tue, 29 Jul 2025 13:42:22 +0000 (09:42 -0400)]
Suppress uninitialized-variable warning.
In the wake of commit 80aa9848b, a few compilers think that
postgresAcquireSampleRowsFunc's "reltuples" might be used
uninitialized. The logic is visibly correct, both before
and after that change; presumably what happened here is that
the previous presence of a setjmp() in the function stopped
them from attempting any flow analysis at all. Add a dummy
initialization to silence the warning.
Reported-by: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAExHW5tkerCufA_F6oct5dMJ61N+yVrVgYXL7M8dD-5_zXjrDw@mail.gmail.com
Add regression test for background worker restart after crash.
Previously, if a background worker crashed and the server restarted
with restart_after_crash enabled, the worker was not restarted
as expected. This issue was fixed by commit b5d084c5353,
which ensures that background workers without the never-restart flag
are correctly restarted after a crash-and-restart cycle.
To guard against regressions, this commit adds a test that verifies
a background worker successfully restarts in such a scenario.
Michael Paquier [Tue, 29 Jul 2025 08:03:07 +0000 (17:03 +0900)]
Handle timeout in PostgreSQL::Test::Cluster::is_alive()
This commit adds an extra --timeout=PG_TEST_TIMEOUT_DEFAULT to the call
of pg_isready done in is_alive(), so as it is possible to have more
leverage with the call on machines constrained on resources.
By default the timeout is 180s, and it can be changed depending on the
environment where the tests are run.
Per buildfarm member mamba, where the default timeout of 3s used by
pg_isready has proved that it may not be enough as the postmaster may
not have the time it needs to reply to a ping request.
Reported-by: Alexander Lakhin <exclusion@gmail.com> Reviewed-by: Nazir Bilal Yavuz <byavuz81@gmail.com>
Discussion: https://postgr.es/m/29b637df-f818-4b52-986a-f11ba28300e9@gmail.com
This commit documents differences in the definition of word separators for
the initcap function between libc and ICU locale providers.
Backpatch to all supported branches.
David Rowley [Tue, 29 Jul 2025 03:18:01 +0000 (15:18 +1200)]
Display Memoize planner estimates in EXPLAIN
There've been a few complaints that it can be overly difficult to figure
out why the planner picked a Memoize plan. To help address that, here we
adjust the EXPLAIN output to display the following additional details:
1) The estimated number of cache entries that can be stored at once
2) The estimated number of unique lookup keys that we expect to see
3) The number of lookups we expect
4) The estimated hit ratio
Technically #4 can be calculated using #1, #2 and #3, but it's not a
particularly obvious calculation, so we opt to display it explicitly.
The original patch by Lukas Fittl only displayed the hit ratio, but
there was a fear that might lead to more questions about how that was
calculated. The idea with displaying all 4 is to be transparent which
may allow queries to be tuned more easily. For example, if #2 isn't
correct then maybe extended statistics or a manual n_distinct estimate can
be used to help fix poor plan choices.
Author: Ilia Evdokimov <ilya.evdokimov@tantorlabs.com>
Author: Lukas Fittl <lukas@fittl.com> Reviewed-by: David Rowley <dgrowleyml@gmail.com> Reviewed-by: Andrei Lepikhov <lepihov@gmail.com> Reviewed-by: Robert Haas <robertmhaas@gmail.com>
Discussion: https://postgr.es/m/CAP53Pky29GWAVVk3oBgKBDqhND0BRBN6yTPeguV_qSivFL5N_g%40mail.gmail.com
Tom Lane [Mon, 28 Jul 2025 20:50:41 +0000 (16:50 -0400)]
Avoid regression in the size of XML input that we will accept.
This mostly reverts commit 6082b3d5d, "Use xmlParseInNodeContext
not xmlParseBalancedChunkMemory". It turns out that
xmlParseInNodeContext will reject text chunks exceeding 10MB, while
(in most libxml2 versions) xmlParseBalancedChunkMemory will not.
The bleeding-edge libxml2 bug that we needed to work around a year
ago is presumably no longer a factor, and the argument that
xmlParseBalancedChunkMemory is semi-deprecated is not enough to
justify a functionality regression. Hence, go back to doing it
the old way.
Reported-by: Michael Paquier <michael@paquier.xyz>
Author: Michael Paquier <michael@paquier.xyz> Co-authored-by: Erik Wienhold <ewie@ewie.name> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/aIGknLuc8b8ega2X@paquier.xyz
Backpatch-through: 13
Robert Haas [Mon, 28 Jul 2025 15:15:47 +0000 (11:15 -0400)]
Remove misleading hint for "unexpected data beyond EOF" error.
Commit ffae5cc5a6024b4e25ec920ed5c4dfac649605f8 added this hint in 2006,
but it's now obsolete and doesn't reflect what users should really check
in this situation. We were not able to agree on a new hint, so just delete
the existing one and update the comments to mention one possibility that
is known to cause problems of this kind: something other than PostgreSQL
is modifying files in the PostgreSQL data directory.
Author: Jakub Wartak <jakub.wartak@enterprisedb.com> Reviewed-by: Robert Haas <rhaas@postgresql.org> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Christoph Berg <myon@debian.org>
Discussion: https://postgr.es/m/CAKZiRmxNbcaL76x=09Sxf7aUmrRQJBf8drzDdUHo+j9_eM+VMg@mail.gmail.com
As a result, an attempt to set synchronous_standby_names to any value
that does not parse resulted in a generic "parser failed." message
rather than anything more specific. This fixes that.
Michael Paquier [Sun, 27 Jul 2025 23:15:11 +0000 (08:15 +0900)]
Fix performance regression with flush of pending fixed-numbered stats
The callback added in fc415edf8ca8 used to check if there is any pending
data to flush for fixed-numbered statistics, done by looping across all
the builtin and custom stats kinds with a call to have_fixed_pending_cb,
is proving to able to show in workloads that do not report any stats
(read-only, no function calls, no WAL, no IO, etc). The code used in
v17 was cheaper than that what HEAD has introduced, relying on three
boolean checks for WAL, SLRU and IO stats.
This commit switches the code to use a more efficient approach than fc415edf8ca8, with a single boolean flag that can be switched to "true"
by any fixed-numbered stats kinds to force pgstat_report_stat() to go
through one round of reports. The flag is reset by pgstat_report_stat()
once a full round of reports is done. The flag being false means that
fixed-numbered stats kinds saw no activity, and that there is no pending
data to flush.
ac000fca743e took one step in improving the performance by reducing the
number of stats kinds that the backend can hold. This commit takes a
more drastic step by bringing back the code efficiency to what it was
before v18 with a cheap check at the beginning of pgstat_report_stat()
for its fast-exit path.
The callback have_static_pending_cb is removed as an effect of all that.
Process sync requests incrementally in AbsorbSyncRequests
If the number of sync requests is big enough, the palloc() call in
AbsorbSyncRequests() will attempt to allocate more than 1 GB of memory,
resulting in failure. This can lead to an infinite loop in the checkpointer
process, as it repeatedly fails to absorb the pending requests.
This commit introduces the following changes to cope with this problem:
1. Turn pending checkpointer requests array in shared memory into a bounded
ring buffer.
2. Limit maximum ring buffer size to 10M items.
3. Make AbsorbSyncRequests() process requests incrementally in 10K batches.
Even #2 makes the whole queue size fit the maximum palloc() size of 1 GB.
of continuous lock holding.
This commit is for master only. Simpler fix, which just limits a request
queue size to 10M, will be backpatched.
