postgres_fdw: Revert unstable tests for postgres_fdw.application_name.
Commit 449ab63505 added the tests that check that postgres_fdw.application_name
GUC works as expected. But they were unstable and caused some buildfarm
members to report the failure. This commit reverts those unstable tests.
Reported-by: Tom Lane as per buildfarm
Discussion: https://postgr.es/m/3220909.1631054766@sss.pgh.pa.us
Amit Kapila [Wed, 8 Sep 2021 06:20:37 +0000 (11:50 +0530)]
Invalidate relcache for publications defined for all tables.
Updates/Deletes on a relation were allowed even without replica identity
after we define the publication for all tables. This would later lead to
an error on subscribers. The reason was that for such publications we were
not invalidating the relcache and the publication information for
relations was not getting rebuilt. Similarly, we were not invalidating the
relcache after dropping of such publications which will prohibit
Updates/Deletes without replica identity even without any publication.
Author: Vignesh C and Hou Zhijie Reviewed-by: Hou Zhijie, Kyotaro Horiguchi, Amit Kapila
Backpatch-through: 10, where it was introduced
Discussion: https://postgr.es/m/CALDaNm0pF6zeWqCA8TCe2sDuwFAy8fCqba=nHampCKag-qLixg@mail.gmail.com
Michael Paquier [Wed, 8 Sep 2021 03:02:30 +0000 (12:02 +0900)]
Introduce GUC shared_memory_size
This runtime-computed GUC shows the size of the server's main shared
memory area, taking into account the amount of shared memory allocated
by extensions as this is calculated after processing
shared_preload_libraries.
Andres Freund [Tue, 7 Sep 2021 18:56:13 +0000 (11:56 -0700)]
windows: Only consider us to be running as service if stderr is invalid.
Previously pgwin32_is_service() would falsely return true when postgres is
started from somewhere within a service, but not as a service. That is
e.g. always the case with windows docker containers, which some CI services
use to run windows tests in.
When postgres falsely thinks its running as a service, no messages are
writting to stdout / stderr. That can be very confusing and causes a few tests
to fail.
To fix additionally check if stderr is invalid in pgwin32_is_service(). For
that to work in backend processes, pg_ctl is changed to pass down handles so
that postgres can do the same check (otherwise "default" handles are created).
While this problem exists in all branches, there have been no reports by
users, the prospective CI usage currently is only for master, and I am not a
windows expert. So doing the change in only master for now seems the sanest
approach.
Author: Andres Freund <andres@anarazel.de> Reviewed-By: Magnus Hagander <magnus@hagander.net>
Discussion: https://postgr.es/m/20210305185752.3up5eq2eanb7ofmb@alap3.anarazel.de
postgres_fdw: Allow application_name of remote connection to be set via GUC.
This commit adds postgres_fdw.application_name GUC which specifies
a value for application_name configuration parameter used
when postgres_fdw establishes a connection to a foreign server.
This GUC setting always overrides application_name option of
the foreign server object. This GUC is useful when we want to
specify our own application_name per remote connection.
Previously application_name of a remote connection could be set
basically only via options of a server object. But which meant that
every session connecting to the same foreign server basically
should use the same application_name. Also if we want to change
the setting, we had to execute "ALTER SERVER ... OPTIONS ..." command.
It was inconvenient.
AIX: Fix missing libpq symbols by respecting SHLIB_EXPORTS.
We make each AIX shared library export all globals found in .o files
that originate in the library. That doesn't include symbols acquired by
-lpgcommon_shlib. That is good on average, but it became a problem for
libpq when commit e6afa8918c461c1dd80c5063a950518fa4e950cd moved five
official libpq API symbols into src/common. Fix this by implementing
the SHLIB_EXPORTS mechanism for AIX, so affected libraries export the
same symbols that they export on Linux. This reintroduces symbols
pg_encoding_to_char, pg_utf_mblen, pg_char_to_encoding,
pg_valid_server_encoding, and pg_valid_server_encoding_id. Back-patch
to v13, where the aforementioned commit first appeared. While a minor
release is usually the wrong time to add or remove symbol exports in
libpq or libecpg, we should expect users to want each documented symbol.
Add PublicationTable and PublicationRelInfo structs
These encapsulate a relation when referred from replication DDL.
Currently they don't do anything useful (they're just wrappers around
RangeVar and Relation respectively) but in the future they'll be used to
carry column lists.
Extracted from a larger patch by Rahila Syed.
Author: Rahila Syed <rahilasyed90@gmail.com> Reviewed-by: Álvaro Herrera <alvherre@alvh.no-ip.org> Reviewed-by: Tomas Vondra <tomas.vondra@enterprisedb.com> Reviewed-by: Amit Kapila <amit.kapila16@gmail.com>
Discussion: https://postgr.es/m/CAH2L28vddB_NFdRVpuyRBJEBWjz4BSyTB=_ektNRH8NJ1jf95g@mail.gmail.com
Tom Lane [Mon, 6 Sep 2021 15:43:44 +0000 (11:43 -0400)]
Fix actively-misleading comments about the contents of struct pg_tm.
pgtime.h documented the PG interpretation of tm_mon right alongside
the POSIX interpretation of tm_year, with no hint that neither
comment was correct throughout our code.
Perhaps someday we ought to switch to using two separate struct
definitions to provide a clearer indication of which semantics are
in use where. But I fear the tedium-versus-safety-gain tradeoff
would not be very good.
Tom Lane [Mon, 6 Sep 2021 15:03:56 +0000 (11:03 -0400)]
Make timetz_zone() stable, and correct a bug for DYNTZ abbreviations.
