Testing done in 2011 by Tom Lane concluded that this is a win on Intel Xeons
and AMD Opterons, but it was not changed back then, because of an old
comment in tas() that suggested that it's a huge loss on older Opterons.
However, didn't have separate TAS() and TAS_SPIN() macros back then, so the
comment referred to doing a non-locked initial test even on the first
access, in uncontended case. I don't have access to older Opterons, but I'm
pretty sure that doing an initial unlocked test is unlikely to be a loss
while spinning, even though it might be for the first access.

We probably should do the same on 32-bit x86, but I'm afraid of changing it
without any testing. Hence just add a note to the x86 implementation
suggesting that we probably should do the same there.

Using the infrastructure provided by this patch, it's possible either
to wait for the startup of a dynamically-registered background worker,
or to poll the status of such a worker without waiting.  In either
case, the current PID of the worker process can also be obtained.
As usual, worker_spi is updated to demonstrate the new functionality.

Patch by me.  Review by Andres Freund.

As noted by Tom Lane, commit 813fb031
was overly optimistic about how safe it is to concurrently change
enumsortorder values under MVCC catalog scan semantics.  Restore
some of the previous text, with hopefully-correct adjustments for
the new state of play.

We already did this for -t (--table) in 9.3, but missed the other similar
options. For consistency, allow all of them to be specified multiple times.

Unfortunately it's too late to sneak this into 9.3, so commit to master
only.

Andres Freund; bug detected by valgrind

Christophe Pettus

The previous coding in plancache.c essentially used 10% of the estimated
runtime as its cost estimate for planning.  This can be pretty bogus,
especially when the estimated runtime is very small, such as in a simple
expression plan created by plpgsql, or a simple INSERT ... VALUES.

While we don't have a really good handle on how planning time compares
to runtime, it seems reasonable to use an estimate based on the number of
relations referenced in the query, with a rather large multiplier.  This
patch uses 1000 * cpu_operator_cost * (nrelations + 1), so that even a
trivial query will be charged 1000 * cpu_operator_cost for planning.
This should address the problem reported by Marc Cousin and others that
9.2 and up prefer custom plans in cases where the planning time greatly
exceeds what can be saved.

The backup will not work (without a logarchive, and that's the whole
point of -x) in this case, this patch just changes it to throw an
error instead of crashing when this happens.

Noticed and diagnosed by TAKATSUKA Haruka

It's possible that inlining of SQL functions (or perhaps other changes?)
has exposed typmod information not known at parse time.  In such cases,
Vars generated by query_planner might have valid typmod values while the
original grouping columns only have typmod -1.  This isn't a semantic
problem since the behavior of grouping only depends on type not typmod,
but it breaks locate_grouping_columns' use of tlist_member to locate the
matching entry in query_planner's result tlist.

We can fix this without an excessive amount of new code or complexity by
relying on the fact that locate_grouping_columns only gets called when
make_subplanTargetList has set need_tlist_eval == false, and that can only
happen if all the grouping columns are simple Vars.  Therefore we only need
to search the sub_tlist for a matching Var, and we can reasonably define a
"match" as being a match of the Var identity fields
varno/varattno/varlevelsup.  The code still Asserts that vartype matches,
but ignores vartypmod.

Per bug #8393 from Evan Martin.  The added regression test case is
basically the same as his example.  This has been broken for a very long
time, so back-patch to all supported branches.

We should account for the per-group hashtable entry overhead when
considering whether to use a hash aggregate to implement DISTINCT.  The
comparable logic in choose_hashed_grouping() gets this right, but I think
I omitted it here in the mistaken belief that there would be no overhead
if there were no aggregate functions to be evaluated.  This can result in
more than 2X underestimate of the hash table size, if the tuples being
aggregated aren't very wide.  Per report from Tomas Vondra.

This bug is of long standing, but per discussion we'll only back-patch into
9.3.  Changing the estimation behavior in stable branches seems to carry too
much risk of destabilizing plan choices for already-tuned applications.

Per suggestion from Vik Fearing

This change will only apply to mingw compilers, and has been found
necessary by late versions of the mingw-w64 compiler. It's the same as
what is done elsewhere for the Microsoft compilers.

If this doesn't upset older compilers in the buildfarm, it will be
backpatched to 9.1.

Problem reported by Michael Cronenworth, although not his patch.

Backpatch to 9.3.

Pavel Stehule

sscanf isn't the easiest thing to use for exact pattern checks ...
also, don't use strncmp where strcmp will do.

Instead of deleting all files in stats_temp_directory and the permanent
directory on a crash, only remove those files that match the pattern of
files we actually write in them, to avoid possibly clobbering existing
unrelated contents of the temporary directory.  Per complaint from Jeff
Janes, and subsequent discussion, starting at message
CAMkU=1z9+7RsDODnT4=cDFBRBp8wYQbd_qsLcMtKEf-oFwuOdQ@mail.gmail.com

Also, fix a bug in the same routine to avoid removing files from the
permanent directory twice (instead of once from that directory and then
from the temporary directory), also per report from Jeff Janes, in
message
CAMkU=1wbk947=-pAosDMX5VC+sQw9W4ttq6RM9rXu=MjNeEQKA@mail.gmail.com

This keeps the usual trigger file name unchanged from 9.2, avoiding nasty
issues if you use a pre-9.3 pg_ctl binary with a 9.3 server or vice versa.
The fallback behavior of creating a full checkpoint before starting up is now
triggered by a file called "fallback_promote". That can be useful for
debugging purposes, but we don't expect any users to have to resort to that
and we might want to remove that in the future, which is why the fallback
mechanism is undocumented.