Reported-by: Ekaterina Sokolova <e.sokolova@postgrespro.ru>
Discussion: https://postgr.es/m/db4534f83a22a29ab5ee2566ad86ca92%40postgrespro.ru
Author: Maxim Orlov <orlovmg@gmail.com> Co-authored-by: Xuneng Zhou <xunengzhou@gmail.com> Reviewed-by: Andres Freund <andres@anarazel.de> Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi> Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Michael Paquier [Sun, 27 Jul 2025 08:48:47 +0000 (17:48 +0900)]
Add assertions for all the required index AM callbacks
Similar checks are done for the mandatory table AM callbacks. A portion
of the index AM callbacks are optional and can be NULL; the rest is
mandatory and is documented as such in the documentation and in amapi.h.
These checks are useful to detect quickly if all the mandatory callbacks
are defined when implementing a new index access method, as the
assertions are run when loading the AM.
Author: Japin Li <japinli@hotmail.com>
Discussion: https://postgr.es/m/ME0P300MB0445795D31CEAB92C58B41FDB651A@ME0P300MB0445.AUSP300.PROD.OUTLOOK.COM
Tom Lane [Fri, 25 Jul 2025 20:37:29 +0000 (16:37 -0400)]
Silence leakage complaint about postgres_fdw's InitPgFdwOptions.
Valgrind complains that the PQconninfoOption array returned by libpq
is leaked. We apparently believed that we could suppress that warning
by storing that array's address in a static variable. However, modern
C compilers are bright enough to optimize the static variable away.
We could escalate that arms race by making the variable global.
But on the whole it seems better to revise the code so that it
can free libpq's result properly. The only thing that costs
us is copying the parameter-name keywords; which seems like a
pretty negligible cost in a function that runs at most once per
process.
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Discussion: https://postgr.es/m/2976982.1748049023@sss.pgh.pa.us
Tom Lane [Fri, 25 Jul 2025 20:36:44 +0000 (16:36 -0400)]
Run pgindent on the changes of the previous patch.
This step can be checked mechanically.
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Discussion: https://postgr.es/m/2976982.1748049023@sss.pgh.pa.us
Tom Lane [Fri, 25 Jul 2025 20:31:43 +0000 (16:31 -0400)]
Reap the benefits of not having to avoid leaking PGresults.
Remove a bunch of PG_TRY constructs, de-volatilize related
variables, remove some PQclear calls in error paths.
Aside from making the code simpler and shorter, this should
provide some marginal performance gains.
For ease of review, I did not re-indent code within the removed
PG_TRY constructs. That'll be done in a separate patch.
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Discussion: https://postgr.es/m/2976982.1748049023@sss.pgh.pa.us
Tom Lane [Fri, 25 Jul 2025 20:30:00 +0000 (16:30 -0400)]
Create infrastructure to reliably prevent leakage of PGresults.
Commit 232d8caea fixed a case where postgres_fdw could lose track
of a PGresult object, resulting in a process-lifespan memory leak.
But I have little faith that there aren't other potential PGresult
leakages, now or in future, in the backend modules that use libpq.
Therefore, this patch proposes infrastructure that makes all
PGresults returned from libpq act as though they are palloc'd
in the CurrentMemoryContext (with the option to relocate them to
another context later). This should greatly reduce the risk of
careless leaks, and it also permits removal of a bunch of code
that attempted to prevent such leaks via PG_TRY blocks.
This patch adds infrastructure that wraps each PGresult in a
"libpqsrv_PGresult" that provides a memory context reset callback
to PQclear the PGresult. Code using this abstraction is inherently
memory-safe to the same extent as we are accustomed to in most backend
code. Furthermore, we add some macros that automatically redirect
calls of the libpq functions concerned with PGresults to use this
infrastructure, so that almost no source-code changes are needed to
wheel this infrastructure into place in all the backend code that
uses libpq.
Perhaps in future we could create similar infrastructure for
PGconn objects, but there seems less need for that.
This patch just creates the infrastructure and makes relevant code
use it, including reverting 232d8caea in favor of this mechanism.
A good deal of follow-on simplification is possible now that we don't
have to be so cautious about freeing PGresults, but I'll put that in
a separate patch.
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Matheus Alcantara <matheusssilv97@gmail.com>
Discussion: https://postgr.es/m/2976982.1748049023@sss.pgh.pa.us
This flag was effectively a no-op in EXEC_BACKEND (ie, Windows)
builds, because it was kept in the process-local HTAB struct,
and it could only ever become set in the postmaster's copy.
The simplest fix is to move it to the shared HASHHDR struct.
We could keep a copy in HTAB as well, as we do with keysize
and some other fields, but the "too much contention" argument
doesn't seem to apply here: we only examine isfixed during
element_alloc(), which had better not get hit very often for
a shared hashtable.
This oversight dates to 7c797e719 which invented the option.
But back-patching doesn't seem appropriate given the lack of
field complaints. If there is anyone running an affected
workload on Windows, they might be unhappy about the behavior
changing in a minor release.
Author: Aidar Imamov <a.imamov@postgrespro.ru> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/4d0cb35ff01c5c74d2b9a582ecb73823@postgrespro.ru
Refactor grammar to create opt_utility_option_list
This changes the grammar for REINDEX, CHECKPOINT, CLUSTER, ANALYZE/ANALYSE;
they still accept the same options as before, but the grammar is written
differently for convenience of future development.
Fix background worker not restarting after crash-and-restart cycle.
Previously, if a background worker crashed (e.g., due to a SIGKILL) and
the server restarted due to restart_after_crash being enabled,
the worker was not restarted as expected. Background workers without
the never-restart flag should automatically restart in this case.
This issue was introduced in commit 28a520c0b77, which failed to reset
the rw_pid field in the RegisteredBgWorker struct for the crashed worker.
This commit fixes the problem by resetting rw_pid for all eligible
background workers during the crash-and-restart cycle.
Back-patched to v18, where the bug was introduced.
Bug fix patches were proposed by Andrey Rudometov and ChangAo Chen,
but this commit uses a different approach.
Michael Paquier [Fri, 25 Jul 2025 07:17:13 +0000 (16:17 +0900)]
Fix assertion failure with latch wait in single-user mode
LatchWaitSetPostmasterDeathPos, the latch event position for the
postmaster death event, is initialized under IsUnderPostmaster.
WaitLatch() considered it as a valid wait target in single-user mode
(!IsUnderPostmaster), which was incorrect.
One code path found to fail with an assertion failure is a database drop
in single-user mode while waiting in WaitForProcSignalBarrier() after
the drop.
Michael Paquier [Fri, 25 Jul 2025 02:17:48 +0000 (11:17 +0900)]
Lower bounds related to pgstats kinds
This commit changes stats kinds to have the following bounds, making
their handling in core cheaper by default:
- PGSTAT_KIND_CUSTOM_MIN 128 -> 24
- PGSTAT_KIND_MAX 256 -> 32
The original numbers were rather high, and showed an impact on
performance in pgstat_report_stat() for the case of simple queries with
its early-exit path if there are no pending statistics to flush. This
logic will be improved more in a follow-up commit to bring the
performance of pgstat_report_stat() on par with v17 and older versions.
Lowering the bounds is a change worth doing on its own, independently of
the other improvement.
These new numbers should be enough to leave some room for the following
years for built-in and custom stats kinds, with stable ID numbers. At
least that should be enough to start with this facility for extension
developers. It can be always increased in the tree depending on the
requirements wanted.
Per discussion with Andres Freund and Bertrand Drouvot.
This function is declared as returning a uint8, but it returns a
bool in one code path. To fix, return (uint8) 0 instead of false
there. This should behave exactly the same as before, but it might
prevent future compiler complaints.
Amit Kapila [Thu, 24 Jul 2025 09:05:32 +0000 (09:05 +0000)]
Fix duplicate transaction replay during pg_createsubscriber.
Previously, the tool could replay the same transaction twice, once during
recovery, then again during replication after the subscriber was set up.
This occurred because the same recovery_target_lsn was used both to
finalize recovery and to start replication. If
recovery_target_inclusive = true, the transaction at that LSN would be
applied during recovery and then sent again by the publisher leading to
duplication.
