Prototypes for functions implementing V1-callable functions are no
longer necessary.

Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Thomas Munro <thomas.munro@enterprisedb.com>

Using offsetof() with a run-time computed argument is not allowed in
either C or C++.  Apparently, gcc allows it, but g++ doesn't.

Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Reviewed-by: Thomas Munro <thomas.munro@enterprisedb.com>

Now that we track initial privileges on extension objects and changes to
those permissions, we can drop the superuser() checks from the various
functions which are part of the pgstattuple extension and rely on the
GRANT system to control access to those functions.

Since a pg_upgrade will preserve the version of the extension which
existed prior to the upgrade, we can't simply modify the existing
functions but instead need to create new functions which remove the
checks and update the SQL-level functions to use the new functions
(and to REVOKE EXECUTE rights on those functions from PUBLIC).

Thanks to Tom and Andres for adding support for extensions to follow
update paths (see: 40b449ae), allowing this patch to be much smaller
since no new base version script needed to be included.

Approach suggested by Noah.

Reviewed by Michael Paquier.

This patch just exposes COPY's FROM PROGRAM option in contrib/file_fdw.
There don't seem to be any security issues with that that are any worse
than what already exist with file_fdw and COPY; as in the existing cases,
only superusers are allowed to control what gets executed.

A regression test case might be nice here, but choosing a 100% portable
command to run is hard.  (We haven't got a test for COPY FROM PROGRAM
itself, either.)

Corey Huinker and Adam Gomaa, reviewed by Amit Langote

Discussion: <CADkLM=dGDGmaEiZ=UDepzumWg-CVn7r8MHPjr2NArj8S3TsROQ@mail.gmail.com>

That makes the view a lot less disruptive to use on a production system.
Without the locks, you don't get a consistent snapshot across all buffers,
but that's OK. It wasn't a very useful guarantee in practice.

Ivan Kartyshov, reviewed by Tomas Vondra and Robert Haas.

Discusssion: <f9d6cab2-73a7-7a84-55a8-07dcb8516ae5@postgrespro.ru>

It's always been possible to create a zero-dimensional cube by converting
from a zero-length float8 array, but cube_in failed to accept the '()'
representation that cube_out produced for that case, resulting in a
dump/reload hazard.  Make it accept the case.  Also fix a couple of
other places that didn't behave sanely for zero-dimensional cubes:
cube_size would produce 1.0 when surely the answer should be 0.0,
and g_cube_distance risked a divide-by-zero failure.

Likewise, it's always been possible to create cubes containing float8
infinity or NaN coordinate values, but cube_in couldn't parse such input,
and cube_out produced platform-dependent spellings of the values.  Convert
them to use float8in_internal and float8out_internal so that the behavior
will be the same as for float8, as we recently did for the core geometric
types (cf commit 50861cd6).  As in that commit, I don't pretend that this
patch fixes all insane corner-case behaviors that may exist for NaNs, but
it's a step forward.

(This change allows removal of the separate cube_1.out and cube_3.out
expected-files, as the platform dependency that previously required them
is now gone: an underflowing coordinate value will now produce an error
not plus or minus zero.)

Make errors from cube_in follow project conventions as to spelling
("invalid input syntax for cube" not "bad cube representation")
and errcode (INVALID_TEXT_REPRESENTATION not SYNTAX_ERROR).

Also a few marginal code cleanups and comment improvements.

Tom Lane, reviewed by Amul Sul

Discussion: <15085.1472494782@sss.pgh.pa.us>

LibreSSL defines OPENSSL_VERSION_NUMBER to claim that it is version 2.0.0,
but it doesn't have the functions added in OpenSSL 1.1.0. Add autoconf
checks for the individual functions we need, and stop relying on
OPENSSL_VERSION_NUMBER.

Backport to 9.5 and 9.6, like the patch that broke this. In the
back-branches, there are still a few OPENSSL_VERSION_NUMBER checks left,
to check for OpenSSL 0.9.8 or 0.9.7. I left them as they were - LibreSSL
has all those functions, so they work as intended.

Per buildfarm member curculio.

Discussion: <2442.1473957669@sss.pgh.pa.us>

Otherwise, users who have configured shared_buffers >= 256GB won't
be able to use this module.  There probably aren't many of those, but
it doesn't hurt anything to fix it so that it works.