Historically, timetz_zone() has used time(NULL) as the reference point
for deciding whether DST is active. That means its result can change
intra-statement, requiring it to be marked VOLATILE (cf. 35979e6c3).
But that definition is pretty inconsistent with the way we deal with
timestamps elsewhere. Let's make it use the transaction start time
("now()") as the reference point instead. That lets it be marked
STABLE, and also saves a kernel call per invocation.
While at it, remove the function's use of pg_time_t and pg_localtime.
Those are inconsistent with the other code in this area, which indeed
created a bug: timetz_zone() delivered completely wrong answers if
the zone was specified by a dynamic TZ abbreviation. (We need to do
something about that in the back branches, but the fix will look
different from this.)
Michael Paquier [Mon, 6 Sep 2021 01:59:20 +0000 (10:59 +0900)]
Move the shared memory size calculation to its own function
This change refactors the shared memory size calculation in
CreateSharedMemoryAndSemaphores() to its own function. This is intended
for use in a future change related to the setup of huge pages and shared
memory with some GUCs, while useful on its own for extensions.
Tom Lane [Sat, 4 Sep 2021 20:29:08 +0000 (16:29 -0400)]
Further portability tweaks for float4/float8 hash functions.
Attempting to make hashfloat4() look as much as possible like
hashfloat8(), I'd figured I could replace NaNs with get_float4_nan()
before widening to float8. However, results from protosciurus
and topminnow show that on some platforms that produces a different
bit-pattern from get_float8_nan(), breaking the intent of ce773f230.
Rearrange so that we use the result of get_float8_nan() for all NaN
cases. As before, back-patch.
Revert "Avoid creating archive status ".ready" files too early"
This reverts commit 515e3d84a0b5 and equivalent commits in back
branches. This solution to the problem has a number of problems, so
we'll try again with a different approach.
Tom Lane [Sat, 4 Sep 2021 01:04:44 +0000 (21:04 -0400)]
Remove arbitrary MAXPGPATH limit on command lengths in pg_ctl.
Replace fixed-length command buffers with psprintf() calls. We didn't
have anything as convenient as psprintf() when this code was written,
but now that we do, there's little reason for the limitation to
stand. Removing it eliminates some corner cases where (for example)
starting the postmaster with a whole lot of options fails.
Most individual file names that pg_ctl deals with are still restricted
to MAXPGPATH, but we've seldom had complaints about that limitation
so long as it only applies to one filename.
Tom Lane [Fri, 3 Sep 2021 20:38:55 +0000 (16:38 -0400)]
Disallow creating an ICU collation if the DB encoding won't support it.
Previously this was allowed, but the collation effectively vanished
into the ether because of the way lookup_collation() works: you could
not use the collation, nor even drop it. Seems better to give an
error up front than to leave the user wondering why it doesn't work.
(Because this test is in DefineCollation not CreateCollation, it does
not prevent pg_import_system_collations from creating ICU collations,
regardless of the initially-chosen encoding.)
Per bug #17170 from Andrew Bille. Back-patch to v10 where ICU support
was added.
Tom Lane [Fri, 3 Sep 2021 14:01:02 +0000 (10:01 -0400)]
Fix portability issue in tests from commit ce773f230.
Modern POSIX seems to require strtod() to accept "-NaN", but there's
nothing about NaN in SUSv2, and some of our oldest buildfarm members
don't like it. Let's try writing it as -'NaN' instead; that seems
to produce the same result, at least on Intel hardware.
Currently the pc files use hard coded paths for "includedir" and
"libdir."
Example:
Cflags: -I/usr/include
Libs: -L/usr/lib -lpq
This is not very fortunate when cross compiling inside a buildroot,
where the includes and libs are inside a staging directory, because
this introduces host paths into the build:
checking for pkg-config... /builder/shared-workdir/build/sdk/staging_dir/host/bin/pkg-config
checking for PostgreSQL libraries via pkg_config... -L/usr/lib <----
This commit addresses this by doing the following two things:
1. Instead of hard coding the paths in "Cflags" and "Libs"
"${includedir}" and "${libdir}" are used. Note: these variables
can be overriden on the pkg-config command line
("--define-variable=libdir=/some/path").
2. Add the variables "prefix" and "exec_prefix". If "includedir"
and/or "libdir" are using these then construct them accordingly.
This is done because buildroots (for instance OpenWrt) tend to
rename the real pkg-config and call it indirectly from a script
that sets "prefix", "exec_prefix" and "bindir", like so:
Like this the paths can be forced into the staging directory when
using a buildroot setup:
checking for pkg-config... /home/sk/tmp/openwrt/staging_dir/host/bin/pkg-config
checking for PostgreSQL libraries via pkg_config... -L/home/sk/tmp/openwrt/staging_dir/target-mips_24kc_musl/usr/lib
Author: Sebastian Kemper <sebastian_ml@gmx.net> Co-authored-by: Peter Eisentraut <peter.eisentraut@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/flat/20200305213827.GA25135%40darth.lan
Since ea53100d5 (PostgreSQL 12), the shipped pkg-config files have
been broken for statically linking libpq because libpgcommon and
libpgport are missing. This patch adds those two missing private
dependencies (in a non-hardcoded way).
Reported-by: Filip Gospodinov <f@gospodinov.ch>
Discussion: https://www.postgresql.org/message-id/flat/c7108bde-e051-11d5-a234-99beec01ce2a@gospodinov.ch
Tom Lane [Thu, 2 Sep 2021 22:53:10 +0000 (18:53 -0400)]
In count_usable_fds(), duplicate stderr not stdin.
We had a complaint that the postmaster fails to start if the invoking
program closes stdin. That happens because count_usable_fds expects
to be able to dup(0), and if it can't, we conclude there are no free
FDs and go belly-up. So far as I can find, though, there is no other
place in the server that touches stdin, and it's not unreasonable to
expect that a daemon wouldn't use that file.
