The buildfarm's CLOBBER_CACHE_ALWAYS animals aren't happy with some
of the test cases added in commit 4b93f579.  There are two different
problems:

* In two places, a different CONTEXT stack is shown because the error
is detected in a different place, due to recompiling an expression
from scratch rather than re-using a previously cached plan for it.
I fixed these via the expedient of hiding the CONTEXT stack altogether.

* In one place, a test expected to fail (because a cached plan hadn't
been updated) actually succeeds (because the forced recompile makes
it good).  I couldn't think of a simple workaround for this, so I've
just commented out that test step altogether.

I have hopes of improving things enough that both of these kluges can
be reverted eventually.  The first one is the same kind of problem
previously discussed at
https://postgr.es/m/31545.1512924904@sss.pgh.pa.us
but there was insufficient agreement about how to fix it, so we
just hacked around the output instability (commit 9edc97b7).
The second issue should be fixed by allowing the plan to be rebuilt
when a type conflict is detected.  But for today, let's just make the
buildfarm green again.

Some older compilers otherwise sometimes complain about undefined
values, even though the return value should not be used in the
overflow case.  We assume that any decent compiler will optimize away
the unnecessary assignment in performance critical cases.

We do not want to restrain the returned value to a specific value,
e.g. 0 or the wrapped-around value, because some fast ways to
implement overflow detecting math do not easily allow for
that (e.g. msvc intrinsics).  As the function documentation already
documents the returned value in case of intrinsics to be
implementation defined, no documentation has to be updated.

Per complaint from Tom Lane and his buildfarm member prairiedog.

Author: Andres Freund
Discussion: https://postgr.es/m/18169.1513958454@sss.pgh.pa.us

All of these are false positives, but in each case a fair amount of
analysis is needed to see that, and it's not too surprising that not all
compilers are smart enough.  (In particular, in the logtape.c case, a
compiler lacking the knowledge provided by the Assert would almost surely
complain, so that this warning will be seen in any non-assert build.)

Some of these are of long standing while others are pretty recent,
but it only seems worth fixing them in HEAD.

Jaime Casanova, tweaked a bit by me

Discussion: https://postgr.es/m/CAJGNTeMcYAMJdPAom52dppLMtF-UnEZi0dooj==75OEv1EoBZA@mail.gmail.com

Per commit e748e902, we appear to have little or no coverage in the
buildfarm of machines that will dump core when asked to printf a
null string pointer.  Let's try to improve that situation by adding
an assertion that will make src/port/snprintf.c behave that way.
Since it's just an assertion, it won't break anything in production
builds, but it will help developers find this type of oversight.

Note that while our buildfarm coverage of machines that use that
snprintf implementation is pretty thin on the Unix side (apparently
amounting only to gaur/pademelon), all of the MSVC critters use it.

Discussion: https://postgr.es/m/156b989dbc6fe7c4d3223cf51da61195@postgrespro.ru

plpython_error_callback() reported the name of the function associated
with the topmost PL/Python execution context.  This was not merely
wrong if there were nested PL/Python contexts, but it risked a core
dump if the topmost one is an inline code block rather than a named
function.  That will have proname = NULL, and so we were passing a NULL
pointer to snprintf("%s").  It seems that none of the PL/Python-testing
machines in the buildfarm will dump core for that, but some platforms do,
as reported by Marina Polyakova.

Investigation finds that there actually is an existing regression test
that used to prove that the behavior was wrong, though apparently no one
had noticed that it was printing the wrong function name.  It stopped
showing the problem in 9.6 when we adjusted psql to not print CONTEXT
by default for NOTICE messages.  The problem is masked (if your platform
avoids the core dump) in error cases, because PL/Python will throw away
the originally generated error info in favor of a new traceback produced
at the outer level.

Repair by using ErrorContextCallback.arg to pass the correct context to
the error callback.  Add a regression test illustrating correct behavior.

Back-patch to all supported branches, since they're all broken this way.

Discussion: https://postgr.es/m/156b989dbc6fe7c4d3223cf51da61195@postgrespro.ru

These features were never implemented previously for composite or record
variables ... not that the documentation admitted it, so there's no doc
updates here.

