This GUC variable works much like idle_in_transaction_session_timeout,
in that it kills sessions that have waited too long for a new client
query.  But it applies when we're not in a transaction, rather than
when we are.

Li Japin, reviewed by David Johnston and Hayato Kuroda, some
fixes by me

Discussion: https://postgr.es/m/763A0689-F189-459E-946F-F0EC4458980B@hotmail.com

A very common usage pattern is that we set a timeout that we don't
expect to reach, cancel it after a little bit, and later repeat.
With the original implementation of timeout.c, this results in one
setitimer() call per timeout set or cancel.  We can do a lot better
by being lazy about changing the timeout interrupt request, namely:
(1) never cancel the outstanding interrupt, even when we have no
active timeout events;
(2) if we need to set an interrupt, but there already is one pending
at or before the required time, leave it alone.  When the interrupt
happens, the signal handler will reschedule it at whatever time is
then needed.

For example, with a one-second setting for statement_timeout, this
method results in having to interact with the kernel only a little
more than once a second, no matter how many statements we execute
in between.  The mainline code might never call setitimer() at all
after the first time, while each time the signal handler fires,
it sees that the then-pending request is most of a second away,
and that's when it sets the next interrupt request for.  Each
mainline timeout-set request after that will observe that the time
it wants is past the pending interrupt request time, and do nothing.

This also works pretty well for cases where a few different timeout
lengths are in use, as long as none of them are very short.  But
that describes our usage well.

Idea and original patch by Thomas Munro; I fixed a race condition
and improved the comments.

Discussion: https://postgr.es/m/CA+hUKG+o6pbuHBJSGnud=TadsuXySWA7CCcPgCt2QE9F6_4iHQ@mail.gmail.com

This commit introduces a view pg_stat_progress_copy, reporting progress
of COPY commands.  This allows rough estimates how far a running COPY
progressed, with the caveat that the total number of bytes may not be
available in some cases (e.g. when the input comes from the client).

Author: Josef Šimánek
Reviewed-by: Fujii Masao, Bharath Rupireddy, Vignesh C, Matthias van de Meent
Discussion: https://postgr.es/m/CAFp7QwqMGEi4OyyaLEK9DR0+E+oK3UtA4bEjDVCa4bNkwUY2PQ@mail.gmail.com
Discussion: https://postgr.es/m/CAFp7Qwr6_FmRM6pCO0x_a0mymOfX_Gg+FEKet4XaTGSW=LitKQ@mail.gmail.com

This module provides a function test_regex() that is functionally
rather like regexp_matches(), but with additional debugging-oriented
options and additional output.  The debug options are somewhat obscure;
they are chosen to match the API of the test harness that Henry Spencer
wrote way-back-when for use in Tcl.  With this, we can import all the
test cases that Spencer wrote originally, even for regex functionality
that we don't currently expose in Postgres.  This seems necessary
because we can no longer rely on Tcl to act as upstream and verify
any fixes or improvements that we make.

In addition to Spencer's tests, I added a few for lookbehind
constraints (which we added in 2015, and Tcl still hasn't absorbed)
that are modeled on his tests for lookahead constraints.  After looking
at code coverage reports, I also threw in a couple of tests to more
fully exercise our "high colormap" logic.

According to my testing, this brings the check-world coverage
for src/backend/regex/ from 71.1% to 86.7% of lines.
(coverage.postgresql.org shows a slightly different number,
which I think is because it measures a non-assert build.)

Discussion: https://postgr.es/m/2873268.1609732164@sss.pgh.pa.us

Forced cache invalidation (CLOBBER_CACHE_ALWAYS) has been impractical
to use for testing in PostgreSQL because it's so slow and because it's
toggled on/off only at build time.  It is helpful when hunting bugs in
any code that uses the sycache/relcache because causes cache
invalidations to be injected whenever it would be possible for an
invalidation to occur, whether or not one was really pending.

Address this by providing run-time control over cache clobber
behaviour using the new debug_invalidate_system_caches_always GUC.
Support is not compiled in at all unless assertions are enabled or
CLOBBER_CACHE_ENABLED is explicitly defined at compile time.  It
defaults to 0 if compiled in, so it has negligible effect on assert
build performance by default.

When support is compiled in, test code can now set
debug_invalidate_system_caches_always=1 locally to a backend to test
specific queries, functions, extensions, etc.  Or tests can toggle it
globally for a specific test case while retaining normal performance
during test setup and teardown.

For backwards compatibility with existing test harnesses and scripts,
debug_invalidate_system_caches_always defaults to 1 if
CLOBBER_CACHE_ALWAYS is defined, and to 3 if CLOBBER_CACHE_RECURSIVE
is defined.

