It was just syntax and catalogs, you couldn't actually do anything useful
with it. Remove it, so that we have less code to maintain, until it's time
to merge this stuff from upstream again when we merge with PostgreSQL 8.4.
It's probably easier to merge this back at that point than maintain this
backported version in the meanwhile. Less effort now, until we reach that
point, and once we get to the point in 8.4 that we merge this in, we'll
have all the preceding patches applied already, so it should merge quite
smoothly.

until we change the catalog. specifically pg_enum typarray, and
ev_enabled

tgconstraint has been added to the pg_trigger catalog

This is the first merge commit in an effort to catch up with PostgreSQL 8.3.
This batch corresponds to about two weeks worth of upstream commits from
REL8_2_0, towards REL8_3_0.

I changed the OID of int8dec function to 3546, as the old OID (2857) clashed
with the new Postgres function, pg_stat_get_backend_xact_start. 3546 is used
for int8dec in the upstream (9.3) anyway.

Most of the conflicts below were trivial in nature, and I'm not sure if
listing them in commit message is useful for anything, but doesn't hurt I
guess..

Conflicts:
	configure
	configure.in
	contrib/hstore/hstore_op.c
	contrib/tsearch2/dict_syn.c
	doc/FAQ
	doc/FAQ_AIX
	doc/FAQ_HPUX
	doc/FAQ_IRIX
	doc/FAQ_MINGW
	doc/FAQ_SCO
	doc/FAQ_Solaris
	doc/TODO
	doc/src/FAQ/FAQ.html
	doc/src/FAQ/TODO.html
	doc/src/sgml/Makefile
	doc/src/sgml/config.sgml
	doc/src/sgml/docguide.sgml
	doc/src/sgml/installation.sgml
	doc/src/sgml/monitoring.sgml
	doc/src/sgml/ref/psql-ref.sgml
	doc/src/sgml/stylesheet-hh.xsl
	doc/src/sgml/stylesheet.dsl
	doc/src/sgml/stylesheet.xsl
	doc/src/sgml/xfunc.sgml
	src/backend/access/transam/xact.c
	src/backend/access/transam/xlog.c
	src/backend/catalog/system_views.sql
	src/backend/executor/execMain.c
	src/backend/executor/nodeLimit.c
	src/backend/executor/spi.c
	src/backend/nodes/copyfuncs.c
	src/backend/nodes/equalfuncs.c
	src/backend/nodes/outfuncs.c
	src/backend/nodes/readfuncs.c
	src/backend/optimizer/geqo/geqo_eval.c
	src/backend/optimizer/path/joinrels.c
	src/backend/optimizer/plan/initsplan.c
	src/backend/optimizer/plan/subselect.c
	src/backend/optimizer/util/joininfo.c
	src/backend/parser/parse_agg.c
	src/backend/port/win32/socket.c
	src/backend/postmaster/pgstat.c
	src/backend/tcop/Makefile
	src/backend/tcop/postgres.c
	src/backend/utils/adt/pgstatfuncs.c
	src/backend/utils/misc/guc.c
	src/bin/pg_controldata/pg_controldata.c
	src/bin/pg_dump/win32.mak
	src/bin/pg_resetxlog/pg_resetxlog.c
	src/include/catalog/catversion.h
	src/include/catalog/pg_control.h
	src/include/executor/executor.h
	src/include/nodes/execnodes.h
	src/include/nodes/primnodes.h
	src/include/optimizer/joininfo.h
	src/include/pgstat.h
	src/port/dirmod.c
	src/template/linux
	src/timezone/tznames/Asia.txt
	src/timezone/tznames/Default

This makes merging and diffing with upstream easier. There were so many
GPDB-added GUCs in guc.c that "git merge" didn't handle it very well, and
you easily got merge conflicts that were error-prone to fix manually.

I kept a few GUCs in guc.c that are not present in REL8_2_0, because they
were backported from later PostgreSQL versions, rather than being
completely GPDB-specific.

Clean up the diff between guc.c and upstream, by removing unnecessary
includes, reordering a few GUCs to match upstream better, and similar
cosmetic changes.

I had to move MAX_KILOBYTES from guc.c to guc_tables.h, because guc_gp.c
needs it too. I also had to export assign_msglvl() function from guc.c, so
that it can be called from other assign_* functions from guc_gp.c. And I
had to add the code to process the ConfigureNames*_gp arrays in addition
to the upstream ones. These things increase the diff against upstream, but
not by too much, IMHO. I also copy-pasted constants like MS_PER_S and
KB_PER_MB, which seems safe as those are natural constants, and should
never change.

