This commit ensures that the wait interval in the replay delay loop
waiting for an amount of time defined by recovery_min_apply_delay is
correctly handled on reload, recalculating the delay if this GUC value
is updated, based on the timestamp of the commit record being replayed.
The previous behavior would be problematic for example with replay
still waiting even if the delay got reduced or just cancelled. If the
apply delay was increased to a larger value, the wait would have just
respected the old value set, finishing earlier.
Author: Soumyadeep Chakraborty, Ashwin Agrawal
Reviewed-by: Kyotaro Horiguchi, Michael Paquier
Discussion: https://postgr.es/m/CAE-ML+93zfr-HLN8OuxF0BjpWJ17O5dv1eMvSE5jsj9jpnAXZA@mail.gmail.com
Backpatch-through: 9.6
Like the ARM case, just use gcc's __sync_lock_test_and_set();
that will compile into AMOSWAP.W.AQ which does what we need.
At some point it might be worth doing some work on atomic ops
for RISC-V, but this should be enough for a creditable port.
Back-patch to all supported branches, just in case somebody
wants to try them on RISC-V.
Marek Szuba
Discussion: https://postgr.es/m/dea97b6d-f55f-1f6d-9109-504aa7dfa421@gentoo.org
This fixes a bug in simplehash.h which caused an incorrect size mask to be
used when the hash table grew to SH_MAX_SIZE (2^32). The code was
incorrectly setting the size mask to 0 when the hash tables reached the
maximum possible number of buckets. This would result always trying to
use the 0th bucket causing an infinite loop of trying to grow the hash
table due to there being too many collisions.
Seemingly it's not that common for simplehash tables to ever grow this big
as this bug dates back to v10 and nobody seems to have noticed it before.
However, probably the most likely place that people would notice it would
be doing a large in-memory Hash Aggregate with something close to at least
2^31 groups.
After this fix, the code now works correctly with up to within 98% of 2^32
groups and will fail with the following error when trying to insert any
more items into the hash table:
ERROR: hash table size exceeded
However, the work_mem (or hash_mem_multiplier in newer versions) settings
will generally cause Hash Aggregates to spill to disk long before reaching
that many groups. The minimal test case I did took a work_mem setting of
over 192GB to hit the bug.
simplehash hash tables are used in a few other places such as Bitmap Index
Scans, however, again the size that the hash table can become there is
also limited to work_mem and it would take a relation of around 16TB
(2^31) pages and a very large work_mem setting to hit this. With smaller
work_mem values the table would become lossy and never grow large enough
to hit the problem.
Author: Yura Sokolov
Reviewed-by: David Rowley, Ranier Vilela
Discussion: https://postgr.es/m/b1f7f32737c3438136f64b26f4852b96@postgrespro.ru
Backpatch-through: 10, where simplehash.h was added
It's hard to disable ASLR on current macOS releases, for testing with
-DEXEC_BACKEND. You could already set the environment variable
PG_SHMEM_ADDR to something not likely to collide with mappings created
earlier in process startup. Let's also provide a default value that
works on current releases and architectures, for developer convenience.
As noted in the pre-existing comment, this is a horrible hack, but
-DEXEC_BACKEND is only used by Unix-based PostgreSQL developers for
testing some otherwise Windows-only code paths, so it seems excusable.
Back-patch to all supported branches.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/20210806032944.m4tz7j2w47mant26%40alap3.anarazel.de
As a result of confusion about whether the "char" type is signed or
unsigned, scans for index searches like "col < 'x'" or "col <= 'x'"
would start at the middle of the index not the left end, thus missing
many or all of the entries they should find. Fortunately, this
is not a symptom of index corruption. It's only the search logic
that is broken, and we can fix it without unpleasant side-effects.
Per report from Jason Kim. This has been wrong since btree_gin's
beginning, so back-patch to all supported branches.
Discussion: https://postgr.es/m/20210810001649.htnltbh7c63re42p@jasonk.me
In ec34040af I added a mention that there was no point in setting
maintenance_work_limit to anything higher than 1GB for vacuum, but that
was incorrect as ginInsertCleanup() also looks at what
maintenance_work_mem is set to during VACUUM and that's not limited to
1GB.