Backpatch to 9.4, where MemoryContextAllocHuge was introduced.  The
same problem exists in older branches, but there's no easy way to
fix it there.

KaiGai Kohei

Changes needed to build at all:

- Check for SSL_new in configure, now that SSL_library_init is a macro.
- Do not access struct members directly. This includes some new code in
  pgcrypto, to use the resource owner mechanism to ensure that we don't
  leak OpenSSL handles, now that we can't embed them in other structs
  anymore.
- RAND_SSLeay() -> RAND_OpenSSL()

Changes that were needed to silence deprecation warnings, but were not
strictly necessary:

- RAND_pseudo_bytes() -> RAND_bytes().
- SSL_library_init() and OpenSSL_config() -> OPENSSL_init_ssl()
- ASN1_STRING_data() -> ASN1_STRING_get0_data()
- DH_generate_parameters() -> DH_generate_parameters()
- Locking callbacks are not needed with OpenSSL 1.1.0 anymore. (Good
  riddance!)

Also change references to SSLEAY_VERSION_NUMBER with OPENSSL_VERSION_NUMBER,
for the sake of consistency. OPENSSL_VERSION_NUMBER has existed since time
immemorial.

Fix SSL test suite to work with OpenSSL 1.1.0. CA certificates must have
the "CA:true" basic constraint extension now, or OpenSSL will refuse them.
Regenerate the test certificates with that. The "openssl" binary, used to
generate the certificates, is also now more picky, and throws an error
if an X509 extension is specified in "req_extensions", but that section
is empty.

Backpatch to all supported branches, per popular demand. In back-branches,
we still support OpenSSL 0.9.7 and above. OpenSSL 0.9.6 should still work
too, but I didn't test it. In master, we only support 0.9.8 and above.

Patch by Andreas Karlsson, with additional changes by me.

Discussion: <20160627151604.GD1051@msg.df7cb.de>

Add a location field to the DefElem struct, used to parse many utility
commands.  Update various error messages to supply error position
information.

To propogate the error position information in a more systematic way,
create a ParseState in standard_ProcessUtility() and pass that to
interested functions implementing the utility commands.  This seems
better than passing the query string and then reassembling a parse state
ad hoc, which violates the encapsulation of the ParseState type.

Reviewed-by: Pavel Stehule <pavel.stehule@gmail.com>

When building libpq, ip.c and md5.c were symlinked or copied from
src/backend/libpq into src/interfaces/libpq, but now that we have a
directory specifically for routines that are shared between the server and
client binaries, src/common/, move them there.

Some routines in ip.c were only used in the backend. Keep those in
src/backend/libpq, but rename to ifaddr.c to avoid confusion with the file
that's now in common.

Fix the comment in src/common/Makefile to reflect how libpq actually links
those files.

There are two more files that libpq symlinks directly from src/backend:
encnames.c and wchar.c. I don't feel compelled to move those right now,
though.

Patch by Michael Paquier, with some changes by me.

Discussion: <69938195-9c76-8523-0af8-eb718ea5b36e@iki.fi>

Where possible, use palloc or pg_malloc instead; otherwise, insert
explicit NULL checks.

Generally speaking, these are places where an actual OOM is quite
unlikely, either because they're in client programs that don't
allocate all that much, or they're very early in process startup
so that we'd likely have had a fork() failure instead.  Hence,
no back-patch, even though this is nominally a bug fix.

Michael Paquier, with some adjustments by me

Discussion: <CAB7nPqRu07Ot6iht9i9KRfYLpDaF2ZuUv5y_+72uP23ZAGysRg@mail.gmail.com>

The previous API for this function had it returning a malloc'd string.
That meant that callers had to check for NULL return, which few of them
were doing, and it also meant that callers had to remember to free()
the string later, which required extra logic in most cases.

Instead, make simple_prompt() write into a buffer supplied by the caller.
Anywhere that the maximum required input length is reasonably small,
which is almost all of the callers, we can just use a local or static
array as the buffer instead of dealing with malloc/free.

A fair number of callers used "pointer == NULL" as a proxy for "haven't
requested the password yet".  Maintaining the same behavior requires
adding a separate boolean flag for that, which adds back some of the
complexity we save by removing free()s.  Nonetheless, this nets out
at a small reduction in overall code size, and considerably less code
than we would have had if we'd added the missing NULL-return checks
everywhere they were needed.

