They can include sys/sdt.h from SystemTap, which itself contains C++
code and so won't compile with a C++ compiler under extern "C" linkage.

Careless use of TopMemoryContext for I/O function data meant that repeated
use of spi_prepare and spi_freeplan would leak memory at the session level,
as per report from Christian Schröder.  In addition, spi_prepare
leaked a lot of transient data within the current plperl function's SPI
Proc context, which would be a problem for repeated use of spi_prepare
within a single plperl function call; and it wasn't terribly careful
about releasing permanent allocations in event of an error, either.

In passing, clean up some copy-and-pasteos in query-lookup error messages.

Alex Hunsaker and Tom Lane

Per gripe from Andres Freund

This is a possibly vain attempt to fix a buffering issue
observed for some MSVC builds.

Thom Brown and me.

The new file in src/port needs to be listed in Mkvcbuild.pm as well.

Fujii Masao and me.

In copy-out mode, the frontend should not send any messages until the
backend has finished streaming, by sending a CopyDone message. I'm not sure
if it would be legal for the client to send a new query before receiving the
CopyDone message from the backend, but trying to support that would require
bigger changes to the backend code structure.

Fixes an assertion failure reported by Fujii Masao.

This includes backend "COPY TO/FROM PROGRAM '...'" syntax, and corresponding
psql \copy syntax. Like with reading/writing files, the backend version is
superuser-only, and in the psql version, the program is run in the client.

In the passing, the psql \copy STDIN/STDOUT syntax is subtly changed: if you
the stdin/stdout is quoted, it's now interpreted as a filename. For example,
"\copy foo from 'stdin'" now reads from a file called 'stdin', not from
standard input. Before this, there was no way to specify a filename called
stdin, stdout, pstdin or pstdout.

This creates a new function in pgport, wait_result_to_str(), which can
be used to convert the exit status of a process, as returned by wait(3),
to a human-readable string.

Etsuro Fujita, reviewed by Amit Kapila.

parseqatom() failed to check for an error return (NULL result) from its
recursive call to parsebranch(), and in consequence could crash with a
null-pointer dereference after an error return.  This bug has been there
since day one, but wasn't noticed before, probably because most error cases
in parsebranch() didn't actually lead to returning NULL.  Add the missing
error check, and also tweak parsebranch() to exit in a less indirect
fashion after a call to parseqatom() fails.

Report by Tomasz Karlik, fix by me.

Use correct type for "result", fix bogus strftime argument, don't use
unnecessary static variables, improve comments.

Andres Freund and Tom Lane

This got missed in commit 8396447c.

Andres Freund

The backend grammar treats STDIN and STDOUT completely interchangeable, so
that the above accepted. Arguably that was a mistake the backend grammar,
but it's not ecpg's business to second guess that.

There's no harm in excessive quoting per se, but it makes the strings nicer
to read. The values can get quite unwieldy, when they're first quoted within
within single-quotes when included in the connection string, and then all
the single-quotes are escaped when the connection string is passed as a
shell argument.

Like with pg_basebackup and pg_receivexlog, it's a bit strange to call the
option -d/--dbname, when in fact you cannot pass a database name in it.

Original patch by Amit Kapila, heavily modified by me.

You could already pass a database name just by passing it as the last
option, without -d. This is an alias for that, like the -d/--dbname option
in psql and many other client applications. For consistency.

The previous commit didn't work on MSVC editions earlier than
Visual Studio 2011, apparently. This works by copying files into the
contrib directory, and making provision to clean them up, which should
work on all editions.

Without this, there's no way to pass arbitrary libpq connection parameters
to these applications. It's a bit strange that the option is called
-d/--dbname, when in fact you can *not* pass a database name in it, but it's
consistent with other client applications where a connection string is also
passed using -d.

Original patch by Amit Kapila, heavily modified by me.

It is obviously no longer true.

rmgrdesc.c is not auto-generated now, though it apparently was the last
time the Makefile was updated.

