Now that 8.3 is not supported, we can use a CTE and not temp tables.
This allows for auto-oid assignment protection in a future patch.

Reverts commits 73d78e11 and
b0488e5c.  Also reverts pg_upgrade
changes.

This trims down the code, and is in preparation for hardening pg_upgrade
against auto-oid assignment.

Add to commit 73d78e11

Also adjust pg_upgrade to not use this method for optional TOAST table
creation.

Patch by Fabrízio de Royes Mello

Report by Pavel Raiskup

Also, remove OID preservation mention, mention non-text dump formats

Backpatch through 9.4

There's no point in setting up a context error callback when doing
conditional lock acquisition, because we never actually wait and so the
user wouldn't be able to see the context message anywhere.  In fact,
this is more in line with what ConditionalXactLockTableWait is doing.

Backpatch to 9.4, where this was added.

The symbol was added by 71901ab6; the original code was introduced by
6868ed74.  Development of both overlapped which is why we apparently
failed to notice.

This is a (very slight) behavior change, so I'm not backpatching this to
9.4 for now, even though the symbol does exist there.

Event triggers want to know the OID of the interesting object created,
which is the main type.  The array created as part of the operation is
just a subsidiary object which is not of much interest.

Other DDL commands are already returning the OID, which is required for
future additional event trigger work.  This is merely making these
commands in line with the rest of utility command support.

Add a succint comment explaining why it's correct to change the
persistence in this way.  Also s/loggedness/persistence/ because native
speakers didn't like the latter term.

Fabrízio and Álvaro

Thomas Munro

Some of the many error messages introduced in 458857cc missed 'FROM
unpackaged'. Also e016b724 and 45ffeb7e forgot to quote extension
version numbers.

Backpatch to 9.1, just like 458857cc which introduced the messages. Do
so because the error messages thrown when the wrong command is copy &
pasted aren't easy to understand.

We also don't track PREPARE, nor do we track planning time in general, so
let's ignore DEALLOCATE as well for consistency.

Backpatch to 9.4, but not further than that. Although it seems unlikely that
anyone is relying on the current behavior, this is a behavioral change.

Fabien Coelho

CheckConstraintFetch() leaked a cstring in the caller's context for each
CHECK constraint expression it copied into the relcache.  Ordinarily that
isn't problematic, but it can be during CLOBBER_CACHE testing because so
many reloads can happen during a single query; so complicate the code
slightly to allow freeing the cstring after use.  Per testing on buildfarm
member barnacle.

This is exactly like the leak fixed in AttrDefaultFetch() by commit
078b2ed2.  (Yes, this time I did look for
other instances of the same coding pattern :-(.)  Like that patch, no
back-patch, since it seems unlikely that there's any problem except under
very artificial test conditions.

BTW, it strikes me that both of these places would require further work
comparable to commit ab8c84db, if we ever
supported defaults or check constraints on system catalogs: they both
assume they are copying into an empty relcache data structure, and that
conceivably wouldn't be the case during recursive reloading of a system
catalog.  This does not seem worth worrying about for the moment, since
there is no near-term prospect of supporting any such thing.  So I'll
just note the possibility for the archives' sake.

Add a note that some options can be specified multiple times to select
multiple objects to restore.  This replaces the somewhat confusing use
of plurals in the option descriptions themselves.

This enables changing permanent (logged) tables to unlogged and
vice-versa.

(Docs for ALTER TABLE / SET TABLESPACE got shuffled in an order that
hopefully makes more sense than the original.)

Author: Fabrízio de Royes Mello
Reviewed by: Christoph Berg, Andres Freund, Thom Brown
Some tweaking by Álvaro Herrera

Cause the path extraction operators to return their lefthand input,
not NULL, if the path array has no elements.  This seems more consistent
since the case ought to correspond to applying the simple extraction
operator (->) zero times.

Cause other corner cases in field/element/path extraction to return NULL
rather than failing.  This behavior is arguably more useful than throwing
an error, since it allows an expression index using these operators to be
built even when not all values in the column are suitable for the
extraction being indexed.  Moreover, we already had multiple
inconsistencies between the path extraction operators and the simple
extraction operators, as well as inconsistencies between the JSON and
JSONB code paths.  Adopt a uniform rule of returning NULL rather than
throwing an error when the JSON input does not have a structure that
permits the request to be satisfied.

Back-patch to 9.4.  Update the release notes to list this as a behavior
change since 9.3.

