This commit adds the description that "non-exclusive" pg_start_backup
and pg_stop_backup can be executed even during recovery. Previously
it was wrongly documented that those functions are not allowed to be
executed during recovery.

Back-patch to 9.6 where non-exclusive backup API was added.

Discussion: https://postgr.es/m/CAHGQGwEuAYrEX7Yhmf2MCrTK81HDkkg-JqsOUh8zw6+zYC5zzw@mail.gmail.com

The formulas used to calculate size while (de)serializing mvndistinct
and functional dependencies were based on offset() of the structs. But
that is incorrect, because the structures are not copied directly, we
we copy the individual fields directly.

At the moment this works fine, because there is no alignment padding
on any platform we support. But it might break if we ever added some
fields into any of the structs, for example. It's also confusing.

Fixed by reworking the macros to directly sum sizes of serialized
fields. The macros are now useful only for serialiation, so there is
no point in keeping them in the public header file. So make them
private by moving them to the .c files.

Also adds a couple more asserts to check the serialization, and fixes
an incorrect allocation of MVDependency instead of (MVDependency *).

Reported-By: Tom Lane
Discussion: https://postgr.es/m/29785.1555365602@sss.pgh.pa.us

Backpatch-through: 10

The GSSAPI encryption patch neglected to update the protocol
documentation to describe how to set up a GSSAPI encrypted connection
from a client to the server, so fix that by adding the appropriate
documentation to protocol.sgml.

The tests added for encryption support were overly long and couldn't be
run in parallel due to race conditions; this was largely because each
test was setting up its own KDC to perform the tests.  Instead, merge
the authentication tests and the encryption tests into the original
test, where we only create one KDC to run the tests with.  Also, have
the tests check what the server's opinion is of the connection and if it
was GSS authenticated or encrypted using the pg_stat_gssapi view.

In passing, fix the libpq label for GSSENC-Mode to be consistent with
the "PGGSSENCMODE" environment variable.

Missing protocol documentation pointed out by Michael Paquier.
Issues with the tests pointed out by Tom Lane and Peter Eisentraut.

Refactored tests and added documentation by me.

Reviewed by Robbie Harwood (protocol documentation) and Michael Paquier
(rework of the tests).

For amcanreorderby scans the nodeIndexscan.c's reorder queue holds
heap tuples, but the underlying table likely does not. Before this fix
we'd return different types of slots, depending on whether the tuple
came from the reorder queue, or from the index + table.

While that could be fixed by signalling that the node doesn't return a
fixed type of slot, it seems better to instead remove the separate
slot for the reorder queue, and use ExecForceStoreHeapTuple() to store
tuples from the queue. It's not particularly common to need
reordering, after all.

This reverts most of the iss_ReorderQueueSlot related changes to
nodeIndexscan.c made in 1a0586de, except that now
ExecForceStoreHeapTuple() is used instead of ExecStoreHeapTuple().

Noticed when testing zheap against the in-core version of tableam.

Author: Andres Freund

As reported by Tom, when ExecStoreMinimalTuple() had to perform a
conversion to store the minimal tuple in the slot, it forgot to
respect the shouldFree flag, and leaked the tuple into the current
memory context if true.  Fix that by freeing the tuple in that case.

Looking at the relevant code made me (Andres) realize that not having
the shouldFree parameter to ExecForceStoreHeapTuple() was a bad
idea. Some callers had to locally implement the necessary logic, and
in one case it was missing, creating a potential per-group leak in
non-hashed aggregation.

The choice to not free the tuple in ExecComputeStoredGenerated() is
not pretty, but not introduced by this commit - I'll start a separate
discussion about it.

Reported-By: Tom Lane
Discussion: https://postgr.es/m/366.1555382816@sss.pgh.pa.us

HoldPinnedPortals() did things in the wrong order: it must not mark
a portal autoHeld until it's been successfully held.  Otherwise,
a failure while persisting the portal results in a server crash
because we think the portal is in a good state when it's not.

Also add a check that portal->status is READY before attempting to
hold a pinned portal.  We have such a check before the only other
use of HoldPortal(), so it seems unwise not to check it here.

Lastly, rethink the responsibility for where to call HoldPinnedPortals.
The comment for it imagined that it was optional for any individual PL
to call it or not, but that cannot be the case: if some outer level of
procedure has a pinned portal, failing to persist it when an inner
procedure commits is going to be trouble.  Let's have SPI do it instead
of the individual PLs.  That's not a complete solution, since in theory
a PL might not be using SPI to perform commit/rollback, but such a PL
is going to have to be aware of lots of related requirements anyway.
(This change doesn't cause an API break for any external PLs that might
be calling HoldPinnedPortals per the old regime, because calling it
twice during a commit or rollback sequence won't hurt.)