Move gpfdist regress cases 'exttab1' and 'custom_format' to extension/gpfdist
directory.

Enhance pg_regress to call needed perl scripts from correct directory.

  1) Short flags for output file and dumping filespaces were erroneously both assigned to -f.
     Dumping filespaces is now properly assigned to -F.
  2) Tablespaces were being dumped even when filespaces flag was not set.
     Now, tablespaces are only dumped when their corresponding filespaces are dumped.

Authors: Jimmy Yih and James McAtamney

Fix brackets missing issue while describe views having ordered-set aggregates

Closes #81

Currently, readfast.c / outfast.c are mostly copy-pasted from readfuncs.c /
outfuncs.c. That's a merge hazard: if a new field is added to a struct in
upstream, and it's added to readfuncs.c and outfuncs.c, we would need to
manually do the same in the readfast.c/outfast.c. If the patch applies
cleanly, we will not notice, and we'll have a bug of omission.

Refactor the code so that all those node types where the text and binary
functions are identical, the duplicate in [read/out]fast.c is removed, and
the definition [read/out]funcs.c is used to compile the binary version too.
This involves some tricks with #ifdefs and #includes, but cuts a lot of
duplicate code. This should avoid the merge hazard.

We'll still need to maintain the read/out functions whenever we modify a
struct in Greenplum, but that's no different from what needs to be done in
PostgreSQL.

This reduces the diff between readfast.c and readfuncs.c. There are
of course parts of these files that need to be different, but the bulk
of the functions should be identical, with all the differences hidden
in the READ_* macros. This patch reduces the diff and thus makes any
suspicious differences between them more obvious.

gptransfer currently does not validate whether required filespaces are missing on the destination system. If filespace does not exist, it fails and returns stack trace. It should return proper error message.

Prior to doing a gptransfer, gptransfer will now validate filespaces on destination system and exit accordingly with the proper error message.

During a restore on a database without the related schemas (ex. --redirect
to empty database) in which these schemas are set to be dropped, the SET
search_path can be missing before the CREATE TABLE statements in the dump
file. This results in tables being restored in the default schema instead
of their expected schema.

The main issue was that the ArchiveHandle->currSchema was not being reset
when drop schema prior to create was requested in gp_dump. The
ArchiveHandle->currSchema is now properly reset. This was originally fixed
in upstream Postgres, and this commit just ports over that fix to gp_dump.

This splits pg_proc.h into two parts: pg_proc.h and pg_proc_gp.h. After the
split, pg_proc.h contains functions we've inherited from upstream, and is
mostly identical to the upstream version. This makes diffing and merging
with upstream easier. All the upstream functions have been removed from
pg_proc.sql, as they are no longer auto-generated. Instead of modifying
pg_proc.h, catullus.pl is now used to create pg_proc_gp.h, which contains
all the auto-generated DATA entries for the functions in pg_proc.sql.

For the upstream-inherited DATA entries in pg_proc.h, remove the columns
that are not present in upstream. This was needed again to make pg_proc.h
closer to the upstream version. Also, remove auto-generated comments and
revert whitespace changes compared to upstream.

To compensate for the missing columns, introduce a new perl script,
processs_col_defaults.pl that is run at build time, which adds the missing
columns back to any DATA rows that miss them, before they are fed to
genbki.sh. All the GPDB-added columns are annotated with a magic
GPDB_COLUMN_DEFAULT() macro, similar to the CATALOG() and DATA() macros
which tells process_col_defaults.pl what values to insert into the missing
columns.

Remove tidycat2 regression test. It tested that all the functions in
pg_catalog came from pg_proc.sql, but that is no longer true. Only the
GPDB-added ones are now in pg_proc.sql.

This partially reverts commit 4e01c159.
The out-function for PlannedStmt is still needed, for debugging purposes,
even though we never try to read them back in from text format. There is
debug code in serializeNode that serializes a complete plan, and that
gives a WARNING if there is no out-function for PlannedStmt. (I'm not
convinced that code is doing much good, as it's only spewing stuff to the
log under DEBUG5, and at that level it will just get lost in the noise.
But that's a separate discussion). There is an out-function for
PlannedStmt in PostgreSQL 9.3 too, which is where we it was backported
from.