This also fixes some issues concerning enforcing DOMAIN NOT NULL
constraints against plpgsql variables, although I'm not sure that
that topic is completely dealt with.

I created a new plpgsql test file for these features, and moved the
one relevant existing test case into that file.

Tom Lane, reviewed by Daniel Gustafsson

Discussion: https://postgr.es/m/18362.1514605650@sss.pgh.pa.us

Over the years we've accreted quite a few special variables that are
predefined in plpgsql trigger functions.  The cost of initializing these
variables to their defined values turns out to be a significant part of
the runtime of simple triggers; but, undoubtedly, most real-world triggers
never examine the values of most of these variables.

To improve matters, invent the notion of a variable that has a "promise"
attached to it, specifying which of the predetermined values should be
assigned to the variable if anything ever reads it.  This eliminates all
the unneeded startup overhead, in return for a small penalty on accesses
to these variables.

Tom Lane, reviewed by Pavel Stehule

Discussion: https://postgr.es/m/11986.1514407114@sss.pgh.pa.us

Previously, copy_plpgsql_datum did a separate palloc for each variable
needing instance-local storage.  In simple benchmarks this made for a
noticeable fraction of the total runtime.  Improve it by precalculating
the space needed for all of a function's variables and doing just one
palloc for all of them.

In passing, remove PLPGSQL_DTYPE_EXPR from the list of plpgsql "datum"
types, since in fact it has nothing in common with the others, and there
is noplace that needs to discriminate on the basis of dtype between an
expression and any type of datum.  And add comments clarifying which
datum struct fields are generic and which aren't.

Tom Lane, reviewed by Pavel Stehule

Discussion: https://postgr.es/m/11986.1514407114@sss.pgh.pa.us

Formerly, DTYPE_REC was used only for variables declared as "record";
variables of named composite types used DTYPE_ROW, which is faster for
some purposes but much less flexible.  In particular, the ROW code paths
are entirely incapable of dealing with DDL-caused changes to the number
or data types of the columns of a row variable, once a particular plpgsql
function has been parsed for the first time in a session.  And, since the
stored representation of a ROW isn't a tuple, there wasn't any easy way
to deal with variables of domain-over-composite types, since the domain
constraint checking code would expect the value to be checked to be a
tuple.  A lesser, but still real, annoyance is that ROW format cannot
represent a true NULL composite value, only a row of per-field NULL
values, which is not exactly the same thing.

Hence, switch to using DTYPE_REC for all composite-typed variables,
whether "record", named composite type, or domain over named composite
type.  DTYPE_ROW remains but is used only for its native purpose, to
represent a fixed-at-compile-time list of variables, for instance the
targets of an INTO clause.

To accomplish this without taking significant performance losses, introduce
infrastructure that allows storing composite-type variables as "expanded
objects", similar to the "expanded array" infrastructure introduced in
commit 1dc5ebc9.  A composite variable's value is thereby kept (most of
the time) in the form of separate Datums, so that field accesses and
updates are not much more expensive than they were in the ROW format.
This holds the line, more or less, on performance of variables of named
composite types in field-access-intensive microbenchmarks, and makes
variables declared "record" perform much better than before in similar
tests.  In addition, the logic involved with enforcing composite-domain
constraints against updates of individual fields is in the expanded
record infrastructure not plpgsql proper, so that it might be reusable
for other purposes.

In further support of this, introduce a typcache feature for assigning a
unique-within-process identifier to each distinct tuple descriptor of
interest; in particular, DDL alterations on composite types result in a new
identifier for that type.  This allows very cheap detection of the need to
refresh tupdesc-dependent data.  This improves on the "tupDescSeqNo" idea
I had in commit 687f096e: that assigned identifying sequence numbers to
successive versions of individual composite types, but the numbers were not
unique across different types, nor was there support for assigning numbers
to registered record types.

In passing, allow plpgsql functions to accept as well as return type
"record".  There was no good reason for the old restriction, and it
was out of step with most of the other PLs.

