Used to say just "could not read password from file "...": Success", which
isn't very informative.

Mats Erik Andersson. Backpatch to all supported versions.

The "file mode" bits in the tar file header is not supposed to include the
file type bits, e.g. S_IFREG or S_IFDIR. The file type is stored in a
separate field. This isn't a problem in practice, all tar programs ignore
the extra bits, but let's be tidy.

This came up in a discussion around bug #11949, reported by Hendrik Grewe,
although this doesn't fix the issue with tar --append. That turned out to be
a bug in GNU tar. Schilly's tartest program revealed this defect in the tar
created by pg_basebackup.

This problem goes as far as we we've had pg_basebackup, but since this
hasn't caused any problems in practice, let's be conservative and fix in
master only.

After commit da34731b, these files are not generated files anymore.

Adam Brightwell

Michael Paquier

It was an oversight in the original commit.

Also note in the sample config file that changing wal_log_hints requires a
restart.

Michael Paquier. Backpatch to 9.4, where wal_log_hints was added.

The old formula was completely bogus with the new WAL record format.

This can happen if an error occurs in a standalone backend.  This bug
was introduced by commit 2bd9e412.

Reported by Álvaro Herrera.

Author: Peter Geoghegan

PGXS computes srcdir from VPATH, PostgreSQL proper computes VPATH from
srcdir, and doing both results in an error from make.  Conditionalize so
only one of these takes effect.

These changes were originally submitted as "adds support for VPATH with
USE_PGXS", but they are not necessary for VPATH support, so they just
add more lines of code for no reason.

The user can just set VPATH directly.  There is no need to invent
another variable.

dblink and postgres_fdw use SHLIB_PREREQS = submake-libpq to build libpq
first.  This doesn't work in a PGXS build, because there is no libpq to
build.  So just omit setting SHLIB_PREREQS in this case.

Note that PGXS users can still use SHLIB_PREREQS (although it is not
documented).  The problem here is only that contrib modules can be built
in-tree or using PGXS, and the prerequisite is only applicable in the
former case.

Commit 6697aa2b previously attempted to
address this by creating a somewhat fake submake-libpq target in
Makefile.global.  That was not the right fix, and it was also done in a
nonportable way, so revert that.

Since this is not something that a user should change,
pg_config_manual.h was an inappropriate place for it.

In initdb.c, remove the use of the macro, because utils/guc.h can't be
included by non-backend code.  But we hardcode all the other
configuration file names there, so this isn't a disaster.

In the spirit of d34b48a0

Per buildfarm member guaibasaurus, via Stefan Kaltenbrunner.

Transactions can now set their commit timestamp directly as they commit,
or an external transaction commit timestamp can be fed from an outside
system using the new function TransactionTreeSetCommitTsData().  This
data is crash-safe, and truncated at Xid freeze point, same as pg_clog.

This module is disabled by default because it causes a performance hit,
but can be enabled in postgresql.conf requiring only a server restart.

A new test in src/test/modules is included.

Catalog version bumped due to the new subdirectory within PGDATA and a
couple of new SQL functions.

Authors: Álvaro Herrera and Petr Jelínek

Reviewed to varying degrees by Michael Paquier, Andres Freund, Robert
Haas, Amit Kapila, Fujii Masao, Jaime Casanova, Simon Riggs, Steven
Singer, Peter Eisentraut

check-world failed in a completely clean tree, because src/test/modules
fail to build unless errcodes.h is generated first.  To fix this,
install a dependency in src/test/modules' Makefile so that the necessary
file is generated.  Even with this, running "make check" within
individual module subdirs will still fail because the dependency is not
considered there, but this case is less interesting and would be messier
to fix.

check-world still failed with the above fix in place, this time because
dummy_seclabel used LOAD to load the dynamic library, which doesn't work
because the @libdir@ (expanded by the makefile) is expanded to the final
install path, not the temporary installation directory used by make
check.  To fix, tweak things so that CREATE EXTENSION can be used
instead, which solves the problem because the library path is expanded
by the backend, which is aware of the true libdir.

Make the error messages issued by array_in() uniformly follow the style
	ERROR: malformed array literal: "actual input string"
	DETAIL: specific complaint here
and rewrite many of the specific complaints to be clearer.

The immediate motivation for doing this is a complaint from Josh Berkus
that json_to_record() produced an unintelligible error message when
dealing with an array item, because it tries to feed the JSON-format
array value to array_in().  Really it ought to be smart enough to
perform JSON-to-Postgres array conversion, but that's a future feature
not a bug fix.  In the meantime, this change is something we agreed
we could back-patch into 9.4, and it should help de-confuse things a bit.

The logical decoding patchset introduced PROC_IN_LOGICAL_DECODING flag
PGXACT flag, that allows such backends to be skipped when computing
the xmin horizon/snapshots. That's fine and sensible for walsenders
streaming out logical changes, but not at all fine for SQL backends
doing logical decoding. If the latter set that flag any change they
have performed outside of logical decoding will not be regarded as
visible - which e.g. can lead to that change being vacuumed away.

Note that not setting the flag for SQL backends isn't particularly
bothersome - the SQL backend doesn't do streaming, so it only runs for
a limited amount of time.

Per buildfarm member 'tick' and Alvaro.

Backpatch to 9.4, where logical decoding was introduced.