The new name was originally my typo, but per discussion it seems like a
better name anyway.  So make the code match the docs, not vice versa.

Louis-Claude Canon and Josh Kupershmidt

Erik Rijkers

... as well a update copyrights statements to 2013.

Noted by Thom Brown and Peter Geoghegan

Harmless, but it's certainly better like this.

Noticed by Andres Freund

This program relies on rm_desc backend routines and the xlogreader
infrastructure to emit human-readable rendering of WAL records.

Author: Andres Freund, with many reworks by Álvaro
Reviewed (in a much earlier version) by Peter Eisentraut

Include eval costs of local conditions in remote-estimate mode, and don't
assume the remote eval cost is zero in local-estimate mode.  (The best
we can do with that at the moment is to assume a seqscan, which may well
be wildly pessimistic ... but zero won't do at all.)

To get a reasonable local estimate, we need to know the relpages count
for the remote rel, so improve the ANALYZE code to fetch that rather
than just setting the foreign table's relpages field to zero.

Apparently, they need -DBUILDING_DLL for the Assert() declarations to
work correctly.

The optimization to not retrieve unnecessary columns wasn't smart enough.
Noted by Thom Brown.

On reflection this method seems to be exposing an unreasonable amount of
implementation detail.  It wouldn't matter when talking to a remote server
of the identical Postgres version, but it seems likely to make things worse
not better if the remote is a different version with different casting
infrastructure.  Instead adopt ruleutils.c's policy of regurgitating the
cast as it was originally specified; including not showing it at all, if
it was implicit to start with.  (We must do that because for some datatypes
explicit and implicit casts have different semantics.)

The only place we depended on that was in sending numeric type OIDs in
PQexecParams; but we can replace that usage with explicitly casting
each Param symbol in the query string, so that the types are specified
to the remote by name not OID.  This makes no immediate difference but
will be essential if we ever hope to support use of non-builtin types.

We must still initialize minRecoveryPoint if we start straight with archive
recovery, e.g when recovering from a normal base backup taken with
pg_start/stop_backup. Otherwise we never consider the system consistent.

Set the remote session's search path to exactly "pg_catalog" at session
start, then schema-qualify only names that aren't in that schema.  This
greatly reduces clutter in the generated SQL commands, as seen in the
regression test changes.  Per discussion.

Also, rethink use of FirstNormalObjectId as the "built-in object" cutoff
--- FirstBootstrapObjectId is safer, since the former will accept
objects in information_schema for instance.

If you create a base backup using an atomic filesystem snapshot, and try to
perform PITR starting from that base backup, or if you just kill a master
server and create recovery.conf to put it into standby mode, we don't know
how far we need to recover before reaching consistency. Normally in crash
recovery, we replay all the WAL present in pg_xlog, and assume that we're
consistent after that. And normally in archive recovery, minRecoveryPoint,
backupEndRequired, or backupEndPoint is set in the control file, indicating
how far we need to replay to reach consistency. But if the server was
previously up and running normally, and you kill -9 it or take an atomic
filesystem snapshot, none of those fields are set in the control file.

The solution is to perform crash recovery first, replaying all the WAL in
pg_xlog. After that's done, we assume that the system is consistent like in
normal crash recovery, and switch to archive recovery mode after that.

Per report from Kyotaro HORIGUCHI. In his scenario, recovery.conf was
created after "pg_ctl stop -m i". I'm not sure we need to support that exact
scenario, but we should support backing up using a filesystem snapshot,
which looks identical.

This issue goes back to at least 9.0, where hot standby was introduced and
we started to track when consistency is reached. In 9.1 and 9.2, we would
open up for hot standby too early, and queries could briefly see an
inconsistent state. But 9.2 made it more visible, as we started to PANIC if
we see a reference to a non-existing page during recovery, if we've already
reached consistency. This is a fairly big patch, so back-patch to 9.2 only,
where the issue is more visible. We can consider back-patching further after
this has received some more testing in 9.2 and master.