To prevent this, we now set recovery_target_inclusive = false. This
ensures the transaction at recovery_target_lsn is not reapplied during
recovery, avoiding duplication when replication begins.
Bug #18897 Reported-by: Zane Duffield <duffieldzane@gmail.com>
Author: Shlok Kyal <shlok.kyal.oss@gmail.com> Reviewed-by: vignesh C <vignesh21@gmail.com> Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Backpatch-through: 17, where it was introduced
Discussion: https://postgr.es/m/18897-d3db67535860dddb@postgresql.org
Michael Paquier [Thu, 24 Jul 2025 06:41:18 +0000 (15:41 +0900)]
Introduce field tracking cached plan type in PlannedStmt
PlannedStmt gains a new field, called CachedPlanType, able to track if a
given plan tree originates from the cache and if we are dealing with a
generic or custom cached plan.
This field can be used for monitoring or statistical purposes, in the
executor hooks, for example, based on the planned statement attached to
a QueryDesc. A patch is under discussion for pg_stat_statements to
provide an equivalent of the counters in pg_prepared_statements for
custom and generic plans, to provide a more global view of such data, as
this data is now restricted to the current session.
The concept introduced in this commit is useful on its own, and has been
extracted from a larger patch by the same author.
Author: Sami Imseih <samimseih@gmail.com> Reviewed-by: Andrei Lepikhov <lepihov@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CAA5RZ0uFw8Y9GCFvafhC=OA8NnMqVZyzXPfv_EePOt+iv1T-qQ@mail.gmail.com
Ensure the test waits for the apply worker to exit after disabling the
subscription. This is necessary to safely enable the retain_dead_tuples
option. Also added a similar wait in another part of the test to prevent
unintended apply worker activity that could lead to test failures
post-subscription disable.
Tom Lane [Wed, 23 Jul 2025 19:44:29 +0000 (15:44 -0400)]
Fix build breakage on Solaris-alikes with late-model GCC.
Solaris has never bothered to add "const" to the second argument
of PAM conversation procs, as all other Unixen did decades ago.
This resulted in an "incompatible pointer" compiler warning when
building --with-pam, but had no more serious effect than that,
so we never did anything about it. However, as of GCC 14 the
case is an error not warning by default.
To complicate matters, recent OpenIndiana (and maybe illumos
in general?) *does* supply the "const" by default, so we can't
just assume that platforms using our solaris template need help.
What we can do, short of building a configure-time probe,
is to make solaris.h #define _PAM_LEGACY_NONCONST, which
causes OpenIndiana's pam_appl.h to revert to the traditional
definition, and hopefully will have no effect anywhere else.
Then we can use that same symbol to control whether we include
"const" in the declaration of pam_passwd_conv_proc().
Bug: #18995 Reported-by: Andrew Watkins <awatkins1966@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/18995-82058da9ab4337a7@postgresql.org
Backpatch-through: 13
lwlock.c, lwlock.h, and wait_event_names.txt each contain a list of
built-in LWLock tranches. It is easy to miss one or the other when
adding or removing tranches, and discrepancies have adverse effects
(e.g., breaking JOINs between pg_stat_activity and pg_wait_events).
This commit moves the lists of built-in tranches in lwlock.{c,h} to
lwlocklist.h and adds a cross-check to the script that generates
lwlocknames.h. If the lists do not match exactly, building will
fail.
Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/aHpOgwuFQfcFMZ/B%40ip-10-97-1-34.eu-west-3.compute.internal
Commit 70e81861fadd split out xlogrecovery.c/h and moved some enums
related to recovery targets to xlogrecovery.h. However, it seems that
the enum RecoveryTargetAction was inadvertently left out by that commit.
This commit moves it to xlogrecovery.h for consistency.
Amit Kapila [Wed, 23 Jul 2025 02:56:00 +0000 (02:56 +0000)]
Preserve conflict-relevant data during logical replication.
Logical replication requires reliable conflict detection to maintain data
consistency across nodes. To achieve this, we must prevent premature
removal of tuples deleted by other origins and their associated commit_ts
data by VACUUM, which could otherwise lead to incorrect conflict reporting
and resolution.
This patch introduces a mechanism to retain deleted tuples on the
subscriber during the application of concurrent transactions from remote
nodes. Retaining these tuples allows us to correctly ignore concurrent
updates to the same tuple. Without this, an UPDATE might be misinterpreted
as an INSERT during resolutions due to the absence of the original tuple.
Additionally, we ensure that origin metadata is not prematurely removed by
vacuum freeze, which is essential for detecting update_origin_differs and
delete_origin_differs conflicts.
To support this, a new replication slot named pg_conflict_detection is
created and maintained by the launcher on the subscriber. Each apply
worker tracks its own non-removable transaction ID, which the launcher
aggregates to determine the appropriate xmin for the slot, thereby
retaining necessary tuples.
Conflict information retention (deleted tuples and commit_ts) can be
enabled per subscription via the retain_conflict_info option. This is
disabled by default to avoid unnecessary overhead for configurations that
do not require conflict resolution or logging.
During upgrades, if any subscription on the old cluster has
retain_conflict_info enabled, a conflict detection slot will be created to
protect relevant tuples from deletion when the new cluster starts.
This is a foundational work to correctly detect update_deleted conflict
which will be done in a follow-up patch.
David Rowley [Wed, 23 Jul 2025 00:02:55 +0000 (12:02 +1200)]
Use strchr instead of strstr for single-char lookups
Compilers such as gcc and clang seem to perform this rewrite
automatically when the lookup string is known at compile-time to contain
a single character. The MSVC compiler does not seem apply the same
optimization, and the code being adjusted here is within an #ifdef WIN32,
so it seems worth adjusting this with the assumption that strchr() will be
slightly more performant.
There are a couple more instances in contrib/fuzzystrmatch that this
commit could also have adjusted. After some discussion, we deemed those
not important enough to bother with.
Michael Paquier [Tue, 22 Jul 2025 23:18:36 +0000 (08:18 +0900)]
ecpg: Improve error detection around ecpg_strdup()
Various code paths of the ECPG code did not check for memory allocation
failures, including the specific case where ecpg_strdup() considers a
NULL value given in input as a valid behavior. strdup() returning
itself NULL on failure, there was no way to make the difference between
what could be valid and what should fail.
With the different cases in mind, ecpg_strdup() is redesigned and gains
a new optional argument, giving its callers the possibility to
differentiate allocation failures and valid cases where the caller is
giving a NULL value in input. Most of the ECPG code does not expect a
NULL value, at the exception of ECPGget_desc() (setlocale) and
ECPGconnect(), like dbname being unspecified, with repeated strdup
calls.
The code is adapted to work with this new routine. Note the case of
ecpg_auto_prepare(), where the code order is switched so as we handle
failures with ecpg_strdup() before manipulating any cached data,
avoiding inconsistencies.
This class of failure is unlikely a problem in practice, so no backpatch
is done. Random OOM failures in ECPGconnect() could cause the driver to
connect to a different server than the one wanted by the caller, because
it could fallback to default values instead of the parameters defined
depending on the combinations of allocation failures and successes.
Author: Evgeniy Gorbanev <gorbanyoves@basealt.ru> Co-authored-by: Aleksander Alekseev <aleksander@tigerdata.com> Reviewed-by: Álvaro Herrera <alvherre@kurilemu.de> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/a6b193c1-6994-4d9c-9059-aca4aaf41ddd@basealt.ru
Remove translation marker from libpq-be-fe-helpers.h.
Commit 112faf1378e introduced a translation marker in libpq-be-fe-helpers.h,
but this caused build failures on some platforms—such as the one reported
by buildfarm member indri—due to linker issues with dblink. This is the same
problem previously addressed in commit 213c959a294.