Here I attempt to make it more clear that the limitation is only around
the number of dead tuple identifiers that we can collect during VACUUM.
I've also added a note to autovacuum_work_mem to mention this limitation.
I didn't do that in ec34040af as I'd had some wrong-headed ideas about
just limiting the maximum value for that GUC to 1GB.
Author: David Rowley
Discussion: https://postgr.es/m/CAApHDvpGwOAvunp-E-bN_rbAs3hmxMoasm5pzkYDbf36h73s7w@mail.gmail.com
Backpatch-through: 9.6, same as ec34040af
Since commit e462856a7a, pg_upgrade automatically creates a script to
update extensions, so mention that instead of ALTER EXTENSION.
Backpatch-through: 9.6
Copy-and-pasteo in 665c5855e, evidently. The 9.6 docs toolchain
whined about duplicate index entries, though our modern toolchain
doesn't. In any case, these GUCs surely are not about the
default settings of these values.
Rather than trying to pick table aliases that won't conflict with
any possible user-defined matview column name, adjust the queries'
syntax so that the aliases are only used in places where they can't be
mistaken for column names. Mostly this consists of writing "alias.*"
not just "alias", which adds clarity for humans as well as machines.
We do have the issue that "SELECT alias.*" acts differently from
"SELECT alias", but we can use the same hack ruleutils.c uses for
whole-row variables in SELECT lists: write "alias.*::compositetype".
We might as well revert to the original aliases after doing this;
they're a bit easier to read.
Like 75d66d10e, back-patch to all supported branches.
Discussion: https://postgr.es/m/2488325.1628261320@sss.pgh.pa.us
Formerly, the numeric code tested whether an integer value of a larger
type would fit in a smaller type by casting it to the smaller type and
then testing if the reverse conversion produced the original value.
That's perfectly fine, except that it caused a test failure on
buildfarm animal castoroides, most likely due to a compiler bug.
Instead, do these tests by comparing against PG_INT16/32_MIN/MAX. That
matches existing code in other places, such as int84(), which is more
widely tested, and so is less likely to go wrong.
While at it, add regression tests covering the numeric-to-int8/4/2
conversions, and adjust the recently added tests to the style of
434ddfb79a (on the v11 branch) to make failures easier to diagnose.
Per buildfarm via Tom Lane, reviewed by Tom Lane.
Discussion: https://postgr.es/m/2394813.1628179479%40sss.pgh.pa.us
postgres_fdw imported generated columns from the remote tables as plain
columns, and caused failures like "ERROR: cannot insert a non-DEFAULT
value into column "foo"" when inserting into the foreign tables, as it
tried to insert values into the generated columns. To fix, we do the
following under the assumption that generated columns in a postgres_fdw
foreign table are defined so that they represent generated columns in
the underlying remote table:
* Send DEFAULT for the generated columns to the foreign server on insert
or update, not generated column values computed on the local server.
* Add to postgresImportForeignSchema() an option "import_generated" to
include column generated expressions in the definitions of foreign
tables imported from a foreign server. The option is true by default.
The assumption seems reasonable, because that would make a query of the
postgres_fdw foreign table return values for the generated columns that
are consistent with the generated expression.
While here, fix another issue in postgresImportForeignSchema(): it tried
to include column generated expressions as column default expressions in
the foreign table definitions when the import_default option was enabled.
Per bug #16631 from Daniel Cherniy. Back-patch to v12 where generated
columns were added.
Discussion: https://postgr.es/m/16631-e929fe9db0ffc7cf%40postgresql.org
This fixes a long-standing bug when using to_char() to format a
numeric value in scientific notation -- if the value's exponent is
less than -NUMERIC_MAX_DISPLAY_SCALE-1 (-1001), it produced a
division-by-zero error.
The reason for this error was that get_str_from_var_sci() divides its
input by 10^exp, which it produced using power_var_int(). However, the
underflow test in power_var_int() causes it to return zero if the
result scale is too small. That's not a problem for power_var_int()'s
only other caller, power_var(), since that limits the rscale to 1000,
but in get_str_from_var_sci() the exponent can be much smaller,
requiring a much larger rscale. Fix by introducing a new function to
compute 10^exp directly, with no rscale limit. This also allows 10^exp
to be computed more efficiently, without any numeric multiplication,
division or rounding.