CLOBBER_CACHE_ENABLED is now visible in pg_config_manual.h, as is the
related RECOVER_RELATION_BUILD_MEMORY setting for the relcache.

Author: Craig Ringer <craig.ringer@2ndquadrant.com>
Discussion: https://www.postgresql.org/message-id/flat/CAMsr+YF=+ctXBZj3ywmvKNUjWpxmuTuUKuv-rgbHGX5i5pLstQ@mail.gmail.com

The deadlocks that the recovery conflict on lock is involved in can
happen between hot-standby backends and the startup process.
If a backend takes an access exclusive lock on the table and which
finally triggers the deadlock, that deadlock can be detected
as expected. On the other hand, previously, if the startup process
took an access exclusive lock and which finally triggered the deadlock,
that deadlock could not be detected and could remain even after
deadlock_timeout passed. This is a bug.

The cause of this bug was that the code for handling the recovery
conflict on lock didn't take care of deadlock case at all. It assumed
that deadlocks involving the startup process and backends were able
to be detected by the deadlock detector invoked within backends.
But this assumption was incorrect. The startup process also should
have invoked the deadlock detector if necessary.

To fix this bug, this commit makes the startup process invoke
the deadlock detector if deadlock_timeout is reached while handling
the recovery conflict on lock. Specifically, in that case, the startup
process requests all the backends holding the conflicting locks to
check themselves for deadlocks.

Back-patch to v9.6. v9.5 has also this bug, but per discussion we decided
not to back-patch the fix to v9.5. Because v9.5 doesn't have some
infrastructure codes (e.g., 37c54863) that this bug fix patch depends on.
We can apply those codes for the back-patch, but since the next minor
version release is the final one for v9.5, it's risky to do that. If we
unexpectedly introduce new bug to v9.5 by the back-patch, there is no
chance to fix that. We determined that the back-patch to v9.5 would give
more risk than gain.

Author: Fujii Masao
Reviewed-by: Bertrand Drouvot, Masahiko Sawada, Kyotaro Horiguchi
Discussion: https://postgr.es/m/4041d6b6-cf24-a120-36fa-1294220f8243@oss.nttdata.com

Per report by Ajin Cherian in email:
https://postgr.es/m/CAFPTHDYnRKDvzgDxoMn_CKqXA-D0MtrbyJvfvjBsO4G=UHDXkg@mail.gmail.com

The default value of recovery_target_timeline was changed in v12,
but the description about the default behavior of that was not updated.

Back-patch to v12 where the default behavior of recovery_target_timeline
was changed.

Author: Benoit Lobréau
Reviewed-by: Fujii Masao
Discussion: https://postgr.es/m/CAPE8EZ7c3aruEmM24GYkj8y8WmHKD1m9TtPtgCF0nQ3zw4LCkQ@mail.gmail.com

This was previously part of the section dedicated to less common
options, but it is an option commonly used these days.

Author: Michael Banck
Reviewed-by: Stephen Frost, Michael Paquier
Discussion: https://postgr.es/m/d7938aca4d4ea8e8c72c33bd75efe9f8218fe390.camel@credativ.de

I momentarily forgot that the "owner" column wouldn't be stable
in the buildfarm.  Oh well, these tests weren't very valuable
anyway.

Discussion: https://postgr.es/m/20201130165436.GX24052@telsasoft.com

Formerly, TOAST objects were unconditionally suppressed, but since
\d is able to print them it's not very clear why these variants
should not.  Instead, use the same rules as for system catalogs:
they can be seen if you write the 'S' modifier or a table name
pattern.  (In practice, since hardly anybody would keep pg_toast
in their search_path, it's really down to whether you use a pattern
that can match pg_toast.*.)

No docs change seems necessary because the docs already say that
this happens for "system objects"; we're just classifying TOAST
tables as being that.

Justin Pryzby, reviewed by Laurenz Albe

Discussion: https://postgr.es/m/20201130165436.GX24052@telsasoft.com

Aside from being queriable via SHOW, this value is sent to the client
immediately at session startup, and again later on if the server gets
promoted to primary during the session.  The immediate report will be
used in an upcoming patch to avoid an extra round trip when trying to
connect to a primary server.

Haribabu Kommi, Greg Nancarrow, Tom Lane; reviewed at various times
by Laurenz Albe, Takayuki Tsunakawa, Peter Smith.