As a simple improvement, let's dup FD 2 (stderr) instead. Unlike stdin,
it *is* reasonable for us to expect that stderr be open; even if we are
configured not to touch it, common libraries such as libc might try to
write error messages there.
Per gripe from Mario Emmenlauer. Given the lack of previous complaints,
I'm not excited about pushing this into stable branches, but it seems
OK to squeeze it into v14.
Tom Lane [Thu, 2 Sep 2021 21:24:41 +0000 (17:24 -0400)]
Fix float4/float8 hash functions to produce uniform results for NaNs.
The IEEE 754 standard allows a wide variety of bit patterns for NaNs,
of which at least two ("NaN" and "-NaN") are pretty easy to produce
from SQL on most machines. This is problematic because our btree
comparison functions deem all NaNs to be equal, but our float hash
functions know nothing about NaNs and will happily produce varying
hash codes for them. That causes unexpected results from queries
that hash a column containing different NaN values. It could also
produce unexpected lookup failures when using a hash index on a
float column, i.e. "WHERE x = 'NaN'" will not find all the rows
it should.
To fix, special-case NaN in the float hash functions, not too much
unlike the existing special case that forces zero and minus zero
to hash the same. I arranged for the most vanilla sort of NaN
(that coming from the C99 NAN constant) to still have the same
hash code as before, to reduce the risk to existing hash indexes.
I dithered about whether to back-patch this into stable branches,
but ultimately decided to do so. It's a clear improvement for
queries that hash internally. If there is anybody who has -NaN
in a hash index, they'd be well advised to re-index after applying
this patch ... but the misbehavior if they don't will not be much
worse than the misbehavior they had before.
Commit a4205fa00 moved setting conn to NULL directly after the call
to PQfinish, but the original conn = NULL; remained a few lines down.
Fix by removing the superfluous assignment.
Amit Kapila [Thu, 2 Sep 2021 02:43:46 +0000 (08:13 +0530)]
Optimize fileset usage in apply worker.
Use one fileset for the entire worker lifetime instead of using
separate filesets for each streaming transaction. Now, the
changes/subxacts files for every streaming transaction will be
created under the same fileset and the files will be deleted
after the transaction is completed.
This patch extends the BufFileOpenFileSet and BufFileDeleteFileSet
APIs to allow users to specify whether to give an error on missing
files.
Author: Dilip Kumar, based on suggestion by Thomas Munro Reviewed-by: Hou Zhijie, Masahiko Sawada, Amit Kapila
Discussion: https://postgr.es/m/E1mCC6U-0004Ik-Fs@gemulon.postgresql.org
Use COPY FREEZE in pgbench for faster benchmark table population.
While populating the pgbench_accounts table, plain COPY was
unconditionally used. By changing it to COPY FREEZE, the time for
VACUUM is significantly reduced, thus the total time of "pgbench -i"
is also reduced. This only happens if pgbench runs against PostgreSQL
14 or later because COPY FREEZE in previous versions of PostgreSQL does
not bring the benefit. Also if partitioning is used, COPY FREEZE
cannot be used. In this case plain COPY will be used too.
Author: Tatsuo Ishii
Discussion: https://postgr.es/m/20210308.143907.2014279678657453983.t-ishii@gmail.com Reviewed-by: Fabien COELHO, Laurenz Albe, Peter Geoghegan, Dean Rasheed
Tom Lane [Wed, 1 Sep 2021 20:21:13 +0000 (16:21 -0400)]
Refactor postgresImportForeignSchema to avoid code duplication.
Avoid repeating fragments of the query we're building, along the
same lines as recent cleanup in pg_dump. I got annoyed about this
because aa769f80e broke my pending patch to change postgres_fdw's
collation handling, due to each of us having incompletely done
this same refactoring. Let's finish that job in hopes of having
a more stable base.
Tomas Vondra [Wed, 1 Sep 2021 15:41:54 +0000 (17:41 +0200)]
Identify simple column references in extended statistics
Until now, when defining extended statistics, everything except a plain
column reference was treated as complex expression. So for example "a"
was a column reference, but "(a)" would be an expression. In most cases
this does not matter much, but there were a couple strange consequences.
For example
CREATE STATISTICS s ON a FROM t;
would fail, because extended stats require at least two columns. But
CREATE STATISTICS s ON (a) FROM t;
would succeed, because that requirement does not apply to expressions.
Moreover, that statistics object is useless - the optimizer will always
use the regular statistics collected for attribute "a".
So do a bit more work to identify those expressions referencing a single
column, and translate them to a simple column reference. Backpatch to
14, where support for extended statistics on expressions was introduced.
pgbench: Fix bug in measurement of disconnection delays.
When -C/--connect option is specified, pgbench establishes and closes
the connection for each transaction. In this case pgbench needs to
measure the times taken for all those connections and disconnections,
to include the average connection time in the benchmark result.
But previously pgbench could not measure those disconnection delays.
To fix the bug, this commit makes pgbench measure the disconnection
delays whenever the connection is closed at the end of transaction,
if -C/--connect option is specified.
Back-patch to v14. Per discussion, we concluded not to back-patch to v13
or before because changing that behavior in stable branches would
surprise users rather than providing benefits.
Amit Kapila [Wed, 1 Sep 2021 04:48:23 +0000 (10:18 +0530)]
Fix the random test failure in 001_rep_changes.
The check to test whether the subscription workers were restarting after a
change in the subscription was failing. The reason was that the test was
assuming the walsender started before it reaches the 'streaming' state and
the walsender was exiting due to an error before that. Now, the walsender
was erroring out before reaching the 'streaming' state because it tries to
acquire the slot before the previous walsender has exited.
