StrategyGetBuffer() has proven to be a bottleneck in a number of
buffer acquisition heavy workloads. To some degree this has already
been alleviated by 5d7962c6, but it still can be quite a heavy
bottleneck.  The problem is that in unfortunate usage patterns a
single StrategyGetBuffer() call will have to look at a large number of
buffers - in turn making it likely that the process will be put to
sleep while still holding the spinlock.

Replace most of the usage of the buffer_strategy_lock spinlock for the
clock sweep by a atomic nextVictimBuffer variable. That variable,
modulo NBuffers, is the current hand of the clock sweep. The buffer
clock-sweep then only needs to acquire the spinlock after a
wraparound. And even then only in the process that did the wrapping
around. That alleviates nearly all the contention on the relevant
spinlock, although significant contention on the cacheline can still
exist.

Reviewed-By: Robert Haas and Amit Kapila

Discussion: 20141010160020.GG6670@alap3.anarazel.de,
    20141027133218.GA2639@awork2.anarazel.de

The old LWLock implementation had the problem that concurrent lock
acquisitions required exclusively acquiring a spinlock. Often that
could lead to acquirers waiting behind the spinlock, even if the
actual LWLock was free.

The new implementation doesn't acquire the spinlock when acquiring the
lock itself. Instead the new atomic operations are used to atomically
manipulate the state. Only the waitqueue, used solely in the slow
path, is still protected by the spinlock. Check lwlock.c's header for
an explanation about the used algorithm.

For some common workloads on larger machines this can yield
significant performance improvements. Particularly in read mostly
workloads.

Reviewed-By: Amit Kapila and Robert Haas
Author: Andres Freund

Discussion: 20130926225545.GB26663@awork2.anarazel.de

Besides being shorter and much easier to read it changes the logic in
LWLockRelease() to release all shared lockers when waking up any. This
can yield some significant performance improvements - and the fairness
isn't really much worse than before, as we always allowed new shared
lockers to jump the queue.

Hiding context messages usually is not a good idea - except for rather
verbose debugging/development utensils like LOG_DEBUG. There the
amount of repeated context messages just bloat the log without adding
information.

Back-patch to 9.4, where this problem was added.

Exposing compression and decompression APIs of pglz makes possible its
use by extensions and contrib modules. pglz_decompress contained a call
to elog to emit an error message in case of corrupted data. This function
is changed to return a status code to let its callers return an error instead.

This commit is required for upcoming WAL compression feature so that
the WAL reader facility can decompress the WAL data by using pglz_decompress.

Michael Paquier

For some reason this seems to have been missed when the lists in
src/timezone/tznames/ were first constructed.  We can't put it in Default
because of the conflict with US CST, but we should certainly list it among
the alternative entries in Asia.txt.  (I checked for other oversights, but
all the other abbreviations that are in current use according to the IANA
files seem to be accounted for.)  Noted while responding to bug #12326.

fe263d11 changed the REINDEX logic so that those fields are not used at all,
but forgot to remove them.

Sawada Masahiko

MSVC doesn't realize ereport(ERROR) doesn't return.

David Rowley

Per buildfarm, this test case does not yield consistent results.
I don't think it's useful enough to figure out a workaround, either.

This reverts commit 1826987a.

The overall design was deemed unacceptable, in discussion following the
previous commit message; we might find some parts of it still
salvageable, but I don't want to be on the hook for fixing it, so let's
wait until we have a new patch.

This allows access to get_object_address from SQL, which is useful to
obtain OID addressing information from data equivalent to that emitted
by the parser.  This is necessary infrastructure of a project to let
replication systems propagate object dropping events to remote servers,
where the schema might be different than the server originating the
DROP.

This patch also adds support for OBJECT_DEFAULT to get_object_address;
that is, it is now possible to refer to a column's default value.

Catalog version bumped due to the new function.

Reviewed by Stephen Frost, Heikki Linnakangas, Robert Haas, Andres
Freund, Abhijit Menon-Sen, Adam Brightwell.