Discussion: https://postgr.es/m/CAF3+xM+8-ztOkaV9gHiJ3wfgENTq97QcjXQt+rbFQ6F7oNzt9A@mail.gmail.com

Reported-by: "Tang, Haiying"

Discussion: https://postgr.es/m/bbbccf7a3c2d436e85d45869d612fd6b@G08CNEXMBPEKD05.g08.fujitsu.local

Author: "Tang, Haiying"

Backpatch-through: 9.5

Commit bc43b7c2 used fabs() directly on an int variable, which
apparently requires an explicit cast on some platforms.

Per buildfarm.

In power_var_int(), the computation of the number of significant
digits to use in the computation used log(Abs(exp)), which isn't safe
because Abs(exp) returns INT_MIN when exp is INT_MIN. Use fabs()
instead of Abs(), so that the exponent is cast to a double before the
absolute value is taken.

Back-patch to 9.6, where this was introduced (by 7d9a4737).

Discussion: https://postgr.es/m/CAEZATCVd6pMkz=BrZEgBKyqqJrt2xghr=fNc8+Z=5xC6cgWrWA@mail.gmail.com

Commit a271a1b5 added the capability to allow decoding at prepare time.
This adds an isolation testcase to test that decoding happens at commit
time when the consistent snapshot state is reached after prepare but
before commit prepared.

Author: Ajin Cherian
Reviewed-by: Amit Kapila
Discussion:
https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru
https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com

Author: Craig Ringer <craig.ringer@2ndquadrant.com>
Discussion: https://www.postgresql.org/message-id/flat/CAMsr+YF=+ctXBZj3ywmvKNUjWpxmuTuUKuv-rgbHGX5i5pLstQ@mail.gmail.com

Bring heap and hash rmgr desc output in line with nbtree and GiST desc
output by using the name latestRemovedXid for all fields that output the
contents of the latestRemovedXid field from the WAL record's C struct
(stop using local variants).

This seems like a clear improvement because latestRemovedXid is a symbol
name that already appears across many different source files, and so is
probably much more recognizable.

Discussion: https://postgr.es/m/CAH2-Wzkt_Rs4VqPSCk87nyjPAAEmWL8STU9zgET_83EF5YfrLw@mail.gmail.com

Author: Peter Smith
Discussion: https://postgr.es/m/CAHut+PsReyuvww_Fn1NN_Vsv0wBP1bnzuhzRFr_2=y1nNZrG7w@mail.gmail.com

Author: Zhijie Hou
Reviewed-by: Sawada Masahiko
Discussion: https://postgr.es/m/ba88bb58aaf14284abca16aec04bf279@G08CNEXMBPEKD05.g08.fujitsu.local

Instead describe the action that the list effects, or just use "list"
where the meaning is obvious from context.

Author: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue%40alap3.anarazel.de

Author: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue%40alap3.anarazel.de

We agreed to remove this terminology and use something more descriptive.

Discussion: https://postgr.es/m/20200615182235.x7lch5n6kcjq4aue%40alap3.anarazel.de

If the substring start index and length overflow when added together,
substring() misbehaved, either throwing a bogus "negative substring
length" error on a case that should succeed, or failing to complain that
a negative length is negative (and instead returning the whole string,
in most cases).  Unsurprisingly, the text, bytea, and bit variants of
the function all had this issue.  Rearrange the logic to ensure that
negative lengths are always rejected, and add an overflow check to
handle the other case.

Also install similar guards into detoast_attr_slice() (nee
heap_tuple_untoast_attr_slice()), since it's far from clear that
no other code paths leading to that function could pass it values
that would overflow.

Patch by myself and Pavel Stehule, per bug #16804 from Rafi Shamim.

Back-patch to v11.  While these bugs are old, the common/int.h
infrastructure for overflow-detecting arithmetic didn't exist before
commit 4d6ad312, and it doesn't seem like these misbehaviors are bad
enough to justify developing a standalone fix for the older branches.

Discussion: https://postgr.es/m/16804-f4eeeb6c11ba71d4@postgresql.org

Cleanup for commit dee663f7.
Reported-by: Tomas Vondra <tomas.vondra@enterprisedb.com>
Discussion: https://postgr.es/m/CA+hUKGLJ=84YT+NvhkEEDAuUtVHMfQ9i-N7k_o50JmQ6Rpj_OQ@mail.gmail.com

Performance issues with the preceding patch to re-implement array
element assignment within pl/pgsql led me to realize that the read/write
parameter mechanism is misdesigned.  Instead of requiring the assignment
source expression to be such that *all* its references to the target
variable could be passed as R/W, we really want to identify *one*
reference to the target variable to be passed as R/W, allowing any other
ones to be passed read/only as they would be by default.  As long as the
R/W reference is a direct argument to the top-level (hence last to be
executed) function in the expression, there is no harm in R/O references
being passed to other lower parts of the expression.  Nor is there any
use-case for more than one argument of the top-level function being R/W.