In passing, improve the die messages so that it is easier to investigate
the failures in the future if any.
Reported-by: Michael Paquier, as per buildfarm
Author: Ajin Cherian Reviewed-by: Masahiko Sawada, Amit Kapila
Backpatch-through: 10, where this test was introduced
Discussion: https://postgr.es/m/YRnhFxa9bo73wfpV@paquier.xyz
Peter Geoghegan [Wed, 1 Sep 2021 03:37:18 +0000 (20:37 -0700)]
VACUUM VERBOSE: Don't report "pages removed".
It doesn't make any sense to report this information, since VACUUM
VERBOSE reports on heap relation truncation directly. This was an
oversight in commit 7ab96cf6, which made VACUUM VERBOSE output a little
more consistent with nearby autovacuum-specific log output. Adjust
comments that describe how this is supposed to work in passing.
Also bring truncation-related VACUUM VERBOSE output in line with the
convention established for VACUUM VERBOSE output by commit f4f4a649.
Author: Peter Geoghegan <pg@bowt.ie>
Backpatch: 14-, where VACUUM VERBOSE's output changed.
Michael Paquier [Wed, 1 Sep 2021 02:48:08 +0000 (11:48 +0900)]
Refactor one conversion of SQLSTATE to string in elog.c
unpack_sql_state() has been introduced in d46bc44 to refactor the
unpacking of a SQLSTATE into a string, but it forgot one code path when
sending error reports to clients that could make use of it. This
changes the code to also use unpack_sql_state() there, simplifying a bit
the code.
Author: Peter Smith
Discussion: https://postgr.es/m/CAHut+PuYituuD1-VVZUNcmCQuc3ZzZMPoO57POgm8tnXOkwJAA@mail.gmail.com
Michael Paquier [Wed, 1 Sep 2021 01:28:01 +0000 (10:28 +0900)]
Add PostgresNode::command_fails_like()
This is useful to test for a command failure with some default
connection parameters associated to a node, in combination with checks
on error patterns expected. This routine will be used by an upcoming
future patch, but could be also plugged into some of the existing
tests.
Tomas Vondra [Tue, 31 Aug 2021 22:42:32 +0000 (00:42 +0200)]
Don't print extra parens around expressions in extended stats
The code printing expressions for extended statistics doubled the
parens, producing results like ((a+1)), which is unnecessary and not
consistent with how we print expressions elsewhere.
Fixed by tweaking the code to produce just a single set of parens.
Reported by Mark Dilger, fix by me. Backpatch to 14, where support for
extended statistics on expressions was added.
Reported-by: Mark Dilger
Discussion: https://postgr.es/m/20210122040101.GF27167%40telsasoft.com
Peter Geoghegan [Tue, 31 Aug 2021 21:59:52 +0000 (14:59 -0700)]
Remove unneeded old_rel_pages VACUUM state field.
The field hasn't been used since commit 3d351d91, which redefined
pg_class.reltuples to be -1 before the first VACUUM or ANALYZE.
Also rename a local variable of the same name ("old_rel_pages"). This is
used by relation truncation to represent the original relation size at
the start of the ongoing VACUUM operation. Rename it to orig_rel_pages,
since that's a lot clearer. (This name matches similar nearby code.)
John Naylor [Tue, 31 Aug 2021 18:15:22 +0000 (14:15 -0400)]
Mark the timestamptz variant of date_bin() as stable
Previously, it was immutable by lack of marking. This is not
correct, since the time zone could change.
Bump catversion
Discussion: https://www.postgresql.org/message-id/CAFBsxsG2UHk8mOWL0tca%3D_cg%2B_oA5mVRNLhDF0TBw980iOg5NQ%40mail.gmail.com
Backpatch to v14, when this function came in
Tom Lane [Tue, 31 Aug 2021 19:04:05 +0000 (15:04 -0400)]
In pg_dump, avoid doing per-table queries for RLS policies.
For no particularly good reason, getPolicies() queried pg_policy
separately for each table. We can collect all the policies in
a single query instead, and attach them to the correct TableInfo
objects using findTableByOid() lookups. On the regression
database, this reduces the number of queries substantially, and
provides a visible savings even when running against a local
server.
Per complaint from Hubert Depesz Lubaczewski. Since this is such
a simple fix and can have a visible performance benefit, back-patch
to all supported branches.
Tom Lane [Tue, 31 Aug 2021 17:53:33 +0000 (13:53 -0400)]
Cache the results of format_type() queries in pg_dump.
There's long been a "TODO: there might be some value in caching
the results" annotation on pg_dump's getFormattedTypeName function;
but we hadn't gotten around to checking what it was costing us to
repetitively look up type names. It turns out that when dumping the
current regression database, about 10% of the total number of queries
issued are duplicative format_type() queries. However, Hubert Depesz
Lubaczewski reported a not-unusual case where these account for over
half of the queries issued by pg_dump. Individually these queries
aren't expensive, but when network lag is a factor, they add up to a
problem. We can very easily add some caching to getFormattedTypeName
to solve it.
Since this is such a simple fix and can have a visible performance
benefit, back-patch to all supported branches.
Tomas Vondra [Tue, 31 Aug 2021 17:21:29 +0000 (19:21 +0200)]
Rename the role in stats_ext to have regress_ prefix
Commit 5be8ce82e8 added a new role to the stats_ext regression suite,
but the role name did not start with regress_ causing failures when
running with ENFORCE_REGRESSION_TEST_NAME_RESTRICTIONS. Fixed by
renaming the role to start with the expected regress_ prefix.