Tom Lane, reviewed by Pavel Stehule

Discussion: https://postgr.es/m/8962.1514399547@sss.pgh.pa.us

This also makes procedures work in psql's \ef and \sf commands.
Reported-by: Pavel Stehule <pavel.stehule@gmail.com>

Instead of issuing a reload after pg_hba.conf changes between test
cases, run a full restart.  With a reload, an error in the new
pg_hba.conf is ignored and the tests will continue to run with the old
settings, invalidating the subsequent test cases.  With a restart, a
faulty pg_hba.conf will lead to the test being aborted, which is what
we'd rather want.

Author: Masahiko Sawada <sawada.mshk@gmail.com>

The modern way is to use a missing_ok argument instead of two separate
almost-identical routines, so do that.

Author: Michaël Paquier
Reviewed-by: Álvaro Herrera
Discussion: https://postgr.es/m/20180201063212.GE6398@paquier.xyz

The previous code failed to realize that this setting effectively
disables parallelism, and would crash if it decided to attempt
parallelism anyway.  Instead, treat it as a disabling condition.

Kyotaro Horiguchi, who also reported the issue. Reviewed by Michael
Paquier and Peter Geoghegan.

Discussion: http://postgr.es/m/20180209.170635.256350357.horiguchi.kyotaro@lab.ntt.co.jp

Noticed while reviewing nearby text

Also add comment on why exit/quit are not documented.

Discussion: https://postgr.es/m/20180202053928.GA13472@momjian.us

pg_dump supposed that a stats object necessarily shares the same schema
as its underlying table, and that it doesn't have a separate owner.
These things may have been true during early development of the feature,
but they are not true as of v10 release.

Failure to track the object's schema separately turns out to have only
limited consequences, because pg_get_statisticsobjdef() always schema-
qualifies the target object name in the generated CREATE STATISTICS command
(a decision out of step with the rest of ruleutils.c, but I digress).
Therefore the restored object would be in the right schema, so that the
only problem is that the TOC entry would be mislabeled as to schema.  That
could lead to wrong decisions for schema-selective restores, for example.

The ownership issue is a bit more serious: not only was the TOC entry
potentially mislabeled as to owner, but pg_dump didn't bother to issue an
ALTER OWNER command at all, so that after restore the stats object would
continue to be owned by the restoring superuser.

A final point is that decisions as to whether to dump a stats object or
not were driven by whether the underlying table was dumped or not.  While
that's not wrong on its face, it won't scale nicely to the planned future
extension to cross-table statistics.  Moreover, that design decision comes
out of the view of stats objects as being auxiliary to a particular table,
like a rule or trigger, which is exactly where the above problems came
from.  Since we're now treating stats objects more like independent objects
in their own right, they ought to behave like standalone objects for this
purpose too.  So change to using the generic selectDumpableObject() logic
for them (which presently amounts to "dump if containing schema is to be
dumped").

Along the way to fixing this, restructure so that getExtendedStatistics
collects the identity info (only) for all extended stats objects in one
query, and then for each object actually being dumped, we retrieve the
definition in dumpStatisticsExt.  This is necessary to ensure that
schema-qualification in the generated CREATE STATISTICS command happens
with respect to the search path that pg_dump will now be using at restore
time (ie, the schema the stats object is in, not that of the underlying
table).  It's probably also significantly faster in the typical scenario
where only a minority of tables have extended stats.

Back-patch to v10 where extended stats were introduced.

Discussion: https://postgr.es/m/18272.1518328606@sss.pgh.pa.us

Prematurely freeing the EState used to evaluate CALL arguments led, in some
cases, to passing dangling pointers to the procedure.  This was masked in
trivial cases because the argument pointers would point to Const nodes in
the original expression tree, and in some other cases because the result
value would end up in the standalone ExprContext rather than in memory
belonging to the EState --- but that wasn't exactly high quality
programming either, because the standalone ExprContext was never
explicitly freed, breaking assorted API contracts.

In addition, using a separate EState for each argument was just silly.

So let's use just one EState, and one ExprContext, and make the latter
belong to the former rather than be standalone, and clean up the EState
(and hence the ExprContext) post-call.

While at it, improve the function's commentary a bit.