Hence, rewrite that logic to identify one single Param that references
the target variable, and make only that Param pass a read/write
reference, not any other Params referencing the target variable.

Discussion: https://postgr.es/m/4165684.1607707277@sss.pgh.pa.us

In the wake of the previous commit, we don't really need this anymore,
since array assignment is primarily handled by the core code.

The only way that that code could still be reached is that a GET
DIAGNOSTICS target variable could be an array element.  But that
doesn't seem like a particularly essential feature.  I'd added it
in commit 55caaaeb, but just because it was easy not because
anyone had actually asked for it.  Hence, revert that patch and
then remove the now-unreachable stuff.  (If we really had to,
we could probably reimplement GET DIAGNOSTICS using the new
assignment machinery; but the cost/benefit ratio looks very poor,
and it'd likely be a bit slower.)

Note that PLPGSQL_DTYPE_RECFIELD remains.  It's possible that we
could get rid of that too, but maintaining the existing behaviors
for RECORD-type variables seems like it might be difficult.  Since
there's not any functional limitation in those code paths as there
was in the ARRAYELEM code, I've not pursued the idea.

Discussion: https://postgr.es/m/4165684.1607707277@sss.pgh.pa.us

Invent new RawParseModes that allow the core grammar to handle
pl/pgsql expressions and assignments directly, and thereby get rid
of a lot of hackery in pl/pgsql's parser.  This moves a good deal
of knowledge about pl/pgsql into the core code: notably, we have to
invent a CoercionContext that matches pl/pgsql's (rather dubious)
historical behavior for assignment coercions.  That's getting away
from the original idea of pl/pgsql as an arm's-length extension of
the core, but really we crossed that bridge a long time ago.

The main advantage of doing this is that we can now use the core
parser to generate FieldStore and/or SubscriptingRef nodes to handle
assignments to pl/pgsql variables that are records or arrays.  That
fixes a number of cases that had never been implemented in pl/pgsql
assignment, such as nested records and array slicing, and it allows
pl/pgsql assignment to support the datatype-specific subscripting
behaviors introduced in commit c7aba7c1.

There are cosmetic benefits too: when a syntax error occurs in a
pl/pgsql expression, the error report no longer includes the confusing
"SELECT" keyword that used to get prefixed to the expression text.
Also, there seem to be some small speed gains.

Discussion: https://postgr.es/m/4165684.1607707277@sss.pgh.pa.us

This patch essentially allows gram.y to implement a family of related
syntax trees, rather than necessarily always parsing a list of SQL
statements.  raw_parser() gains a new argument, enum RawParseMode,
to say what to do.  As proof of concept, add a mode that just parses
a TypeName without any other decoration, and use that to greatly
simplify typeStringToTypeName().

In addition, invent a new SPI entry point SPI_prepare_extended() to
allow SPI users (particularly plpgsql) to get at this new functionality.
In hopes of making this the last variant of SPI_prepare(), set up its
additional arguments as a struct rather than direct arguments, and
promise that future additions to the struct can default to zero.
SPI_prepare_cursor() and SPI_prepare_params() can perhaps go away at
some point.

Discussion: https://postgr.es/m/4165684.1607707277@sss.pgh.pa.us

Author: Justin Pryzby
Discussion: https://postgr.es/m/X/Ff7jfnvJUab013@paquier.xyz

This patch allows PREPARE-time decoding of two-phase transactions (if the
output plugin supports this capability), in which case the transactions
are replayed at PREPARE and then committed later when COMMIT PREPARED
arrives.

Now that we decode the changes before the commit, the concurrent aborts
may cause failures when the output plugin consults catalogs (both system
and user-defined).

We detect such failures with a special sqlerrcode
ERRCODE_TRANSACTION_ROLLBACK introduced by commit 7259736a and stop
decoding the remaining changes. Then we rollback the changes when rollback
prepared is encountered.

Author: Ajin Cherian and Amit Kapila based on previous work by Nikhil Sontakke and Stas Kelvich
Reviewed-by: Amit Kapila, Peter Smith, Sawada Masahiko, Arseny Sher, and Dilip Kumar
Tested-by: Takamichi Osumi
Discussion:
https://postgr.es/m/02DA5F5E-CECE-4D9C-8B4B-418077E2C010@postgrespro.ru
https://postgr.es/m/CAMGcDxeqEpWj3fTXwqhSwBdXd2RS9jzwWscO-XbeCfso6ts3+Q@mail.gmail.com

Backpatch-through: 9.5

Try to be clearer about what computation is actually happening here.