Discussion: https://postgr.es/m/CAEZATCWhojfH4whaqgUKBe8D5jNHB8ytzemL-PnRx+KCTyMXmg@mail.gmail.com
As of commit 84f5c2908, executing SQL commands (via SPI or otherwise)
requires having either an active Portal, or a caller-established
active snapshot. We were simply Assert'ing that that's the case.
But we've now had a couple different reports of people testing
extensions that didn't meet this requirement, and were confused by
the resulting crash. Let's convert the Assert to a test-and-elog,
in hopes of making the issue clearer for extension authors.
Per gripes from Liu Huailing and RekGRpth. Back-patch to v11,
like the prior commit.
Discussion: https://postgr.es/m/OSZPR01MB6215671E3C5956A034A080DFBEEC9@OSZPR01MB6215.jpnprd01.prod.outlook.com
Discussion: https://postgr.es/m/17035-14607d308ac8643c@postgresql.org
This fixes a couple of related problems that arise when raising
numbers to very large powers.
Firstly, when raising a negative number to a very large integer power,
the result should be well-defined, but the previous code would only
cope if the exponent was small enough to go through power_var_int().
Otherwise it would throw an internal error, attempting to take the
logarithm of a negative number. Fix this by adding suitable handling
to the general case in power_var() to cope with negative bases,
checking for integer powers there.
Next, when raising a (positive or negative) number whose absolute
value is slightly less than 1 to a very large power, the result should
approach zero as the power is increased. However, in some cases, for
sufficiently large powers, this would lose all precision and return 1
instead of 0. This was due to the way that the local_rscale was being
calculated for the final full-precision calculation:
local_rscale = rscale + (int) val - ln_dweight + 8
The first two terms on the right hand side are meant to give the
number of significant digits required in the result ("val" being the
estimated result weight). However, this failed to account for the fact
that rscale is clipped to a maximum of NUMERIC_MAX_DISPLAY_SCALE
(1000), and the result weight might be less then -1000, causing their
sum to be negative, leading to a loss of precision. Fix this by
forcing the number of significant digits calculated to be nonnegative.
It's OK for it to be zero (when the result weight is less than -1000),
since the local_rscale value then includes a few extra digits to
ensure an accurate result.
Finally, add additional underflow checks to exp_var() and power_var(),
so that they consistently return zero for cases like this where the
result is indistinguishable from zero. Some paths through this code
already returned zero in such cases, but others were throwing overflow
errors.
Dean Rasheed, reviewed by Yugo Nagata.
Discussion: http://postgr.es/m/CAEZATCW6Dvq7+3wN3tt5jLj-FyOcUgT5xNoOqce5=6Su0bCR0w@mail.gmail.com
Buildfarm shows that this test has a further failure mode when a
checkpoint starts earlier than expected, so we detect a "checkpoint
completed" line that's not the one we want. Change the config to try
and prevent this.
Per buildfarm
Author: Kyotaro Horiguchi <horikyota.ntt@gmail.com>
Discussion: https://postgr.es/m/20210729.162038.534808353849568395.horikyota.ntt@gmail.com
pg_verifybackup needs by default pg_waldump to check after a range of
WAL segments required for a backup, except if --no-parse-wal is
specified. The code checked for the presence of the binary pg_waldump
in an installation and reported an error, but it forgot to properly
exit(). This could lead to confusing errors reported.
Reviewed-by: Robert Haas, Fabien Coelho
Discussion: https://postgr.es/m/YQDMdB+B68yePFeT@paquier.xyz
Backpatch-through: 13
If a file is truncated, we must update minRecoveryPoint. Once a file is
truncated, there's no going back; it would not be safe to stop recovery
at a point earlier than that anymore.
Commit 7bffc9b7bf changed xact_redo_commit() so that it updates
minRecoveryPoint on truncation, but forgot to change xact_redo_abort().
Back-patch to all supported versions.