In passing, clean up the API comment for simple_prompt() and get rid
of a very-unnecessary malloc/free in its Windows code path.

This is nominally a bug fix, but it does not seem worth back-patching,
because the actual risk of an OOM failure in any of these places seems
pretty tiny, and all of them are client-side not server-side anyway.

This patch is by me, but it owes a great deal to Michael Paquier
who identified the problem and drafted a patch for fixing it the
other way.

Discussion: <CAB7nPqRu07Ot6iht9i9KRfYLpDaF2ZuUv5y_+72uP23ZAGysRg@mail.gmail.com>

OpenSSL officially only supports 1.0.1 and newer. Some OS distributions
still provide patches for 0.9.8, but anything older than that is not
interesting anymore. Let's simplify things by removing compatibility code.

Andreas Karlsson, with small changes by me.

I found that half a dozen (nearly 5%) of our AllocSetContextCreate calls
had typos in the context-sizing parameters.  While none of these led to
especially significant problems, they did create minor inefficiencies,
and it's now clear that expecting people to copy-and-paste those calls
accurately is not a great idea.  Let's reduce the risk of future errors
by introducing single macros that encapsulate the common use-cases.
Three such macros are enough to cover all but two special-purpose contexts;
those two calls can be left as-is, I think.

While this patch doesn't in itself improve matters for third-party
extensions, it doesn't break anything for them either, and they can
gradually adopt the simplified notation over time.

In passing, change TopMemoryContext to use the default allocation
parameters.  Formerly it could only be extended 8K at a time.  That was
probably reasonable when this code was written; but nowadays we create
many more contexts than we did then, so that it's not unusual to have a
couple hundred K in TopMemoryContext, even without considering various
dubious code that sticks other things there.  There seems no good reason
not to let it use growing blocks like most other contexts.

Back-patch to 9.6, mostly because that's still close enough to HEAD that
it's easy to do so, and keeping the branches in sync can be expected to
avoid some future back-patching pain.  The bugs fixed by these changes
don't seem to be significant enough to justify fixing them further back.

Discussion: <21072.1472321324@sss.pgh.pa.us>

You can use ALTER FOREIGN TABLE SET WITH OIDS on a foreign table, but the
oid column read out as zeros, because the postgres_fdw didn't know about
it. Teach postgres_fdw how to fetch it.

Etsuro Fujita, with an additional test case by me.

Discussion: <56E90A76.5000503@lab.ntt.co.jp>

Etsuro Fujita

Emre Hasegeli

Patch: <CAE2gYzzF24ZHWqkMukkHwqa0otbES9Rex22LrjQUNbi=oKziNQ@mail.gmail.com>

As implemented, -e ran an EXPLAIN but then discarded the output, which
certainly seems pointless.  Make it print to stdout instead.  It's been
like that forever, so back-patch to all supported branches.

Daniel Gustafsson, reviewed by Andreas Scherbaum

Patch: <B97BDCB7-A3B3-4734-90B5-EDD586941629@yesql.se>

We implement a dozen or so parameterless functions that the SQL standard
defines special syntax for.  Up to now, that was done by converting them
into more or less ad-hoc constructs such as "'now'::text::date".  That's
messy for multiple reasons: it exposes what should be implementation
details to users, and performance is worse than it needs to be in several
cases.  To improve matters, invent a new expression node type
SQLValueFunction that can represent any of these parameterless functions.

Bump catversion because this changes stored parsetrees for rules.

Discussion: <30058.1463091294@sss.pgh.pa.us>

In blinsert(), cope with the possibility that a page we pull from the
notFullPage list is marked BLOOM_DELETED.  This could happen if VACUUM
recently marked it deleted but hasn't (yet) updated the metapage.
We can re-use such a page safely, but we *must* reinitialize it so that
it's no longer marked deleted.

Fix blvacuum() so that it updates the notFullPage list even if it's
going to update it to empty.  The previous "optimization" of skipping
the update seems pretty dubious, since it means that the next blinsert()
will uselessly visit whatever pages we left in the list.

Uniformly treat PageIsNew pages the same as deleted pages.  This should
allow proper recovery if a crash occurs just after relation extension.

Properly use vacuum_delay_point, not assorted ad-hoc CHECK_FOR_INTERRUPTS
calls, in the blvacuum() main loop.