Discussion: https://postgr.es/m/29173.1518282748@sss.pgh.pa.us

CALL statements cannot support sub-SELECTs in the arguments of the called
procedure, since they just use ExecEvalExpr to evaluate such arguments.
Teach transformSubLink() to reject the case, as it already does for other
contexts in which subqueries are not supported.

In passing, s/EXPR_KIND_CALL/EXPR_KIND_CALL_ARGUMENT/ to make that enum
symbol line up more closely with the phrasing of the error messages it is
associated with.  And fix someone's weak grasp of English grammar in the
preceding EXPR_KIND_PARTITION_EXPRESSION addition.  Also update an
incorrect comment in resolve_unique_index_expr (possibly it was correct
when written, but nowadays transformExpr definitely does reject SRFs here).

Per report from Pavel Stehule --- but this resolves only one of the bugs
he mentions.

Discussion: https://postgr.es/m/CAFj8pRDxOwPPzpA8i+AQeDQFj7bhVw-dR2==rfWZ3zMGkm568Q@mail.gmail.com

We can now create indexes more easily than before, so update this
chapter to use the simpler instructions.

After an idea of Amit Langote.  I (Álvaro) opted to do more invasive
surgery and remove the previous suggestion to create per-partition
indexes, which his patch left in place.

Discussion: https://postgr.es/m/eafaaeb1-f0fd-d010-dd45-07db0300f645@lab.ntt.co.jp
Author: Amit Langote, Álvaro Herrera

This makes life easier for extension authors.

Metin Doslu

Discussion: http://postgr.es/m/CAL1dPcfa45o1dC-c4t-48v0OZE6oy4ChJhObrtkK8mzNfXqDTA@mail.gmail.com

Otherwise, we can end up with the flag set when the timeout is
actually disabled, leading to misbehavior.  Commit
f8e5f156 introduced this bug.

Reported by Peter Eisentraut.  Analysis and fix by Thomas Munro,
tweaked by me.

Discussion: http://postgr.es/m/6a909374-2602-7136-8c70-397330a418f3@2ndquadrant.com

Even after commit 882ea509, some
buildfarm members are still failing in the postgres_fdw tests.
Try to fix that by disabling use of remote statistics for some
test cases.

Etsuro Fujita

Discussion: http://postgr.es/m/5A7D76CF.8080601@lab.ntt.co.jp

Atsushi Torikoshi

Discussion: http://postgr.es/m/1b056262-4bc0-a982-c899-bb67a0a7fd52@lab.ntt.co.jp

Doing so causes EXPLAIN ANALYZE to show trigger statistics multiple
times.  Commit 2f178441 seems to
be to blame for this.

Amit Langote, revieed by Amit Khandekar, Etsuro Fujita, and me.

When the previously-chosen plan was non-partial, all pa_finished
flags for partial plans are now set, and pa_next_plan has not yet
been set to INVALID_SUBPLAN_INDEX, the previous code could go into
an infinite loop.

Report by Rajkumar Raghuwanshi.  Patch by Amit Khandekar and me.
Review by Kyotaro Horiguchi.

Discussion: http://postgr.es/m/CAJ3gD9cf43z78qY=U=H0HvOEN341qfRO-vLpnKPSviHeWgJQ5w@mail.gmail.com

Instead of using the psql/libpq connection string as the displayed test
name and relying on "notes" and source code comments to explain the
tests, give the tests self-explanatory names, like we do elsewhere.
Reviewed-by: Michael Paquier <michael.paquier@gmail.com>
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>

Commit 1bc0100d added these tests,
but they're not stable enough to survive in the buildfarm.  Remove
CTIDs from the output in the hopes of fixing that.

Commit 0bf3ae88 allowed direct
foreign table modification; instead of fetching each row, updating
it locally, and then pushing the modification back to the remote
side, we would instead do all the work on the remote server via a
single remote UPDATE or DELETE command.  However, that commit only
enabled this optimization when join tree consisted only of the
target table.

This change allows the same optimization when an UPDATE statement
has a FROM clause or a DELETE statement has a USING clause.  This
works much like ordinary foreign join pushdown, in that the tables
must be on the same remote server, relevant parts of the query
must be pushdown-safe, and so forth.