Backpatch-through: 10, same as the new regression test
Discussion: https://postgr.es/m/1F238937-7CC2-4703-A1B1-6DC225B8978A%40enterprisedb.com
Tomas Vondra [Tue, 31 Aug 2021 16:03:05 +0000 (18:03 +0200)]
Fix lookup error in extended stats ownership check
When an ownership check on extended statistics object failed, the code
was calling aclcheck_error_type to report the failure, which is clearly
wrong, resulting in cache lookup errors. Fix by calling aclcheck_error.
This issue exists since the introduction of extended statistics, so
backpatch all the way back to PostgreSQL 10. It went unnoticed because
there were no tests triggering the error, so add one.
Reported-by: Mark Dilger
Backpatch-through: 10, where extended stats were introduced
Discussion: https://postgr.es/m/1F238937-7CC2-4703-A1B1-6DC225B8978A%40enterprisedb.com
Tom Lane [Tue, 31 Aug 2021 16:02:36 +0000 (12:02 -0400)]
Fix missed lock acquisition while inlining new-style SQL functions.
When starting to use a query parsetree loaded from the catalogs,
we must begin by applying AcquireRewriteLocks(), to obtain the same
relation locks that the parser would have gotten if the query were
entered interactively, and to do some other cleanup such as dealing
with later-dropped columns. New-style SQL functions are just as
subject to this rule as other stored parsetrees; however, of the
places dealing with such functions, only init_sql_fcache had gotten
the memo. In particular, if we successfully inlined a new-style
set-returning SQL function that contained any relation references,
we'd either get an assertion failure or attempt to use those
relation(s) sans locks.
I also added AcquireRewriteLocks calls to fmgr_sql_validator and
print_function_sqlbody. Desultory experiments didn't demonstrate any
failures in those, but I suspect that I just didn't try hard enough.
Certainly we don't expect nearby code paths to operate without locks.
On the same logic of it-ought-to-have-the-same-effects-as-the-old-code,
call pg_rewrite_query() in fmgr_sql_validator, too. It's possible
that neither code path there needs to bother with rewriting, but
doing the analysis to prove that is beyond my goals for today.
map and grep are not intended to be used as mutators, iterating
with side-effects should be done with for or foreach loops. This
fixes the one occurrence of the pattern, and bumps the perlcritic
policy to severity 5 for the map and grep policies.
Author: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org> Reviewed-by: Michael Paquier <michael@paquier.xyz> Reviewed-by: Andrew Dunstan <andrew@dunslane.net> Reviewed-by: Julien Rouhaud <rjuju123@gmail.com>
Discussion: https://postgr.es/m/87fsvzhhc4.fsf@wibble.ilmari.org
Michael Paquier [Tue, 31 Aug 2021 01:19:38 +0000 (10:19 +0900)]
Refactor one use of IDENTIFY_SYSTEM in WAL streaming code of pg_basebackup
0c013e0 has done a large refactoring to unify all the code paths using
replication commands, but forgot one code path doing WAL streaming that
checks the validity of a cluster connecting to with IDENTIFY_SYSTEM.
There is a generic routine able to handle that, so make use of it in
this code path. This impacts pg_receivewal and pg_basebackup.
Alvaro Herrera [Mon, 30 Aug 2021 18:01:29 +0000 (14:01 -0400)]
psql: Fix name quoting on extended statistics
Per our message style guidelines, for human consumption we quote
qualified names as a whole rather than each part separately; but commits bc085205c8a4 introduced a deviation for extended statistics and a4d75c86bf15 copied it. I don't agree with this policy applying to
names shown by psql, but that's a poor reason to deviate from the
practice only in two obscure corners, so make said corners use the same
style as everywhere else.
Backpatch to 14. The first of these is older, but I'm not sure we want
to destabilize the psql output in the older branches for such a small
thing.
Fujii Masao [Mon, 30 Aug 2021 12:35:24 +0000 (21:35 +0900)]
pgbench: Avoid unnecessary measurement of connection delays.
Commit 547f04e734 changed pgbench so that it used the measurement result
of connection delays in its benchmark report only when -C/--connect option
is specified. But previously those delays were unnecessarily measured
even when that option is not specified. Which was a waste of cycles.
This commit improves pgbench so that it avoids such unnecessary measurement.
Back-patch to v14 where commit 547f04e734 first appeared.
Amit Kapila [Mon, 30 Aug 2021 03:44:31 +0000 (09:14 +0530)]
Fix incorrect error code in StartupReplicationOrigin().
ERRCODE_CONFIGURATION_LIMIT_EXCEEDED was used for checksum failure, use
ERRCODE_DATA_CORRUPTED instead.
Reported-by: Tatsuhito Kasahara
Author: Tatsuhito Kasahara
Backpatch-through: 9.6, where it was introduced
Discussion: https://postgr.es/m/CAP0=ZVLHtYffs8SOWcFJWrBGoRzT9QQbk+_aP+E5AHLNXiOorA@mail.gmail.com
Amit Kapila [Mon, 30 Aug 2021 03:15:35 +0000 (08:45 +0530)]
Refactor sharedfileset.c to separate out fileset implementation.
Move fileset related implementation out of sharedfileset.c to allow its
usage by backends that don't want to share filesets among different
processes. After this split, fileset infrastructure is used by both
sharedfileset.c and worker.c for the named temporary files that survive
across transactions.
Author: Dilip Kumar, based on suggestion by Andres Freund Reviewed-by: Hou Zhijie, Masahiko Sawada, Amit Kapila
Discussion: https://postgr.es/m/E1mCC6U-0004Ik-Fs@gemulon.postgresql.org
Michael Paquier [Mon, 30 Aug 2021 00:46:20 +0000 (09:46 +0900)]
Add more tab completion support for ALTER TABLE ADD in psql
This includes the detection of new patterns for various constraint
types, with the addition of USING INDEX for unique indexes of a table
on primary keys and unique constraints.
