Files changed:
    modified:   src/backend/cdb/motion/ic_common.c
    modified:   src/backend/executor/spi.c
    modified:   src/backend/nodes/outfuncs.c
    modified:   src/backend/optimizer/path/costsize.c
    modified:   src/backend/storage/file/compress_zlib.c

Note: The warning in function _outScanInfo() of outfuncs.c is temporally
    fixed and would be treated as dead code to be removed soon.

Thanks to Heikki Linnakangas' comments.

 * a bug was introduced by a feature that looked for partition tables that were moved to other schemas.
 * fixed here by qualifying with schema name, tightening the criteria
 * added Behave test

Refactor test_with_orca.py and build_with_orca.py
Extract the common piece out as GporcaCommon package.

Disable gpfdist tests with --disable-gpfdist option for configure,
because GPORCA doesn't impact the management utilities like gpfdist.

When user dumps database to Data Domain Boost server, storage
unit and backup directory must be already created and specified,
previously, we hard coded the storage unit to "GPDB" and user
had no option to use others.

This commit adds --ddboost-storage-unit option, which allows
user to dynamically specify storage unit for dump and restore.

This commits allows user to have storage unit information
statically saved into configure file in their cluster host.

This commit added storage unit option into gpmfr for replicating
and recovering dump copies, in which case it uses identical storage
unit and backup directory between primary and secondary DDBoost server.

--ddboost-storage-unit option takes higher priority than using
statically configured storage unit.

Authors:
Pengcheng Tang, Marbin Tan, Nikhil Kak
Lawrence Hamel, Stephen Wu, Chris Hajas, Chumki Roy

Now s3ext could recognize then decompress gzip encoded files
automatically, doesn't require any extra parameter, configuration or
extended filename.

I don't see anything in this test that would require a huge number of rows.
Half a million should be more than enough to show up in reltuples/relpages.

Almost all tests in bfv_legacy were also in qp_misc_rio. I kept the ones
in qp_misc_rio, except for a few that have different output with ORCA.
This way, only bfv_legacy needs to have an alternative expected output file
for ORCA.

Test 31 was identical to test 25.

This exact same test case, with some extra EXPLAINs, is in the
co_nestloop_idxscan regression test.

Instead of pushing the responsibility of rescanning down to each different
kind of external table implement rescanning in fileam.c in a generic
fashion, by closing and reopening the underlying "url". This gets us rescan
support for custom and EXECUTE-type external tables, which was missing
before, and also makes the code simpler.

There are no known cases where the rescan support is currently needed (hence
no test case included), because the planner puts Materialize nodes on top of
external scans, but in principle every plan node is supposed to be
rescannable. I tested this by reverting the previous patch that fixed
using external scans in a subplan; without that patch, an external table
in a subplan would get rescanned.

ParallelizeCorrelatedSubPlanMutator() turns each Scan on a base relation
into a "Result - Material - Broadcast - Scan" pattern, but it missed
ExternalScans. External tables are supposed to be treated as distributed,
i.e. each segment holds different part of the external table, so they
need to be treated like regular tables.

Long time ago, a hack was put in place in GPDB to use the "scan" slot,
instead of the "outer" slot which is used in the upstream, to hold the
result of an Agg or Window plan node's child. It's not clear to me why
that was done. There was even a comment in fix_upper_expr() saying we
wouldn't need it if we just fixed the executor to not contain that hack,
and there was also a TODO comment in CMappingColIdVarPlStmt.cpp about
that.

Everything seems to work without those hacks, so revert this thing back
to the way it works in the upstream. This is simpler in its own right,
and also reduces our diff vs. upstream, which will make merging easier
in the future.

This was raised in #691 and was identified as a bug in upstream as well. The
patch has now been committed to upstream, this is a backport with Greenplum
versions of Flex/Perl maintained. See below for upstream commit message.

    commit 7d7b1292
    Author: Tom Lane <tgl@sss.pgh.pa.us>
    Date:   Mon May 2 11:18:10 2016 -0400

    Fix configure's incorrect version tests for flex and perl.

    awk's equality-comparison operator is "==" not "=".  We got this right
    in many places, but not in configure's checks for supported version
    numbers of flex and perl.  It hadn't been noticed because unsupported
    versions are so old as to be basically extinct in the wild, and because
    the only consequence is whether or not a WARNING flies by during
    configure.

