PredicateLockTuple() has a fast exit if tuple was written by the current
transaction, as in that case it already has a lock.  This check can be
performed using TransactionIdIsCurrentTransactionId() instead of
SubTransGetTopmostTransaction(), to avoid any chance of having to hit the
disk.

Author: Ashwin Agrawal, based on a suggestion from Andres Freund
Reviewed-by: Thomas Munro
Discussion: https://postgr.es/m/CALfoeiv0k3hkEb3Oqk%3DziWqtyk2Jys1UOK5hwRBNeANT_yX%2Bng%40mail.gmail.com

If the passed in xid is the current top transaction, we can do a fast
check and exit early.  This should work well for the current heap but
also works very well for proposed AMs that don't use a separate xid
for subtransactions.

Author: Ashwin Agrawal, based on a suggestion from Andres Freund
Reviewed-by: Thomas Munro
Discussion: https://postgr.es/m/CALfoeiv0k3hkEb3Oqk%3DziWqtyk2Jys1UOK5hwRBNeANT_yX%2Bng%40mail.gmail.com

During Drop Database, it is better to error out before allowing other
sessions to exit and forcefully terminating autovacuum workers.  All the
other errors except for checking subscriptions are already done before.

Author: Amit Kapila
Discussion: https://postgr.es/m/CAA4eK1+qhLkCYG2oy9xug9ur_j=G2wQNRYAyd+-kZfZ1z42pLw@mail.gmail.com

After altering a subscription, we should wait until the updated table
sync data has been fetched by the subscriber.

The previous statement that using a passphrase disables the ability to
change the server's SSL configuration without a server restart was no
longer completely true since the introduction of
ssl_passphrase_command_supports_reload.

Remove one sentence that was deemed misleading.

Discussion: https://www.postgresql.org/message-id/flat/E1iEgSp-0004R5-2E%40gemulon.postgresql.org

This happens when we add a replica identity column on a subscriber
that does not yet exist on the publisher, according to the mapping
maintained by the subscriber.  Code that checks whether the target
relation on the subscriber is updatable would check the replica
identity attribute bitmap with a column number -1, which would result
in an error.  To fix, skip such columns in the bitmap lookup and
consider the relation not updatable.  The result is consistent with
the rule that the replica identity columns on the subscriber must be a
subset of those on the publisher, since if the column doesn't exist on
the publisher, the column set on the subscriber can't be a subset.
Reported-by: Tim Clarke <tim.clarke@minerva.info>
Analyzed-by: Jehan-Guillaume de Rorthais <jgdr@dalibo.com>
Discussion: https://www.postgresql.org/message-id/flat/a9139c29-7ddd-973b-aa7f-71fed9c38d75%40minerva.info

The buildfarm has turned red after 1858b105 because VPATH builds need to
use "@abs_srcdir@" and not "@abs_builddir@" for paths coming directly
from the source tree.  The input file of the new test got that right,
but not the output file.

Per complaints from several buildfarm animals, including desmoxytes and
culicidae.  I have also reproduced the error by myself.

This stresses the error handling of COPY inside SPI which does not
support the operation using stdin or stdout, and these scenarios were
not tested up to now.

Author: Mark Dilger
Discussion: https://postgr.es/m/a6e9b130-7fd5-387b-4ec5-89bda24373ab@gmail.com

Not all AMs use HeapTuples internally, making it inconvenient to pass
a HeapTuple. As the index callbacks really only need the TID, not the
full tuple, modify callback to only take ItemPointer.

Author: Ashwin Agrawal
Reviewed-By: Andres Freund
Discussion: https://postgr.es/m/CALfoeis6=8ehuR=VNtHvj3z16cYfCwPdTcpaxU+sfSUJ5QgR3g@mail.gmail.com

Add some support for automatically showing backtraces in certain error
situations in the server.  Backtraces are shown on assertion failure;
also, a new setting backtrace_functions can be set to a list of C
function names, and all ereport()s and elog()s from the mentioned
functions will have backtraces generated.  Finally, the function
errbacktrace() can be manually added to an ereport() call to generate a
backtrace for that call.

