Foremost, it should go to stdout.

The recent cleanup of GUC assign hooks got rid of the kludge of using
"unknown" as a magic value for timezone and log_timezone.  But I forgot
to update the documentation to match, as noted by Martin Pitt.

Discard any collation aliases that match the built-in pg_collation entries
(ie, "default", "C", "POSIX").  Such aliases would be refused by a CREATE
COLLATION command, but since initdb is injecting them via a simple INSERT,
it has to make the corresponding check for itself.  Per Martin Pitt's
report of funny behavior in a machine that had a bogus "C.UTF-8" locale.

Also, use E'' syntax for the output of escape_quotes, as per its header
comment.

that we report the libpq connection failure string.  Per suggestion from
Robert Haas.

three-value boolean logic.

Backpatch to 9.0.X since we just got another bug report about this
today.

This doesn't work as expected because the isolationtester program requires
libpq to already be installed.  While it works when you've already installed
libpq, having to already have done "make install" defeats most of the point
of a check with a temp installation.  And there are weird corner cases if
the dynamic linker picks up an old libpq.so from system library directories.
Remove the target (or more precisely, make it print a helpful message) so
people don't expect the case to work.

by 3, but that is it OK.

Remove random system #includes in favor of using postgres_fe.h.  (The
alternative to that is letting this module grow its own configuration
testing ability...)

Also fix the "make clean" target to actually clean things up.

Per local testing.

README.links to explain xref properly.

Also report the error message when the post-pg_ctl connection fails.

Per private bug report from EnterpriseDB.

Dan Ports, per head-scratching from Simon Riggs and myself.

Also adjust some error message capitalization for consistency.

KaiGai Kohei

With some compilers such as Clang and ICC emulating GCC, using a
version string of the form "GCC $version" can be quite misleading.
Also, a great while ago, the version output from gcc --version started
including the string "gcc", so it is redundant to repeat that.  In
order to support ancient GCC versions, we now prefix the result with
"GCC " only if the version output does not start with a letter.

The SSI patch inserted a call of RegisterPredicateLockingXid into
GetNewTransactionId, which was a bad idea on a couple of grounds.  First,
it's not necessary to hold XidGenLock while manipulating that shared
memory, and doing so is bad because XidGenLock is a high-contention lock
that should be held for as short a time as possible.  (Not to mention that
it adds an entirely unnecessary deadlock hazard, since we must take
SerializableXactHashLock as well.)  Second, the specific place where it was
put was between extending CLOG and advancing nextXid, which could result in
unpleasant behavior in case of a failure there.  Pull the call out to
AssignTransactionId, which is much safer and arguably better from a
modularity standpoint too.

There is more work to do to clean up the failure-before-advancing-nextXid
issue, but that is a separate change that will need to be back-patched.
So for the moment I just want to make GetNewTransactionId look the same as
it did in prior versions.

These were labeled with precedences just to avoid attaching explicit
precedences to the productions in which they were the last terminal symbol.
Since a terminal symbol precedence marking can affect many other things
too, it seems like better practice to attach precedence labels to the
productions, and not mark the terminal symbols.

Ideally we'd also remove the precedence attached to NULL_P, but it turns
out that we are actually depending on that having a precedence higher than
POSTFIXOP, else we get a shift/reduce conflict for postfix operators in
b_expr.  (Which more or less proves my point about these markings having a
high risk of unexpected consequences.)  For the moment, move NULL_P into
the set of keywords grouped with IDENT, so that at least it will act
similarly to non-keywords; and document the interaction.

For consistency with other tools, put the options before further usage
information.

In pg_standby, remove the supposedly deprecated -l option from the
given example invocation.

Instead of just saying "is not a table", specifically state that
indexes aren't supported on *foreign* tables.

Per gripe from Grzegorz Szpetkowski.

Also, change the subsection heading from "Lexical Precedence" (which is
a contradiction in terms) to "Operator Precedence".

pg_basebackup doesn't need to police the format of port numbers.
libpq already does that.

Tabs replaced by spaces.

Per http://joomla.aws.continuent.com/community/lab-projects/sequoia

Greg Smith

Greg Smith, after a suggestion of James Bruce

David Fetter

After finding an EXISTS or ANY sub-select that can be converted to a
semi-join or anti-join, we should recurse into the body of the sub-select.
This allows cases such as EXISTS-within-EXISTS to be optimized properly.
The original coding would leave the lower sub-select as a SubLink, which
is no better and often worse than what we can do with a join.  Per example
from Wayne Conrad.

Back-patch to 8.4.  There is a related issue in older versions' handling
of pull_up_IN_clauses, but they're lame enough anyway about the whole area
that it seems not worth the extra work to try to fix.

Greg Smith