Melanie Plageman [Tue, 25 Nov 2025 21:39:56 +0000 (16:39 -0500)]
Assert that cutoffs are provided if freezing will be attempted
heap_page_prune_and_freeze() requires the caller to initialize
PruneFreezeParams->cutoffs so that the function can correctly evaluate
whether tuples should be frozen. This requirement previously existed
only in comments and was easy to miss, especially after “cutoffs” was
converted from a direct function parameter to a field of the newly
introduced PruneFreezeParams struct (added in 1937ed70621). Adding an
assert makes this requirement explicit and harder to violate.
Also, fix a minor typo while we're at it.
Author: Chao Li <li.evan.chao@gmail.com> Reviewed-by: Melanie Plageman <melanieplageman@gmail.com>
Discussion: https://postgr.es/m/0AC177F5-5E26-45EE-B273-357C51212AC5%40gmail.com
Álvaro Herrera [Tue, 25 Nov 2025 18:08:36 +0000 (19:08 +0100)]
pg_dump tests: don't put dumps in stdout
This bloats the regression log files for no reason.
Backpatch to 18; no further only because it fails to apply cleanly.
(It's just whitespace change that conflicts, but I don't think this
warrants more effort than this.)
Álvaro Herrera [Tue, 25 Nov 2025 17:20:06 +0000 (18:20 +0100)]
Improve test case stability
Given unlucky timing, some of the new tests added by commit bc32a12e0db2
can fail spuriously. We haven't seen such failures yet in buildfarm,
but allegedly we can prevent them with this tweak.
While at it, remove an unused injection point I (Álvaro) added.
Author: Mihail Nikalayeu <mihailnikalayeu@gmail.com>
Discussion: https://postgr.es/m/CADzfLwUc=jtSUEaQCtyt8zTeOJ-gHZ8=w_KJsVjDOYSLqaY9Lg@mail.gmail.com
Discussion: https://postgr.es/m/CADzfLwV5oQq-Vg_VmG_o4SdL6yHjDoNO4T4pMtgJLzYGmYf74g@mail.gmail.com
Peter Eisentraut [Tue, 25 Nov 2025 15:50:34 +0000 (16:50 +0100)]
gen_guc_tables.pl: Validate required GUC fields before code generation
Previously, gen_guc_tables.pl would emit "Use of uninitialized value"
warnings if required fields were missing in guc_parameters.dat (for
example, when an integer or real GUC omitted the 'max' value). The
resulting error messages were unclear and did not identify which GUC
entry was problematic.
Add explicit validation of required fields depending on the parameter
type, and fail with a clear and specific message such as:
guc_parameters.dat:1909: error: entry "max_index_keys" of type "int" is missing required field "max"
No changes to generated guc_tables.c.
Author: Chao Li <lic@highgo.com> Reviewed-by: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org> Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://www.postgresql.org/message-id/flat/CAEoWx2%3DoP4LgHi771_OKhPPUS7B-CTqCs%3D%3DuQcNXWrwBoAm5Vg%40mail.gmail.com
The issue occurred because the replication slot was not released in the
slotsync worker when a slot synchronization cycle was skipped. This skip
happened because the required WAL was not received and flushed on the
standby server. As a result, in the next cycle, when attempting to acquire
the slot, an assertion failure was triggered.
Amit Kapila [Tue, 25 Nov 2025 06:47:49 +0000 (06:47 +0000)]
Add slotsync skip statistics.
This patch adds two new columns to the pg_stat_replication_slots view:
slotsync_skip_count - the total number of times a slotsync operation was
skipped.
slotsync_skip_at - the timestamp of the most recent skip.
These additions provide better visibility into replication slot
synchronization behavior.
A future patch will introduce the slotsync_skip_reason column in
pg_replication_slots to capture the reason for skip.
Peter Eisentraut [Tue, 25 Nov 2025 05:26:49 +0000 (06:26 +0100)]
Remove obsolete comment
This comment should probably have been moved to pg_locale_libc.c in
commit 66ac94cdc79 (2024), but upon closer examination it was already
completely obsolete then.
The first part of the comment has been obsolete since commit 85feb77aa09 (2017), which required that the system provides the
required wide-character functions.
The second part has been obsolete since commit e9931bfb751 (2024),
which eliminated code paths depending on the global LC_CTYPE setting.
Andres Freund [Mon, 24 Nov 2025 22:37:09 +0000 (17:37 -0500)]
lwlock: Fix, currently harmless, bug in LWLockWakeup()
Accidentally the code in LWLockWakeup() checked the list of to-be-woken up
processes to see if LW_FLAG_HAS_WAITERS should be unset. That means that
HAS_WAITERS would not get unset immediately, but only during the next,
unnecessary, call to LWLockWakeup().
Luckily, as the code stands, this is just a small efficiency issue.
However, if there were (as in a patch of mine) a case in which LWLockWakeup()
would not find any backend to wake, despite the wait list not being empty,
we'd wrongly unset LW_FLAG_HAS_WAITERS, leading to potentially hanging.
While the consequences in the backbranches are limited, the code as-is
confusing, and it is possible that there are workloads where the additional
wait list lock acquisitions hurt, therefore backpatch.
Tom Lane [Mon, 24 Nov 2025 22:00:16 +0000 (17:00 -0500)]
Improve detection of implicitly-temporary views.
We've long had a practice of making views temporary by default if they
reference any temporary tables. However the implementation was pretty
incomplete, in that it only searched for RangeTblEntry references to
temp relations. Uses of temporary types, regclass constants, etc
were not detected even though the dependency mechanism considers them
grounds for dropping the view. Thus a view not believed to be temp
could silently go away at session exit anyhow.
To improve matters, replace the ad-hoc isQueryUsingTempRelation()
logic with use of the dependency-based infrastructure introduced by
commit 572c40ba9. This is complete by definition, and it's less code
overall.
While we're at it, we can also extend the warning NOTICE (or ERROR
in the case of a materialized view) to mention one of the temp
objects motivating the classification of the view as temp, as was
done for functions in 572c40ba9.
Author: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Jim Jones <jim.jones@uni-muenster.de>
Discussion: https://postgr.es/m/19cf6ae1-04cd-422c-a760-d7e75fe6cba9@uni-muenster.de
Jacob Champion [Mon, 24 Nov 2025 18:01:30 +0000 (10:01 -0800)]
Reorganize pqcomm.h a bit
Group the PG_PROTOCOL() codes, add a comment to AuthRequest now that the
AUTH_REQ codes live in a different header, and make some small
adjustments to spacing and comment style for the sake of scannability.