The previous representation using a boolean column for each attribute
would not scale as well as we want to add further attributes.

Extra auxilliary functions are added to go along with this change, to
make up for the lost convenience of access of the old representation.

Catalog version bumped due to change in catalogs and the new functions.

Author: Adam Brightwell, minor tweaks by Álvaro
Reviewed by: Stephen Frost, Andres Freund, Álvaro Herrera

Apart from enabling comments on domain constraints, this enables a
future project to replicate object dropping to remote servers: with the
current mechanism there's no way to distinguish between the two types of
constraints, so there's no way to know what to drop.

Also added support for the domain constraint comments in psql's \dd and
pg_dump.

Catalog version bumped due to the change in ObjectType enum.

This allows it to be used with ALTER ROLE SET.

Although the old setting of PGC_BACKEND prevented changes after session
start, after discussion it was more useful to allow ALTER ROLE SET
instead and just document that changes during a session have no effect.
This is similar to how session_preload_libraries works already.

An alternative would be to change things to allow PGC_BACKEND and
PGC_SU_BACKEND settings to be changed by ALTER ROLE SET.  But that might
need further research (e.g., log_connections would probably not work).

based on patch by Kyotaro Horiguchi

json_agg was originally designed to aggregate records. However, it soon
became clear that it is useful for aggregating all kinds of values and
that's what we have on 9.3 and 9.4, and in head for it and jsonb_agg.
The documentation suggested otherwise, so this fixes it.

We have other general-purpose data structures in src/backend/lib, so it
seems like a better home for the red-black tree as well.

This makes the functions much nicer to read and edit, and also makes
debugging easier.

This performs slightly better, uses less memory, and needs slightly less
code in GiST, than the Red-Black tree previously used.

Reviewed by Peter Geoghegan

Explain that you have to use "VARIADIC ARRAY[]" to pass an empty array
to a variadic parameter position.  This was already implicit in the text
but it seems better to spell it out.

Per a suggestion from David Johnston, though I didn't use his proposed
wording.  Back-patch to all supported branches.

XLogFileInit() returns a file descriptor, which needs to be closed. The leak
was short-lived, since the startup process exits shortly afterwards, but it
was clearly a bug, nevertheless.

Per Coverity report.

Report by Noah Misch

Add "normal" and "original" flags as output columns to the
pg_event_trigger_dropped_objects() function.  With this it's possible to
distinguish which objects, among those listed, need to be explicitely
referenced when trying to replicate a deletion.

This is necessary so that the list of objects can be pruned to the
minimum necessary to replicate the DROP command in a remote server that
might have slightly different schema (for instance, TOAST tables and
constraints with different names and such.)

Catalog version bumped due to change of function definition.

Reviewed by: Abhijit Menon-Sen, Stephen Frost, Heikki Linnakangas,
Robert Haas.

We used time(null) to set a TimestampTz field, which gave bogus results.
Noticed while looking at pg_xlogdump output.

Backpatch to 9.3 and above, where the fast promotion was introduced.

In LWLockRelease() (and in 9.4+ LWLockUpdateVar()) we release enqueued
waiters using PGSemaphoreUnlock(). As there are other sources of such
unlocks backends only wake up if MyProc->lwWaiting is set to false;
which is only done in the aforementioned functions.

Before this commit there were dangers because the store to lwWaitLink
could become visible before the store to lwWaitLink. This could both
happen due to compiler reordering (on most compilers) and on some
platforms due to the CPU reordering stores.

The possible consequence of this is that a backend stops waiting
before lwWaitLink is set to NULL. If that backend then tries to
acquire another lock and has to wait there the list could become
corrupted once the lwWaitLink store is finally performed.

Add a write memory barrier to prevent that issue.

Unfortunately the barrier support has been only added in 9.2. Given
that the issue has not knowingly been observed in praxis it seems
sufficient to prohibit compiler reordering using volatile for 9.0 and
9.1. Actual problems due to compiler reordering are more likely
anyway.