To fix the issue, this commit removes the translation marker from
libpq-be-fe-helpers.h, following the approach used in 213c959a294.
It also removes the associated gettext_noop() calls added in commit 112faf1378e, as they are no longer needed.
While reviewing this, a gettext_noop() call was also found in
contrib/basic_archive. Since contrib modules don't support translation,
this call has been removed as well.
Andres Freund [Tue, 22 Jul 2025 12:30:52 +0000 (08:30 -0400)]
aio: Fix assertion, clarify README
The assertion wouldn't have triggered for a long while yet, but this won't
accidentally fail to detect the issue if/when it occurs.
Author: Matthias van de Meent <boekewurm+postgres@gmail.com>
Discussion: https://postgr.es/m/CAEze2Wj-43JV4YufW23gm=Uwr7Lkj+p0yKctKHxNm1rwFC+_DQ@mail.gmail.com
Backpatch-through: 18
Amit Kapila [Tue, 22 Jul 2025 06:00:21 +0000 (06:00 +0000)]
Doc: Fix logical replication examples.
The definition of \dRp+ was modified in commit 7054186c4e. This patch
updates the column list and row filter examples to align with the revised
definition.
Author: Shlok Kyal <shlok.kyal.oss@gmail.com>
Reviewed by: Peter Smith <smithpb2250@gmail.com>
Backpatch-through: 18, where it was introduced
Discussion: https://postgr.es/m/CANhcyEUvqkSO6b9zi_fs_BBPEge5acj4mf8QKmq2TX-7axa7EQ@mail.gmail.com
Michael Paquier [Tue, 22 Jul 2025 05:34:15 +0000 (14:34 +0900)]
doc: Inform about aminsertcleanup optional NULLness
This index AM callback has been introduced in c1ec02be1d79 and it is
optional, currently only being used by BRIN. Optional callbacks are
documented with NULL as possible value in amapi.h and indexam.sgml, but
this callback has missed this part of the description.
Reported-by: Peter Smith <smithpb2250@gmail.com> Reviewed-by: Japin Li <japinli@hotmail.com>
Discussion: https://postgr.es/m/CAHut+PvgYcPmPDi1YdHMJY5upnyGRpc0N8pk1xNB11xDSBwNog@mail.gmail.com
Backpatch-through: 17
Log remote NOTICE, WARNING, and similar messages using ereport().
Previously, NOTICE, WARNING, and similar messages received from remote
servers over replication, postgres_fdw, or dblink connections were printed
directly to stderr on the local server (e.g., the subscriber). As a result,
these messages lacked log prefixes (e.g., timestamp), making them harder
to trace and correlate with other log entries.
This commit addresses the issue by introducing a custom notice receiver
for replication, postgres_fdw, and dblink connections. These messages
are now logged via ereport(), ensuring they appear in the logs with proper
formatting and context, which improves clarity and aids in debugging.
Michael Paquier [Tue, 22 Jul 2025 05:00:00 +0000 (14:00 +0900)]
ecpg: Fix NULL pointer dereference during connection lookup
ECPGconnect() caches established connections to the server, supporting
the case of a NULL connection name when a database name is not specified
by its caller.
A follow-up call to ECPGget_PGconn() to get an established connection
from the cached set with a non-NULL name could cause a NULL pointer
dereference if a NULL connection was listed in the cache and checked for
a match. At least two connections are necessary to reproduce the issue:
one with a NULL name and one with a non-NULL name.
Author: Aleksander Alekseev <aleksander@tigerdata.com>
Discussion: https://postgr.es/m/CAJ7c6TNvFTPUTZQuNAoqgzaSGz-iM4XR61D7vEj5PsQXwg2RyA@mail.gmail.com
Backpatch-through: 13
Richard Guo [Tue, 22 Jul 2025 02:21:36 +0000 (11:21 +0900)]
Reduce "Var IS [NOT] NULL" quals during constant folding
In commit b262ad440, we introduced an optimization that reduces an IS
[NOT] NULL qual on a NOT NULL column to constant true or constant
false, provided we can prove that the input expression of the NullTest
is not nullable by any outer joins or grouping sets. This deduction
happens quite late in the planner, during the distribution of quals to
rels in query_planner. However, this approach has some drawbacks: we
can't perform any further folding with the constant, and it turns out
to be prone to bugs.
Ideally, this deduction should happen during constant folding.
However, the per-relation information about which columns are defined
as NOT NULL is not available at that point. This information is
currently collected from catalogs when building RelOptInfos for base
or "other" relations.
This patch moves the collection of NOT NULL attribute information for
relations before pull_up_sublinks, storing it in a hash table keyed by
relation OID. It then uses this information to perform the NullTest
deduction for Vars during constant folding. This also makes it
possible to leverage this information to pull up NOT IN subqueries.
Note that this patch does not get rid of restriction_is_always_true
and restriction_is_always_false. Removing them would prevent us from
reducing some IS [NOT] NULL quals that we were previously able to
reduce, because (a) the self-join elimination may introduce new IS NOT
NULL quals after constant folding, and (b) if some outer joins are
converted to inner joins, previously irreducible NullTest quals may
become reducible.
Author: Richard Guo <guofenglinux@gmail.com> Reviewed-by: Robert Haas <robertmhaas@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAMbWs4-bFJ1At4btk5wqbezdu8PLtQ3zv-aiaY3ry9Ymm=jgFQ@mail.gmail.com
Richard Guo [Tue, 22 Jul 2025 02:20:40 +0000 (11:20 +0900)]
Centralize collection of catalog info needed early in the planner
There are several pieces of catalog information that need to be
retrieved for a relation during the early stage of planning. These
include relhassubclass, which is used to clear the inh flag if the
relation has no children, as well as a column's attgenerated and
default value, which are needed to expand virtual generated columns.
More such information may be required in the future.
Currently, these pieces of catalog data are collected in multiple
places, resulting in repeated table_open/table_close calls for each
relation in the rangetable. This patch centralizes the collection of
all required early-stage catalog information into a single loop over
the rangetable, allowing each relation to be opened and closed only
once.
Author: Richard Guo <guofenglinux@gmail.com> Reviewed-by: Robert Haas <robertmhaas@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAMbWs4-bFJ1At4btk5wqbezdu8PLtQ3zv-aiaY3ry9Ymm=jgFQ@mail.gmail.com
Richard Guo [Tue, 22 Jul 2025 02:19:17 +0000 (11:19 +0900)]
Expand virtual generated columns before sublink pull-up
Currently, we expand virtual generated columns after we have pulled up
any SubLinks within the query's quals. This ensures that the virtual
generated column references within SubLinks that should be transformed
into joins are correctly expanded. This approach works well and has
posed no issues.
In an upcoming patch, we plan to centralize the collection of catalog
information needed early in the planner. This will help avoid
repeated table_open/table_close calls for relations in the rangetable.
Since this information is required during sublink pull-up, we are
moving the expansion of virtual generated columns to occur beforehand.
To achieve this, if any EXISTS SubLinks can be pulled up, their
rangetables are processed just before pulling them up.
Author: Richard Guo <guofenglinux@gmail.com> Reviewed-by: Robert Haas <robertmhaas@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAMbWs4-bFJ1At4btk5wqbezdu8PLtQ3zv-aiaY3ry9Ymm=jgFQ@mail.gmail.com
pg_dump: include comments on not-null constraints on domains, too
Commit e5da0fe3c22b introduced catalog entries for not-null constraints
on domains; but because commit b0e96f311985 (the original work for
catalogued not-null constraints on tables) forgot to teach pg_dump to
process the comments for them, this one also forgot. Add that now.
We also need to teach repairDependencyLoop() about the new type of
constraints being possible for domains.