Davide S. reported that json_agg() sometimes produced multiple trailing
right brackets.  This turns out to be because json_agg_finalfn() attaches
the final right bracket, and was doing so by modifying the aggregate state
in-place.  That's verboten, though unfortunately it seems there's no way
for nodeAgg.c to check for such mistakes.

Fix that back to 9.3 where the broken code was introduced.  In 9.4 and
HEAD, likewise fix json_object_agg(), which had copied the erroneous logic.
Make some cosmetic cleanups as well.

Get rid of PG_FUNCTION_INFO_V1() macros, which are quite inappropriate
for built-in functions (possibly leftovers from testing as a loadable
module?).  Also, fix gratuitous inconsistency between SQL-level and
C-level names of the minmax support functions.

Michael Paquier

Double-dash additions suggested by Peter Geoghegan

Michael Paquier

We were not checking to see if the supplied dscale was valid for the given
digit array when receiving binary-format numeric values.  While dscale can
validly be more than the number of nonzero fractional digits, it shouldn't
be less; that case causes fractional digits to be hidden on display even
though they're there and participate in arithmetic.

Bug #12053 from Tommaso Sala indicates that there's at least one broken
client library out there that sometimes supplies an incorrect dscale value,
leading to strange behavior.  This suggests that simply throwing an error
might not be the best response; it would lead to failures in applications
that might seem to be working fine today.  What seems the least risky fix
is to truncate away any digits that would be hidden by dscale.  This
preserves the existing behavior in terms of what will be printed for the
transmitted value, while preventing subsequent arithmetic from producing
results inconsistent with that.

In passing, throw a specific error for the case of dscale being outside
the range that will fit into a numeric's header.  Before you got "value
overflows numeric format", which is a bit misleading.

Back-patch to all supported branches.

Rather than have the core security_label regression test depend on the
dummy_seclabel module, have that part of the test be executed by
dummy_seclabel itself directly.  This simplifies the testing rig a bit;
in particular it should silence the problems from the MSVC buildfarm
phylum, which haven't yet gotten taught how to install src/test/modules.

We expose a function IsValidJsonNumber that internally calls the lexer
for json numbers. That allows us to use the same test everywhere,
instead of inventing a broken test for hstore conversions. The new
function is also used in datum_to_json, replacing the code that is now
moved to the new function.

Backpatch to 9.3 where hstore_to_json_loose was introduced.

It seems likely that any SSL implementation will need a similar call, not
just OpenSSL.

Ian Barwick

Extracted from pending inet selectivity patch.  The rest of it isn't
quite ready to commit, but we might as well push this part so the patch
doesn't have to track the moving target of pg_operator.h.

Coverity complained that the "else" added to fillPGconn() was unreachable,
which it was.  Remove the dead code.  In passing, rearrange the tests so as
not to bother trying to fetch values for options that can't be assigned.

Pre-9.3 did not have that issue, but it did have a "return" that should be
"goto oom_error" to ensure that a suitable error message gets filled in.

This is advance preparation for introducing even more test modules; the
easy solution is to add them to contrib, but that's bloated enough that
it seems a good time to think of something different.

Moved modules are dummy_seclabel, test_shm_mq, test_parser and
worker_spi.

(test_decoding was also a candidate, but there was too much opposition
to moving that one.  We can always reconsider later.)

Back-patch to 9.4, like the feature's removal.

Apart from ignoring "hostaddr" set to the empty string, this behaves
identically to its predecessor.  Back-patch to 9.4, where the original
commit first appeared.

Reviewed by Fujii Masao.

This reverts commit 9f80f483.  The
function returned the raw value of a connection parameter, a task served
by PQconninfo().  The next commit will reimplement the psql \conninfo
change that way.  Back-patch to 9.4, where that commit first appeared.

The original definitions were leaving no room for cross-type operators,
so queries that compared a column of one type against something of a
different type were not taking advantage of the index.  Fix by making
the opfamilies more like the ones for Btree, and include a few
cross-type operator classes.

Catalog version bumped.

Per complaints from Hubert Lubaczewski, Mark Wong, Heikki Linnakangas.

Multixacts are now maintained during recovery, but the README didn't get
the memo.  Backpatch to 9.3, where the divergence was introduced.

This patch adds a function that replaces a bms_membership() test followed
by a bms_singleton_member() call, performing both the test and the
extraction of a singleton set's member in one scan of the bitmapset.
The performance advantage over the old way is probably minimal in current
usage, but it seems worthwhile on notational grounds anyway.

David Rowley

This patch adds a way of iterating through the members of a bitmapset
nondestructively, unlike the old way with bms_first_member().  While
bms_next_member() is very slightly slower than bms_first_member()
(at least for typical-size bitmapsets), eliminating the need to palloc
and pfree a temporary copy of the target bitmapset is a significant win.
So this method should be preferred in all cases where a temporary copy
would be necessary.

Tom Lane, with suggestions from Dean Rasheed and David Rowley

This function was initially coded on the assumption that it would not be
performance-critical, but that turns out to be wrong in workloads that
are heavily dependent on the speed of plpgsql functions.  Speed it up by
hard-coding the comparison rules, thereby avoiding palloc/pfree traffic
from creating and immediately freeing an OverrideSearchPath object.
Per report from Scott Marlowe.