Reviewed-by: Jelte Fennema-Nio <postgres@jeltef.nl> Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://postgr.es/m/CAOYmi%2B%3D6zg4oXXOQtifrVao_YKiujTDa3u6bxnU08r0FsSig4g%40mail.gmail.com
Jacob Champion [Mon, 24 Nov 2025 17:59:54 +0000 (09:59 -0800)]
postgres: Use pg_{add,mul}_size_overflow()
The backend implementations of add_size() and mul_size() can now make
use of the APIs provided in common/int.h.
Reviewed-by: Chao Li <li.evan.chao@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CAOYmi%2B%3D%2BpqUd2MUitvgW1pAJuXgG_TKCVc3_Ek7pe8z9nkf%2BAg%40mail.gmail.com
Jacob Champion [Mon, 24 Nov 2025 17:59:38 +0000 (09:59 -0800)]
Add pg_add_size_overflow() and friends
Commit 600086f47 added (several bespoke copies of) size_t addition with
overflow checks to libpq. Move this to common/int.h, along with
its subtraction and multiplication counterparts.
pg_neg_size_overflow() is intentionally omitted; I'm not sure we should
add SSIZE_MAX to win32_port.h for the sake of a function with no
callers.
Reviewed-by: Chao Li <li.evan.chao@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CAOYmi%2B%3D%2BpqUd2MUitvgW1pAJuXgG_TKCVc3_Ek7pe8z9nkf%2BAg%40mail.gmail.com
Álvaro Herrera [Mon, 24 Nov 2025 16:03:10 +0000 (17:03 +0100)]
Fix infer_arbiter_index during concurrent index operations
Previously, we would only consider indexes marked indisvalid as usable
for INSERT ON CONFLICT. But that's problematic during CREATE INDEX
CONCURRENTLY and REINDEX CONCURRENTLY, because concurrent transactions
would end up with inconsistents lists of inferred indexes, leading to
deadlocks and spurious errors about unique key violations (because two
transactions are operating on different indexes for the speculative
insertion tokens). Change this function to return indexes even if
invalid. This fixes the spurious errors and deadlocks.
Because such indexes might not be complete, we still need uniqueness to
be verified in a different way. We do that by requiring that at least
one index marked valid is part of the set of indexes returned. It is
that index that is going to help ensure that the inserted tuple is
indeed unique.
This does not fix similar problems occurring with partitioned tables or
with named constraints. These problems will be fixed in follow-up
commits.
We have no user report of this problem, even though it exists in all
branches. Because of that and given that the fix is somewhat tricky, I
decided not to backpatch for now.
Author: Mihail Nikalayeu <mihailnikalayeu@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz> Reviewed-by: Álvaro Herrera <alvherre@kurilemu.de>
Discussion: https://postgr.es/m/CANtu0ogv+6wqRzPK241jik4U95s1pW3MCZ3rX5ZqbFdUysz7Qw@mail.gmail.com
Michael Paquier [Mon, 24 Nov 2025 10:33:51 +0000 (19:33 +0900)]
Move isolation test index-killtuples to src/test/modules/index/
index-killtuples test depends on the contrib modules btree_gin and
btree_gist, which would not be installed in a temporary installation
with an execution of the main isolation test suite like this one:
make -C src/test/isolation/ check
src/test/isolation/ should not depend on contrib/, and EXTRA_INSTALL has
no effect in this case as this test suite uses its own Makefile rules.
This commit moves index-killtuples into its new module, called "index",
whose name looks like the best fit there can be as it depends on more
than one index AM. btree_gin and btree_gist are now pulled in the
temporary installation with EXTRA_INSTALL. The test is renamed to
"killtuples", for simplicity.
Author: Nazir Bilal Yavuz <byavuz81@gmail.com> Suggested-by: Andres Freund <andres@anarazel.de> Suggested-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/aKJsWedftW7UX1WM@paquier.xyz
Peter Eisentraut [Mon, 24 Nov 2025 06:36:49 +0000 (07:36 +0100)]
C11 alignas instead of unions -- extended alignments
This replaces some uses of pg_attribute_aligned() with the standard
alignas() for cases where extended alignment (larger than max_align_t)
is required.
This patch stipulates that all supported compilers must support
alignments up to PG_IO_ALIGN_SIZE, but that seems pretty likely.
We can then also desupport the case where direct I/O is disabled
because pg_attribute_aligned is not supported.
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/46f05236-d4d4-4b4e-84d4-faa500f14691%40eisentraut.org
Michael Paquier [Mon, 24 Nov 2025 05:29:15 +0000 (14:29 +0900)]
pg_buffercache: Add pg_buffercache_os_pages
ba2a3c2302f has added a way to check if a buffer is spread across
multiple pages with some NUMA information, via a new view
pg_buffercache_numa that depends on pg_buffercache_numa_pages(), a SQL
function. These can only be queried when support for libnuma exists,
generating an error if not.
However, it can be useful to know how shared buffers and OS pages map
when NUMA is not supported or not available. This commit expands the
capabilities around pg_buffercache_numa:
- pg_buffercache_numa_pages() is refactored as an internal function,
able to optionally process NUMA. Its SQL definition prior to this
commit is still around to ensure backward-compatibility with v1.6.
- A SQL function called pg_buffercache_os_pages() is added, able to work
with or without NUMA.
- The view pg_buffercache_numa is redefined to use
pg_buffercache_os_pages().
- A new view is added, called pg_buffercache_os_pages. This ignores
NUMA for its result processing, for a better efficiency.
The implementation is done so as there is no code duplication between
the NUMA and non-NUMA views/functions, relying on one internal function
that does the job for all of them. The module is bumped to v1.7.
Author: Bertrand Drouvot <bertranddrouvot.pg@gmail.com> Reviewed-by: Mircea Cadariu <cadariu.mircea@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/Z/fFA2heH6lpSLlt@ip-10-97-1-34.eu-west-3.compute.internal
David Rowley [Mon, 24 Nov 2025 04:00:01 +0000 (17:00 +1300)]
Fix incorrect IndexOptInfo header comment
The comment incorrectly indicated that indexcollations[] stored
collations for both key columns and INCLUDE columns, but in reality it
only has elements for the key columns. canreturn[] didn't get a mention,
so add that while we're here.