Per bug #15703 from Julian Schauder.  Back-patch to v11 where this code
came in.

Discussion: https://postgr.es/m/15703-c12c5bc0ea34ba26@postgresql.org

Author: Justin Pryzby
Discussion: https://postgr.es/m/20190330224333.GQ5815@telsasoft.com

This was missing from 88bdbd3f.

Author: Christoph Berg, Justin Pryzby
Discussion: https://postgr.es/m/20190403215938.GA26375@alvherre.pgsql
Discussion: https://postgr.es/m/20190330224333.GQ5815@telsasoft.com

If contrib/pageinspect is not installed, this causes the test checking
the minimum recovery point to fail.  The point is that the dependency
with pageinspect is not really necessary as the test does also all
checks with an offline cluster by scanning directly the on-disk pages,
which is enough for the purpose of the test.

Per complaint from Tom Lane.

Author: Michael Paquier
Discussion: https://postgr.es/m/17806.1555566345@sss.pgh.pa.us

When such a trigger returns the old row version, it naturally get
stored in the slot for the trigger result. When a table AMs doesn't
store HeapTuples internally, ExecBRUpdateTriggers() frees the old row
version passed to triggers - but before this fix it might still be
referenced by the slot holding the new tuple.

Noticed when running the out-of-core zheap AM against the in-core
version of tableam.

Author: Andres Freund

If a FOR ALL TABLES publication exists, temporary and unlogged tables
are ignored for publishing changes.  But CheckCmdReplicaIdentity()
would still check in that case that such a table has a replica
identity set before accepting updates.  To fix, have
GetRelationPublicationActions() return that such a table publishes no
actions.

Discussion: https://www.postgresql.org/message-id/f3f151f7-c4dd-1646-b998-f60bd6217dd3@2ndquadrant.com

Reported-By: Michael Paquier, Michel Pelletier
Discussion:
   https://postgr.es/m/20190410025531.GA2728@paquier.xyz
   https://postgr.es/m/CACxu=v+u_QTeFqdajCHv3i4QmzV_63arVb57R19dSKtThdSLkQ@mail.gmail.com

I (Andres) missed this in 578b2297, the removal of WITH OIDS
support.

Author: Daniel Verite
Discussion: https://postgr.es/m/f06e9735-3717-4904-8c95-47d0b9c3bb10@manitou-mail.org

I (Andres) missed these in 578b2297.

Author: Justin Pryzby, editorialized a bit by Andres Freund
Reviewed-By: Daniel Verite, Andres Freund
Discussion: https://postgr.es/m/20190408002847.GA904@telsasoft.com

* Remove one unnecessary pg_class join in SQL command.  Not needed,
  because we use a regclass cast instead.

* Doc: refer to "partitioned relations" rather than specifically tables,
  since indexes are also displayed.

* Rename "On table" column to "Table", for consistency with \di.

Author: Justin Pryzby
Discussion: https://postgr.es/m/20190407212525.GB10080@telsasoft.com

Nothing was shown previously.

Previously, include actions include_dir, include_if_exists, and include
listed commented-out values which were not the defaults, which is
inconsistent with other entries.  Instead, replace them with '', which
is the default value.

Reported-by: Emanuel Araújo

Discussion: https://postgr.es/m/CAMuTAkYMx6Q27wpELDR3_v9aG443y7ZjeXu15_+1nGUjhMWOJA@mail.gmail.com

Backpatch-through: 9.4

The previous paragraph trying to explain --check, --link, and no --link
modes and the various points of failure was too complex.  Instead, use
bullet lists and sublists.

Reported-by: Daniel Gustafsson

Discussion: https://postgr.es/m/qtqiv7hI87s_Xvz5ZXHCaH-1-_AZGpIDJowzlRjF3-AbCr3RhSNydM_JCuJ8DE4WZozrtxhIWmyYTbv0syKyfGB6cYMQitp9yN-NZMm-oAo=@yesql.se

Backpatch-through: 9.4

The buildfarm points out that UINT64_FORMAT might not work with sscanf;
it's calibrated for our printf implementation, which might not agree
with the platform-supplied sscanf.  Fall back to just accepting an
unsigned long, which is already more than the documentation promises.

