Due to the support of special characters, we made some changes in the format
of appendonly tables' state file, which inevitably create extra tables into the
dirty table lists for incremental backup.

Use the "template1" instead of the restoring database for connection url

Authors: Pengcheng Tang, Christopher Hajas

- targeted_dispatch

If gpfdist is built with ssl and does not use --ssl
it will now print the correct message.

Authors: Marbin Tan & Larry Hamel

the cleanup of EState

Below is the error condition that was being filtered by the cdbfast test
framework.  I have added an ignore section around the update that is
expected to fail in the planner scenario.

*** ./expected/DML_over_joins.out2016-03-27 16:44:49.000000000 -0700
--- ./results/DML_over_joins.out2016-03-27 16:44:49.000000000 -0700
***************
*** 1342,1348 ****
(2 rows)

update sales_par set region = 'new_region' where id in (select s.b from s, r where s.a = r.b) and day in (select a from r);
! ERROR:  moving tuple from partition "sales_par_1_prt_usa" to partition "sales_par_1_prt_other_regions" not supported
select sales_par.* from sales_par where id in (select s.b from s, r where s.a = r.b) and day in (select a from r);
id | year | month | day | region
----+------+-------+-----+--------
--- 1342,1348 ----
(2 rows)

update sales_par set region = 'new_region' where id in (select s.b from s, r where s.a = r.b) and day in (select a from r);
! ERROR:  moving tuple from partition "sales_par_1_prt_europe" to partition "sales_par_1_prt_other_regions" not supported
select sales_par.* from sales_par where id in (select s.b from s, r where s.a = r.b) and day in (select a from r);
id | year | month | day | region
----+------+-------+-----+--------

======================================================================

This enhances the existing "sirv_functions" ICG tests.  This is
accomplished by migrating the cdbfast regression "list_queries" test
schedule's "sirv_functions" test suite to ICG.

Here are some characteristics of the test suite:

  Schedule file: greenplum_schedule
  Schema: sirv_functions
  Set to run in parallel: yes
  Orca specific test output: no
  Test execution time: <1 minute
  Pivotal Tracker Story: #113967799

Migrate the cdbfast "list_queries" test schedule "DML_over_joins" to
ICG.

Here are some characteristics of the test suite:

  Schedule file: greenplum_schedule
  Schema: DML_over_joins
  Set to run in parallel: yes
  Orca specific test output: yes
  Test execution time: <1 minute
  Pivotal Tracker Story: #113967799

gpdbrestore now performs pre validation of tables (also partition tables)
in the dump file, this fix applies to ddboost, netbackup, and other remote
dump location.

Also made some refactoring of the code.

Authors:
Pengcheng Tang, Nikhil Kak,
Lawrence Hamel, Marbin Victor Tan

This closes #570

- unify source-code of #518 commit according to upstream
- add regression tests (previously stored in TINC)
- fix parameter type of the function AddNewRelationType()

    Thanks to Heikki Linnakangas for suggesting

Squashed commit of the following:

commit db752c0c916f36479b0c5049c671538565294c25
Author: Nikos Armenatzoglou <narmenatzoglou@gmail.com>
Date:   Thu Mar 24 14:01:38 2016 -0700

    Adding check to avoid unit test failures when sizeof(long) = 8

commit 1c39c58279cf73af92811ed1b2e6f2bc22d01414
Author: Shreedhar Hardikar <shardikar@pivotal.io>
Date:   Thu Mar 17 19:42:47 2016 +0000

    --with-codegen-prefix instead of CMAKE_PREFIX_PATH

commit 77978e92444580bc4e84dc3f6b3e366a09e92dba
Author: Shreedhar Hardikar <shardikar@pivotal.io>
Date:   Thu Mar 10 14:57:48 2016 +0000

    Make codegen work with GPDB build system and other minor fixes

      * Update CMakeLists to build SUBSYS.o
      * Update CMakeLists to build release build by default
      * Update CMakeLists to add tests in a loop.
      * Add codegen and cmake specific .gitignore
      * Update license and READMEs
      * Move codegen utils related header files back into src/backend/codegen/include/codegen
Signed-off-by: NNikos Armenatzoglou <nikos.armenatzoglou@gmail.com>

commit 1b61c12f4eb082f49a66c28957c22810e98a5074
Author: Shreedhar Hardikar <shardikar@pivotal.io>
Date:   Wed Mar 16 21:06:28 2016 +0000

    Remove codegen specific gtest source tree and use the one in gpAux.
Signed-off-by: NNikos Armenatzoglou <nikos.armenatzoglou@gmail.com>

commit 13bb2dcf3465be0ef240e316a149b2d8a125a9e9
Author: Shreedhar Hardikar <shardikar@pivotal.io>
Date:   Thu Mar 10 02:51:29 2016 +0000

    Set up GPDB to use codegen utilities

      * Move in codegen utility source and include files to GPDB specific locations.
      * Configure codegen with --enable-codegen and use gpcodegen namespace
Signed-off-by: NNikos Armenatzoglou <nikos.armenatzoglou@gmail.com>

commit 9a4e40f931f71cb7a82e15b51745944116bc10ab
Author: Craig Chasseur <spoonboy42@gmail.com>
Date:   Mon Nov 2 09:10:37 2015 -0800

    Introduce codegen utilties into GPDB.