Tom Lane [Sun, 23 Nov 2025 20:02:55 +0000 (15:02 -0500)]
Issue a NOTICE if a created function depends on any temp objects.
We don't have an official concept of temporary functions. (You can
make one explicitly in pg_temp, but then you have to explicitly
schema-qualify it on every call.) However, until now we were quite
laissez-faire about whether a non-temporary function could depend on
a temporary object, such as a temp table or view. If one does,
it will silently go away at end of session, due to the automatic
DROP ... CASCADE on the session's temporary objects. People have
complained that that's surprising; however, we can't really forbid
it because other people (including our own regression tests) rely
on being able to do it. Let's compromise by emitting a NOTICE
at CREATE FUNCTION time. This is somewhat comparable to our
ancient practice of emitting a NOTICE when forcing a view to
become temp because it depends on temp tables.
Along the way, refactor recordDependencyOnExpr() so that the
dependencies of an expression can be combined with other
dependencies, instead of being emitted separately and perhaps
duplicatively.
We should probably make the implementation of temp-by-default
views use the same infrastructure used here, but that's for
another patch. It's unclear whether there are any other object
classes that deserve similar treatment.
Author: Jim Jones <jim.jones@uni-muenster.de> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/19cf6ae1-04cd-422c-a760-d7e75fe6cba9@uni-muenster.de
Michael Paquier [Sun, 23 Nov 2025 04:37:42 +0000 (13:37 +0900)]
pg_buffercache: Remove unused fields from BufferCacheNumaRec
These fields have been added in commit ba2a3c2302f, and have never been
used. While on it, this commit moves a comment that was out of place,
improving it.
Tom Lane [Sun, 23 Nov 2025 00:33:34 +0000 (19:33 -0500)]
Add SupportRequestInlineInFrom planner support request.
This request allows a support function to replace a function call
appearing in FROM (typically a set-returning function) with an
equivalent SELECT subquery. The subquery will then be subject
to the planner's usual optimizations, potentially allowing a much
better plan to be generated. While the planner has long done this
automatically for simple SQL-language functions, it's now possible
for extensions to do it for functions outside that group.
Notably, this could be useful for functions that are presently
implemented in PL/pgSQL and work by generating and then EXECUTE'ing
a SQL query.
Author: Paul A Jungwirth <pj@illuminatedcomputing.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/09de6afa-c33d-4d94-a5cb-afc6cea0d2bb@illuminatedcomputing.com
Peter Eisentraut [Sat, 22 Nov 2025 08:40:00 +0000 (09:40 +0100)]
Add range_minus_multi and multirange_minus_multi functions
The existing range_minus function raises an exception when the range is
"split", because then the result can't be represented by a single range.
For example '[0,10)'::int4range - '[4,5)' would be '[0,4)' and '[5,10)'.
This commit adds new set-returning functions so that callers can get
results even in the case of splits. There is no risk of an exception for
multiranges, but a set-returning function lets us handle them the same
way we handle ranges.
Both functions return zero results if the subtraction would give an
empty range/multirange.
The main use-case for these functions is to implement UPDATE/DELETE FOR
PORTION OF, which must compute the application-time of "temporal
leftovers": the part of history in an updated/deleted row that was not
changed. To preserve the untouched history, we will implicitly insert
one record for each result returned by range/multirange_minus_multi.
Using a set-returning function will also let us support user-defined
types for application-time update/delete in the future.
Author: Paul A. Jungwirth <pj@illuminatedcomputing.com> Reviewed-by: Peter Eisentraut <peter@eisentraut.org> Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/ec498c3d-5f2b-48ec-b989-5561c8aa2024%40illuminatedcomputing.com
Thomas Munro [Sat, 22 Nov 2025 07:51:16 +0000 (20:51 +1300)]
jit: Adjust AArch64-only code for LLVM 21.
LLVM 21 changed the arguments of RTDyldObjectLinkingLayer's
constructor, breaking compilation with the backported
SectionMemoryManager from commit 9044fc1d.
Andrew Dunstan [Fri, 21 Nov 2025 22:08:46 +0000 (17:08 -0500)]
Add 'make check-tests' behavior to the meson based builds
There was no easy way to run specific tests in the meson based builds.
Author: Nazir Bilal Yavuz <byavuz81@gmail.com> Reviewed-by: Ashutosh Bapat <ashutosh.bapat.oss@gmail.com> Reviewed-by: Jian He <jian.universality@gmail.com>
Discussion: postgr.es/m/CAExHW5tK-QqayUN0%2BN3MF5bjV6vLKDCkRuGwoDJwc7vGjwCygQ%40mail.gmail.com
Peter Eisentraut [Fri, 21 Nov 2025 15:49:40 +0000 (16:49 +0100)]
Remove useless casts to (void *)
Their presence causes (small) risks of hiding actual type mismatches
or silently discarding qualifiers. Some have been missed in 7f798aca1d5 and some are new ones along the same lines.
Use strtoi64() in pgbench, replacing its open-coded implementation
Makes the code a little simpler.
The old implementation accepted trailing whitespace, but that was
unnecessary. Firstly, its sibling function for parsing decimals,
strtodouble(), does not accept trailing whitespace. Secondly, none of
the callers can pass a string with trailing whitespace to it.
In the passing, check specifically for ERANGE before printing the "out
of range" error. On some systems, strtoul() and strtod() return EINVAL
on an empty or all-spaces string, and "invalid input syntax" is more
appropriate for that than "out of range". For the existing
strtodouble() function this is purely academical because it's never
called with errorOK==false, but let's be tidy. (Perhaps we should
remove the dead codepaths altogether, but I'll leave that for another
day.)
Reviewed-by: Chao Li <li.evan.chao@gmail.com> Reviewed-by: Yuefei Shi <shiyuefei1004@gmail.com> Reviewed-by: Neil Chen <carpenter.nail.cz@gmail.com>
Discussion: https://www.postgresql.org/message-id/861dd5bd-f2c9-4ff5-8aa0-f82bdb75ec1f@iki.fi
Peter Eisentraut [Fri, 21 Nov 2025 12:01:30 +0000 (13:01 +0100)]
Update timezone to C99
This reverts changes done in PostgreSQL over the upstream code to
avoid relying on C99 <stdint.h> and <inttypes.h>.