Backpatch-through: 17 Co-authored-by: jian he <jian.universality@gmail.com> Co-authored-by: Álvaro Herrera <alvherre@kurilemu.de> Reported-by: jian he <jian.universality@gmail.com>
Discussion: https://postgr.es/m/CACJufxF-0bqVR=j4jonS6N2Ka6hHUpFyu3_3TWKNhOW_4yFSSg@mail.gmail.com
doc: Document reopen of output file via SIGHUP in pg_recvlogical.
When pg_recvlogical receives a SIGHUP signal, it closes the current
output file and reopens a new one. This is useful since it allows us to
rotate the output file by renaming the current file and sending a SIGHUP.
This behavior was previously undocumented. This commit adds
the missing documentation.
Tom Lane [Sat, 19 Jul 2025 18:23:02 +0000 (14:23 -0400)]
Mostly-cosmetic adjustments to estimate_multivariate_bucketsize().
The only practical effect of these changes is to avoid a useless
list_copy() operation when there is a single hashclause. That's
never going to make any noticeable performance difference, but
the code is arguably clearer this way, especially if we take the
opportunity to add some comments so that readers don't have to
reverse-engineer the usage of these local variables. Also add
some braces for better/more consistent style.
Author: Tender Wang <tndrwang@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAHewXNnHBOO9NEa=NBDYOrwZL4oHu2NOcTYvqyNyWEswo8f5OQ@mail.gmail.com
This commit is only for HEAD and v18, where the test has been removed.
It also incorporates improvements below to stability and coverage of the
original test, which were already backpatched to v17.
- Add one pg_logical_emit_message() call to force the creation of a record
that spawns across two pages.
- Make the logic wait for the checkpoint completion.
Author: Alexander Korotkov <akorotkov@postgresql.org> Co-authored-by: Hayato Kuroda <kuroda.hayato@fujitsu.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Backpatch-through: 18
Improve the stability of the recovery test 047_checkpoint_physical_slot
Currently, the comments in 047_checkpoint_physical_slot. It shows an
incomplete intention to wait for checkpoint completion before performing
an immediate database stop. However, an immediate node stop can occur both
before and after checkpoint completion. Both cases should work correctly.
But we would like the test to be more stable and deterministic. This is why
this commit makes this test explicitly wait for the checkpoint completion
log message.
Discussion: https://postgr.es/m/CAPpHfdurV-j_e0pb%3DUFENAy3tyzxfF%2ByHveNDNQk2gM82WBU5A%40mail.gmail.com
Discussion: https://postgr.es/m/aHXLep3OaX_vRTNQ%40paquier.xyz
Author: Alexander Korotkov <akorotkov@postgresql.org> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Backpatch-through: 17
Fix infinite wait when reading a partially written WAL record
If a crash occurs while writing a WAL record that spans multiple pages, the
recovery process marks the page with the XLP_FIRST_IS_OVERWRITE_CONTRECORD
flag. However, logical decoding currently attempts to read the full WAL
record based on its expected size before checking this flag, which can lead
to an infinite wait if the remaining data is never written (e.g., no activity
after crash).
This patch updates the logic first to read the page header and check for
the XLP_FIRST_IS_OVERWRITE_CONTRECORD flag before attempting to reconstruct
the full WAL record. If the flag is set, decoding correctly identifies
the record as incomplete and avoids waiting for WAL data that will never
arrive.
Discussion: https://postgr.es/m/CAAKRu_ZCOzQpEumLFgG_%2Biw3FTa%2BhJ4SRpxzaQBYxxM_ZAzWcA%40mail.gmail.com
Discussion: https://postgr.es/m/CALDaNm34m36PDHzsU_GdcNXU0gLTfFY5rzh9GSQv%3Dw6B%2BQVNRQ%40mail.gmail.com
Author: Vignesh C <vignesh21@gmail.com> Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com> Reviewed-by: Dilip Kumar <dilipbalaut@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Backpatch-through: 13
Michael Paquier [Sat, 19 Jul 2025 06:03:14 +0000 (15:03 +0900)]
Check status of nodes after regression test run in 027_stream_regress
This commit improves the recovery TAP test 027_stream_regress so as
regression diffs are printed only if both the primary and the standby
are still alive after the main regression test suite finishes, relying
on d4c9195eff41 to do the job.
Particularly, a crash of the primary could scribble the contents
reported with mostly useless data, as the diffs would refer to query
that failed to run, not necessarily the cause of the crash.
Suggested-by: Andres Freund <andres@anarazel.de>
Author: Nazir Bilal Yavuz <byavuz81@gmail.com>
Discussion: https://postgr.es/m/CAN55FZ1D6KXvjSs7YGsDeadqCxNF3UUhjRAfforzzP0k-cE=bA@mail.gmail.com
Michael Paquier [Sat, 19 Jul 2025 05:38:52 +0000 (14:38 +0900)]
Add PostgreSQL::Test::Cluster::is_alive()
This new routine acts as a wrapper of pg_isready, that can be run on a
node to check its connection status. This will be used in a recovery
test in a follow-up commit.
Suggested-by: Andres Freund <andres@anarazel.de>
Author: Nazir Bilal Yavuz <byavuz81@gmail.com>
Discussion: https://postgr.es/m/CAN55FZ1D6KXvjSs7YGsDeadqCxNF3UUhjRAfforzzP0k-cE=bA@mail.gmail.com
Tom Lane [Fri, 18 Jul 2025 20:42:02 +0000 (16:42 -0400)]
Speed up byteain by not parsing traditional-style input twice.
Instead of laboriously computing the exact output length, use strlen
to get an upper bound cheaply. (This is still O(N) of course, but
the constant factor is a lot less.) This will typically result in
overallocating the output datum, but that's of little concern since
it's a short-lived allocation in just about all use-cases.
A simple microbenchmark showed about 40% speedup for long input
strings.
While here, make some cosmetic cleanups and add a test case that
covers the double-backslash code path in byteain and byteaout.
Author: Steven Niu <niushiji@gmail.com> Reviewed-by: Kirill Reshke <reshkekirill@gmail.com> Reviewed-by: Stepan Neretin <slpmcf@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/ca315729-140b-426e-81a6-6cd5cfe7ecc5@gmail.com
SELECT pg_catalog.lo_create('5432');
ALTER LARGE OBJECT 5432 OWNER TO alice;
GRANT SELECT ON LARGE OBJECT 5432 TO bob;
for each large object. This is particularly slow at restore time,
especially when there are tens or hundreds of millions of large
objects. From reports and personal experience, such slow restores
seem to be most painful when encountered during pg_upgrade. This
commit teaches pg_dump to instead dump pg_largeobject_metadata and
the corresponding pg_shdepend rows when in binary upgrade mode,
i.e., pg_dump now generates commands like
COPY pg_catalog.pg_shdepend (dbid, classid, objid, objsubid, refclassid, refobjid, deptype) FROM stdin;
5 2613 5432 0 1260 16384 o
5 2613 5432 0 1260 16385 a
\.
Testing indicates the COPY approach can be significantly faster.
To do any better, we'd probably need to find a way to copy/link
pg_largeobject_metadata's files during pg_upgrade, which would be
limited to upgrades from >= v16 (since commit 7b378237aa changed
the storage format for aclitem, which is used for
pg_largeobject_metadata.lomacl).
Note that this change only applies to binary upgrade mode (i.e.,
dumps initiated by pg_upgrade) since it inserts rows directly into
catalogs. Also, this optimization can only be used for upgrades
from >= v12 because pg_largeobject_metadata was created WITH OIDS
in older versions, which prevents pg_dump from handling
pg_largeobject_metadata.oid properly. With some extra effort, it
might be possible to support upgrades from older versions, but the
added complexity didn't seem worth it to support versions that will
have been out-of-support for nearly 3 years by the time this change
is released.