Discussion: 20140210134625.GA15246@awork2.anarazel.de

gcc's -Wempty-body warns about the current usage when compiling
postgres without --enable-cassert.

The possibility that constant subexpressions of a CASE might be evaluated
at planning time was touched on in 9.17.1 (CASE expressions), but it really
ought to be explained in 4.2.14 (Expression Evaluation Rules) which is the
primary discussion of such topics.  Add text and an example there, and
revise the <note> under CASE to link there.

Back-patch to all supported branches, since it's acted like this for a
long time (though 9.2+ is probably worse because of its more aggressive
use of constant-folding via replanning of nominally-prepared statements).
Pre-9.4, also back-patch text added in commit 0ce627d4 about CASE versus
aggregate functions.

Tom Lane and David Johnston, per discussion of bug #12273.

Per Tom Lane

Author: Jim Nasby, some kibitzing by Heikki Linnankangas.
Discussion leading to current behavior and precise wording fueled by
thoughts from Robert Haas and Andres Freund.

Previously, if you wanted anything besides C-string hash keys, you had to
specify a custom hashing function to hash_create().  Nearly all such
callers were specifying tag_hash or oid_hash; which is tedious, and rather
error-prone, since a caller could easily miss the opportunity to optimize
by using hash_uint32 when appropriate.  Replace this with a design whereby
callers using simple binary-data keys just specify HASH_BLOBS and don't
need to mess with specific support functions.  hash_create() itself will
take care of optimizing when the key size is four bytes.

This nets out saving a few hundred bytes of code space, and offers
a measurable performance improvement in tidbitmap.c (which was not
exploiting the opportunity to use hash_uint32 for its 4-byte keys).
There might be some wins elsewhere too, I didn't analyze closely.

In future we could look into offering a similar optimized hashing function
for 8-byte keys.  Under this design that could be done in a centralized
and machine-independent fashion, whereas getting it right for keys of
platform-dependent sizes would've been notationally painful before.

For the moment, the old way still works fine, so as not to break source
code compatibility for loadable modules.  Eventually we might want to
remove tag_hash and friends from the exported API altogether, since there's
no real need for them to be explicitly referenced from outside dynahash.c.

Teodor Sigaev and Tom Lane

Two changes:

1. When copying a WAL segment from old timeline to create the first segment
on the new timeline, only copy up to the point where the timeline switch
happens, and zero-fill the rest. This avoids corner cases where we might
think that the copied WAL from the previous timeline belong to the new
timeline.

2. If the timeline switch happens at a segment boundary, don't copy the
whole old segment to the new timeline. It's pointless, because it's 100%
identical to the old segment.

st_changecount protocol needs the memory barriers to ensure that
the apparent order of execution is as it desires. Otherwise,
for example, the CPU might rearrange the code so that st_changecount
is incremented twice before the modification on a machine with
weak memory ordering. This surprising result can lead to bugs.

This commit introduces the macros to load and store st_changecount
with the memory barriers. These are called before and after
PgBackendStatus entries are modified or copied into private memory,
in order to prevent CPU from reordering PgBackendStatus access.

Per discussion on pgsql-hackers, we decided not to back-patch this
to 9.4 or before until we get an actual bug report about this.

Patch by me. Review by Robert Haas.

In generate_series_step_numeric(), the variables "start_num"
and "stop_num" may be potentially freed until the next call.
So they should be put in the location which can survive across calls.
But previously they were not, and which could cause incorrect
behavior of generate_series(numeric, numeric). This commit fixes
this problem by copying them on multi_call_memory_ctx.

Andrew Gierth

Also add a comment about why regreesion.* aren't listed in .gitignore.

Jim Nasby

CRC computation is now done in XLogRecordAssemble.

Back-patch to 9.0 (all supported versions).  This is mere
future-proofing in the context of the master branch, but commit
f6dc6dd5 requires it of older branches.

Etsuro Fujita