In passing, there were a few other minor and cosmetic changes that I
left in to improve alignment with upstream, including some C11 feature
use (_Noreturn).
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://www.postgresql.org/message-id/flat/9ad2749f-77ab-4ecb-a321-1ca915480b05%40eisentraut.org
Peter Eisentraut [Fri, 21 Nov 2025 08:57:06 +0000 (09:57 +0100)]
C11 alignas instead of unions
This changes a few union members that only existed to ensure
alignments and replaces them with the C11 alignas specifier.
This change only uses fundamental alignments (meaning approximately
alignments of basic types), which all C11 compilers must support.
There are opportunities for similar changes using extended alignments,
for example in PGIOAlignedBlock, but these are not necessarily
supported by all compilers, so they are kept as a separate change.
Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/46f05236-d4d4-4b4e-84d4-faa500f14691%40eisentraut.org
Masahiko Sawada [Thu, 20 Nov 2025 22:50:27 +0000 (14:50 -0800)]
Use "COPY table TO" for partitioned tables in initial table synchronization.
Commit 4bea91f added support for "COPY table TO" with partitioned
tables. This commit enhances initial table synchronization in logical
replication to use "COPY table TO" for partitioned tables if possible,
instead of "COPY (SELECT ...) TO" variant, improving performance.
Melanie Plageman [Thu, 20 Nov 2025 22:36:40 +0000 (17:36 -0500)]
Split PruneFreezeParams initializers to one field per line
This conforms more closely with the style of other struct initializers
in the code base. Initializing multiple fields on a single line is
unpopular in part because pgindent won't permit a space after the comma
before the next field's period.
Melanie Plageman [Thu, 20 Nov 2025 16:40:36 +0000 (11:40 -0500)]
Update PruneState.all_[visible|frozen] earlier in pruning
During pruning and freezing in phase I of vacuum, we delay clearing
all_visible and all_frozen in the presence of dead items. This allows
opportunistic freezing if the page would otherwise be fully frozen,
since those dead items are later removed in vacuum phase III.
To move the VM update into the same WAL record that
prunes and freezes tuples, we must know whether the page will
be marked all-visible/all-frozen before emitting WAL. Previously we
waited until after emitting WAL to update all_visible/all_frozen to
their correct values.
The only barrier to updating these flags immediately after deciding
whether to opportunistically freeze was that while emitting WAL for a
record freezing tuples, we use the pre-corrected value of all_frozen to
compute the snapshot conflict horizon. By determining the conflict
horizon earlier, we can update the flags immediately after making the
opportunistic freeze decision.
This is required to set the VM in the XLOG_HEAP2_PRUNE_VACUUM_SCAN
record emitted by pruning and freezing.
Melanie Plageman [Thu, 20 Nov 2025 15:58:34 +0000 (10:58 -0500)]
Keep all_frozen updated in heap_page_prune_and_freeze
Previously, we relied on all_visible and all_frozen being used together
to ensure that all_frozen was correct, but it is better to keep both
fields updated.
Future changes will separate their usage, so we should not depend on
all_visible for the validity of all_frozen.
Melanie Plageman [Thu, 20 Nov 2025 15:30:43 +0000 (10:30 -0500)]
Refactor heap_page_prune_and_freeze() parameters into a struct
heap_page_prune_and_freeze() had accumulated an unwieldy number of input
parameters and upcoming work to handle VM updates in this function will
add even more.
Introduce a new PruneFreezeParams struct to group the function’s input
parameters, improving readability and maintainability.
doc: Document how to run a subset of regress tests
This patch was originally submitted a year ago, but never
ended up getting committed. It was later brought up again
on a recent thread on the same subject.
Original patch by Paul A Jungwirth with some wordsmithing
by me based on the review from the original thread.
Author: Paul A. Jungwirth <pj@illuminatedcomputing.com> Reviewed-by: Daniel Gustafsson <daniel@yesql.se> Reviewed-by: Viktor Holmberg <v@viktorh.net> Reviewed-by: Nazir Bilal Yavuz <byavuz81@gmail.com>
Discussion: https://postgr.es/m/CA+renyXB5jYG9r5-CaDc4g607EB398QwTk_efEXTzarrO8bPzw@mail.gmail.com
Discussion: https://postgr.es/m/CACJufxHOcmeTkoh2CxFHKv9GRnp9sLVzN=LZhqTgvqT++PXZNQ@mail.gmail.com
Tomas Vondra [Thu, 20 Nov 2025 11:51:58 +0000 (12:51 +0100)]
Handle EPERM in pg_numa_init
When running in Docker, the container may not have privileges needed by
get_mempolicy(). This is called by numa_available() in libnuma, but
versions prior to 2.0.19 did not expect that. The numa_available() call
seemingly succeeds, but then we get unexpected failures when trying to
query status of pages:
postgres =# select * from pg_shmem_allocations_numa;
ERROR: XX000: failed NUMA pages inquiry status: Operation not
permitted
LOCATION: pg_get_shmem_allocations_numa, shmem.c:691
The best solution is to call get_mempolicy() first, and proceed to
numa_available() only when it does not fail with EPERM. Otherwise we'd
need to treat older libnuma versions as insufficient, which seems a bit
too harsh, as this only affects containerized systems.
Fix by me, based on suggestions by Christoph. Backpatch to 18, where the
NUMA functions were introduced.
Reported-by: Christoph Berg <myon@debian.org> Reviewed-by: Christoph Berg <myon@debian.org>
Discussion: https://postgr.es/m/aPDZOxjrmEo_1JRG@msg.df7cb.de
Backpatch-through: 18
Peter Eisentraut [Thu, 20 Nov 2025 06:47:48 +0000 (07:47 +0100)]
Remove obsolete cast
The upstream timezone code uses a bool variable as an array subscript.
Back when PostgreSQL's bool was char, this would have caused a warning
from gcc -Wchar-subscripts, which is included in -Wall. But this has
been obsolete since probably commit d26a810ebf9, but certainly since
bool is now the C standard bool. So we can remove this deviation from
the upstream code, to make future code merges simpler.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://www.postgresql.org/message-id/flat/9ad2749f-77ab-4ecb-a321-1ca915480b05%40eisentraut.org
Fujii Masao [Thu, 20 Nov 2025 00:18:51 +0000 (09:18 +0900)]
doc: Update pg_upgrade documentation to match recent description changes.