Peter Geoghegan

The option is called --tablespace-mapping, not --tablespace.

Amit Kapila

Previously, we would first create the symlinks the way they are in the
original system, and at the end replace them with the mapped symlinks.
That never really made much sense, so now we create the symlink pointing
to the correct location to begin with, so that there's no need to fix
them at the end.

The old coding didn't work correctly on Windows, because Windows junction
points look more like directories than files, and ought to be removed with
rmdir rather than unlink. Also, it incorrectly used "%d" rather than "%u"
to print an Oid, but that's gone now.

Report and patch by Amit Kapila, with minor changes by me. Reviewed by
MauMau. Backpatch to 9.4, where the --tablespace feature was added.

As 'ALTER TABLESPACE .. MOVE ALL' really didn't change the tablespace
but instead changed objects inside tablespaces, it made sense to
rework the syntax and supporting functions to operate under the
'ALTER (TABLE|INDEX|MATERIALIZED VIEW)' syntax and to be in
tablecmds.c.

Pointed out by Alvaro, who also suggested the new syntax.

Back-patch to 9.4.

The new column shows how many backends have a buffer pinned. That can
be useful during development or to diagnose production issues
e.g. caused by vacuum waiting for cleanup locks.

To handle upgrades transparently - the extension might be used in
views - deal with callers expecting the old number of columns.

Reviewed by Fujii Masao and Rajeev rastogi.

We have had INT64_FORMAT and UINT64_FORMAT for a long time, but that's not
good enough if you want something more exotic, like "%20lld".

Abhijit Menon-Sen, per Andres Freund's suggestion.

Cover some cases I omitted before, such as null and empty-string
elements in the path array.  This exposes another inconsistency:
json_extract_path complains about empty path elements but
jsonb_extract_path does not.

jsonb's #> operator segfaulted (dereferencing a null pointer) if the RHS
was a zero-length array, as reported in bug #11207 from Justin Van Winkle.
json's #> operator returns NULL in such cases, so for the moment let's
make jsonb act likewise.

Also add a bunch of regression test queries memorializing the -> and #>
operators' behavior for this and other corner cases.

There is a good argument for changing some of these behaviors, as they
are not very consistent with each other, and throwing an error isn't
necessarily a desirable behavior for operators that are likely to be
used in indexes.  However, everybody can agree that a core dump is the
Wrong Thing, and we need test cases even if we decide to change their
expected output later.

Allows the logging to print the entire text of failed queries, rather
than a truncated version.

While the space is optional, it seems nicer to be consistent with what
you get if you do "SET search_path=...". SET always normalizes the
separator to be comma+space.

Christoph Martin

This reverts commit 083d29c6.

The commit changed the code so that it causes an errors when
IDENTIFY_SYSTEM returns three columns. But which prevents us
from using the replication-related utilities against the server
with older version. This is not what we want. For that
compatibility, we allow the utilities to receive three columns
as the result of IDENTIFY_SYSTEM eventhough it actually returns
four columns in 9.4 or later.

Pointed out by Andres Freund.

5a991ef8 added new column into
the result of IDENTIFY_SYSTEM command. But it was not reflected into
several codes checking that result. Specifically though the number of
columns in the result was increased to 4, it was still compared with 3
in some replication codes.

Back-patch to 9.4 where the number of columns in IDENTIFY_SYSTEM
result was increased.

Report from Michael Paquier

Programs need execute permission on a DLL file to load it.  MSYS
"install" ignores the mode argument, and our Cygwin build statically
links libpq into programs.  That explains the lack of buildfarm trouble.
Back-patch to 9.0 (all supported versions).

In support of this, have the MSVC build follow GNU make in preferring
GNUmakefile over Makefile when a directory contains both.

Michael Paquier, reviewed by MauMau.

strncmp() is a specialized API unsuited for routine copying into
fixed-size buffers.  On a system where the length of a single filename
can exceed MAXPGPATH, the pg_archivecleanup change prevents a simple
crash in the subsequent strlen().  Few filesystems support names that
long, and calling pg_archivecleanup with untrusted input is still not a
credible use case.  Therefore, no back-patch.

David Rowley

The aboriginal sample placed connection parameters in
groupOfUniqueNames/uniqueMember.  OpenLDAP, at least as early as version
2.4.23, rejects uniqueMember entries that do not conform to the syntax
for a distinguished name.  Use device/description, which is free-form.
Back-patch to 9.4 for web site visibility.