Per bug #16797 from Dana Burd.

Discussion: https://postgr.es/m/16797-f264b0b980b53b8b@postgresql.org

On further reflection it seems better to call PageGetMaxOffsetNumber()
after acquiring a buffer lock on the page.  This shouldn't really
matter, but doing it this way is cleaner.

Follow-up to commit 42288174.

Backpatch: 12-, just like commit 42288174

The logic for determining the latest removed XID for the purposes of
generating recovery conflicts in REDO routines was subtly broken.  It
failed to follow links from HOT chains, and so failed to consider all
relevant heap tuple headers in some cases.

To fix, expand the loop that deals with LP_REDIRECT line pointers to
also deal with HOT chains.  The new version of the loop is loosely based
on a similar loop from heap_prune_chain().

The impact of this bug is probably quite limited, since the horizon code
necessarily deals with heap tuples that are pointed to by LP_DEAD-set
index tuples.  The process of setting LP_DEAD index tuples (e.g. within
the kill_prior_tuple mechanism) is highly correlated with opportunistic
pruning of pointed-to heap tuples.  Plus the question of generating a
recovery conflict usually comes up some time after index tuple LP_DEAD
bits were initially set, unlike heap pruning, where a latestRemovedXid
is generated at the point of the pruning operation (heap pruning has no
deferred "would-be page split" style processing that produces conflicts
lazily).

Only backpatch to Postgres 12, the first version where this logic runs
during original execution (following commit 558a9165).  The index
latestRemovedXid mechanism has had the same bug since it first appeared
over 10 years ago (in commit a760893d), but backpatching to all
supported versions now seems like a bad idea on balance.  Running the
new improved code during recovery seems risky, especially given the lack
of complaints from the field.

Author: Peter Geoghegan <pg@bowt.ie>
Discussion: https://postgr.es/m/CAH2-Wz=Eib393+HHcERK_9MtgNS7Ew1HY=RDC_g6GL46zM5C6Q@mail.gmail.com
Backpatch: 12-

The behavior of cross-type comparisons among date/time data types was
not really explained anywhere.  You could probably infer it if you
recognized the applicability of comments elsewhere about datatype
conversions, but it seems worthy of explicit documentation.

Per bug #16797 from Dana Burd.

Discussion: https://postgr.es/m/16797-f264b0b980b53b8b@postgresql.org

Commit 7ca37fb0 removed regress_putenv from the regress.so library,
so reloading a SQL function dependent on that would not work.
Fix similarly to 52202bb3.

Per buildfarm.

This commit preserves the logic of multirange_in() but makes it more clear
what's going on.  Also, this commit fixes the compiler warning spotted by the
buildfarm.

Reported-by: Tom Lane
Discussion: https://postgr.es/m/2246043.1609290699%40sss.pgh.pa.us

Since at least 2001 we've used putenv() and avoided setenv(), on the
grounds that the latter was unportable and not in POSIX.  However,
POSIX added it that same year, and by now the situation has reversed:
setenv() is probably more portable than putenv(), since POSIX now
treats the latter as not being a core function.  And setenv() has
cleaner semantics too.  So, let's reverse that old policy.

This commit adds a simple src/port/ implementation of setenv() for
any stragglers (we have one in the buildfarm, but I'd not be surprised
if that code is never used in the field).  More importantly, extend
win32env.c to also support setenv().  Then, replace usages of putenv()
with setenv(), and get rid of some ad-hoc implementations of setenv()
wannabees.

Also, adjust our src/port/ implementation of unsetenv() to follow the
POSIX spec that it returns an error indicator, rather than returning
void as per the ancient BSD convention.  I don't feel a need to make
all the call sites check for errors, but the portability stub ought
to match real-world practice.

Discussion: https://postgr.es/m/2065122.1609212051@sss.pgh.pa.us

Attempt to get selectivity estimation for @> (anymultirange, anyrange) operator
caused an error in buildfarm, because this operator was missed in switch()
of calc_hist_selectivity().  Fix that and also make regression tests reliably
check that selectivity estimation for (multi)ranges doesn't fall.  Previously,
whether we test selectivity estimation for (multi)ranges depended on
whether autovacuum managed to gather concurrently to the test.

Reported-by: Michael Paquier
Discussion: https://postgr.es/m/X%2BwmgjRItuvHNBeV%40paquier.xyz