Commit 792353f7d52 updated the pg_dump and pg_dumpall documentation to
clarify which statistics are not included in their output. The pg_upgrade
documentation contained a nearly identical description, but it was not updated
at the same time.
This commit updates the pg_upgrade documentation to match those changes.
Backpatch to v18, where commit 792353f7d52 was backpatched to.
Fujii Masao [Thu, 20 Nov 2025 00:14:02 +0000 (09:14 +0900)]
Add HINT listing valid encodings to encode() and decode() errors.
This commit updates encode() and decode() so that when an invalid encoding
is specified, their error message includes a HINT listing all valid encodings.
This helps users quickly see which encodings are supported without needing
to consult the documentation.
Author: Shinya Sugamoto <shinya34892@gmail.com> Reviewed-by: Chao Li <lic@highgo.com> Reviewed-by: Daniel Gustafsson <daniel@yesql.se> Reviewed-by: Masahiko Sawada <sawada.mshk@gmail.com> Reviewed-by: Peter Eisentraut <peter@eisentraut.org> Reviewed-by: Nathan Bossart <nathandbossart@gmail.com> Reviewed-by: Fujii Masao <masao.fujii@gmail.com>
Discussion: https://postgr.es/m/CAAe3y+99sfPv8UDF1VM-rC1i5HBdqxUh=2HrbJJFm2+i=1OwOw@mail.gmail.com
Thomas Munro [Wed, 19 Nov 2025 21:23:44 +0000 (10:23 +1300)]
Drop support for MSVCRT's float formatting quirk.
Commit f1885386 added code to remove an unnecessary leading zero from
the exponent in a float formatted by the system snprintf(). The C
standard doesn't allow unnecessary digits beyond two, and the tests pass
without this on Windows' modern UCRT (required since commit 1758d424).
Thomas Munro [Wed, 19 Nov 2025 21:04:11 +0000 (10:04 +1300)]
Drop support for MSVCRT's %I64 format strings.
MSVCRT predated C99 and invented non-standard placeholders for 64-bit
numbers, and then later used them in standard macros when C99
<inttypes.h> arrived. The macros just use %lld etc when building with
UCRT, so there should be no way for our interposed sprintf.c code to
receive the pre-standard kind these days. Time to drop the code that
parses them.
That code was in fact already dead when commit 962da900 landed, as we'd
disclaimed MSVCRT support a couple of weeks earlier in commit 1758d424,
but patch development overlapped and the history of these macros hadn't
been investigated.
Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://postgr.es/m/4d8b1a67-aab2-4429-b44b-f03988095939%40eisentraut.org
Tom Lane [Wed, 19 Nov 2025 18:22:12 +0000 (13:22 -0500)]
Speed up eqjoinsel() with lots of MCV entries.
If both sides of the operator have most-common-value statistics,
eqjoinsel wants to check which MCVs have matches on the other side.
Formerly it did this with a dumb compare-all-the-entries loop,
which had O(N^2) behavior for long MCV lists. When that code was
written, twenty-plus years ago, that seemed tolerable; but nowadays
people frequently use much larger statistics targets, so that the
O(N^2) behavior can hurt quite a bit.
To add insult to injury, when asked for semijoin semantics, the
entire comparison loop was done over, even though we frequently
know that it will yield exactly the same results.
To improve matters, switch to using a hash table to perform the
matching. Testing suggests that depending on the data type, we may
need up to about 100 MCVs on each side to amortize the extra costs
of setting up the hash table and performing hash-value computations;
so continue to use the old looping method when there are fewer MCVs
than that.
Also, refactor so that we don't repeat the matching work unless
we really need to, which occurs only in the uncommon case where
eqjoinsel_semi decides to truncate the set of inner MCVs it
considers. The refactoring also got rid of the need to use the
presented operator's commutator. Real-world operators that are
using eqjoinsel should pretty much always have commutators, but
at the very least this saves a few syscache lookups.
Author: Ilia Evdokimov <ilya.evdokimov@tantorlabs.com> Co-authored-by: David Geier <geidav.pg@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/20ea8bf5-3569-4e46-92ef-ebb2666debf6@tantorlabs.com
Print new OldestXID value in pg_resetwal when it's being changed
Commit 74cf7d46a91d added the --oldest-transaction-id option to
pg_resetwal, but forgot to update the code that prints all the new
values that are being set. Fix that.
Tom Lane [Tue, 18 Nov 2025 17:56:55 +0000 (12:56 -0500)]
Don't allow CTEs to determine semantic levels of aggregates.
The fix for bug #19055 (commit b0cc0a71e) allowed CTE references in
sub-selects within aggregate functions to affect the semantic levels
assigned to such aggregates. It turns out this broke some related
cases, leading to assertion failures or strange planner errors such
as "unexpected outer reference in CTE query". After experimenting
with some alternative rules for assigning the semantic level in
such cases, we've come to the conclusion that changing the level
is more likely to break things than be helpful.
Therefore, this patch undoes what b0cc0a71e changed, and instead
installs logic to throw an error if there is any reference to a
CTE that's below the semantic level that standard SQL rules would
assign to the aggregate based on its contained Var and Aggref nodes.
(The SQL standard disallows sub-selects within aggregate functions,
so it can't reach the troublesome case and hence has no rule for
what to do.)
Perhaps someone will come along with a legitimate query that this
logic rejects, and if so probably the example will help us craft
a level-adjustment rule that works better than what b0cc0a71e did.
I'm not holding my breath for that though, because the previous
logic had been there for a very long time before bug #19055 without
complaints, and that bug report sure looks to have originated from
fuzzing not from real usage.
Like b0cc0a71e, back-patch to all supported branches, though
sadly that no longer includes v13.
Bug: #19106 Reported-by: Kamil Monicz <kamil@monicz.dev>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/19106-9dd3668a0734cd72@postgresql.org
Backpatch-through: 14
Nathan Bossart [Tue, 18 Nov 2025 16:28:36 +0000 (10:28 -0600)]
Switch from tabs to spaces in postgresql.conf.sample.
This file is written for 8-space tabs, since we expect that most
users who edit their configuration files use 8-space tabs.
However, most of PostgreSQL is written for 4-space tabs, and at
least one popular web interface defaults to 4-space tabs. Rather
than trying to standardize on a particular tab width for this file,
let's just switch to spaces.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/aReNUKdMgKxLqmq7%40nathan
Álvaro Herrera [Tue, 18 Nov 2025 15:13:29 +0000 (16:13 +0100)]
Log a note at program start when running in dry-run mode
Users might get some peace of mind knowing their data is not being
destroyed or whatever.
