heap_update needs to clear any existing "all visible" flag on
the old tuple's page (and on the new page too, if different).
Per coding rules, to do this it must acquire pin on the appropriate
visibility-map page while not holding exclusive buffer lock;
which creates a race condition since someone else could set the
flag whenever we're not holding the buffer lock.  The code is
supposed to handle that by re-checking the flag after acquiring
buffer lock and retrying if it became set.  However, one code
path through heap_update itself, as well as one in its subroutine
RelationGetBufferForTuple, failed to do this.  The end result,
in the unlikely event that a concurrent VACUUM did set the flag
while we're transiently not holding lock, is a non-recurring
"PANIC: wrong buffer passed to visibilitymap_clear" failure.

This has been seen a few times in the buildfarm since recent VACUUM
changes that added code paths that could set the all-visible flag
while holding only exclusive buffer lock.  Previously, the flag
was (usually?) set only after doing LockBufferForCleanup, which
would insist on buffer pin count zero, thus preventing the flag
from becoming set partway through heap_update.  However, it's
clear that it's heap_update not VACUUM that's at fault here.

What's less clear is whether there is any hazard from these bugs
in released branches.  heap_update is certainly violating API
expectations, but if there is no code path that can set all-visible
without a cleanup lock then it's only a latent bug.  That's not
100% certain though, besides which we should worry about extensions
or future back-patch fixes that could introduce such code paths.

I chose to back-patch to v12.  Fixing RelationGetBufferForTuple
before that would require also back-patching portions of older
fixes (notably 0d1fe9f7), which is more code churn than seems
prudent to fix a hypothetical issue.

Discussion: https://postgr.es/m/2247102.1618008027@sss.pgh.pa.us

Introduced in commit 0aa8a01d.

Author: Peter Smith
Discussion: https://postgr.es/m/CAHut+Ps8rkVHKWyjg09Fb1PaVG5-EhoFTFJ9OZMF4HPzDSXfew@mail.gmail.com

With the Oracle Developer Studio 12.6 compiler, #line directives alter
the current source file location for purposes of #include "..."
directives.  Hence, a VPATH build failed with 'cannot find include file:
"specscanner.c"'.  With two exceptions, parser-containing directories
already add "-I. -I$(srcdir)"; eliminate the exceptions.  Back-patch to
9.6 (all supported versions).

It doesn't support "\(foo\)*" like a POSIX "sed" implementation does;
see the Autoconf manual.  Back-patch to 9.6 (all supported versions).

Commit 6f38d4da failed to heed a warning about the stability of the
value pointed to by "otid".  The caller is allowed to pass in a pointer to
newtup->t_self, which will be updated during the execution of the
function.  Instead, the SSI check should use the value we copy into
oldtup.t_self near the top of the function.

Not a live bug, because newtup->t_self doesn't really get updated until
a bit later, but it was confusing and broke the rule established by the
comment.

Back-patch to 13.
Reported-by: Tom Lane <tgl@sss.pgh.pa.us>
Discussion: https://postgr.es/m/2689164.1618160085%40sss.pgh.pa.us

These got introduced in 6568cef2.

Reported-by: Noah Misch
Discussion: https://postgr.es/m/20210404220802.GA728316@rfd.leadboat.com

collate.icu.utf8.sql was exercising the recording of a collation
dependency for an enum comparison expression, but such an expression
should never have had any collation dependency in the first place.
After I fixed that in commit c402b02b, the test started failing.
We don't need to test that scenario anymore, so just remove the
now-useless test steps.

(This test case is new in HEAD, so no need to back-patch.)

Discussion: https://postgr.es/m/3044030.1618261159@sss.pgh.pa.us
Discussion: https://postgr.es/m/HK0PR01MB22744393C474D503E16C8509F4709@HK0PR01MB2274.apcprd01.prod.exchangelabs.com

There are hacks in parse_coerce.c to push down a requested coercion
to below any CollateExpr that may appear.  However, we did that even
if the requested data type is non-collatable, leading to an invalid
expression tree in which CollateExpr is applied to a non-collatable
type.  The fix is just to drop the CollateExpr altogether, reasoning
that it's useless.

