Functions can be declared with parameters with default values or
expressions.  The default expressions are used as parameter value
if the parameter is not explicitly specified in a function call.
All parameters after a parameter with default value have to be
parameters with default values as well.

It allows user to invoke a UDF without setting all the parameters.
Two examples to demo its usage:

    CREATE FUNCTION dfunc1(text DEFAULT 'Hello', text DEFAULT 'World')
        RETURNS text AS $$
        SELECT $1 || ', ' || $2;
        $$ LANGUAGE SQL;
    SELECT dfunc1();  -- 'Hello, World'
    SELECT dfunc1('Hi');  -- 'Hi, World'
    SELECT dfunc1('Hi', 'Beijing');  -- 'Hi, Beijing'

    CREATE FUNCTION dfunc2(id int4, t timestamp DEFAULT now())
        RETURNS text AS $$
        SELECT 'Time for id:' || $1 || ' is ' || $2;
        $$ LANGUAGE SQL;
    SELECT dfunc2(24);  -- 'Time for id:24 is 2016-01-07 14:38'

NOTE: The default change set is ported from from PostgreSQL 8.4,
    original commits:
    '517ae403'
    '455dffbb'

The file was being included in the Windows installers, which caused the
Windows build to fail once the file was removed.

Closes #303

This patch merges the fix for plpgsql simple expression evaluation mechanism
to ensure it behaves safely in the presence of subtransactions.

 Conflicts:
	src/pl/plpgsql/src/pl_exec.c
	src/pl/plpgsql/src/pl_handler.c
	src/pl/plpgsql/src/plpgsql.h

Forgot to update this in previous commit, when I added a test query to the
test.

Some test queries in the TINC/cdbfast test suite were failing with that
error, after merging the equivalence classes path. The problem was that
the Motion node's hash key an incorrect equivalence class member, one that
wasn't in the target list, instead of one that was.

To fix, make cdbpullup_findPathKeyExprInTargetList() return the actual Expr
that it finds, not just the EquivalenceClass containing it. The logic in
cdbpathlocus_get_partkey_exprs() that tried to find the right member was
not correct: it's not enough to check that the relids overlap.

This brings the code in question closer to the way it was before the
equivalence classes merge.

The cdbdistributedxidmap.c file was not in Makefile and was not compiled
(and would throw errors if you tried).

All the commits in this set were either doc commits or had already been cherry
picked into GPDB so barring a small pg_sleep() diff in the tests, another merge
without side effects.

 Conflicts:
	configure
	doc/src/sgml/Makefile
	doc/src/sgml/docguide.sgml
	src/backend/access/nbtree/nbtinsert.c
	src/backend/postmaster/postmaster.c
	src/backend/utils/adt/selfuncs.c
	src/port/exec.c

I removed pathnode_copy_node() in an earlier commit, but missed this
comment. If someone needs to copy a Path in the future, they'll have to
come up with a way to do it at that point, but there hasn't been any need
for that so far. (copyObject() should work, but it makes a deep copy, while
the old pathnode_copy_node() function made a shallow copy.

These really don't belong here.

The compiler was wrong, the variable is only used if 'ismeta', and it is
always initialized in that case. Silence the warning by initializing it
anyway.

_outEquivalenceClass and _outEquivalenceMember were unused, when included
into outfast.c. We have no need to serialize them when shipping a finished
Plan to segments, so just ifdef them out to silence the compiler. (The
regular textual "out" functions, for debugging purposes, work normally.)

It was quite clearly broken, as it didn't do anything with the new_tle
that it chased up, and the end result was just the same as calling
canonicalize_pathkeys() on the input.

When compiling gpfdist on Ubuntu, the compile fails because of the compiler
flag to treat all warnings as errors. You get a warning when calling
freopen and not checking the result of the call.

Added a check for NULL of the redirect of stdout and stderr on init. If it
fails, it will print a message and return out of init with a -1.

This set of patches were all already cherry-picked into PGDB or was rejected
due to being Doc patches.