Author: Peter Smith <smithpb2250@gmail.com> Reviewed-by: Álvaro Herrera <alvherre@kurilemu.de>
Discussion: https://postgr.es/m/CAHut+PsvQJQnQO0KT0S2oegenkvJ8FUuY-QS5syyqmT24R2xFQ@mail.gmail.com
We need separate pairing heaps for different WaitLSNType's, because there
might be waiters for different LSN's at the same time. However, one process
can wait only for one type of LSN at a time. So, no need for inHeap
and heapNode fields to be arrays.
Discussion: https://postgr.es/m/CAPpHfdsBR-7sDtXFJ1qpJtKiohfGoj%3DvqzKVjWxtWsWidx7G_A%40mail.gmail.com
Author: Alexander Korotkov <aekorotkov@gmail.com> Reviewed-by: Xuneng Zhou <xunengzhou@gmail.com>
Michael Paquier [Tue, 18 Nov 2025 06:46:43 +0000 (15:46 +0900)]
pg_buffercache: Fix incorrect result cast for relforknumber
pg_buffercache_pages.relforknumber is defined as an int2, but its value
was stored with ObjectIdGetDatum() rather than Int16GetDatum() in the
result record.
Amit Kapila [Tue, 18 Nov 2025 03:58:55 +0000 (03:58 +0000)]
Rename two columns in pg_stat_subscription_stats.
This patch renames the sync_error_count column to sync_table_error_count
in the pg_stat_subscription_stats view. The new name makes the purpose
explicit now that a separate column exists to track sequence
synchronization errors.
Additionally, the column seq_sync_error_count is renamed to
sync_seq_error_count to maintain a consistent naming pattern, making it
easier for users to group, and query synchronization related counters.
Author: Vignesh C <vignesh21@gmail.com> Reviewed-by: Peter Smith <smithpb2250@gmail.com> Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/CALDaNm3WwJmz=-4ybTkhniB-Nf3qmFG9Zx1uKjyLLoPF5NYYXA@mail.gmail.com
Amit Kapila [Tue, 18 Nov 2025 03:48:00 +0000 (03:48 +0000)]
Doc: Use <structfield> markup for sequence fields.
Following commit 980a855c5c, update documentation to use <structfield> for
sequence columns. Previously, these were incorrectly marked up as <literal>.
Author: Peter Smith <smithpb2250@gmail.com> Reviewed-by: Chao Li <li.evan.chao@gmail.com>
Discussion: https://postgr.es/m/CAHut+PtpDMUE3Kd1p=1ff9pw2HMbgQCpowE_0Hd6gs5v2pKfQg@mail.gmail.com
Masahiko Sawada [Mon, 17 Nov 2025 21:22:20 +0000 (13:22 -0800)]
Use streaming read I/O in BRIN vacuum scan.
This commit implements streaming read I/O for BRIN vacuum
scans. Although BRIN indexes tend to be relatively small by design,
performance tests have shown performance improvements.
Tom Lane [Mon, 17 Nov 2025 20:24:34 +0000 (15:24 -0500)]
Fix pg_crc32c_armv8_choose.c to build with ancient glibc releases.
If you go back as far as the RHEL7 era, <sys/auxv.h> does not provide
the HWCAPxxx macros needed with elf_aux_info or getauxval, so you need
to get those from the kernel header <asm/hwcap.h> instead. We knew
that for the 32-bit case but failed to extrapolate to the 64-bit case.
Oversight in commit aac831caf.
Reported-by: GaoZengqi <pgf00a@gmail.com>
Author: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAFmBtr3Av62-jBzdhFkDHXJF9vQmNtSnH2upwODjnRcsgdTytw@mail.gmail.com
Backpatch-through: 18
Tom Lane [Mon, 17 Nov 2025 18:54:52 +0000 (13:54 -0500)]
Clean up match_orclause_to_indexcol().
Remove bogus stripping of RelabelTypes: that can result in building
an output SAOP tree with incorrect exposed exprType for the operands,
which might confuse polymorphic operators. Moreover it demonstrably
prevents folding some OR-trees to SAOPs when the RHS expressions
have different base types that were coerced to the same type by
RelabelTypes.
Reduce prohibition on type_is_rowtype to just disallow type RECORD.
We need that because otherwise we would happily fold multiple RECORD
Consts into a RECORDARRAY Const even if they aren't the same record
type. (We could allow that perhaps, if we checked that they all have
the same typmod, but the case doesn't seem worth that much effort.)
However, there is no reason at all to disallow the transformation
for named composite types, nor domains over them: as long as we can
find a suitable array type we're good.
Remove some assertions that seem rather out of place (it's not
this code's duty to verify that the RestrictInfo structure is
sane). Rewrite some comments.
The issues with RelabelType stripping seem severe enough to
back-patch this into v18 where the code was introduced.
Author: Tender Wang <tndrwang@gmail.com> Co-authored-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CAHewXN=aH7GQBk4fXU-WaEeVmQWUmBAeNyBfJ3VKzPphyPKUkQ@mail.gmail.com
Backpatch-through: 18
Fujii Masao [Mon, 17 Nov 2025 14:24:39 +0000 (23:24 +0900)]
doc: Document default values for some pg_recvlogical options.
The documentation did not previously mention the default values for
the --fsync-interval and --plugin options, even though pg_recvlogical --help
shows them. This omission made it harder for users to understand
the tool's behavior from the documentation alone.
This commit adds the missing default value descriptions for both options
to the pg_recvlogical documentation.
Author: Chao Li <lic@highgo.com> Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Discussion: https://postgr.es/m/CAEoWx2kt8m7wV39_zOBds5SNXx9EAkDqb5cPshk7Bxw6Js4Zpg@mail.gmail.com
PostgreSQL 18 deprecated password_encryption='md5', but the
comments for this GUC in the sample configuration file did
not mention the deprecation. Update comments with a notice
to make as many users as possible aware of it. Also add a
comment to the related md5_password_warnings GUC while there.