This bug is ten years old, dating to the original addition of
COLLATE support.  The lack of field complaints suggests that there
aren't a lot of user-visible consequences.  We noticed the problem
because it would trigger an assertion in DefineVirtualRelation if
the invalid structure appears as an output column of a view; however,
in a non-assert build, you don't see a crash just a (subtly incorrect)
complaint about applying collation to a non-collatable type.  I found
that by putting the incorrect structure further down in a view, I could
make a view definition that would fail dump/reload, per the added
regression test case.  But CollateExpr doesn't do anything at run-time,
so this likely doesn't lead to any really exciting consequences.

Per report from Yulin Pei.  Back-patch to all supported branches.

Discussion: https://postgr.es/m/HK0PR01MB22744393C474D503E16C8509F4709@HK0PR01MB2274.apcprd01.prod.exchangelabs.com

This could write wrong output into the cluster deletion script if a
database OID exceeds the signed 32-bit range.

broken by 37d2ff38

Commit 8ff1c946 extended TRUNCATE command so that it can also truncate
foreign tables. But it forgot to support tab-complete for TRUNCATE on
foreign tables. That is, previously tab-complete for TRUNCATE displayed
only the names of regular tables.

This commit improves tab-complete for TRUNCATE so that it displays also
the names of foreign tables.

Author: Fujii Masao
Reviewed-by: Bharath Rupireddy
Discussion: https://postgr.es/m/551ed8c1-f531-818b-664a-2cecdab99cd8@oss.nttdata.com

This GUC has already been classified as LOGGING_WHAT, but its location
in postgresql.conf.sample and the documentation did not reflect that, so
fix those inconsistencies.

Author: Justin Pryzby
Discussion: https://postgr.es/m/20210404012546.GK6592@telsasoft.com

Updated documentation for new messages added for streaming of in-progress
transactions, as well as changes made to the existing messages. It also
updates the information of protocol versions supported for logical
replication.

Author: Ajin Cherian
Reviewed-by: Amit Kapila, Peter Smith, Euler Taveira
Discussion: https://postgr.es/m/CAFPTHDYHN9m=MZZct-B=BYg_TETvv+kXvL9RD2DpaBS5pGxGYg@mail.gmail.com

Using Roman numbers (via "RM" or "rm") for a conversion to calculate a
number of months has never considered the case of negative numbers,
where a conversion could easily cause out-of-bound memory accesses.  The
conversions in themselves were not completely consistent either, as
specifying 12 would result in NULL, but it should mean XII.

This commit reworks the conversion calculation to have a more
consistent behavior:
- If the number of months and years is 0, return NULL.
- If the number of months is positive, return the exact month number.
- If the number of months is negative, do a backward calculation, with
-1 meaning December, -2 November, etc.

Reported-by: Theodor Arsenij Larionov-Trichkin
Author: Julien Rouhaud
Discussion: https://postgr.es/m/16953-f255a18f8c51f1d5@postgresql.org
backpatch-through: 9.6

Coverity complained about possible overflow in expressions like
	intresult = tm->tm_sec * 1000000 + fsec;
on the grounds that the multiplication would happen in 32-bit
arithmetic before widening to the int64 result.  I think these
are all false positives because of the limited possible range of
tm_sec; but nonetheless it seems silly to spell it like that when
nearby lines have the identical computation written with a 64-bit
constant.

... or more accurately, with an LL constant, which is not project
style.  Make all of these use INT64CONST(), as we do elsewhere.

This is all new code from a2da77cd, so no need for back-patch.

We'd previously noted the need for coping with Windows headers
that provide some other definition of macro "ERROR" than elog.h
does.  It turns out that R also wants to define ERROR, and
WARNING too.  PL/R has been working around this in a hacky way
that broke when we recently changed the numeric value of ERROR.
To let them have a more future-proof solution, provide an
alternate macro PGWARNING for WARNING, and make PGERROR visible
always, not only when #ifdef WIN32.