 Conflicts:
	contrib/intarray/_int.h
	doc/TODO
	doc/src/FAQ/TODO.html
	doc/src/sgml/Makefile
	doc/src/sgml/docguide.sgml
	doc/src/sgml/installation.sgml
	src/backend/libpq/be-secure.c
	src/backend/port/win32/socket.c
	src/backend/postmaster/pgstat.c
	src/backend/utils/adt/xml.c
	src/include/port/win32.h
	src/interfaces/ecpg/ChangeLog
	src/interfaces/ecpg/ecpglib/extern.h
	src/interfaces/ecpg/ecpglib/typename.c
	src/interfaces/libpq/fe-secure.c
	src/interfaces/libpq/libpq-int.h
	src/pl/plperl/expected/plperl.out
	src/pl/plperl/plperl.c
	src/pl/plperl/sql/plperl.sql
	src/timezone/ialloc.c
	src/timezone/zic.c
	src/tools/msvc/Project.pm
	src/tools/msvc/build.bat
	src/tools/msvc/mkvcbuild.pl

This merge only brings a small doc update and a bugfix in detaosting
access to bytea field in pg_trigger.tgargs. The main changes in this
set of patches were XML related which we already have (from 9.1) and
the introduction of tablespaces for temporary objects which GPDB
already have with filespaces for temporary objects.

 Conflicts:
	doc/TODO
	doc/src/FAQ/TODO.html
	doc/src/sgml/config.sgml
	doc/src/sgml/datatype.sgml
	src/backend/commands/indexcmds.c
	src/backend/commands/tablecmds.c
	src/backend/commands/tablespace.c
	src/backend/commands/trigger.c
	src/backend/executor/execMain.c
	src/backend/executor/execQual.c
	src/backend/parser/gram.y
	src/backend/parser/keywords.c
	src/backend/parser/parse_expr.c
	src/backend/storage/file/fd.c
	src/backend/utils/adt/xml.c
	src/backend/utils/misc/postgresql.conf.sample
	src/include/utils/guc.h
	src/include/utils/xml.h
	src/pl/plpython/plpython.c
	src/test/regress/expected/xml.out
	src/test/regress/expected/xml_1.out
	src/test/regress/sql/xml.sql

I'm not 100% the code is doing the right thing even with this: If the
pathkey's operator family doesn't match the join clause's operator family,
the output isn't truly ordered the same way. This is not an issue for the
case that I included in the regression test, where the join happens on a
constant, but I can't figure out whether this might be an issue with some
other query.

When the combocids patch was backported from PostgreSQL 8.3 (c3983003), the
changes to change the cache key of compiled functions were left out for
some reason. Those changes are necessary: calling HeapTupleHeaderGetCMin()
on the pg_proc tuple will trip an assertion, if the tuple has been moved by
VACUUM FULL, and therefore HEAP_MOVED flags on the tuple are set. To fix,
backport the missing parts of the upstream combocid patch.

The comment in apply_shareinput_xslice() said that we process top-plan
first, to ensure that we find the producer of a shared scan before its
consumers, when the top plan contains a producer for subplans. But the
situation can just as well be the other way 'round.

No new test case added, because running the existing with_clause test
with gp_cte_sharing=on trips the assertion.

Fixes github issue #287.

My previous commit (3d0a4806), which was solving the issue described in pull request #276, was braking some ICG tests in some operating systems and we reverted it. This is a complete fix for the same issue.

Currently to check previous runtimes it takes multiple grep commands
to summarize the log file. This change adds all the information to a
single line making it far easier to check historical runtimes. The
historical runtimes can be used to estimate the required maintenance
window for future runs.

This merge adds --tablespaces-only and --roles-only to pg_dumpall. Since
both GPDB and upstream have used -r in pg_dumpall, this commits disables
-r and instead issues an error to avoid confusion.

 Conflicts:
	doc/TODO
	doc/src/FAQ/TODO.html
	doc/src/sgml/ref/pg_dumpall.sgml
	src/bin/pg_dump/pg_backup_archiver.h
	src/bin/pg_dump/pg_dumpall.c
	src/include/c.h

This patch prevents WAL logging when COPY is done in the same transaction
that created it. fba8113c is also cherry
picked into this merge since it has fixups to this patch as well as
related WAL skipping changes to CLUSTER.