Author: Dagfinn Ilmari Mannsåker
Discussion: https://postgr.es/m/87bl6ehhpl.fsf@wibble.ilmari.org
Tom Lane [Sun, 29 Aug 2021 16:48:49 +0000 (12:48 -0400)]
Doc: add a little about LACON execution to src/backend/regex/README.
I wrote this while thinking about a possible optimization, but it's
a useful description of the existing code regardless of whether the
optimization ever happens. So push it separately.
Alvaro Herrera [Sat, 28 Aug 2021 19:58:23 +0000 (15:58 -0400)]
Keep stats up to date for partitioned tables
In the long-going saga for analyze on partitioned tables, one thing I
missed while reverting 0827e8af70f4 is the maintenance of analyze count
and last analyze time for partitioned tables. This is a mostly trivial
change that enables users assess the need for invoking manual ANALYZE on
partitioned tables.
This patch, posted by Justin and modified a bit by me (Álvaro), can be
mostly traced back to Hosoya-san, though any problems introduced with
the scissors are mine.
Alvaro Herrera [Sat, 28 Aug 2021 16:04:15 +0000 (12:04 -0400)]
psql \dX: reference regclass with "pg_catalog." prefix
Déjà vu of commit fc40ba1296a7, for another backslash command.
Strictly speaking this isn't a bug, but since all references to catalog
objects are schema-qualified, we might as well be consistent. The
omission first appeared in commit ad600bba0422 and replicated in a4d75c86bf15; backpatch to 14.
Alvaro Herrera [Sat, 28 Aug 2021 15:45:47 +0000 (11:45 -0400)]
psql \dP: reference regclass with "pg_catalog." prefix
Strictly speaking this isn't a bug, but since all references to catalog
objects are schema-qualified, we might as well be consistent. The
omission first appeared in commit 1c5d9270e339, so backpatch to 12.
Noah Misch [Sat, 28 Aug 2021 06:33:23 +0000 (23:33 -0700)]
Fix data loss in wal_level=minimal crash recovery of CREATE TABLESPACE.
If the system crashed between CREATE TABLESPACE and the next checkpoint,
the result could be some files in the tablespace unexpectedly containing
no rows. Affected files would be those for which the system did not
write WAL; see the wal_skip_threshold documentation. Before v13, a
different set of conditions governed the writing of WAL; see v12's
<sect2 id="populate-pitr">. (The v12 conditions were broader in some
ways and narrower in others.) Users may want to audit non-default
tablespaces for unexpected short files. The bug could have truncated an
index without affecting the associated table, and reindexing the index
would fix that particular problem.
This fixes the bug by making create_tablespace_directories() more like
TablespaceCreateDbspace(). create_tablespace_directories() was
recursively removing tablespace contents, reasoning that WAL redo would
recreate everything removed that way. That assumption holds for other
wal_level values. Under wal_level=minimal, the old approach could
delete files for which no other copy existed. Back-patch to 9.6 (all
supported versions).
Reviewed by Robert Haas and Prabhat Sahu. Reported by Robert Haas.
Tom Lane [Fri, 27 Aug 2021 23:42:42 +0000 (19:42 -0400)]
Count SP-GiST index scans in pg_stat statistics.
Somehow, spgist overlooked the need to call pgstat_count_index_scan().
Hence, pg_stat_all_indexes.idx_scan and equivalent columns never
became nonzero for an SP-GiST index, although the related per-tuple
counters worked fine.
This fix works a bit differently from other index AMs, in that the
counter increment occurs in spgrescan not spggettuple/spggetbitmap.
It looks like this won't make the user-visible semantics noticeably
different, so I won't go to the trouble of introducing an is-this-
the-first-call flag just to make the counter bumps happen in the
same places.
Per bug #17163 from Christian Quest. Back-patch to all supported
versions.
Stephen Frost [Fri, 27 Aug 2021 23:23:14 +0000 (19:23 -0400)]
Use maintenance_io_concurrency for ANALYZE prefetch
When prefetching pages for ANALYZE, we should be using
maintenance_io_concurrenty (by calling
get_tablespace_maintenance_io_concurrency(), not
get_tablespace_io_concurrency()).
ANALYZE prefetching was introduced in c6fc50c, so back-patch to 14.
Stephen Frost [Fri, 27 Aug 2021 22:25:26 +0000 (18:25 -0400)]
docs: Add command tags for SQL commands
Commit 6c3ffd6 added a couple new predefined roles but didn't properly
wrap the SQL commands mentioned in the description of those roles with
command tags, so add them now.
Backpatch-through: 14 Reported-by: Michael Banck
Discussion: https://postgr.es/m/606d8b1c.1c69fb81.3df04.1a99@mx.google.com
Author: Dave Cramer <davecramer@gmail.com> Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/CADK3HHLZmqAQZ2ByPDQQ9yhGqax36kksq6sDkV0yYzsxw6ipvQ@mail.gmail.com
Peter Geoghegan [Fri, 27 Aug 2021 20:34:00 +0000 (13:34 -0700)]
track_io_timing logging: Don't special case 0 ms.
Adjust track_io_timing related logging code added by commit 94d13d474d.
Make it consistent with other nearby autovacuum and autoanalyze logging
code by removing logic that suppressed zero millisecond outputs.
log_autovacuum_min_duration log output now reliably shows "read:" and
"write:" millisecond-based values in its report (when track_io_timing is
enabled).