Fix broken tuple-counting logic: blvacuum.c counted the number of live
index tuples over again in each scan, leading to VACUUM VERBOSE reporting
some multiple of the actual number of surviving index tuples after any
vacuum that removed any tuples (since they'd be counted in blvacuum, maybe
more than once, and then again in blvacuumcleanup, without ever zeroing the
counter).  It's sufficient to count them in blvacuumcleanup.

stats->estimated_count is a boolean, not a counter, and we don't want
to set it true, so don't add tuple counts to it.

Add a couple of Asserts that we don't overrun available space on a bloom
page.  I don't think there's any bug there today, but the way the
FreeBlockNumberArray size calculation is set up is scarily fragile, and
BloomPageGetFreeSpace isn't much better.  The Asserts should help catch
any future mistakes.

Per investigation of a report from Jeff Janes.  I think the first item
above may explain his report; the other changes were things I noticed
while casting about for an explanation.

Report: <CAMkU=1xEUuBphDwDmB1WjN4+td4kpnEniFaTBxnk1xzHCw8_OQ@mail.gmail.com>

Per discussion, we should provide such functions to replace the lost
ability to discover AM properties by inspecting pg_am (cf commit
65c5fcd3).  The added functionality is also meant to displace any code
that was looking directly at pg_index.indoption, since we'd rather not
believe that the bit meanings in that field are part of any client API
contract.

As future-proofing, define the SQL API to not assume that properties that
are currently AM-wide or index-wide will remain so unless they logically
must be; instead, expose them only when inquiring about a specific index
or even specific index column.  Also provide the ability for an index
AM to override the behavior.

In passing, document pg_am.amtype, overlooked in commit 473b9328.

Andrew Gierth, with kibitzing by me and others

Discussion: <87mvl5on7n.fsf@news-spur.riddles.org.uk>

Don't spell "InvalidOid" as "0".  Initialize method fields in the same
order as amapi.h declares them (and every other AM handler initializes
them).

This oversight could cause logical decoding to fail to decode an outer
transaction containing changes, if a subtransaction had an XID but no
actual changes.  Per bug #14279 from Marko Tiikkaja.  Patch by Marko
based on analysis by Andrew Gierth.

Discussion: <20160804191757.1430.39011@wrigleys.postgresql.org>

Commits 4452000f et al established semantics for NullTest.argisrow that
are a bit different from its initial conception: rather than being merely
a cache of whether we've determined the input to have composite type,
the flag now has the further meaning that we should apply field-by-field
testing as per the standard's definition of IS [NOT] NULL.  If argisrow
is false and yet the input has composite type, the construct instead has
the semantics of IS [NOT] DISTINCT FROM NULL.  Update the comments in
primnodes.h to clarify this, and fix ruleutils.c and deparse.c to print
such cases correctly.  In the case of ruleutils.c, this merely results in
cosmetic changes in EXPLAIN output, since the case can't currently arise
in stored rules.  However, it represents a live bug for deparse.c, which
would formerly have sent a remote query that had semantics different
from the local behavior.  (From the user's standpoint, this means that
testing a remote nested-composite column for null-ness could have had
unexpected recursive behavior much like that fixed in 4452000f.)

In a related but somewhat independent fix, make plancat.c set argisrow
to false in all NullTest expressions constructed to represent "attnotnull"
constructs.  Since attnotnull is actually enforced as a simple null-value
check, this is a more accurate representation of the semantics; we were
previously overpromising what it meant for composite columns, which might
possibly lead to incorrect planner optimizations.  (It seems that what the
SQL spec expects a NOT NULL constraint to mean is an IS NOT NULL test, so
arguably we are violating the spec and should fix attnotnull to do the
other thing.  If we ever do, this part should get reverted.)

Back-patch, same as the previous commit.

Discussion: <10682.1469566308@sss.pgh.pa.us>

That script is incorrect in that it sets the combine function for
max(citext) twice instead of setting the combine function for
max(citext) once and the combine functon for min(citext) once.  The
consequence is that if you install 1.0 or 1.1 and then update to 1.2,
you end up with min(citext) not having a combine function, contrary to
what was intended.  If you install 1.2 directly, you're OK.

Fix things up by defining a new 1.3 version.  Upgrading from 1.2 to
1.3 won't change anything for people who first installed the 1.2
version, but people upgrading from 1.0 or 1.1 will get the right
catalog contents once they reach 1.3.