Reported-by: mengjuan.cmj@alibaba-inc.com
Author: Fujii Masao
Reviewed-by: Heikki Linnakangas
Discussion: https://postgr.es/m/b029fce3-4fac-4265-968e-16f36ff4d075.mengjuan.cmj@alibaba-inc.com
It wasn't all that clear which lock levels, if any, would be held on the
DEFAULT partition during an ATTACH PARTITION operation.
Also, clarify which locks will be taken if the DEFAULT partition or the
table being attached are themselves partitioned tables.
Here I'm only backpatching to v12 as before then we obtained an ACCESS
EXCLUSIVE lock on the partitioned table. It seems much less relevant to
mention which locks are taken on other tables when the partitioned table
itself is locked with an ACCESS EXCLUSIVE lock.
Author: Matthias van de Meent, David Rowley
Discussion: https://postgr.es/m/CAEze2WiTB6iwrV8W_J=fnrnZ7fowW3qu-8iQ8zCHP3FiQ6+o-A@mail.gmail.com
Backpatch-through: 12
This changes the behavior of examining the pg_file_settings view after
changing a config option that requires restart. The user needs to know
that any change of such options does not take effect until a restart,
and this worked correctly if the line is edited without removing it.
However, for the case where the line is removed altogether, the flag
doesn't get set, because a flag was only set in set_config_option, but
that's not called for lines removed. Repair.
(Ref.: commits 62d16c7fc5 and a486e35706)
Author: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/202107262302.xsfdfc5sb7sh@alvherre.pgsql
The error messages using the word "non-negative" are confusing
because it's ambiguous about whether it accepts zero or not.
This commit improves those error messages by replacing it with
less ambiguous word like "greater than zero" or
"greater than or equal to zero".
Also this commit added the note about the word "non-negative" to
the error message style guide, to help writing the new error messages.
When postgres_fdw option fetch_size was set to zero, previously
the error message "fetch_size requires a non-negative integer value"
was reported. This error message was outright buggy. Therefore
back-patch to all supported versions where such buggy error message
could be thrown.
Reported-by: Hou Zhijie
Author: Bharath Rupireddy
Reviewed-by: Kyotaro Horiguchi, Fujii Masao
Discussion: https://postgr.es/m/OS0PR01MB5716415335A06B489F1B3A8194569@OS0PR01MB5716.jpnprd01.prod.outlook.com
Add pg_resetxlog -u option to set the oldest xid in pg_control.
Previously -x set this value be -2 billion less than the -x value.
However, this causes the server to immediately scan all relation's
relfrozenxid so it can advance pg_control's oldest xid to be inside the
autovacuum_freeze_max_age range, which is inefficient and might disrupt
diagnostic recovery. pg_upgrade will use this option to better create
the new cluster to match the old cluster.
Reported-by: Jason Harvey, Floris Van Nee
Discussion: https://postgr.es/m/20190615183759.GB239428@rfd.leadboat.com, 87da83168c644fd9aae38f546cc70295@opammb0562.comp.optiver.com
Author: Bertrand Drouvot
Backpatch-through: 9.6
These have been introduced by 7fbe0c8, and could happen for
pg_basebackup and pg_receivewal.
Per report from Coverity for the ones in walmethods.c, I have spotted
the ones in receivelog.c after more review.
Backpatch-through: 10
The point of introducing the hash_mem_multiplier GUC was to let users
reproduce the old behavior of hash aggregation, i.e. that it could use
more than work_mem at need. However, the implementation failed to get
the job done on Win64, where work_mem is clamped to 2GB to protect
various places that calculate memory sizes using "long int". As
written, the same clamp was applied to hash_mem. This resulted in
severe performance regressions for queries requiring a bit more than
2GB for hash aggregation, as they now spill to disk and there's no
way to stop that.
Getting rid of the work_mem restriction seems like a good idea, but
it's a big job and could not conceivably be back-patched. However,
there's only a fairly small number of places that are concerned with
the hash_mem value, and it turns out to be possible to remove the
restriction there without too much code churn or any ABI breaks.
So, let's do that for now to fix the regression, and leave the
larger task for another day.