Author: Peter Geoghegan <pg@bowt.ie> Reviewed-By: Stephen Frost <sfrost@snowman.net>
Discussion: https://postgr.es/m/CAH2-WznW0FNxSVQMSRazAMYNfZ6DR_gr5WE78hc6E1CBkkJpzw@mail.gmail.com
Backpatch: 14-, where the track_io_timing logging was introduced.
Peter Geoghegan [Fri, 27 Aug 2021 20:08:41 +0000 (13:08 -0700)]
Reorder log_autovacuum_min_duration log output.
This order seems more natural. It starts with details that are
particular to heap and index data structures, and ends with system-level
costs incurred during the autovacuum worker's VACUUM/ANALYZE operation.
Author: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/CAH2-WzkzxK6ahA9xxsOftRtBX_R0swuHZsvo4QUbak1Bz7hb7Q@mail.gmail.com
Backpatch: 14-, which enhanced the log output in various ways.
Peter Geoghegan [Fri, 27 Aug 2021 16:47:16 +0000 (09:47 -0700)]
vacuumlazy.c: Remove unnecessary parentheses.
This was arguably a minor oversight in commit b4af70cb, which cleaned up
the function signatures of functions that modify IndexBulkDeleteResult
variables.
Peter Eisentraut [Fri, 27 Aug 2021 16:15:57 +0000 (18:15 +0200)]
Change Texinfo output to UTF-8
Since the whole documentation tool chain is now UTF-8 and there is an
increasing number of non-ISO-8859-1 characters in the text, keeping
the Texinfo output in ISO 8859-1 just creates unnecessary
complications. Depending on the platform, there are conversion
failures and thus build failures, or weirdly converted characters. By
changing the output to UTF-8, the whole encoding conversion business
is sidestepped.
Tom Lane [Fri, 27 Aug 2021 16:18:58 +0000 (12:18 -0400)]
Handle interaction of regexp's makesearch and MATCHALL more honestly.
Second thoughts about commit 824bf7190: we apply makesearch() to
an NFA after having determined whether it is a MATCHALL pattern.
Prepending ".*" doesn't make it non-MATCHALL, but it does change the
maximum possible match length, and makesearch() failed to update that.
This has no ill effects given the stylized usage of search NFAs, but
it seems like it's better to keep the data structure consistent. In
particular, fixing this allows more honest handling of the MATCHALL
check in matchuntil(): we can now assert that maxmatchall is infinity,
instead of lamely assuming that it should act that way.
In passing, improve the code in dump[c]nfa so that infinite maxmatchall
is printed as "inf" not a magic number.
When looping over the resultset from a SQL query, extracting the field
number before the loop body to avoid repeated calls to PQfnumber is an
established pattern. On very wide tables this can have a performance
impact, but it wont be noticeable in the common case. This fixes a few
queries which were extracting the field number in the loop body.
Amit Kapila [Fri, 27 Aug 2021 03:00:23 +0000 (08:30 +0530)]
Add logical change details to logical replication worker errcontext.
Previously, on the subscriber, we set the error context callback for the
tuple data conversion failures. This commit replaces the existing error
context callback with a comprehensive one so that it shows not only the
details of data conversion failures but also the details of logical change
being applied by the apply worker or table sync worker. The additional
information displayed will be the command, transaction id, and timestamp.
The error context is added to an error only when applying a change but not
while doing other work like receiving data etc.
This will help users in diagnosing the problems that occur during logical
replication. It also can be used for future work that allows skipping a
particular transaction on the subscriber.
Add a CHECK_FOR_INTERRUPTS() call to make heap relation verification
responsive to query cancellations.
Author: Mark Dilger <mark.dilger@enterprisedb.com>
Discussion: https://postgr.es/m/CAH2-Wzk-9RtQgb2QiuLv8j2O0j9tSFKPmmch5nWSZhguUxvbrw%40mail.gmail.com
Backpatch: 14-, where amcheck heap verification was introduced.
John Naylor [Thu, 26 Aug 2021 17:07:34 +0000 (13:07 -0400)]
Extend collection of Unicode combining characters to beyond the BMP
The former limit was perhaps a carryover from an older hand-coded
table. Since commit bab982161 we have enough space in mbinterval to
store larger codepoints, so collect all combining characters.
John Naylor [Thu, 26 Aug 2021 14:53:56 +0000 (10:53 -0400)]
Update display widths as part of updating Unicode
The hardcoded "wide character" set in ucs_wcwidth() was last updated
around the Unicode 5.0 era. This led to misalignment when printing
emojis and other codepoints that have since been designated
wide or full-width.
To fix and keep up to date, extend update-unicode to download the list
of wide and full-width codepoints from the offical sources.
In passing, remove some comments about non-spacing characters that
haven't been accurate since we removed the former hardcoded logic.
Jacob Champion
Reported and reviewed by Pavel Stehule
Discussion: https://www.postgresql.org/message-id/flat/CAFj8pRCeX21O69YHxmykYySYyprZAqrKWWg0KoGKdjgqcGyygg@mail.gmail.com
After I had committed eb0d0d2c7 and 78ab944cd, I decided to add
a sanity check for a "can't happen" scenario just to be cautious.
It turned out that it already happened in the official Unicode source
data, namely that a character can be both wide and a combining
character. This fact renders the aforementioned commits unnecessary,
so revert both of them.
After I had committed eb0d0d2c7 and 78ab944cd, I decided to add
a sanity check for a "can't happen" scenario just to be cautious.
It turned out that it already happened in the official Unicode source
data, namely that a character can be both wide and a combining
character. This fact renders the aforementioned commits unnecessary,
so revert both of them.