Author: Michael Banck <mbanck@gmx.net> Reviewed-by: Daniel Gustafsson <daniel@yesql.se> Reviewed-by: Nathan Bossart <nathandbossart@gmail.com> Reviewed-by: Robert Treat <rob@xzilla.net>
Backpatch-through: 18
Michael Paquier [Mon, 17 Nov 2025 01:44:26 +0000 (10:44 +0900)]
Rework output format of pg_dependencies
The existing format of pg_dependencies uses a single-object JSON
structure, with each key value embedding all the knowledge about the
set attributes tracked, like:
{"1 => 5": 1.000000, "5 => 1": 0.423130}
While this is a very compact format, it is confusing to read and it is
difficult to manipulate the values within the object, particularly when
tracking multiple attributes.
The new output format introduced in this commit is a JSON array of
objects, with:
- A key named "degree", with a float value.
- A key named "attributes", with an array of attribute numbers.
- A key named "dependency", with an attribute number.
The values use the same underlying type as previously when printed, with
a new output format that shows now as follows:
[{"degree": 1.000000, "attributes": [1], "dependency": 5},
{"degree": 0.423130, "attributes": [5], "dependency": 1}]
This new format will become handy for a follow-up set of changes, so as
it becomes possible to inject extended statistics rather than require an
ANALYZE, like in a dump/restore sequence or after pg_upgrade on a new
cluster.
This format has been suggested by Tomas Vondra. The key names are
defined in the header introduced by 1f927cce4498, to ease the
integration of frontend-specific changes that are still under
discussion. (Again a personal note: if anybody comes up with better
name for the keys, of course feel free.)
The bulk of the changes come from the regression tests, where
jsonb_pretty() is now used to make the outputs generated easier to
parse.
Author: Corey Huinker <corey.huinker@gmail.com> Reviewed-by: Jian He <jian.universality@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CADkLM=dpz3KFnqP-dgJ-zvRvtjsa8UZv8wDAQdqho=qN3kX0Zg@mail.gmail.com
Michael Paquier [Mon, 17 Nov 2025 00:52:20 +0000 (09:52 +0900)]
Rework output format of pg_ndistinct
The existing format of pg_ndistinct uses a single-object JSON structure
where each key is itself a comma-separated list of attnums, like:
{"3, 4": 11, "3, 6": 11, "4, 6": 11, "3, 4, 6": 11}
While this is a very compact format, it is confusing to read and it is
difficult to manipulate the values within the object.
The new output format introduced in this commit is an array of objects,
with:
- A key named "attributes", that contains an array of attribute numbers.
- A key named "ndistinct", represented as an integer.
The values use the same underlying type as previously when printed, with
a new output format that shows now as follows:
[{"ndistinct": 11, "attributes": [3,4]},
{"ndistinct": 11, "attributes": [3,6]},
{"ndistinct": 11, "attributes": [4,6]},
{"ndistinct": 11, "attributes": [3,4,6]}]
This new format will become handy for a follow-up set of changes, so as
it becomes possible to inject extended statistics rather than require an
ANALYZE, like in a dump/restore sequence or after pg_upgrade on a new
cluster.
This format has been suggested by Tomas Vondra. The key names are
defined in a new header, to ease with the integration of
frontend-specific changes that are still under discussion. (Personal
note: I am not specifically wedded to these key names, but if there are
better name suggestions for this release, feel free.)
The bulk of the changes come from the regression tests, where
jsonb_pretty() is now used to make the outputs generated easier to
parse.
Author: Corey Huinker <corey.huinker@gmail.com> Reviewed-by: Jian He <jian.universality@gmail.com> Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://postgr.es/m/CADkLM=dpz3KFnqP-dgJ-zvRvtjsa8UZv8wDAQdqho=qN3kX0Zg@mail.gmail.com
Thomas Munro [Sun, 16 Nov 2025 23:01:12 +0000 (12:01 +1300)]
Define PS_USE_CLOBBER_ARGV on GNU/Hurd.
Until d2ea2d310dfdc40328aca5b6c52225de78432e01, the PS_USE_PS_STRINGS
option was used on the GNU/Hurd. As this option got removed and
PS_USE_CLOBBER_ARGV appears to work fine nowadays on the Hurd, define
this one to re-enable process title changes on this platform.
In the 14 and 15 branches, the existing test for __hurd__ (added 25
years ago by commit 209aa77d, removed in 16 by the above commit) is left
unchanged for now as it was activating slightly different code paths and
would need investigation by a Hurd user.
Author: Michael Banck <mbanck@debian.org>
Discussion: https://postgr.es/m/CA%2BhUKGJMNGUAqf27WbckYFrM-Mavy0RKJvocfJU%3DJ2XcAZyv%2Bw%40mail.gmail.com
Backpatch-through: 16
David Rowley [Sun, 16 Nov 2025 23:27:00 +0000 (12:27 +1300)]
Adjust MemSet macro to use size_t rather than long
Likewise for MemSetAligned.
"long" wasn't the most suitable type for these macros as with MSVC in
64-bit builds, sizeof(long) == 4, which is narrower than the processor's
word size, therefore these macros had to perform twice as many loops as
they otherwise might.
Author: David Rowley <dgrowleyml@gmail.com> Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi> Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://postgr.es/m/CAApHDvoGFjSA3aNyVQ3ivbyc4ST=CC5L-_VjEUQ92HbE2Cxovg@mail.gmail.com
David Rowley [Sun, 16 Nov 2025 23:26:41 +0000 (12:26 +1300)]
Get rid of long datatype in CATCACHE_STATS enabled builds
"long" is 32 bits on Windows 64-bit. Switch to a datatype that's 64-bit
on all platforms. While we're there, use an unsigned type as these
fields count things that have occurred, of which it's not possible to
have negative numbers of.
Author: David Rowley <dgrowleyml@gmail.com> Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi> Reviewed-by: Peter Eisentraut <peter@eisentraut.org>
Discussion: https://postgr.es/m/CAApHDvoGFjSA3aNyVQ3ivbyc4ST=CC5L-_VjEUQ92HbE2Cxovg@mail.gmail.com
Michael Paquier [Sun, 16 Nov 2025 23:01:04 +0000 (08:01 +0900)]
Add test for temporary file removal and WITH HOLD cursor
This new test, added in 009_log_temp_files, checks that the temporary
files created by a WITH HOLD cursor are dropped at the end of the
transaction where the transaction has been created.
The portal's executor is shutdown in PersistHoldablePortal(), after for
example some forced detoast, so as the cursor data can be accessed
without requiring a snapshot.
