This includes code for sharedcache lookup and eviction.

We don't use workfile manager as a cache, we only keep workfiles around for
as long as we need them. So "clearing the cache" should always be a no-op.

It was only used in logging. Doesn't seem worth the trouble to keep it
around.

postgres.h is not a system header. Also remove some unnecessary #includes.

Callers to FaultInjector_InjectFaultIfSet() which don't pass neither
databasename nor tablename and that use DDLNotSpecified can instead
use the convenient macro SIMPLE_FAULT_INJECTOR() which cuts down on
the boilerplate in the code. This commit does not bring any changes
in functionality, merely readability.

While in there fixing other things, cleaned up the most visible
deviations from the norm of the file to make it more consistent
internally.

The support for having multiple hashing algorithms for the table
distribution was short circuited over 10 years ago when support for
more generic hashing was introduced together with NULL hashing
policies. Remove the dead code and avoid passing around a parameter
with a single possible value.

The warnings were for a set of unused variables and the use of
memcpy() without including the proper header file.

Backend crash due to PANIC causes postmaster ServerLoop to shutdown all
subprocesses and then restart them.  During 8.3 merge, the condition to restart
was weakened, causing ServerLoop to incorrectly initiate restart of
subprocesses before all existing subprocesses are shutdown.  This led to an
infinite loop in the event of PANIC.

Paired effort with Amil Khanzada <akhanzada@pivotal.io>

Greenplum has limited trigger capabilities which do not work with the current AO
implementation. It was found that after row triggers were being executed for AO
table DMLs using planner resulting in erroneous heap reads. Instead of adding
logic to do AO segfile reads, we disable after row triggers for AO tables as
Greenplum does not really support triggers for AO anyways. The issue with heap
tables started after the 8.3 merge which introduced LockBuffer usage at the
bottom of the GetTupleForTrigger function. LockBuffer assumes that a
MIRROREDLOCK_BUFMGR_LOCK is held so we just needed to release the
MIRROREDLOCK_BUFMGR_LOCK after the buffer lock is released.

1. Enable C++0x features in gcc.
2. Replace ChunkBuffer::chunkData with vector<uint8_t>. Remove init()
   and destroy().
3. Modify S3Service::fetchData, move buffer from Response object
   to data.
4. Modify S3Service::getXMLContext, use reference to avoid copy.
5. Fix unit tests in s3key_reader_test.
Signed-off-by: NPeifeng Qiu <pqiu@pivotal.io>

Any GPDB exception happens in while ORCA generating plan will abort
query.
Signed-off-by: NFoyzur Rahman <frahman@gmail.com>

* Fix Debian Vagrant file

Add more required packages
Fix upgraded Python packages
Add missing options to configure line
Fix ssh known hosts

* Fix ImportError with cffi and setuptools_ext

* latest changes from Dave

This closes #874

Signed-off-by: NKarthikeyan Jambu Rajaraman <karthi.jrk@gmail.com>

Verify that input expression is of type VAR <= CONST
Codegen expression evaluation
Fix wrong return value type
Support constants and variables of Date type
Add comments and fix code style

This reverts commit aae0ad3d.

Replace RegisterExternalFunction with GetOrRegisterExternalFunction using an unordered_map in codegen_utils

* Derive GpCodegenUtils from CodegenUtils for GPDB specific utilities.
* Move ElogWrapper functions to GpCodegenUtils
* Implement GetOrRegisterExternalFunction

* The user can now update the DDBOOST_CONFIG storage unit,
  which needs to be reflected on the help file.

Root cause: DecompressReader requests data from upstream reader, which
might return a very small amount of data, zlib inflate() may not be able
to decompress it.

Fix: fill zlib's chunk as possible as it could.

Test case: DecompressReaderTest.AbleToDecompressFragmentalCompressedData

Also refactor DecompressReader::read().
Signed-off-by: NPeifeng Qiu <pqiu@pivotal.io>
Signed-off-by: NAdam Lee <ali@pivotal.io>

The functions GetPartitionTargetlist() and UpdateGenericExprState()
are not called from anywhere so remove.

of meaningless code.

The correct name is gp_set_proc_affinity, not gp_proc_affinity.
Spotted by Alexey Grishchenko.