Authors: Peter Eisentraut, Álvaro Herrera
Discussion: https://postgr.es/m//5f48cb47-bf1e-05b6-7aae-3bf2cd01586d@2ndquadrant.com
Discussion: https://postgr.es/m/CAMsr+YGL+yfWE=JvbUbnpWtrRZNey7hJ07+zT4bYJdVp4Szdrg@mail.gmail.com

Currently, postgres_fdw does not support preparing a remote transaction
for two-phase commit even in the case where the remote transaction is
read-only, but the old error message appeared to imply that that was not
supported only if the remote transaction modified remote tables.  Change
the message so as to include the case where the remote transaction is
read-only.

Also fix a comment above the message.

Also add a note about the lack of supporting PREPARE TRANSACTION to the
postgres_fdw documentation.

Reported-by: Gilles Darold
Author: Gilles Darold and Etsuro Fujita
Reviewed-by: Michael Paquier and Kyotaro Horiguchi
Backpatch-through: 9.4
Discussion: https://postgr.es/m/08600ed3-3084-be70-65ba-279ab19618a5%40darold.net

Use a separate error message for invalid checkpoint location and
invalid state instead of just "invalid data" for both.
Reviewed-by: Michael Paquier <michael@paquier.xyz>
Discussion: https://www.postgresql.org/message-id/20191107041630.GK1768@paquier.xyz

nbtree index builds once stashed the "minimum key" for a page, which was
used as the basis of the pivot tuple that gets placed in the next level
up (i.e. the tuple that stores the downlink to the page in question).
It doesn't quite work that way anymore, so the "minimum key" terminology
now seems misleading (these days the minimum key is actually a straight
copy of the high key from the left sibling, which is a distinct thing in
subtle but important ways).  Rename this concept to "low key".  This
name is a lot clearer given that there is now a sharp distinction
between pivot and non-pivot tuples.  Also remove comments that describe
obsolete details about how the minimum key concept used to work.

Rather than generating the minus infinity item for the leftmost page on
a level by copying the new item and truncating that copy, simply
allocate a small buffer.  The old approach confusingly created the
impression that the new item had some kind of significance.  This was
another artifact of how things used to work before commits 8224de4f and
dd299df8.

The point is that DELETE triggers cannot modify any values.

Reported-by: Eugen Konkov

Discussion: https://postgr.es/m/919823407.20191029175436@yandex.ru

Backpatch-through: 9.4

Declaring this in the client-visible header ecpglib.h was a pretty
poor decision.  It's not meant to be application-callable (and if
it was, putting it outside the extern "C" { ... } wrapper means
that C++ clients would fail to call it).  And the declaration would
not even compile for a client, anyway, since it would not have the
macro pg_attribute_format_arg().  Fortunately, it seems that no
clients have tried to include this header with ENABLE_NLS defined,
or we'd have gotten complaints about that.  But we have no business
putting such a restriction on client code.

Move the declaration to ecpglib_extern.h, since in fact nothing
outside src/interfaces/ecpg/ecpglib/ needs to call it.

The practical effect of this is just that clients can now safely
#include ecpglib.h while having ENABLE_NLS defined, but that seems
like enough of a reason to back-patch it.

Discussion: https://postgr.es/m/20590.1573069709@sss.pgh.pa.us

SET CONSTRAINTS ... DEFERRED failed on partitioned tables, because of a
sanity check that ensures that the affected constraints have triggers.
On partitioned tables, the triggers are in the leaf partitions, not in
the partitioned relations themselves, so the sanity check fails.
Removing the sanity check solves the problem, because the code needed to
support the case is already there.

Backpatch to 11.

Note: deferred unique constraints are not affected by this bug, because
they do have triggers in the parent partitioned table.  I did not add a
test for this scenario.

Discussion: https://postgr.es/m/20191105212915.GA11324@alvherre.pgsql

This patch adopts the overflow check logic introduced by commit cbdb8b4c
into two more places.  interval_mul() failed to notice if it computed a
new microseconds value that was one more than INT64_MAX, and pgbench's
double-to-int64 logic had the same sorts of edge-case problems that
cbdb8b4c fixed in the core code.