Experienced hackers may remember that prior to v12, pg_upgrade
copied/linked pg_largeobject_metadata's files (see commit 12a53c732c). Besides the aforementioned storage format issues,
this approach failed to transfer the relevant pg_shdepend rows, and
pg_dump still had to generate an lo_create() command per large
object so that creating the dependent comments and security labels
worked. We could perhaps adopt a hybrid approach for upgrades from
v16 and newer (i.e., generate lo_create() commands for each large
object, copy/link pg_largeobject_metadata's files, and COPY the
relevant pg_shdepend rows), but further testing is needed.
Reported-by: Hannu Krosing <hannuk@google.com> Suggested-by: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Hannu Krosing <hannuk@google.com> Reviewed-by: Nitin Motiani <nitinmotiani@google.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAMT0RQSS-6qLH%2BzYsOeUbAYhop3wmQTkNmQpo5--QRDUR%2BqYmQ%40mail.gmail.com
Dean Rasheed [Fri, 18 Jul 2025 08:55:43 +0000 (09:55 +0100)]
Fix concurrent update trigger issues with MERGE in a CTE.
If a MERGE inside a CTE attempts an UPDATE or DELETE on a table with
BEFORE ROW triggers, and a concurrent UPDATE or DELETE happens, the
merge code would fail (crashing in the case of an UPDATE action, and
potentially executing the wrong action for a DELETE action).
This is the same issue that 9321c79c86 attempted to fix, except now
for a MERGE inside a CTE. As noted in 9321c79c86, what needs to happen
is for the trigger code to exit early, returning the TM_Result and
TM_FailureData information to the merge code, if a concurrent
modification is detected, rather than attempting to do an EPQ
recheck. The merge code will then do its own rechecking, and rescan
the action list, potentially executing a different action in light of
the concurrent update. In particular, the trigger code must never call
ExecGetUpdateNewTuple() for MERGE, since that is bound to fail because
MERGE has its own per-action projection information.
Commit 9321c79c86 did this using estate->es_plannedstmt->commandType
in the trigger code to detect that a MERGE was being executed, which
is fine for a plain MERGE command, but does not work for a MERGE
inside a CTE. Fix by passing that information to the trigger code as
an additional parameter passed to ExecBRUpdateTriggers() and
ExecBRDeleteTriggers().
Back-patch as far as v17 only, since MERGE cannot appear inside a CTE
prior to that. Additionally, take care to preserve the trigger ABI in
v17 (though not in v18, which is still in beta).
Bug: #18986 Reported-by: Yaroslav Syrytsia <me@ys.lc>
Author: Dean Rasheed <dean.a.rasheed@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/18986-e7a8aac3d339fa47@postgresql.org
Backpatch-through: 17
Support for deparsing of ArrayCoerceExpr node in contrib/postgres_fdw
When using a prepared statement to select data from a PostgreSQL foreign
table (postgres_fdw) with the "field = ANY($1)" expression, the operation
is not pushed down when an implicit type case is applied, and a generic plan
is used. This commit resolves the issue by supporting the push-down of
ArrayCoerceExpr, which is used in this case. The support is quite
straightforward and similar to other nods, such as RelabelType.
Discussion: https://postgr.es/m/4f0cea802476d23c6e799512ffd17aff%40postgrespro.ru
Author: Alexander Pyhalov <a.pyhalov@postgrespro.ru> Reviewed-by: Maxim Orlov <orlovmg@gmail.com> Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
Tom Lane [Thu, 17 Jul 2025 16:46:38 +0000 (12:46 -0400)]
Fix PQport to never return NULL unless the connection is NULL.
This is the documented behavior, and it worked that way before
v10. However, addition of the connhost[] array created cases
where conn->connhost[conn->whichhost].port is NULL. The rest
of libpq is careful to substitute DEF_PGPORT[_STR] for a null
or empty port string, but we failed to do so here, leading to
possibly returning NULL. As of v18 that causes psql's \conninfo
command to segfault. Older psql versions avoid that, but it's
pretty likely that other clients have trouble with this,
so we'd better back-patch the fix.
In stable branches, just revert to our historical behavior of
returning an empty string when there was no user-given port
specification. However, it seems substantially more useful and
indeed more correct to hand back DEF_PGPORT_STR in such cases,
so let's make v18 and master do that.
Author: Daniele Varrazzo <daniele.varrazzo@gmail.com> Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CA+mi_8YTS8WPZPO0PAb2aaGLwHuQ0DEQRF0ZMnvWss4y9FwDYQ@mail.gmail.com
Backpatch-through: 13
We have an assertion to ensure that a command tag has been assigned by
the time we're done executing, but if we happen to execute a command
with no queries, the assertion would fail. Per discussion, rather than
contort things to get a tag assigned, just remove the assertion.
Oversight in 2f9661311b83. That commit also retained a comment that
explained logic that had been adjacent to it but diffused into various
places, leaving none apt to keep part of the comment. Remove that part,
and rewrite what remains for extra clarity.
Bug: #18984
Backpatch-through: 13 Reported-by: Aleksander Alekseev <aleksander@tigerdata.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Michaël Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/18984-0f4778a6599ac3ae@postgresql.org
This commit adds a note to the pg_overexplain page that describes
how to use it (LOAD, session_preload_libraries, or
shared_preload_libraries). The new text is mostly lifted from the
auto_explain page. We should probably consider centralizing this
information in the future.
While at it, add a missing "module" to the opening sentence.
Reviewed-by: "David G. Johnston" <david.g.johnston@gmail.com> Reviewed-by: Robert Treat <rob@xzilla.net> Reviewed-by: Dean Rasheed <dean.a.rasheed@gmail.com>
Discussion: https://postgr.es/m/aHVWKM8l8kLlZzgv%40nathan
Backpatch-through: 18
Michael Paquier [Thu, 17 Jul 2025 05:08:55 +0000 (14:08 +0900)]
Split regression tests for TOAST compression methods into two files
The regression tests for TOAST compression methods are split into two
independent files: one specific to LZ4 and interactions between two
different TOAST compression methods, now called compression_lz4, and a
second one for the "core" cases where only pglz is required.
This saves 300 lines in diffs coming from the alternate output of
compression.sql, required for builds where lz4 is not available. The
new test is skipped if the build does not support LZ4 compression,
relying on an \if and the values reported in pg_settings for the GUC
default_toast_compression, "lz4" being available only under USE_LZ4.
Another benefit of this split is that this facilitates the addition of
more compression methods for TOAST, which are under discussion.
Note the trick added for the tests of the GUC default_toast_compression,
where VERBOSITY = terse is used to avoid the HINT printing the lists of
values available in the GUC, which are environment-dependent. This
makes compression.sql independent of the availability of LZ4.
The code coverage of toast_compression.c is slightly improved, increased
from 89% to 91%, with one new case covered in lz4_compress_datum() for
incompressible data.
Michael Paquier [Thu, 17 Jul 2025 00:30:26 +0000 (09:30 +0900)]
Fix inconsistent LWLock tranche names for MultiXact*
The terms used in wait_event_names.txt and lwlock.c were inconsistent
for MultiXactOffsetSLRU and MultiXactMemberSLRU, which could cause joins
between pg_wait_events and pg_stat_activity to fail. lwlock.c is
adjusted in this commit to what the historical name of the event has
always been, and what is documented.
The paragraph for introducing INSERT and COPY discussed how a file
could be used for bulk loading with COPY, without actually showing
what the file would look like. This adds a programlisting for the
file contents.
Backpatch to all supported branches since this example has lacked
the file contents since PostgreSQL 7.2.
Author: Daniel Gustafsson <daniel@yesql.se> Reviewed-by: Fujii Masao <masao.fujii@oss.nttdata.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/158017814191.19852.15019251381150731439@wrigleys.postgresql.org
Backpatch-through: 13
Jeff Davis [Wed, 16 Jul 2025 21:13:18 +0000 (14:13 -0700)]
Force LC_COLLATE to C in postmaster.