Oversight in e6c3ba7f; back-patch to v11, as that was.

I noted that some buildfarm members were complaining about %ld being
used to format values that are (probably) declared size_t.  Use %zu
instead, and insert a cast just in case some versions of the GSSAPI
API declare the length field differently.  While at it, clean up
gratuitous differences in wording of equivalent messages, show
the complained-of length in all relevant messages not just some,
include trailing newline where needed, adjust random deviations
from project-standard code layout and message style, etc.

Restore missed "make clean" rule, fix misspelling.

John Naylor

Discussion: https://postgr.es/m/CACPNZCt5B8jDCCGQiFoSuqmg-za_NCy4QDioBTLaNRih9+-bXg@mail.gmail.com

Returning 0 could falsely indicate that there is no problem. NULL
correctly indicates that there is no information about potential
problems.

Also return 0 as numbackends instead of NULL for shared objects (as no
connection can be made to a shared object only).

Author: Julien Rouhaud <rjuju123@gmail.com>
Reviewed-by: Robert Treat <rob@xzilla.net>

When saving a replication slot, failing to close the temporary path used
to save the slot information is considered as a failure and reported as
such.  However the code forgot to leave immediately as other failure
paths do.

Noticed while looking up at this area of the code for another patch.

Transient files and wait events get normally cleaned up when seeing an
exception (be it in the context of a transaction for a backend or
another process like the checkpointer), hence there is little point in
complicating error code paths to do this work.  This shaves a bit of
code, and removes some extra handling with errno which needed to be
preserved during the cleanup steps done.

Reported-by: Masahiko Sawada
Author: Michael Paquier
Reviewed-by: Tom Lane, Masahiko Sawada
Discussion: https://postgr.es/m/CAD21AoDhHYVq5KkXfkaHhmjA-zJYj-e4teiRAJefvXuKJz1tKQ@mail.gmail.com

Per discussion with others, allowing REINDEX INDEX CONCURRENTLY to work
for invalid indexes when working directly on them can have a lot of
value to unlock situations with invalid indexes without having to use a
dance involving DROP INDEX followed by an extra CREATE INDEX
CONCURRENTLY (which would not work for indexes with constraint
dependency anyway).  This also does not create extra bloat on the
relation involved as this works on individual indexes, so let's enable
it.

Note that REINDEX TABLE CONCURRENTLY still bypasses invalid indexes as
we don't want to bloat the number of indexes defined on a relation in
the event of multiple and successive failures of REINDEX CONCURRENTLY.

More regression tests are added to cover those behaviors, using an
invalid index created with CREATE INDEX CONCURRENTLY.

Reported-by: Dagfinn Ilmari Mannsåker, Álvaro Herrera
Author: Michael Paquier
Reviewed-by: Peter Eisentraut, Dagfinn Ilmari Mannsåker
Discussion: https://postgr.es/m/20190411134947.GA22043@alvherre.pgsql

Author: Fujii Masao
Reviewed-By: Alvaro Herrera, Michael Paquier
Discussion: https://postgr.es/m/CAHGQGwHyKt9-xkibVguPzYqKgb_2tdw14Ub1XDTu08kyHMiTQA@mail.gmail.com

Turn on the previously disabled automatic scaling of images in HTML
output.  To avoid images looking too large on nowadays-normal screens,
restrict the width to 75% on such screens.

Some work is still necessary because SVG images without a viewBox
still won't scale, but that will a separate patch.

Discussion: https://www.postgresql.org/message-id/flat/6d2442d1-84a2-36ef-e014-b6d1ece8a139%40postgresql.org

The private data in the WAL reader is already getting set when
allocating it.

Author: Antonin Houska
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/30563.1555329094@localhost

Instead, close that stdin.  Per buildfarm member conchuela.  Back-patch
to 9.6, where the test was introduced.

Discussion: https://postgr.es/m/26478.1555373328@sss.pgh.pa.us

This struct seems to have not gotten the word about preferred
coding style for variable-length arrays.

The regression tests added in commit 7300a699 test cardinality
estimates using a function that extracts the interesting pieces
from the EXPLAIN output, instead of testing the whole plan. That
seems both easier to understand and less fragile, so this applies
the same approach to pre-existing tests of ndistinct coefficients
and functional dependencies.