    Support library for integrating LLVM code-generation and runtime
    compilation with Clang in a larger codebase.
Signed-off-by: NNikos Armenatzoglou <nikos.armenatzoglou@gmail.com>

Change behaviour to use default estimates for relation size
for cases when no statistics are available.

This avoid having heavy statstistics computation as part
of query compilation.

Customer can explicitly using ANALYZE to compute the stats
before running queries to use more accurate stats instead
of default statistics.

Currently, the default size of table is:
- Internal table: 100 pages
- External table: 1000 pages
Signed-off-by: NXin Zhang <xzhang@pivotal.io>

Currently, when relation open fails it emits error message which is very
dev centric plus stack trace gets printed. The message and behavior is
confusing as more hints towards some catastrophic problem. While its
more legitimate due to concurrency so best to communicate the same, plus
avoid printing stacktrace.

This commit adds more filespace and tablespace test coverage in installcheck.
Most of these test additions are modified and inspired from Pivotal's own
internal tests.

Codes of ic_tcp and ic_udp are not maintained any more, ic_udpifc is
the only interconnect type for now. We keep the gp_interconnect_type
guc for backward compatibility.

Type name is a fixed-length string padded by '\0', normal(wrong)
renaming will make cdbhash() getting different values.

Signed-off-by: NXin Zhang <xzhang@pivotal.io>

It seems that the install target for gpAux/extensions which in turn
invoke gpmapreduce installation fell away in the rebase and didn't
make it in the final commit. Re-adding it such that `make install`
also handles gpmapreduce.

Thanks to Xin Zhang for reporting.

- change the path to abs_srcdir

When alter table set distributed command specifying "reorganize=true/false", it
will set reloptions of pg_class on segments but not on master.

missed due to incorrect iterating logic.

This seems to be leftovers from when gpcheckdb was in src/bin and
was written in C++. Nothing references this and the directory is
empty so remove.

Backport below commits from upstream:

    commit adac22bf
    Author: Tom Lane <tgl@sss.pgh.pa.us>
    Date:   Fri Dec 19 05:04:35 2008 +0000

        When we added the ability to have zero-element ARRAY[] constructs by adding an
        explicit cast to show the intended array type, we forgot to teach ruleutils.c
        to print out such constructs properly.  Found by noting bogus output from
        recent changes in polymorphism regression test.

    commit 30137bde6db48a8b8c1ffc736eb239bd7381f04d
    Author: Heikki Linnakangas <heikki.linnakangas@iki.fi>
    Date:   Fri Nov 13 19:48:26 2009 +0000

        A better fix for the "ARRAY[...]::domain" problem. The previous patch worked,
        but the transformed ArrayExpr claimed to have a return type of "domain",
        even though the domain constraint was only checked by the enclosing
        CoerceToDomain node. With this fix, the ArrayExpr is correctly labeled with
        the base type of the domain. Per gripe by Tom Lane.

    commit 6b0706ac
    Author: Tom Lane <tgl@sss.pgh.pa.us>
    Date:   Thu Mar 20 21:42:48 2008 +0000

        Arrange for an explicit cast applied to an ARRAY[] constructor to be applied
        directly to all the member expressions, instead of the previous implementation
        where the ARRAY[] constructor would infer a common element type and then we'd
        coerce the finished array after the fact.  This has a number of benefits,
        one being that we can allow an empty ARRAY[] construct so long as its
        element type is specified by such a cast.

Besides, this commit also adds 'location' field in array related
structures, but they are not actived yet. Thanks to Heikki's suggestion.

This bug was introduced in fd2e045f.

Thanks to Shoaib Lari for pointing this out.

Modifying output of gpcheckcat to include the name of the test on the same line to make grepping easier

This patch slightly changes the way the demo cluster is built.
A gpinitdbsystem return code of 1 (warning) is ignored, and the script
exits with return code 0. Any error code greather than that is returned.
The Makefile is no longer ignoring errors.

This will fix #456.

The data directory and the TCP ports can be changed when "make cluster" is
executed to build the demo cluster:

DATADIRS=/tmp/gpdb-cluster MASTER_PORT=15432 PORT_BASE=25432 make cluster

This fixes #441 and #442

Also make the TCP port for the regression tests configurable.

Closes #255.

These were added by merge commit 60f45387, but we had backported a later
version of this file already, where these static functions had been removed
again.

This saves a little bit of memory when parsing massively partitioned
CREATE TABLE statements.

All of the callers are in places where leaking a few bytes of memory to
the current memory context will do no harm. Either parsing, or processing
a DDL command, or planning. So let's simplify the callers by removing the
argument. That makes the code match the upstream again, which makes merging
easier.

These changes were originally made to reduce the memory consumption when
doing parse analysis on a heavily partitioned table, but the previous
commit provided a more whole-sale solution for that, so we don't need to
nickel-and-dime every allocation anymore.

With a heavily partitioned table, the parse analysis of each CreateStmt
consumes significant amounts of memory. By using a temporary memory
context, we can reclaim some of the garbage left behind by the parsing of
each one. In very quick testing on my laptop, this reduces the memory
consumption of parse analysis of a massively-partitioned CREATE TABLE by
about 10%, but YMMV.

To make this work, transformAttributeEncoding() and its subroutines have
to be more careful to not modify the input CreateStmt, as any new Nodes
stored in it would be allocated in the temporary context and be blown away
at the end of transformCreateStmt. That's not cool, if there are more
partitions to process, which rely on the same CreateStmt.