While we're at it, take the opportunity to move _outPlannedStmt to the
same location it's in in upstream, and remove spurious whitespace
differences.

g++ gave an error: "using typedef-name 'TBMIterateResult' after 'struct'",
breaking the build with ORCA. It looks nicer to use the typedef without the
struct keyword anyway.

_readTupleDescNode function in readfuncs.c had an 'str' argument, but it
was called with zero arguments. The reason that we haven't seen crashes
is that the out/readfuncs.c support for PlannedStmts and TupleDescNodes is
dead code. They are only serialized when sent from QD to QE, and we use
the out/readfast.c functions for that.

Remove the code, and add a few comments to TupleDescNode and PlannedStmt
to clarify this a bit.

There's no particular need to avoid #including other header files in
execnodes.h, so let's not diverge from upstream. Also revert a few
whitespace changes.

enum ExprDoneCond, enum SetFunctionReturnMode and struct ReturnSetInfo
had been moved from execnodes.h to fmgr.h for some reason. I couldn't
figure out why, although I agree that if we were writing this code in a
vacuum, fmgr.h might make more sense, as these structs are needed by
user-defined set-returning functions, to avoid #including so much stuff.
But differences like this create conflicts when merging with upstream,
so let's just stick to what the upstream does.

When the segment dump process crashes, master ends up waiting
infinitely. So adding a fault detection mechanism to periodically
check if the segment dump process is alive.

Fix DDBoost restore to retrieve dump file from Data Domain Server;

DDBoost restore incorrectly assumes that the post schema dump file is
located locally, and failed the restore process.

Gpcrondump now sets the correct schema for schema altered sub partition tables;

If sub partition tables got altered and set to different schema,
gpcrondump incorrectly sets the schema for these modified tables.

Fix the usage of fmtId() to correctly consume the returned value;

fmtId() reuses the same buffer; so before calling fmtId() for the 2nd time,
consume the value in the returned buffer.

Fix restore failure when exchange partition with external tables has 'LOG ERRORS' statement.

If an external table has a "LOG ERRORS" statement in the definition, it explicitly
creates the error table. Hence, restore fails due to name conflicts with the subpartition table name.

Authors : Pengcheng Tang, Nikhil Kak, Isobel Redelmeier

Previously the  classpath was relative
to $GPHOME/lib/postgresql/java/

The implication of this is that in order to reference jars outside
of this directory the classpath will look like ../../../ etc.

This patch retains that behaviour but adds the ability to use absolute
paths.

For security purposes changing the classpath can only be done by a
super user or set in postgresql.conf

There are no translations for Greenplum-specific strings, so it seems like
a disservice to allow anyone to build with --enable-nls.

The PostgreSQL translation files were a bit of a hodge-podge too; they
seemed to be from PostgreSQL 8.4.1, but we still carried the bunch of
incomplete translations that were removed in PostgreSQL 8.4.1. Let's not
pretend that there's any sense in them, and just remove all .po files. If
we want to make NLS to work properly in the future, we can bring them back,
using some sensible versions. Another option would've been to replace them
with the REL8_2_0 versions, which would possibly have made merging with
upstream slightly easier, as you wouldn't get merge conflicts on the
removed files, but those conflicts are easy to deal with. Also, by removing
the files completely, we don't need to decide what to do with the
references to pgsql-bugs@postgresql.org in them (we're trying to replace
all those references with bugs@greenplum.org).

README.debian to remove "ed" dependency.

Replace "ed" with "sed" in ED_PG_CONF function.  ED_PG_CONF and
supporting variables are renamed with "SED" as needed.

send() returns -1 in case of errors and thus we must use ssize_t rather
than size_t in order for the error check to be useful. A next pass over
this should be to check the actual errno of the triggered error to be
able to improve the error message.

This option is very misleading and needs to be removed.
We want to make sure that the user sees the double prompt that the
tool should only be used by Pivotal Support.

Fix issue where, if there is an empty filespace, pt rebuild will
fail the validation backup portion, as it currently assumes that
every filespace is not empty. This causes the whole validation to fail.

Fix assumption in our error output during validation process.
We were assuming that the failure will always be in the default filespace,
which is not the case.

Mark some functions that are only used in xlog.c as static, rename
XLog_OutRec() back to xlog_outrec(), fix comments.

Also update README.txt to match more closely the current way the unit test
Makefiles are written. Looks like it still reflected an old system where you
had to manually list all the *_mock.o files that a test program needs, but
that's not how it works anymore.

Mostly whitespace changes.

Makes diffing and merging easier. I don't see any reason for why it was
moved in the first place.