Avoid dependence on setlocale().
strcoll(), etc., are not called directly; all collation-sensitive
calls should go through pg_locale.c and use the appropriate
provider. By setting LC_COLLATE to C, we avoid accidentally depending
on libc behavior when using a different provider.
No behavior change in the backend, but it's possible that some
extensions will be affected. Such extensions should be updated to use
the pg_locale_t APIs.
Discussion: https://postgr.es/m/9875f7f9-50f1-4b5d-86fc-ee8b03e8c162@eisentraut.org Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Fix dumping of comments on invalid constraints on domains
We skip dumping constraints together with domains if they are invalid
('separate') so that they appear after data -- but their comments were
dumped together with the domain definition, which in effect leads to the
comment being dumped when the constraint does not yet exist. Delay
them in the same way.
Oversight in 7eca575d1c28; backpatch all the way back.
Author: jian he <jian.universality@gmail.com>
Discussion: https://postgr.es/m/CACJufxF_C2pe6J_+nPr6C5jf5rQnbYP8XOKr4HM8yHZtp2aQqQ@mail.gmail.com
Peter Geoghegan [Wed, 16 Jul 2025 17:05:44 +0000 (13:05 -0400)]
nbtree: Use only one notnullkey ScanKeyData.
_bt_first need only store one ScanKeyData struct on the stack for the
purposes of building an IS NOT NULL key based on an implied NOT NULL
constraint. We don't need INDEX_MAX_KEYS-many ScanKeyData structs.
This saves us a little over 2KB in stack space. It's possible that this
has some performance benefit. It also seems simpler and more direct.
It isn't possible for more than a single index attribute to need its own
implied IS NOT NULL key: the first such attribute/IS NOT NULL key always
makes _bt_first stop adding additional boundary keys to startKeys[].
Using INDEX_MAX_KEYS-many ScanKeyData entries was (at best) misleading.
Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Mircea Cadariu <cadariu.mircea@gmail.com>
Discussion: https://postgr.es/m/CAH2-Wzm=1kJMSZhhTLoM5BPbwQNWxUj-ynOEh=89ptDZAVgauw@mail.gmail.com
Jeff Davis [Wed, 16 Jul 2025 16:57:12 +0000 (09:57 -0700)]
pg_dumpall: Skip global objects with --statistics-only or --no-schema.
Previously, pg_dumpall would still dump global objects such as roles
and tablespaces even when --statistics-only or --no-schema was specified.
Since these global objects are treated as schema-level data, they should
be skipped in these cases.
This commit fixes the issue by ensuring that global objects are not
dumped when either --statistics-only or --no-schema is used.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Discussion: https://postgr.es/m/aHVo791guQR6uqwT%40nathan
Backpatch-through: 13
Michael Paquier [Wed, 16 Jul 2025 02:59:22 +0000 (11:59 +0900)]
Refactor non-supported compression error message in toast_compression.c
This code used a NO_LZ4_SUPPORT() macro to issue an error in the code
paths where LZ4 [de]compression is attempted but the build does not
support it. This commit refactors the code to use a more flexible error
message so as it can be used for other compression methods, where the
method is given in input of macro.
pgoutput: Initialize missing default for "origin" parameter.
The pgoutput plugin initializes optional parameters like "binary" with
default values at the start of processing. However, the "origin"
parameter was previously missed and left without explicit initialization.
Although the PGOutputData struct, which holds these settings,
is zero-initialized at allocation (resulting in publish_no_origin field
for "origin" parameter being false by default), this default was not
set explicitly, unlike other parameters.
This commit adds explicit initialization of the "origin" parameter to
ensure consistency and clarity in how defaults are handled.
doc: Document default values for pgoutput options in protocol.sgml.
The pgoutput plugin options are described in the logical streaming
replication protocol documentation, but their default values were
previously not mentioned. This made it less convenient for users,
for example, when specifying those options to use pg_recvlogical
with pgoutput plugin.
This commit adds the explanations of the default values for pgoutput
options to improve clarity and usability.
doc: Fix confusing description of streaming option in START_REPLICATION.
Previously, the documentation described the streaming option as a boolean,
which is outdated since it's no longer a boolean as of protocol version 4.
This could confuse users.
This commit updates the description to remove the "boolean" reference and
clearly list the valid values for the streaming option.
Back-patch to v16, where the streaming option changed to a non-boolean.
doc: Clarify that total_vacuum_time excludes VACUUM FULL.
The last_vacuum and vacuum_count fields in pg_stat_all_tables already
state that they do not include VACUUM FULL. However, total_vacuum_time,
which also excludes VACUUM FULL, did not mention this. This could
mislead users into thinking VACUUM FULL time is included.
To address this, this commit updates the documentation for
pg_stat_all_tables to explicitly state that total_vacuum_time does not
count VACUUM FULL.
Back-patched to v18, where total_vacuum_time was introduced.
Additionally, this commit clarifies that n_ins_since_vacuum also
excludes VACUUM FULL. Although n_ins_since_vacuum was added in v13,
we are not back-patching this change to stable branches, as it is
a documentation improvement, not a bug fix.
Author: Fujii Masao <masao.fujii@gmail.com> Reviewed-by: Robert Treat <rob@xzilla.net> Reviewed-by: David G. Johnston <david.g.johnston@gmail.com> Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at>
Discussion: https://postgr.es/m/2ac375d1-591b-4f1b-a2af-f24335567866@oss.nttdata.com
Backpatch-through: 18
Tom Lane [Tue, 15 Jul 2025 22:53:00 +0000 (18:53 -0400)]
Doc: clarify description of regexp fields in pg_ident.conf.
The grammar was a little shaky and confusing here, so word-smith it
a bit. Also, adjust the comments in pg_ident.conf.sample to use the
same terminology as the SGML docs, in particular "DATABASE-USERNAME"
not "PG-USERNAME".
Back-patch appropriate subsets. I did not risk changing
pg_ident.conf.sample in released branches, but it still seems OK
to change it in v18.
Reported-by: Alexey Shishkin <alexey.shishkin@enterprisedb.com>
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: David G. Johnston <david.g.johnston@gmail.com>
Discussion: https://postgr.es/m/175206279327.3157504.12519088928605422253@wrigleys.postgresql.org
Backpatch-through: 13
Tom Lane [Tue, 15 Jul 2025 22:21:12 +0000 (18:21 -0400)]
Clarify the ra != rb case in compareJsonbContainers().
It's impossible to reach this case with either ra or rb being
WJB_DONE, because our earlier checks that the structure and
length of the inputs match should guarantee that we reach their
ends simultaneously. However, the comment completely fails to
explain this, and the Asserts don't cover it either. The comment
is pretty obscure anyway, so rewrite it, and extend the Asserts
to reject WJB_DONE.
Tom Lane [Tue, 15 Jul 2025 22:11:18 +0000 (18:11 -0400)]
Silence uninitialized-value warnings in compareJsonbContainers().
Because not every path through JsonbIteratorNext() sets val->type,
some compilers complain that compareJsonbContainers() is comparing
possibly-uninitialized values. The paths that don't set it return
WJB_DONE, WJB_END_ARRAY, or WJB_END_OBJECT, so it's clear by
manual inspection that the "(ra == rb)" code path is safe, and
indeed we aren't seeing warnings about that. But the (ra != rb)
case is much less obviously safe. In Assert-enabled builds it
seems that the asserts rejecting WJB_END_ARRAY and WJB_END_OBJECT
persuade gcc 15.x not to warn, which makes little sense because
it's impossible to believe that the compiler can prove of its
own accord that ra/rb aren't WJB_DONE here. (In fact they never
will be, so the code isn't wrong, but why is there no warning?)
Without Asserts, the appearance of warnings is quite unsurprising.
We discussed fixing this by converting those two Asserts into
pg_assume, but that seems not very satisfactory when it's so unclear
why the compiler is or isn't warning: the warning could easily
reappear with some other compiler version. Let's fix it in a less
magical, more future-proof way by changing JsonbIteratorNext()
so that it always does set val->type. The cost of that should be
pretty negligible, and it makes the function's API spec less squishy.