Discussion: https://postgr.es/m/CADK3HHK6iMChd1yoOqssxBn5Z14Zar8Ztr3G-N_fuG7F8YTP3w@mail.gmail.com

The path for *exprs != NIL would misbehave, and likely crash,
since pull_varattnos expects its last argument to be valid
at call.

Found by Coverity --- we have no coverage of this path in
the regression tests.

ExecuteTruncate() filters out the duplicate tables specified
in the TRUNCATE command, for example in the case where "TRUNCATE foo, foo"
is executed. Such duplicate tables obviously don't need to be opened
and closed because they are skipped. But previously it always opened
the tables before checking whether they were duplicated ones or not,
and then closed them if they were. That is, the duplicated tables were
opened and closed unnecessarily.

This commit changes ExecuteTruncate() so that it opens the table
after it confirms that table is not duplicated one, which leads to
avoid unnecessary table open/close.

Do not back-patch because such unnecessary table open/close is not
a bug though it exists in older versions.

Author: Bharath Rupireddy
Reviewed-by: Amul Sul, Fujii Masao
Discussion: https://postgr.es/m/CALj2ACUdBO_sXJTa08OZ0YT0qk7F_gAmRa9hT4dxRcgPS4nsZA@mail.gmail.com

Commit 438fc4a3 prevented the WAL replay from writing
COMMIT_TS_SETTS record. By this change there is no code that
generates COMMIT_TS_SETTS record in PostgreSQL core.
Also we can think that there are no extensions using the record
because we've not received so far any complaints about the issue
that commit 438fc4a3 fixed. Therefore this commit removes
COMMIT_TS_SETTS record and its related code. Even without
this record, the timestamp required for commit timestamp feature
can be acquired from the COMMIT record.

Bump WAL page magic.
Reported-by: lx zou <zoulx1982@163.com>
Author: Fujii Masao
Reviewed-by: Alvaro Herrera
Discussion: https://postgr.es/m/16931-620d0f2fdc6108f1@postgresql.org

Commit c9c41c7a used two different
naming patterns.  Standardize on the majority pattern, which was the
only pattern in the last reviewed version of that commit.

Currently, when the origin is after the input, the result is the
timestamp at the end of the bin, rather than the beginning as
expected.  This puts the result consistently at the beginning of the
bin.

Author: John Naylor <john.naylor@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/CAFBsxsGjLDxQofRfH+d4KSAXxPf3MMevUG7s6EDfdBOvHLDLjw@mail.gmail.com

Per cpluspluscheck.

Adjust docs and description string to note that check_function_bodies
applies to procedures too.  (In hindsight it should have been named
check_routine_bodies, but it seems too late for that now.)

Daniel Westermann

Discussion: https://postgr.es/m/GV0P278MB04834A9EB9A74B036DC7CE49D2739@GV0P278MB0483.CHEP278.PROD.OUTLOOK.COM

There were some comments in nodeFuncs.c that, depending on your
interpretation of the word "result", could lead you to believe that the
comments were badly copied and pasted from somewhere else.  If you thought
of "result" as the return value of the function that the comment is
written in, then you'd be misled.  However, if you'd correctly
interpreted "result" to mean the result type of the given node type,
you'd not have seen any issues.

Here we do a small cleanup to try to prevent any future
misinterpretations.  Per wording suggestion from Tom Lane.

Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/CAApHDvp+Bw=2Qiu5=uXMKfC7gd0+B=4JvexVgGJU=am2g9a1CA@mail.gmail.com

Whitespace between tags is significant, and in some cases it creates
extra vertical space in man pages.  The fix is to remove some newlines
in the markup.

A (relatively minor) annoyance of ErrorResponse/NoticeResponse messages
as printed by PQtrace() is that their length might vary when we move
error messages from one source file to another, one function to another,
or even when their location line numbers change number of digits.