    Daniel Gustafsson noted the problem with respect to the test for flex,
    I found the other by reviewing other awk calls.

This closes #697

This closes #682

Without GUC_GPDB_ADOPT, the values for guc is not going to be dispatched
to the QE processes.

Specically persistent check:
  gp_persistent_relation_node   <=> filesystem

Authors: Marbin Tan & Larry Hamel

Batch size of 8 is too low and each cluster may have a different
system configuration, so we would like to dertermine a default batch
size before running gpcheckcat.

* Add unittest for batch size

* Truncate batch size to be, at maximum, the amount of primaries.
  Batch size can be no longer be larger than the amount of primaries.

* Refactor: Create method main() for gpcheckcat
  In order improve unit testing, move functionality from '__main__'
  to a method.

Authors: Marbin Tan, Larry Hamel, Nikhil Kak

* Remove warning when compiling aosegfiles.c

Use strlcpy instead of strncat to stop compiler complaining. The
original case "strncat(segnumArray, tmp, sizeof(tmp))" is simple
and safe, but the compiler complains much by law [-Wstrncat-size].

Changes to be committed:
	modified:   src/backend/access/aocs/aocssegfiles.c
	modified:   src/backend/access/appendonly/aosegfiles.c

* speed up a bit by replacing strncat with strlcpy

Files changed:
    modified:   src/backend/access/external/url.c
    modified:   src/backend/cdb/cdbhash.c
    modified:   src/backend/cdb/cdbmutate.c
    modified:   src/pl/plpgsql/src/pl_funcs.c

Skipped:
    catcoretable.c:104:26:CatCoreType_int4_array
	because it is an enumerated constant to ensure integrity

In a previous commit, f569c1d1,
gpcheckcat was modified to display a list of tables with missing
attributes.  This commit adds the ability to list tables with
extraneous attributes.

Authors: Chumki Roy and James McAtamney

Fixed a typo in the --with-codegen-prefix description and corrected
spelling in the CMake check error message. No functional changes.

atpxPartAddList() needs a CreateStmt that represents the parent table,
but instead of creating it already in the parser, and adding more details
to it in analyze.c, it's simpler to create it later, in atpxPartAddList(),
where it's actually needed.

The code in InitPostgres() was refactored in PostgreSQL 9.0 so that it
no longer uses FindMyDatabaseByOid() function. We had backpatched the
InitPostgres() changes already, so backpatch the removal of
FindMyDatabaseByOid() as well. Silences a compiler warning.

In _SPI_execute_plan, debug_query_string was set directly to plan->query which is not
memory context safe, it means debug_query_string have a chance to refer to a invalid
address when FATAL/PANIC level error occurs.

A common source of bugs has been that an object gets assigned a different
OID in the master and in segments. A segment should normally never have to
allocate an OID (for catalog objects) on its own, all OIDs should be
allocated in the master, and sent over to the segments. To make such bugs
easier to catch, add a WARNING if an OID is allocated in a segment.

There were some DEBUG1 elogs for the same thing in place already, but the
list of catalogs that need synchronized OIDS wasn't up-do-date, and this new
place for the elog() is less invasive anyway.

When I merged the operator family patch, I missed dispatching the new DDL
commands to segments. Because of that, the segments didn't have information
about operator families. Some operator families would be greated implicitly
by CREATE OPERATOR CLASS, but you wouldn't necessarily get the same
configuration of families and classes as in the master. Things worked pretty
well despite that, because operator families and classes are used for
planning, and planning happens in the master. Nevertheless, we really should
have the operator family information in segments too, in case you run
queries in maintenance mode directly on the segments, or if you execute
functions in segments that need to execute expression that depend on them.
Also, there were no regression tests for the new DDL commands.

If you do CREATE OPERATOR, with a commutator or negator operator that
doesn't exist yet, the system creates a "shell" entry for the non-existent
operator. But those shell operators didn't get the same OID in all segments,
which could lead to strange errors later. I couldn't find a test case
demonstrating actual bugs from that, but it sure seems sketchy. Given that
we take care to synchronize the OID of the primary created operator, surely
we should do the same for all operators.

The out/readfuncs.c support for AlterTableStmt.comptypeArrayOid was
missing. Because of that, the segments didn't get the OID of the composite
type's array type from master, and allocated it on their own.

We were using "complex" as a user defined type, which is no longer the
case once #605 was checked in.