Reported-by: Erik Rijkers <er@xs4all.nl>
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/988bf1bc-3f1f-99f3-bf98-222f1cd9dc5e@xs4all.nl
Discussion: https://postgr.es/m/0c623e8a204187b87b4736792398eaf1@postgrespro.ru
Backpatch-through: 13
Tom Lane [Tue, 15 Jul 2025 20:35:42 +0000 (16:35 -0400)]
Doc: clarify description of current-date/time functions.
Minor wordsmithing of the func.sgml paragraph describing
statement_timestamp() and allied functions: don't switch between
"statement" and "command" when those are being used to mean about
the same thing.
Also, add some text to protocol.sgml describing the perhaps-surprising
behavior these functions have in a multi-statement Query message.
Reported-by: P M <petermittere@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Laurenz Albe <laurenz.albe@cybertec.at> Reviewed-by: David G. Johnston <david.g.johnston@gmail.com>
Discussion: https://postgr.es/m/175223006802.3157505.14764328206246105568@wrigleys.postgresql.org
Backpatch-through: 13
psql: Fix tab-completion after GRANT/REVOKE on LARGE OBJECT and FOREIGN SERVER.
Previously, when pressing Tab after GRANT or REVOKE ... ON LARGE OBJECT
or ON FOREIGN SERVER, TO or FROM was incorrectly suggested by psql's
tab-completion. This was not appropriate, as those clauses are not valid
at that point.
This commit fixes the issue by preventing TO and FROM from being offered
immediately after those specific GRANT/REVOKE statements.
amcheck: Improve error message for partitioned index target.
Previously, amcheck could produce misleading error message when
a partitioned index was passed to functions like bt_index_check().
For example, bt_index_check() with a partitioned btree index produced:
ERROR: expected "btree" index as targets for verification
DETAIL: Relation ... is a btree index.
Reporting "expected btree index as targets" even when the specified
index was a btree was confusing. In this case, the function should fail
since the partitioned index specified is not valid target. This commit
improves the error reporting to better reflect this actual issue. Now,
bt_index_check() with a partitioned index, the error message is:
ERROR: expected index as targets for verification
DETAIL: This operation is not supported for partitioned indexes.
This commit also applies the following minor changes:
- Simplifies index_checkable() by using get_am_name() to retrieve
the access method name.
- Changes index_checkable() from extern to static, as it is only used
in verify_common.c.
- Updates the error code for invalid indexes to
ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE,
aligning with usage in similar modules like pgstattuple.
Michael Paquier [Mon, 14 Jul 2025 00:08:46 +0000 (09:08 +0900)]
psql: Add variable SERVICEFILE
This new psql variable can be used to check which service file has been
used for a connection. Like other variables, this can be set in a
PROMPT or reported by an \echo, like these commands:
\echo :SERVICEFILE
\set PROMPT1 '=(%:SERVICEFILE:)%# '
This relies on commits 092f3c63efc6 and fef6da9e9c87 to retrieve this
information from the connection's PQconninfoOption.
Tom Lane [Sun, 13 Jul 2025 17:52:32 +0000 (13:52 -0400)]
In username-map substitution, cope with more than one \1.
If the system-name field of a pg_ident.conf line is a regex
containing capturing parentheses, you can write \1 in the
user-name field to represent the captured part of the system
name. But what happens if you write \1 more than once?
The only reasonable expectation IMO is that each \1 gets
replaced, but presently our code replaces only the first.
Fix that.
Also, improve the tests for this feature to exercise cases
where a non-empty string needs to be substituted for \1.
The previous testing didn't inspire much faith that it
was verifying correct operation of the substitution code.
Given the lack of field complaints about this, I don't
feel a need to back-patch.
Reported-by: David G. Johnston <david.g.johnston@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAKFQuwZu6kZ8ZPvJ3pWXig+6UX4nTVK-hdL_ZS3fSdps=RJQQQ@mail.gmail.com
Michael Paquier [Sun, 13 Jul 2025 07:52:19 +0000 (16:52 +0900)]
libpq: Add "servicefile" connection option
This commit adds the possibility to specify a service file in a
connection string, using a new option called "servicefile". The parsing
of the service file happens so as things are done in this order of
priority:
- The servicefile connection option.
- Environment variable PGSERVICEFILE.
- Default path, depending on the HOME environment.
Note that in the last default case, we need to fill in "servicefile" for
the connection's PQconninfoOption to let clients know which service file
has been used for the connection. Some TAP tests are added, with a few
tweaks required for Windows when using URIs or connection option values,
for the location paths.
Tom Lane [Sat, 12 Jul 2025 15:50:35 +0000 (11:50 -0400)]
Replace float8 with int in date2isoweek() and date2isoyear().
The values of the "result" variables in these functions are
always integers; using a float8 variable accomplishes nothing
except to incur useless conversions to and from float. While
that wastes a few nanoseconds, these functions aren't all that
time-critical. But it seems worth fixing to remove possible
reader confusion.
Also, in the case of date2isoyear(), "result" is a very poorly
chosen variable name because it is *not* the function's result.
Rename it to "week", and do the same in date2isoweek() for
consistency.
Since this is mostly cosmetic, there seems little need
for back-patch.
Author: Sergey Fukanchik <s.fukanchik@postgrespro.ru> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/6323a-68726500-1-7def9d00@137821581
Andres Freund [Sat, 12 Jul 2025 15:00:44 +0000 (11:00 -0400)]
Remove long-unused TransactionIdIsActive()
TransactionIdIsActive() has not been used since bb38fb0d43c, in 2014. There
are no known uses in extensions either and it's hard to see valid uses for
it. Therefore remove TransactionIdIsActive().
Thomas Munro [Sat, 12 Jul 2025 04:20:11 +0000 (16:20 +1200)]
aio: Fix configuration reload in IO workers.
method_worker.c installed SignalHandlerForConfigReload, but it failed to
actually process reload requests. That hasn't yet produced any concrete
problem reports in terms of GUC changes it should have cared about in
v18, but it was inconsistent.
It did cause problems for a couple of patches in development that need
IO workers to react to ALTER SYSTEM + pg_reload_conf(). Fix extracted
from one of those patches.
Thomas Munro [Sat, 12 Jul 2025 01:47:59 +0000 (13:47 +1200)]
aio: Remove obsolete IO worker ID references.
In an ancient ancestor of this code, the postmaster assigned IDs to IO
workers. Now it tracks them in an unordered array and doesn't know
their IDs, so it might be confusing to readers that it still referred to
their indexes as IDs.
No change in behavior, just variable name and error message cleanup.
Thomas Munro [Sat, 12 Jul 2025 01:43:27 +0000 (13:43 +1200)]
aio: Regularize IO worker internal naming.
Adopt PgAioXXX convention for pgaio module type names. Rename a
function that didn't use a pgaio_worker_ submodule prefix. Rename the
internal submit function's arguments to match the indirectly relevant
function pointer declaration and nearby examples. Rename the array of
handle IDs in PgAioSubmissionQueue to sqes, a term of art seen in the
systems it emulates, also clarifying that they're not IO handle
pointers as the old name might imply.
No change in behavior, just type, variable and function name cleanup.
Thomas Munro [Fri, 11 Jul 2025 23:18:05 +0000 (11:18 +1200)]
Fix stale idle flag when IO workers exit.
Otherwise we could choose a worker that has exited and crash while
trying to wake it up.
Back-patch to 18.
Reported-by: Tomas Vondra <tomas@vondra.me> Reported-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/t5aqjhkj6xdkido535pds7fk5z4finoxra4zypefjqnlieevbg%40357aaf6u525j
projects / thirdparty / postgresql.git / log