Discussion: https://postgr.es/m/dfdac334-9cf2-2597-fb27-f0fb3753f435@2ndquadrant.com

The memcpy() was copying type OIDs in the wrong direction, so the
deserialized MCV list always had them as 0. This is mostly harmless
except when printing the data in pg_mcv_list_items(), in which case
it reported

    ERROR:  cache lookup failed for type 0

Also added a simple regression test for pg_mcv_list_items() function,
printing a single-item MCV list.

Reported-By: Dean Rasheed
Discussion: https://postgr.es/m/CAEZATCX6T0iDTTZrqyec4Cd6b4yuL7euu4=rQRXaVBAVrUi1Cg@mail.gmail.com

Commit 5e1963fb overlooked two places in partbounds.c that now
need to pass a collation identifier to the hash functions for
a partition key column.

Amit Langote, per report from Jesper Pedersen

Discussion: https://postgr.es/m/a620f85a-42ab-e0f3-3337-b04b97e2e2f5@redhat.com

Concurrent autovacuum could result in a change in the order of the
live rows in timestamp_tbl.  While this would not happen with the
default autovacuum parameters, it's fairly easy to hit if
autovacuum_vacuum_threshold is made small enough to allow autovac
to decide to process this table.  That's a stumbling block for trying
to exercise autovacuum aggressively using the core regression tests.

To fix, replace an unqualified DELETE with a TRUNCATE.  There's a
similar DELETE just above (and no order-sensitive queries between),
so this doesn't lose any test coverage and might indeed be argued
to improve it.

Discussion: https://postgr.es/m/17428.1555348950@sss.pgh.pa.us

Checks inside _bt_vacuum_needs_cleanup() allow division by zero to happen when
metad->btm_last_cleanup_num_heap_tuples == 0.  This commit adjusts the
expression so that no division by zero might happen.

Reported-by: Piotr Stefaniak
Discussion: https://postgr.es/m/DB8PR03MB5931C41F7787A95313F08322F22A0%40DB8PR03MB5931.eurprd03.prod.outlook.com
Reviewed-by: Masahiko Sawada
Backpatch-through: 11

Commit 3f2393ed changed ExecCleanupTupleRouting() so that it skipped
cleaning up subplan resultrels before calling EndForeignInsert(), but
that would cause an issue: when those resultrels were foreign tables,
the FDWs would fail to shut down.  Repair by skipping it after calling
EndForeignInsert() as before.

Author: Etsuro Fujita
Reviewed-by: David Rowley and Amit Langote
Discussion: https://postgr.es/m/5CAF3B8F.2090905@lab.ntt.co.jp

The same code is used to handle both text and bytea, but bytea is not
collation-aware, so we shouldn't call get_collation_isdeterministic()
in that case, since that will error out with an invalid collation.
Reported-by: Jeevan Chalke <jeevan.chalke@enterprisedb.com>
Discussion: https://www.postgresql.org/message-id/flat/CAM2%2B6%3DWaf3qJ1%3DyVTUH8_yG-SC0xcBMY%2BSFLhvKKNnWNXSUDBw%40mail.gmail.com

Since Postgres 10, SHOW commands can be triggered with replication
connections in a WAL sender context, however it missed that a
transaction context is needed for syscache lookups.  This commit makes
sure that the syscache lookups can happen correctly by setting a
transaction context when running SHOW commands in a WAL sender.

Superuser-only parameters can be displayed using SHOW commands not only
to superusers, but also to members of system role pg_read_all_settings,
which requires a syscache lookup to check if the connected role is a
member of this system role or not, or the instance crashes.  Superusers
do not need to check the syscache so it worked correctly in this case.

New tests are added to cover this issue.

Reported-by: Alexander Kukushkin
Author: Michael Paquier
Reviewed-by: Álvaro Herrera
Discussion: https://postgr.es/m/15734-2daa8761eeed8e20@postgresql.org
Backpatch-through: 10

Commit c0985099 changed
PostgresNode::get_new_node() to probe 0.0.0.0 instead of 127.0.0.1, but
the new test was less effective for Windows native Perl.  This increased
the failure rate of buildfarm members bowerbird and jacana.  Instead,
test 0.0.0.0 and concrete addresses.  This restores the old level of
defense, but the algorithm is still subject to its longstanding time of
check to time of use race condition.  Back-patch to 9.6, like the
previous change.

Discussion: https://postgr.es/m/GrdLgAdUK9FdyZg8VIcTDKVOkys122ZINEb3CjjoySfGj2KyPiMKTh1zqtRp0TAD7FJ27G-OBB3eplxIB5GhcQH5o8zzGZfp0MuJaXJxVxk=@yesql.se