To make this easier to get right in future, put the guts of the checks
into new macros in c.h, and add commentary about how to use the macros
correctly.

Back-patch to all supported branches, as we did with the previous fix.

Yuya Watari

Discussion: https://postgr.es/m/CAJ2pMkbkkFw2hb9Qb1Zj8d06EhWAQXFLy73St4qWv6aX=vqnjw@mail.gmail.com

The presence of long long int is now implied in the requirement for
C99 and the configure check for the same.

We keep the define hard-coded in ecpg_config.h for backward
compatibility with ecpg-using user code.

Discussion: https://www.postgresql.org/message-id/flat/5cdd6a2b-b2c7-c6f6-344c-a406d5c1a254%402ndquadrant.com

We need to pop the error stack before running the user-supplied
PG_FINALLY code.  Otherwise an error in the cleanup code would end up
at the same sigsetjmp() invocation and result in an infinite error
handling loop.
Reviewed-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://www.postgresql.org/message-id/flat/95a822c3-728b-af0e-d7e5-71890507ae0c%402ndquadrant.com

If there is the WAL page that the continuation WAL record just fits within
(i.e., the continuation record ends just at the end of the page) and
the LSN in such page is specified with -s option, previously pg_waldump
caused an assertion failure. The cause of this assertion failure was that
XLogFindNextRecord() that pg_waldump -s calls mistakenly handled
such special WAL page.

This commit changes XLogFindNextRecord() so that it can handle
such WAL page correctly.

Back-patch to all supported versions.

Author: Andrey Lepikhov
Reviewed-by: Fujii Masao, Michael Paquier
Discussion: https://postgr.es/m/99303554-5dd5-06e6-f943-b3005ccd6edd@postgrespro.ru

Introduce qunique() and qunique_arg(), which can be used after qsort()
and qsort_arg() respectively to remove duplicate values.  Use it where
appropriate.

Author: Thomas Munro
Reviewed-by: Tom Lane (in an earlier version)
Discussion: https://postgr.es/m/CAEepm%3D2vmFTNpAmwbGGD2WaryM6T3hSDVKQPfUwjdD_5XY6vAA%40mail.gmail.com

SPI gets used to build a list of relation OIDs for XML object
generation, and one code path building a list uses SPI_execute() without
looking at errors it produces.  So fix that.

Author: Mark Dilger
Reviewed-by: Michael Paquier, Pavel Stehule
Discussion: https://postgr.es/m/17d30445-4862-7917-170f-84328dcd292d@gmail.com

This allows logging a sample of statements, without incurring excessive
log traffic (which may impact performance).  This can be useful when
analyzing workloads with lots of short queries.

The sampling is configured using two new GUC parameters:

 * log_min_duration_sample - minimum required statement duration

 * log_statement_sample_rate - sample rate (0.0 - 1.0)

Only statements with duration exceeding log_min_duration_sample are
considered for sampling. To enable sampling, both those GUCs have to
be set correctly.

The existing log_min_duration_statement GUC has a higher priority, i.e.
statements with duration exceeding log_min_duration_statement will be
always logged, irrespectedly of how the sampling is configured. This
means only configurations

  log_min_duration_sample < log_min_duration_statement

do actually sample the statements, instead of logging everything.

Author: Adrien Nayrat
Reviewed-by: David Rowley, Vik Fearing, Tomas Vondra
Discussion: https://postgr.es/m/bbe0a1a8-a8f7-3be2-155a-888e661cc06c@anayrat.info

The docs do say which GUCs can be changed only by superusers, but we
forgot to mention this for the new log_transaction_sample_rate. This
GUC was introduced in PostgreSQL 12, so backpatch accordingly.

Author: Adrien Nayrat
Backpatch-through: 12

Avoid creating transiently-inconsistent slot states where possible,
by not setting TTS_FLAG_SHOULDFREE until after the slot actually has
a free'able tuple pointer, and by making sure that we reset tts_nvalid
and related derived state before we replace the tuple contents.  This
would only matter if something were to examine the slot after we'd
suffered some kind of error (e.g. out of memory) while manipulating
the slot.  We typically don't do that, so these changes might just be
cosmetic --- but even if so, it seems like good future-proofing.