Etsuro Fujita, reviewed by Ashutosh Bapat, Rushabh Lathia, and me.
Some formatting corrections by me.

Discussion: http://postgr.es/m/5A57193A.2080003@lab.ntt.co.jp
Discussion: http://postgr.es/m/b9cee735-62f8-6c07-7528-6364ce9347d0@lab.ntt.co.jp

Add explicit collation on the trigger name to avoid locale dependencies.
Also restrict the tables selected, to avoid interference from
concurrently running tests.

- table_constraints.enforced
- triggers.action_order
- triggers.action_reference_old_table
- triggers.action_reference_new_table
Reviewed-by: Michael Paquier <michael.paquier@gmail.com>

Commit 4b0d28de should have updated
this comment, but did not.

Thomas Munro

Discussion: http://postgr.es/m/CAEepm=0iJ8aqQcF9ij2KerAkuHF3SwrVTzjMdm1H4w++nfBf9A@mail.gmail.com

Commit 5ded4bd2 removed the code
for this function, but neglected to remove the prototype and
associated comments.

Dagfinn Ilmari Mannsåker

Discussion: http://postgr.es/m/d8j4lmuxjzk.fsf@dalvik.ping.uio.no

Since we now support the server side handler for git over https (so
we're no longer using the "dumb protocol"), make https the primary
choice for cloning the repository, and the git protocol the secondary
choice.

In passing, also change the links to git-scm.com from http to https.

Reviewed by Stefan Kaltenbrunner and David G. Johnston

This patch adds the ability to use "RANGE offset PRECEDING/FOLLOWING"
frame boundaries in window functions.  We'd punted on that back in the
original patch to add window functions, because it was not clear how to
do it in a reasonably data-type-extensible fashion.  That problem is
resolved here by adding the ability for btree operator classes to provide
an "in_range" support function that defines how to add or subtract the
RANGE offset value.  Factoring it this way also allows the operator class
to avoid overflow problems near the ends of the datatype's range, if it
wishes to expend effort on that.  (In the committed patch, the integer
opclasses handle that issue, but it did not seem worth the trouble to
avoid overflow failures for datetime types.)

The patch includes in_range support for the integer_ops opfamily
(int2/int4/int8) as well as the standard datetime types.  Support for
other numeric types has been requested, but that seems like suitable
material for a follow-on patch.

In addition, the patch adds GROUPS mode which counts the offset in
ORDER-BY peer groups rather than rows, and it adds the frame_exclusion
options specified by SQL:2011.  As far as I can see, we are now fully
up to spec on window framing options.

Existing behaviors remain unchanged, except that I changed the errcode
for a couple of existing error reports to meet the SQL spec's expectation
that negative "offset" values should be reported as SQLSTATE 22013.

Internally and in relevant parts of the documentation, we now consistently
use the terminology "offset PRECEDING/FOLLOWING" rather than "value
PRECEDING/FOLLOWING", since the term "value" is confusingly vague.

Oliver Ford, reviewed and whacked around some by me

Discussion: https://postgr.es/m/CAGMVOdu9sivPAxbNN0X+q19Sfv9edEPv=HibOJhB14TJv_RCQg@mail.gmail.com

Etsuro Fujita

Discussion: http://postgr.es/m/5A7981EA.8020201@lab.ntt.co.jp

LogicalTapeFreeze() may write out its first block when it is dirty but
not full, and then immediately read the first block back in from its
BufFile as a BLCKSZ-width block.  This can only occur in rare cases
where very few tuples were written out, which is currently only
possible with parallel external tuplesorts.  To avoid valgrind
complaints, tell it to treat the tail of logtape.c's buffer as
defined.

Commit 9da0cc35 exposed this problem
but did not create it.  LogicalTapeFreeze() has always tended to write
out some amount of garbage bytes, but previously never wrote less than
one block of data in total, so the problem was masked.

Per buildfarm members lousyjack and skink.

Peter Geoghegan, based on a suggestion from Tom Lane and me.  Some
comment revisions by me.