 Conflicts:
	doc/src/sgml/perform.sgml
	src/backend/access/heap/heapam.c
	src/backend/access/heap/tuptoaster.c
	src/backend/catalog/index.c
	src/backend/commands/copy.c
	src/backend/executor/execMain.c
	src/backend/utils/cache/relcache.c
	src/include/access/tuptoaster.h

Closes 275

The dump process was checking the segments for several kinds of
metadata, such as indexes and triggers, even though they were not
being included in the dump.  The statements that were performing
these useless checks have been modified to only run on the master.

Authors: James McAtamney and Jimmy Yih

This reverts commit 3d0a4806.

Closes #288

This fix prevents gpdb from a segmentation fault in case that: i) a SharedInputScan has a Sort child node, ii) sort spills, and iii) somebody requests to finish the execution during the last stage of the sort (i.e., QueryFinishPending = true). Currently, in the above scenario, SharedInputScan continues the execution and tries to access sort's result_tape, which is NULL, due to QueryFinishPending.

This commit makes a sort node to pretend that there are no results when QueryFinishPending flag is set to true. The commit also  creates a new Fault injector type and includes an ICG test.

This adds the capability to dump operator families. XML and MSVC build
fixes were already merged and kept at master.

 Conflicts:
	doc/src/sgml/xindex.sgml
	src/backend/executor/execScan.c
	src/backend/utils/adt/xml.c
	src/backend/utils/mb/wchar.c
	src/bin/pg_dump/common.c
	src/bin/pg_dump/pg_backup_archiver.c
	src/bin/pg_dump/pg_dump.c
	src/bin/pg_dump/pg_dump.h
	src/include/port/win32.h
	src/test/regress/expected/xml.out
	src/tools/msvc/Project.pm
	src/tools/msvc/Solution.pm
	src/tools/msvc/mkvcbuild.pl

Gpcrondump and gpdbrestore redirect stderr into status file for segments,
errors should be reflected into the master's report file.
Warnings will be given if process completed but errors were generated in
the report file.

Also made some refactor of the code.

This commit also reverts 08f4ada7.

Authors: Pengcheng Tang, Nikhil Kak

systemd IPC cleanup takes effect if gpadmin is not system user

int2vector and aclitem don't have B-tree operators, making them not
mergejoinable. In the merge of the equivalence classes patch, I made
cdb_make_pathkeys_for_expr() more strict on that, and it now throws
an error if it's passed a datatype that doesn't have a mergejoinable
= operator. It used to return a path key with InvalidOid as the sort
operator, but that seems like a recipe for bugs elsehere.

To fix, mandate that all gpdb-hashable datatypes must be mergejoinable.
int2vector and aclitem are not, so forbid using them in DISTRIBUTED BY.

This is a backwards-incompatible change, but hopefully no-one is using
int2vector or aclitem in user-tables, at least not as a distribution key.

Ensure that the comment match what can happen, YAML files with .yml
suffix can also be created from mapred.

This adds CREATE/ALTER/DROP OPERATOR FAMILY commands.

Conflicts:
	doc/src/sgml/ref/allfiles.sgml
	doc/src/sgml/ref/comment.sgml
	doc/src/sgml/ref/drop_opclass.sgml
	doc/src/sgml/reference.sgml
	src/backend/catalog/aclchk.c
	src/backend/commands/opclasscmds.c
	src/backend/parser/gram.y
	src/backend/parser/keywords.c
	src/backend/tcop/utility.c
	src/include/nodes/parsenodes.h

Conflicts:
	contrib/pgbench/pgbench.c
	doc/FAQ_MINGW
	doc/src/sgml/manage-ag.sgml
	src/backend/optimizer/path/costsize.c
	src/backend/postmaster/postmaster.c
	src/include/port/win32.h
	src/port/exec.c

There were failures from these in the Pivotal buildfarm. Some tables were
larger than expected. I'm not sure why - perhaps there was some concurrent
activity that bloats append-only tables - but there have been no functional
changes so these really should be passing.

This is dead, because SubPlan is an Expr node and therefore cannot appear
in a Plan tree.

This was broken, because the "subplan = (SubPlan *) subplan" cast didn't
do anything, and subplan was potentially uninitialized. It meant to say
"subplan = (SubPlan *) plan".

Spotted by Daniel Gustafsson.