Also remove some redundant Asserts, and add a couple for consistency.

Back-patch to v12 where all this code was rewritten.

Discussion: https://postgr.es/m/16095-c3ff2e5283b8dba5@postgresql.org

Since commit d26a810e, we've defined bool as being either _Bool from
<stdbool.h>, or "unsigned char"; but that commit overlooked the fact
that probes.d has "#define bool char".  For consistency, make it say
"unsigned char" instead.  This should be strictly a cosmetic change,
but it seems best to be in sync.

Formally, in the now-normal case where we're using <stdbool.h>, it'd
be better to write "#define bool _Bool".  However, then we'd need
some build infrastructure to inject that configuration choice into
probes.d, and it doesn't seem worth the trouble.  We only use
<stdbool.h> if sizeof(_Bool) is 1, so having DTrace think that
bool parameters are "unsigned char" should be close enough.

Back-patch to v12 where d26a810e came in.

Discussion: https://postgr.es/m/CAA4eK1LmaKO7Du9M9Lo=kxGU8sB6aL8fa3sF6z6d5yYYVe3BuQ@mail.gmail.com

The previous code was allocating more memory than necessary because
the formula used the wrong data type.
Reported-by: Jehan-Guillaume de Rorthais <jgdr@dalibo.com>
Discussion: https://www.postgresql.org/message-id/20191105172918.3e32a446@firost

The cache_plan argument to ri_PlanCheck has not been used since
e8c9fd5f.
Reviewed-by: vignesh C <vignesh21@gmail.com>
Discussion: https://www.postgresql.org/message-id/flat/ec8a8b45-a30b-9193-cd4b-985d60d1497e%402ndquadrant.com

When sending data for logical decoding using the streaming replication
protocol via a WAL sender, the timestamp of the sent write message is
allocated at the beginning of the message when preparing for the write,
and actually computed when the write message is ready to be sent.

The timestamp was getting computed after sending the message.  This
impacts anything using logical decoding, causing for example logical
replication to report mostly NULL for last_msg_send_time in
pg_stat_subscription.

This commit makes sure that the timestamp is computed before sending the
message.  This is wrong since 5a991ef8, so backpatch down to 9.4.

Author: Jeff Janes
Discussion: https://postgr.es/m/CAMkU=1z=WMn8jt7iEdC5sYNaPgAgOASb_OW5JYv-vMdYaJSL-w@mail.gmail.com
Backpatch-through: 9.4

WindowAgg will potentially store large numbers of input rows into
tuplestores to allow access to other rows in the frame. If the input
is coming via an explicit Sort node, then unneeded columns will
already have been discarded (since Sort requests a small tlist); but
there are idioms like COUNT(*) OVER () that result in the input not
being sorted at all, and cases where the input is being sorted by some
means other than a Sort; if we don't request a small tlist, then
WindowAgg's storage requirement is inflated by the unneeded columns.

Backpatch back to 9.6, where the current tlist handling was added.
(Prior to that, WindowAgg would always use a small tlist.)

Discussion: https://postgr.es/m/87a7ator8n.fsf@news-spur.riddles.org.uk

Previously ALTER MATERIALIZED VIEW / FOREIGN TABLE ... RENAME COLUMN ...
returned "ALTER TABLE" as a command tag. This commit fixes them so that
they return "ALTER MATERIALIZED VIEW" and "ALTER FOREIGN TABLE" as
command tags, respectively.

This issue exists in all supported versions, but we don't back-patch this
because it's not enough of a bug to justify taking any compatibility risks for.
Otherwise, the back-patch would cause minor version update to break,
for example, the existing event trigger functions using TG_TAG.

Author: Fujii Masao
Reviewed-by: Ibrar Ahmed
Discussion: https://postgr.es/m/CAHGQGwGUaC03FFdTFoHsCuDrrNvFvNVQ6xyd40==P25WvuBJjg@mail.gmail.com

This commit allows --init-steps option in pgbench to accept "G" character
meaning server-side data generation as an initialization step.
With "G", only limited queries are sent from pgbench client and
then data is actually generated in the server. This might make
the initialization phase faster if the bandwidth between pgbench client
and the server is low.