Author: Mircea Cadariu <cadariu.mircea@gmail.com>
Discussion: https://postgr.es/m/0a666d28-9080-4239-90d6-f6345bb43468@gmail.com
Dean Rasheed [Sun, 16 Nov 2025 22:14:06 +0000 (22:14 +0000)]
Fix Assert failure in EXPLAIN ANALYZE MERGE with a concurrent update.
When instrumenting a MERGE command containing both WHEN NOT MATCHED BY
SOURCE and WHEN NOT MATCHED BY TARGET actions using EXPLAIN ANALYZE, a
concurrent update of the target relation could lead to an Assert
failure in show_modifytable_info(). In a non-assert build, this would
lead to an incorrect value for "skipped" tuples in the EXPLAIN output,
rather than a crash.
This could happen if the concurrent update caused a matched row to no
longer match, in which case ExecMerge() treats the single originally
matched row as a pair of not matched rows, and potentially executes 2
not-matched actions for the single source row. This could then lead to
a state where the number of rows processed by the ModifyTable node
exceeds the number of rows produced by its source node, causing
"skipped_path" in show_modifytable_info() to be negative, triggering
the Assert.
Fix this in ExecMergeMatched() by incrementing the instrumentation
tuple count on the source node whenever a concurrent update of this
kind is detected, if both kinds of merge actions exist, so that the
number of source rows matches the number of actions potentially
executed, and the "skipped" tuple count is correct.
Back-patch to v17, where support for WHEN NOT MATCHED BY SOURCE
actions was introduced.
Fix WaitLSNWakeup() fast-path check for InvalidXLogRecPtr
WaitLSNWakeup() incorrectly returned early when called with
InvalidXLogRecPtr (meaning "wake all waiters"), because the fast-path
check compared minWaitedLSN > 0 without validating currentLSN first.
This caused WAIT FOR LSN commands to wait indefinitely during standby
promotion until random signals woke them.
Add an XLogRecPtrIsValid() check before the comparison so that
InvalidXLogRecPtr bypasses the fast-path and wakes all waiters immediately.
Discussion: https://postgr.es/m/CABPTF7UieOYbOgH3EnQCasaqcT1T4N6V2wammwrWCohQTnD_Lw%40mail.gmail.com
Author: Xuneng Zhou <xunengzhou@gmail.com> Reviewed-by: Alexander Korotkov <aekorotkov@gmail.com>
All GUCs in postgresql.conf.sample should be set to the default
value and be commented out. This syntax was however not tested
for, making omissions easy to miss. Add a test which check all
lines for syntax.
Author: Daniel Gustafsson <daniel@yesql.se> Reviewed-by: Andrew Dunstan <andrew@dunslane.net>
Discussion: https://postgr.es/m/19727040-3EE4-4719-AF4F-2548544113D7@yesql.se
Nathan Bossart [Fri, 14 Nov 2025 19:20:09 +0000 (13:20 -0600)]
Add note about CreateStatistics()'s selective use of check_rights.
Commit 5e4fcbe531 added a check_rights parameter to this function
for use by ALTER TABLE commands that re-create statistics objects.
However, we intentionally ignore check_rights when verifying
relation ownership because this function's lookup could return a
different answer than the caller's. This commit adds a note to
this effect so that we remember it down the road.
Bruce Momjian [Fri, 14 Nov 2025 14:08:53 +0000 (09:08 -0500)]
doc: double-quote use of %f, %p, and %r in literal commands.
Path expansion might expose characters like spaces which would cause
command failure, so double-quote the examples. While %f doesn't need
quoting since it uses a fixed character set, it is best to be
consistent.
Fujii Masao [Fri, 14 Nov 2025 13:40:39 +0000 (22:40 +0900)]
pgbench: Fix assertion failure with multiple \syncpipeline in pipeline mode.
Previously, when pgbench ran a custom script that triggered retriable errors
(e.g., deadlocks) followed by multiple \syncpipeline commands in pipeline mode,
the following assertion failure could occur:
Assertion failed: (res == ((void*)0)), function discardUntilSync, file pgbench.c, line 3594.
The issue was that discardUntilSync() assumed a pipeline sync result
(PGRES_PIPELINE_SYNC) would always be followed by either another sync result
or NULL. This assumption was incorrect: when multiple sync requests were sent,
a sync result could instead be followed by another result type. In such cases,
discardUntilSync() mishandled the results, leading to the assertion failure.
This commit fixes the issue by making discardUntilSync() correctly handle cases
where a pipeline sync result is followed by other result types. It now continues
discarding results until another pipeline sync followed by NULL is reached.
Backpatched to v17, where support for \syncpipeline command in pgbench was
introduced.
Michael Paquier [Fri, 14 Nov 2025 05:37:10 +0000 (14:37 +0900)]
Revert "Drop unnamed portal immediately after execution to completion"
This reverts commit 1fd981f05369, based on concerns that the logging
improvements do not justify the protocol breakage of dropping an unnamed
portal once its execution has completed.
It seems unlikely that one would try to send an execute or describe
message after the portal has been used, but if they do such
post-completion messages would not be able to process as the previous
versions. Let's revert this change for now so as we keep compatibility
and consider a different solution.
Bruce Momjian [Fri, 14 Nov 2025 02:53:13 +0000 (21:53 -0500)]
doc: adjust "Replication Slot" to mention physical & logical
Much of the "Replication Slot" chapter applies to physical and logical
slots, but it was sloppy in mentioning mostly physical slots. This
patch clarified which parts of the text apply to which slot types.
This chapter is referenced from the logical slot/subscriber chapter, so
it needs to do double duty.
Bruce Momjian [Fri, 14 Nov 2025 01:44:00 +0000 (20:44 -0500)]
doc: clarify "physical" replication slot creation on the primary
Previously it was not clear that "physical" replication slots were being
discussed, and that they needed to be created on the primary and not the
standby.
Document that pg_getaddrinfo_all does not accept null hints
While the underlying getaddrinfo call accepts a null pointer for
the hintp parameter, pg_getaddrinfo_all does not. Document this
difference with a comment to make it clear.
Author: Daniel Gustafsson <daniel@yesql.se> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us> Reported-by: Sergey Tatarintsev <s.tatarintsev@postgrespro.ru>
Discussion: https://postgr.es/m/1e5efc94-407e-40b8-8b10-4d25f823c6d7@postgrespro.ru
projects / thirdparty / postgresql.git / log