Report and patch by David Rowley, reviewed by Andreas Karlsson.

Author: Masahiko Sawada

In btree_gin and citext, avoid some not-particularly-interesting
dependencies on the sorting of 'aa'.  In tsearch2, use COLLATE "C" to
remove an uninteresting dependency on locale sort order (and thereby
allow removal of a variant expected-file).

Also, in citext, avoid assuming that lower('I') = 'i'.  This isn't relevant
to Danish but it does fail in Turkish.

To ensure that "make installcheck" can be used safely against an existing
installation, we need to be careful about what global object names
(database, role, and tablespace names) we use; otherwise we might
accidentally clobber important objects.  There's been a weak consensus that
test databases should have names including "regression", and that test role
names should start with "regress_", but we didn't have any particular rule
about tablespace names; and neither of the other rules was followed with
any consistency either.

This commit moves us a long way towards having a hard-and-fast rule that
regression test databases must have names including "regression", and that
test role and tablespace names must start with "regress_".  It's not
completely there because I did not touch some test cases in rolenames.sql
that test creation of special role names like "session_user".  That will
require some rethinking of exactly what we want to test, whereas the intent
of this patch is just to hit all the cases in which the needed renamings
are cosmetic.

There is no enforcement mechanism in this patch either, but if we don't
add one we can expect that the tests will soon be violating the convention
again.  Again, that's not such a cosmetic change and it will require
discussion.  (But I did use a quick-hack enforcement patch to find these
cases.)

Discussion: <16638.1468620817@sss.pgh.pa.us>

The old code used SEQ_MINVALUE to get the smallest int64 value.  This
was done as a convenience to avoid having to deal with INT64_IS_BUSTED,
but that is obsolete now.  Also, it is incorrect because the smallest
int64 value is actually SEQ_MINVALUE-1.  Fix by using PG_INT64_MIN.

We must not push down a foreign join when the foreign tables involved
should be accessed under different user mappings.  Previously we tried
to enforce that rule literally during planning, but that meant that the
resulting plans were dependent on the current contents of the
pg_user_mapping catalog, and we had to blow away all cached plans
containing any remote join when anything at all changed in pg_user_mapping.
This could have been improved somewhat, but the fact that a syscache inval
callback has very limited info about what changed made it hard to do better
within that design.  Instead, let's change the planner to not consider user
mappings per se, but to allow a foreign join if both RTEs have the same
checkAsUser value.  If they do, then they necessarily will use the same
user mapping at runtime, and we don't need to know specifically which one
that is.  Post-plan-time changes in pg_user_mapping no longer require any
plan invalidation.

This rule does give up some optimization ability, to wit where two foreign
table references come from views with different owners or one's from a view
and one's directly in the query, but nonetheless the same user mapping
would have applied.  We'll sacrifice the first case, but to not regress
more than we have to in the second case, allow a foreign join involving
both zero and nonzero checkAsUser values if the nonzero one is the same as
the prevailing effective userID.  In that case, mark the plan as only
runnable by that userID.

The plancache code already had a notion of plans being userID-specific,
in order to support RLS.  It was a little confused though, in particular
lacking clarity of thought as to whether it was the rewritten query or just
the finished plan that's dependent on the userID.  Rearrange that code so
that it's clearer what depends on which, and so that the same logic applies
to both RLS-injected role dependency and foreign-join-injected role
dependency.

Note that this patch doesn't remove the other issue mentioned in the
original complaint, which is that while we'll reliably stop using a foreign
join if it's disallowed in a new context, we might fail to start using a
foreign join if it's now allowed, but we previously created a generic
cached plan that didn't use one.  It was agreed that the chance of winning
that way was not high enough to justify the much larger number of plan
invalidations that would have to occur if we tried to cause it to happen.

In passing, clean up randomly-varying spelling of EXPLAIN commands in
postgres_fdw.sql, and fix a COSTS ON example that had been allowed to
leak into the committed tests.

This reverts most of commits fbe5a3fb and 5d4171d1, which were the
previous attempt at ensuring we wouldn't push down foreign joins that
span permissions contexts.

Etsuro Fujita and Tom Lane

Discussion: <d49c1e5b-f059-20f4-c132-e9752ee0113e@lab.ntt.co.jp>

This is fallout from join pushdown; get_relid_attribute_name can't
handle an attribute number of 0, indicating a whole-row reference,
and shouldn't be called in that case.