This patch does introduce a bit more infrastructure that should help
with the larger task, namely pg_bitutils.h support for working with
size_t values.
Per gripe from Laurent Hasson. Back-patch to v13 where the
behavior change came in.
Discussion: https://postgr.es/m/997817.1627074924@sss.pgh.pa.us
Discussion: https://postgr.es/m/MN2PR15MB25601E80A9B6D1BA6F592B1985E39@MN2PR15MB2560.namprd15.prod.outlook.com
This commit changes the startup process in the standby server so that
it handles the interrupt signals after waiting for wal_retrieve_retry_interval
on the latch and resetting it, before entering another wait on the latch.
This change causes the standby server to promptly handle interrupt signals.
Otherwise, previously, there was the case where the standby needs to
wait extra five seconds to shutdown when the shutdown request arrived
while the startup process was waiting for wal_retrieve_retry_interval
on the latch.
Author: Fujii Masao, but implementation idea is from Soumyadeep Chakraborty
Reviewed-by: Soumyadeep Chakraborty
Discussion: https://postgr.es/m/9d7e6ab0-8a53-ddb9-63cd-289bcb25fe0e@oss.nttdata.com
Per discussion of BUG #17073, back-patch to all supported versions.
Discussion: https://postgr.es/m/17073-1a5fdaed0fa5d4d0@postgresql.org
We have an implementation restriction that PREPARE TRANSACTION can't
handle cases where both session-lifespan and transaction-lifespan locks
are held on the same lockable object. (That's because we'd otherwise
need to acquire a new PROCLOCK entry during post-prepare cleanup, which
is an operation that might fail. The situation can only arise with odd
usages of advisory locks, so removing the restriction is probably not
worth the amount of effort it would take.) AtPrepare_Locks attempted
to enforce this, but its logic was many bricks shy of a load, because
it only detected cases where the session and transaction locks had the
same lockmode. Locks of different modes on the same object would lead
to the rather unhelpful message "PANIC: we seem to have dropped a bit
somewhere".
To fix, build a transient hashtable with one entry per locktag,
not one per locktag + mode, and use that to detect conflicts.
Per bug #17122 from Alexander Pyhalov. This bug is ancient,
so back-patch to all supported branches.
Discussion: https://postgr.es/m/17122-04f3c32098a62233@postgresql.org
We previously took a hard-line attitude that callers should never print
a null string pointer, and doing so is worthy of an assertion failure
or crash. However, we've long since flushed out any easy-to-find bugs
of that nature. What remains is a lot of code that perhaps could fail
that way in hard-to-reach corner cases. For example, in something as
simple as
ereport(ERROR,
(errcode(ERRCODE_UNDEFINED_OBJECT),
errmsg("constraint \"%s\" for table \"%s\" does not exist",
conname, get_rel_name(relid))));
one must wonder whether it's completely guaranteed that get_rel_name
cannot return NULL in this context. If such a situation did occur,
the existing policy converts what might be a pretty minor bug into
a server crash condition. This is not good for robustness.
Hence, let's follow the lead of glibc and print "(null)" instead
of failing. We should, of course, still consider it a bug if that
behavior is reachable in ordinary use; but crashing seems less
desirable than not crashing.
This fix works across-the-board in v12 and up, where we always use
src/port/snprintf.c. Before that, on most platforms we're at the mercy
of the local libc, but it appears that Solaris 10 is the only supported
platform where we'd still get a crash. Most other platforms such as
*BSD, macOS, and Solaris 11 have adopted glibc's behavior at some
point. (AIX and HPUX just print "" not "(null)", but that's close
enough.) I've not checked what Windows' native printf would do, but
it doesn't matter because we've long used snprintf.c on that platform.
In v12 and up, also const-ify related code so that we're not casting
away const on the constant string. This is just neatnik-ism, since
next to no compilers will warn about that.
Discussion: https://postgr.es/m/17098-b960f3616c861f83@postgresql.org
Code inlined by LLVM can crash or fail with "Relocation type not
implemented yet!" if it tries to access thread local variables. Don't
inline such code.
Back-patch to 11, where LLVM arrived. Bug #16696.