Author: Fabien Coelho
Reviewed-by: Anna Endo, Ibrar Ahmed, Fujii Masao
Discussion: https://postgr.es/m/alpine.DEB.2.21.1904061826420.3678@lancre

Reported-by: matthew.alton@gmail.com

Discussion: https://postgr.es/m/157204060717.1042.8194076510523669244@wrigleys.postgresql.org

Backpatch-through: 9.4

There's plenty places in frontend code that could benefit from a
string buffer implementation. Some because it yields simpler and
faster code, and some others because of the desire to share code
between backend and frontend.

While there is a string buffer implementation available to frontend
code, libpq's PQExpBuffer, it is clunkier than stringinfo, it
introduces a libpq dependency, doesn't allow for sharing between
frontend and backend code, and has a higher API/ABI stability
requirement due to being exposed via libpq.

Therefore it seems best to just making StringInfo being usable by
frontend code. There's not much to do for that, except for rewriting
two subsequent elog/ereport calls into others types of error
reporting, and deciding on a maximum string length.

For the maximum string size I decided to privately define MaxAllocSize
to the same value as used in the backend. It seems likely that we'll
want to reconsider this for both backend and frontend code in the not
too far away future.

For now I've left stringinfo.h in lib/, rather than common/, to reduce
the likelihood of unnecessary breakage. We could alternatively decide
to provide a redirecting stringinfo.h in lib/, or just not provide
compatibility.

Author: Andres Freund
Reviewed-By: Kyotaro Horiguchi, Daniel Gustafsson
Discussion: https://postgr.es/m/20190920051857.2fhnvhvx4qdddviz@alap3.anarazel.de

When maintaining or merging patches, one of the most common sources
for conflicts are the list of objects in makefiles. Especially when
the split across lines has been changed on both sides, which is
somewhat common due to attempting to stay below 80 columns, those
conflicts are unnecessarily laborious to resolve.

By splitting, and alphabetically sorting, OBJS style lines into one
object per line, conflicts should be less frequent, and easier to
resolve when they still occur.

Author: Andres Freund
Discussion: https://postgr.es/m/20191029200901.vww4idgcxv74cwes@alap3.anarazel.de

Avoid initial capital, since that's not how we do it.

Discussion: https://postgr.es/m/CACP=ajbrFFYUrLyJBLV8=q+eNCapa1xDEyvXhMoYrNphs-xqPw@mail.gmail.com

For a long time (since commit aed378e8) we have had a policy to log
nothing about a connection if the client disconnects when challenged
for a password.  This is because libpq-using clients will typically
do that, and then come back for a new connection attempt once they've
collected a password from their user, so that logging the abandoned
connection attempt will just result in log spam.  However, this did
not work well for PAM authentication: the bottom-level function
pam_passwd_conv_proc() was on board with it, but we logged messages
at higher levels anyway, for lack of any reporting mechanism.
Add a flag and tweak the logic so that the case is silent, as it is
for other password-using auth mechanisms.

Per complaint from Yoann La Cancellera.  It's been like this for awhile,
so back-patch to all supported branches.

Discussion: https://postgr.es/m/CACP=ajbrFFYUrLyJBLV8=q+eNCapa1xDEyvXhMoYrNphs-xqPw@mail.gmail.com

get_relkind_objtype, and hence get_object_type, failed when applied to a
toast table.  This is not a good thing, because it prevents reporting of
perfectly legitimate permissions errors.  (At present, these functions
are in fact *only* used to determine the ObjectType argument for
acl_error() calls.)  It seems best to have them fall back to returning
OBJECT_TABLE in every case where they can't determine an object type
for a pg_class entry, so do that.

In passing, make some edits to alter.c to make it more obvious that
those calls of get_object_type() are used only for error reporting.
This might save a few cycles in the non-error code path, too.

Back-patch to v11 where this issue originated.

John Hsu, Michael Paquier, Tom Lane

Discussion: https://postgr.es/m/C652D3DF-2B0C-4128-9420-FB5379F6B1E4@amazon.com