Etsuro Fujita, reviewed by Ashutosh Bapat

As pointed out by Ashutosh Bapat, the header comments for this file
say that schema-qualification is needed for all and only those types
outside pg_catalog.  pg_catalog.text is not outside pg_catalog.

After pg_upgrade, it is possible that some tuples' Xmax have multixacts
corresponding to the old installation; such multixacts cannot have
running members anymore.  In many code sites we already know not to read
them and clobber them silently, but at least when VACUUM tries to freeze
a multixact or determine whether one needs freezing, there's an attempt
to resolve it to its member transactions by calling GetMultiXactIdMembers,
and if the multixact value is "in the future" with regards to the
current valid multixact range, an error like this is raised:
    ERROR:  MultiXactId 123 has not been created yet -- apparent wraparound
and vacuuming fails.  Per discussion with Andrew Gierth, it is completely
bogus to try to resolve multixacts coming from before a pg_upgrade,
regardless of where they stand with regards to the current valid
multixact range.

It's possible to get from under this problem by doing SELECT FOR UPDATE
of the problem tuples, but if tables are large, this is slow and
tedious, so a more thorough solution is desirable.

To fix, we realize that multixacts in xmax created in 9.2 and previous
have a specific bit pattern that is never used in 9.3 and later (we
already knew this, per comments and infomask tests sprinkled in various
places, but we weren't leveraging this knowledge appropriately).
Whenever the infomask of the tuple matches that bit pattern, we just
ignore the multixact completely as if Xmax wasn't set; or, in the case
of tuple freezing, we act as if an unwanted value is set and clobber it
without decoding.  This guarantees that no errors will be raised, and
that the values will be progressively removed until all tables are
clean.  Most callers of GetMultiXactIdMembers are patched to recognize
directly that the value is a removable "empty" multixact and avoid
calling GetMultiXactIdMembers altogether.

To avoid changing the signature of GetMultiXactIdMembers() in back
branches, we keep the "allow_old" boolean flag but rename it to
"from_pgupgrade"; if the flag is true, we always return an empty set
instead of looking up the multixact.  (I suppose we could remove the
argument in the master branch, but I chose not to do so in this commit).

This was broken all along, but the error-facing message appeared first
because of commit 8e9a16ab and was partially fixed in a25c2b7c.
This fix, backpatched all the way back to 9.3, goes approximately in the
same direction as a25c2b7c but should cover all cases.

Bug analysis by Andrew Gierth and Álvaro Herrera.

A number of public reports match this bug:
  https://www.postgresql.org/message-id/20140330040029.GY4582@tamriel.snowman.net
  https://www.postgresql.org/message-id/538F3D70.6080902@publicrelay.com
  https://www.postgresql.org/message-id/556439CF.7070109@pscs.co.uk
  https://www.postgresql.org/message-id/SG2PR06MB0760098A111C88E31BD4D96FB3540@SG2PR06MB0760.apcprd06.prod.outlook.com
  https://www.postgresql.org/message-id/20160615203829.5798.4594@wrigleys.postgresql.org

something.* IS NOT NULL means that every attribute of the row is not
NULL, not that the row itself is non-NULL (e.g. because it's coming
from below an outer join.  Use (somevar.*)::pg_catalog.text IS NOT
NULL instead.

Ashutosh Bapat, per a report by Rushabh Lathia.  Reviewed by
Amit Langote and Etsuro Fujita.  Schema-qualification added by me.

Etsuro Fujita

Per buildfarm.  Fortunately, it's not quite too late to squeeze this fix
into the pg_trgm 1.3 update.

There was some very strange code here, dating to commit b525bf77, that
purported to work around an ancient gcc bug by forcing a float4 comparison
to be done as int instead.  Commit 5871b884 broke that when it changed
one side of the comparison to "double" but left the comparison code alone.
Commit f576b17c doubled down on the weirdness by introducing a "volatile"
marker, which had nothing to do with the actual problem.

Guess that the gcc bug, even if it's still present in the wild, was
triggered by comparison of float4's and can be avoided if we store the
result of cnt_sml() into a double before comparing to the double "nlimit".
This will at least work correctly on non-broken compilers, and it's way
more readable.

Per bug #14202 from Greg Navis.  Add a regression test based on his
example.

Report: <20160620115321.5792.10766@wrigleys.postgresql.org>