In an example such as
SELECT * FROM
  i LEFT JOIN LATERAL (SELECT * FROM j WHERE i.n = j.n) j ON true;
it is safe to pull up the LATERAL subquery into its parent, but we must
then treat the "i.n = j.n" clause as a qual clause of the LEFT JOIN.  The
previous coding in deconstruct_recurse mistakenly labeled the clause as
"is_pushed_down", resulting in wrong semantics if the clause were applied
at the join node, as per an example submitted awhile ago by Jeremy Evans.
To fix, postpone processing of such clauses until we return back up to
the appropriate recursion depth in deconstruct_recurse.

In addition, tighten the is-safe-to-pull-up checks in is_simple_subquery;
we previously missed the possibility that the LATERAL subquery might itself
contain an outer join that makes lateral references in lower quals unsafe.

A regression test case equivalent to Jeremy's example was already in my
commit of yesterday, but was giving the wrong results because of this
bug.  This patch fixes the expected output for that, and also adds a
test case for the second problem.

When upgrading from servers of versions 9.2 and older, and MultiXactIds
have been used in the old server beyond the first page (that is, 2048
multis or more in the default 8kB-page build), pg_upgrade would set the
next multixact offset to use beyond what has been allocated in the new
cluster.  This would cause a failure the first time the new cluster
needs to use this value, because the pg_multixact/offsets/ file wouldn't
exist or wouldn't be large enough.  To fix, ensure that the transient
server instances launched by pg_upgrade extend the file as necessary.

Per report from Jesse Denardo in
CANiVXAj4c88YqipsyFQPboqMudnjcNTdB3pqe8ReXqAFQ=HXyA@mail.gmail.com

Etsuro Fujita

This was added as part of the attempt to support unlogged matviews
along with a populated status.  It got missed when unlogged
support was removed pre-commit.

Noticed by Noah Misch.  Back-patched to 9.3 branch.

The planner largely failed to consider the possibility that a
PlaceHolderVar's expression might contain a lateral reference to a Var
coming from somewhere outside the PHV's syntactic scope.  We had a previous
report of a problem in this area, which I tried to fix in a quick-hack way
in commit 4da6439b, but Antonin Houska
pointed out that there were still some problems, and investigation turned
up other issues.  This patch largely reverts that commit in favor of a more
thoroughly thought-through solution.  The new theory is that a PHV's
ph_eval_at level cannot be higher than its original syntactic level.  If it
contains lateral references, those don't change the ph_eval_at level, but
rather they create a lateral-reference requirement for the ph_eval_at join
relation.  The code in joinpath.c needs to handle that.

Another issue is that createplan.c wasn't handling nested PlaceHolderVars
properly.

In passing, push knowledge of lateral-reference checks for join clauses
into join_clause_is_movable_to.  This is mainly so that FDWs don't need
to deal with it.

This patch doesn't fix the original join-qual-placement problem reported by
Jeremy Evans (and indeed, one of the new regression test cases shows the
wrong answer because of that).  But the PlaceHolderVar problems need to be
fixed before that issue can be addressed, so committing this separately
seems reasonable.

Backpatch to 9.3.

Backpatch to 9.3, of course.

Per discussion on pgsql-hackers.

Michael Paquier, slightly modified by me.  Original suggestion
from Amit Kapila.

Register atexit() server shutdown if pg_ctl successfully started the
server, but we can't connect to it.

Backpatch to 9.3.

Pavel Raiskup

Was accidentally allowed, but not documented and lacked support
for rename or drop once created.

Per report from Noah Misch.

Timeline IDs are unsigned ints everywhere, except the replication parser
treated them as signed ints.

Avoid using the term "updatable" in confusing ways.  Suggest a trigger
first, before a rule.

The planner logic that attempted to make a preliminary estimate of the
ph_needed levels for PlaceHolderVars seems to be completely broken by
lateral references.  Fortunately, the potential join order optimization
that this code supported seems to be of relatively little value in
practice; so let's just get rid of it rather than trying to fix it.

Getting rid of this allows fairly substantial simplifications in
placeholder.c, too, so planning in such cases should be a bit faster.

Issue noted while pursuing bugs reported by Jeremy Evans and Antonin
Houska, though this doesn't in itself fix either of their reported cases.
What this does do is prevent an Assert crash in the kind of query
illustrated by the added regression test.  (I'm not sure that the plan for
that query is stable enough across platforms to be usable as a regression
test output ... but we'll soon find out from the buildfarm.)

Back-patch to 9.3.  The problem case can't arise without LATERAL, so
no need to touch older branches.

This syntax is supported by the parser, but is non-standard.

_Not_ backpatched to 9.3 in case we change our minds.

We don't want to prevent an extension which creates a matview from
being installed in pg_catalog.

Issue was raised by Hitoshi Harada.
Backpatched to 9.3.

Update emacs.samples with new configuration snippets that match pgindent
et al. formatting more accurately and follow Emacs Lisp best practices
better.

Add .dir-locals.el with a subset of that configuration for casual
editing and viewing.
Reviewed-by: Dimitri Fontaine <dimitri@2ndQuadrant.fr>
Reviewed-by: Noah Misch <noah@leadboat.com>

We've seen multiple cases of people looking at the postmaster's original
stderr output to try to diagnose problems, not realizing/remembering that
their logging configuration is set up to send log messages somewhere else.
This seems particularly likely to happen in prepackaged distributions,
since many packagers patch the code to change the factory-standard logging
configuration to something more in line with their platform conventions.

In hopes of reducing confusion, emit a LOG message about this at the point
in startup where we are about to switch log output away from the original
stderr, providing a pointer to where to look instead.  This message will
appear as the last thing in the original stderr output.  (We might later
also try to emit such link messages when logging parameters are changed
on-the-fly; but that case seems to be both noticeably harder to do nicely,
and much less frequently a problem in practice.)

Per discussion, back-patch to 9.3 but not further.