Peter Eisentraut [Thu, 26 Aug 2021 06:48:07 +0000 (08:48 +0200)]
Fix handling of partitioned index in RelationGetNumberOfBlocksInFork()
Since a partitioned index doesn't have storage, getting the number of
blocks from it will not give sensible results. Existing callers
already check that they don't call it that way, so there doesn't
appear to be a live problem. But for correctness, handle
RELKIND_PARTITIONED_INDEX together with the other non-storage
relkinds.
Reviewed-by: Michael Paquier <michael@paquier.xyz> Reviewed-by: Alvaro Herrera <alvherre@alvh.no-ip.org>
Discussion: https://www.postgresql.org/message-id/1d3a5fbe-f48b-8bea-80da-9a5c4244aef9@enterprisedb.com
John Naylor [Fri, 20 Aug 2021 15:10:17 +0000 (11:10 -0400)]
Change mbbisearch to return the character range
Add a width field to mbinterval and have mbbisearch return a
pointer to the found range rather than just bool for success.
A future commit will add another width besides zero, and this
will allow that to use the same search.
Reviewed by Jacob Champion
Discussion: https://www.postgresql.org/message-id/CAFBsxsGOCpzV7c-f3a8ADsA1n4uZ%3D8puCctQp%2Bx7W0vgkv%3Dw%2Bg%40mail.gmail.com
Tom Lane [Wed, 25 Aug 2021 15:06:34 +0000 (11:06 -0400)]
Remove redundant test.
The condition "context_start < context_end" is strictly weaker
than "context_end - context_start >= 50", so we don't need both.
Oversight in commit ffd3944ab, noted by tanghy.fnst.
In passing, line-wrap a nearby test to make it more readable.
Robert Haas [Wed, 25 Aug 2021 12:32:04 +0000 (08:32 -0400)]
Fix broken snapshot handling in parallel workers.
Pengchengliu reported an assertion failure in a parallel woker while
performing a parallel scan using an overflowed snapshot. The proximate
cause is that TransactionXmin was set to an incorrect value. The
underlying cause is incorrect snapshot handling in parallel.c.
In particular, InitializeParallelDSM() was unconditionally calling
GetTransactionSnapshot(), because I (rhaas) mistakenly thought that
was always retrieving an existing snapshot whereas, at isolation
levels less than REPEATABLE READ, it's actually taking a new one. So
instead do this only at higher isolation levels where there actually
is a single snapshot for the whole transaction.
By itself, this is not a sufficient fix, because we still need to
guarantee that TransactionXmin gets set properly in the workers. The
easiest way to do that seems to be to install the leader's active
snapshot as the transaction snapshot if the leader did not serialize a
transaction snapshot. This doesn't affect the results of future
GetTrasnactionSnapshot() calls since those have to take a new snapshot
anyway; what we care about is the side effect of setting TransactionXmin.
Report by Pengchengliu. Patch by Greg Nancarrow, except for some comment
text which I supplied.
Peter Eisentraut [Wed, 25 Aug 2021 09:54:58 +0000 (11:54 +0200)]
psql: Make cancel test more timing robust
The previous coding relied on the PID file appearing and the query
starting "fast enough", which can fail on slow machines. Also, there
might have been an undocumented interference between alarm and
IPC::Run. This new coding doesn't rely on any of these concurrency
mechanisms. Instead, we wait unitl the PID file is complete before
proceeding, and then also wait until the sleep query is registered by
the server.
Michael Paquier [Wed, 25 Aug 2021 06:16:31 +0000 (15:16 +0900)]
Fix incorrect merge in ECPG code with DECLARE
The same condition was repeated twice when comparing the connection used
by existing declared statement with the one coming from a fresh DECLARE
statement. This had no consequences, but let's keep the code clean.
Oversight in f576de1.
Author: Shenhao Wang
Discussion: https://postgr.es/m/OSBPR01MB42149653BC0AB0A49D23C1B8F2C69@OSBPR01MB4214.jpnprd01.prod.outlook.com
Backpatch-through: 14
Amit Kapila [Wed, 25 Aug 2021 04:23:07 +0000 (09:53 +0530)]
Fix toast rewrites in logical decoding.
Commit 325f2ec555 introduced pg_class.relwrite to skip operations on
tables created as part of a heap rewrite during DDL. It links such
transient heaps to the original relation OID via this new field in
pg_class but forgot to do anything about toast tables. So, logical
decoding was not able to skip operations on internally created toast
tables. This leads to an error when we tried to decode the WAL for the
next operation for which it appeared that there is a toast data where
actually it didn't have any toast data.
To fix this, we set pg_class.relwrite for internally created toast tables
as well which allowed skipping operations on them during logical decoding.
Author: Bertrand Drouvot Reviewed-by: David Zhang, Amit Kapila
Backpatch-through: 11, where it was introduced
Discussion: https://postgr.es/m/b5146fb1-ad9e-7d6e-f980-98ed68744a7c@amazon.com
Fujii Masao [Wed, 25 Aug 2021 02:46:25 +0000 (11:46 +0900)]
Avoid using ambiguous word "positive" in error message.
There are two identical error messages about valid value of modulus for
hash partition, in PostgreSQL source code. Commit 0e1275fb07 improved
only one of them so that ambiguous word "positive" was avoided there,
and forgot to improve the other. This commit improves the other.
Which would reduce translator burden.
Fujii Masao [Wed, 25 Aug 2021 02:43:56 +0000 (11:43 +0900)]
Improve error message about valid value for distance in phrase operator.
The distance in phrase operator must be an integer value between zero
and MAXENTRYPOS inclusive. But previously the error message about
its valid value included the information about its upper limit
but not lower limit (i.e., zero). This commit improves the error message
so that it also includes the information about its lower limit.
Back-patch to v9.6 where full-text phrase search was supported.
projects / thirdparty / postgresql.git / log