Author: Dmitry Marakasov <amdmi3@amdmi3.ru>
Reviewed-by: Andres Freund <andres@anarazel.de>
Discussion: https://postgr.es/m/16696-29d944a33801fbfe@postgresql.org
In postgresql.conf, memory and file size GUCs can be specified with "B"
(bytes) as of b06d8e58b. Likewise, time GUCs can be specified with "us"
(microseconds) as of caf626b2c. Update postgres.conf.sample to reflect
that fact.
Pavel Luzanov
Backpatch to v12, which is the earliest version that allows both of
these units. A separate commit will document the "B" case for v11.
Discussion: https://www.postgresql.org/message-id/flat/f10d16fc-8fa0-1b3c-7371-cb3a35a13b7a%40postgrespro.ru
If an error was raised during our initial attempt to check whether
a successfully-compiled expression is "simple", subsequent calls of
exec_stmt_execsql would suppose that stmt->mod_stmt was already computed
when it had not been. This could lead to assertion failures in debug
builds; in production builds the effect would typically be to act as
if INTO STRICT had been specified even when it had not been. Of course
that only matters if the subsequent attempt to execute the expression
succeeds, so that the problem can only be reached by fixing a failure
in some referenced, inline-able SQL function and then retrying the
calling plpgsql function in the same session.
(There might be even-more-obscure ways to change the expression's
behavior without changing the plpgsql function, but that one seems
like the only one people would be likely to hit in practice.)
The most foolproof way to fix this would be to arrange for
exec_prepare_plan to not set expr->plan until we've finished the
subsidiary simple-expression check. But it seems hard to do that
without creating reference-count leak issues. So settle for documenting
the hazard in a comment and fixing exec_stmt_execsql to test separately
for whether it's computed stmt->mod_stmt. (That adds a test-and-branch
per execution, but hopefully that's negligible in context.) In v11 and
up, also fix exec_stmt_call which had a variant of the same issue.
Per bug #17113 from Alexander Lakhin. Back-patch to all
supported branches.
Discussion: https://postgr.es/m/17113-077605ce00e0e7ec@postgresql.org
The logic handling the opening of new WAL segments was fuzzy when using
--compress if a partial, non-compressed, segment with the same base name
existed in the repository storing those files. In this case, using
--compress would cause the code to first check for the existence and the
size of a non-compressed segment, followed by the opening of a new
compressed, partial, segment. The code was accidentally working
correctly on most platforms as the buildfarm has proved, except
bowerbird where gzflush() could fail in this code path. It is wrong
anyway to take the code path used pre-padding when creating a new
partial, non-compressed, segment, so let's fix it.
Note that this issue exists when users mix successive runs of
pg_receivewal with or without compression, as discovered with the tests
introduced by ffc9dda.
While on it, this refactors the code so as code paths that need to know
about the ".gz" suffix are down from four to one in walmethods.c, easing
a bit the introduction of new compression methods. This addresses a
second issue where log messages generated for an unexpected failure
would not show the compressed segment name involved, which was
confusing, printing instead the name of the non-compressed equivalent.
Reported-by: Georgios Kokolatos
Discussion: https://postgr.es/m/YPDLz2x3o1aX2wRh@paquier.xyz
Backpatch-through: 10
We don't allow to create replication slot_name as an empty string ('') via
SQL API pg_create_logical_replication_slot() but it is allowed to be set
via Alter Subscription command. This will lead to apply worker repeatedly
keep trying to stream data via slot_name '' and the user is not allowed to
create the slot with that name.
Author: Japin Li
Reviewed-By: Ranier Vilela, Amit Kapila
Backpatch-through: 10, where it was introduced
Discussion: https://postgr.es/m/MEYP282MB1669CBD98E721C77CA696499B61A9@MEYP282MB1669.AUSP282.PROD.OUTLOOK.COM
Heikki Linnakangas
Michael Paquier
Tom Lane
David Rowley
Thomas Munro
Peter Eisentraut
Bruce Momjian
Dean Rasheed
Etsuro Fujita
John Naylor
Alvaro Herrera
Fujii Masao
Amit Kapila