To avoid having to adjust expected files for some tests, make the
regress mode of PQtrace() suppress the length word of NoticeResponse and
ErrorResponse messages.

Discussion: https://postgr.es/m/20210402023010.GA13563@alvherre.pgsqlReviewed-by: Tom Lane <tgl@sss.pgh.pa.us>

Reported-by: Daniel Westermann (DWE) <daniel.westermann@dbi-services.com>
Discussion: https://postgr.es/m/GV0P278MB0483490FEAC879DCA5ED583DD2739%40GV0P278MB0483.CHEP278.PROD.OUTLOOK.COM

Typos, corrections and language improvements in the docs, and a few in
code comments too.
Reported-by: Justin Pryzby <pryzby@telsasoft.com>
Discussion: https://postgr.es/m/20210409033703.GP6592%40telsasoft.com

Reported-by: Justin Pryzby <pryzby@telsasoft.com>
Author: John Naylor <john.naylor@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/CAFBsxsEEm1nuhZmfVQxvu_i3nDDEuvNJ_WMrDo9whFD_jusp-A@mail.gmail.com

Commit 0827e8af added parameters for autovacuum to support
partitioned tables, but didn't add any docs.  Add them.

Discussion: https://postgr.es/m/20210408213051.GL6592@telsasoft.com

When commit 0827e8af added auto-analyze support for partitioned
tables, it included code to obtain reltuples for the partitioned table
as a number of catalog accesses to read pg_class.reltuples for each
partition.  That's not only very inefficient, but also problematic
because autovacuum doesn't hold any locks on any of those tables -- and
doesn't want to.  Replace that code with a read of pg_class.reltuples
for the partitioned table, and make sure ANALYZE and TRUNCATE properly
maintain that value.

I found no code that would be affected by the change of relpages from
zero to non-zero for partitioned tables, and no other code that should
be maintaining it, but if there is, hopefully it'll be an easy fix.

Per buildfarm.

Author: Álvaro Herrera <alvherre@alvh.no-ip.org>
Reviewed-by: Zhihong Yu <zyu@yugabyte.com>
Discussion: https://postgr.es/m/1823909.1617862590@sss.pgh.pa.us

Author: Daniel Westermann
Backpatch-through: 9.6
Discussion: https://postgr.es/m/GV0P278MB0483A7AA85BAFCC06D90F453D2739@GV0P278MB0483.CHEP278.PROD.OUTLOOK.COM

Comment fixes are applied on HEAD, and documentation improvements are
applied on back-branches where needed.

Author: Justin Pryzby
Discussion: https://postgr.es/m/20210408164008.GJ6592@telsasoft.com
Backpatch-through: 9.6

Per complaint from Tom Lane

Discussion: https://postgr.es/m/1922884.1617909599@sss.pgh.pa.us

Oversight in commit a3027e1e.

Several buildfarm critters that don't usually produce such
warnings are complaining about e717a9a1.  I think it's
actually safe, but move initialization to silence the warning.

Ignore parallel workers in pg_stat_statements
  Oversight in 4f0b0966 which exposed queryid in parallel workers.
  Counters are aggregated by the main backend process so parallel workers
  would report duplicated activity, and could also report activity for the
  wrong entry as they are only aware of the top level queryid.

Fix thinko in pg_stat_get_activity when retrieving the queryid.

Remove unnecessary call to pgstat_report_queryid().

Reported-by: Amit Kapila, Andres Freund, Thomas Munro

Discussion: https://postgr.es/m/20210408051735.lfbdzun5zdlax5gd@alap3.anarazel.de p634GTSOqnDW86Owrn6qDAVosC5dJjXjp7BMfc5Gz1Q@mail.gmail.com

Author: Julien Rouhaud

v1.9 is already new in this version of PostgreSQL, so turn it into just
one change.

Author: Julien Rohaud
Discussion: https://postgr.es/m/20210408120505.7zinijtdexbyghvb@nol

Thinko in commit 323cbe7c, per complaint from BF animal locust's older
GCC compiler.