PathNameOpenFile failed to ensure that the correct value of errno was
returned to its caller after a failure (because it incorrectly supposed
that free() can never change errno).  In some cases this would result
in a user-visible failure because an expected ENOENT errno was replaced
with something else.  Bogus EINVAL failures have been observed on OS X,
for example.

There were also a couple of places that could mangle an important value
of errno if FDDEBUG was defined.  While the usefulness of that debug
support is highly debatable, we might as well make it safe to use,
so add errno save/restore logic to the DO_DB macro.

Per bug #8167 from Nelson Minar, diagnosed by RhodiumToad.
Back-patch to all supported branches.

If we're going to quote a well-known pangram, we should quote it
accurately.  Per gripe from Thom Brown.

The behavior is that the required sequence is created locally, which is
appropriate because the default expression will be evaluated locally.
Per gripe from Brad Nicholson that this case was refused with a confusing
error message.  We could have improved the error message but it seems
better to just allow the case.

Also, remove ALTER TABLE's arbitrary prohibition against being applied to
foreign tables, which was pretty inconsistent considering we allow it for
views, sequences, and other relation types that aren't even called tables.
This is needed to avoid breaking pg_dump, which sometimes emits column
defaults using separate ALTER TABLE commands.  (I think this can happen
even when the default is not associated with a sequence, so that was a
pre-existing bug once we allowed column defaults for foreign tables.)

If OID wraparound should occur while in standalone mode (unlikely but
possible), we want to advance the counter to FirstNormalObjectId not
FirstBootstrapObjectId.  Otherwise, user objects might be created with OIDs
in the system-reserved range.  That isn't immediately harmful but it poses
a risk of conflicts during future pg_upgrade operations.

Noted by Andres Freund.  Back-patch to all supported branches, since all of
them are supported sources for pg_upgrade operations.

In a construct like "select plain_function(set_returning_function(...))",
the plain function is applied to each output row of the SRF successively.
If some of the SRF outputs are NULL, and the plain function is strict,
you'd expect to get NULL results for such rows ... but what actually
happened was that such rows were omitted entirely from the result set.
This was due to confusion of this case with what should happen for nested
set-returning functions; a strict SRF is indeed supposed to yield an empty
set for null input.  Per bug #8150 from Erwin Brandstetter.

Although this has been broken forever, we're not back-patching because
of the possibility that some apps out there expect the incorrect behavior.
This change should be listed as a possible incompatibility in the 9.3
release notes.

The existing code in NUM_numpart_from_char has hard-wired logic to treat
'.' as decimal point, even when we're using a locale-aware format string
and the locale says that '.' is the thousands separator.  This results in
clearly wrong answers in FM mode (where we must be able to identify the
decimal point location), as per bug report from Patryk Kordylewski.

Since the initialization code in NUM_prepare_locale already sets up
Np->decimal as either the locale decimal-point string or "." depending
on which decimal-point format code was used, there's really no need to
have any extra logic at all in NUM_numpart_from_char: we only need to
test for a match to Np->decimal.

(Note: AFAICS there's nothing in here that explicitly checks for thousands
separators --- rather, any unmatched character is silently skipped over.
That's pretty bogus IMO but it's not the issue being complained of.)

This is a longstanding bug, but it's possible that some existing apps
are depending on '.' being recognized as decimal point even when using
a D format code.  Hence, no back-patch.  We should probably list this
as a potential incompatibility in the 9.3 release notes.

Looks like some versions of the buildfarm script try to set the port via
--port in $EXTRA_REGRESS_OPTS.  Override that ...

Previously, the port number used in this test script was hard-wired at
pg_upgrade's default of 50432; which is not so great because parallel build
runs might conflict.  Commit 3d53173e
removed this setting for the postmasters started by the script proper
(not by pg_upgrade), which didn't do anything to fix that problem and also
guaranteed a failure if there was a live postmaster at the build's default
port number.  Instead, select a non-conflicting temporary port number in
the same way that pg_regress.c does.  (Its method isn't entirely
bulletproof, but given the lack of complaints I'm not going to worry
about that today.)

In passing, unset MAKEFLAGS and MAKELEVEL to avoid problems with the
script's internal invocations of make, for the same reason pg_regress.c
does: it could cause problems in a parallel make.

The 9.2 patch that added argument name support in SQL-language functions
missed updating a parenthetical comment about that in the CREATE FUNCTION
reference page.  Noted by Erwin Brandstetter.

This case doesn't normally happen, because the planner usually clamps
all row estimates to at least one row; but I found that it can arise
when dealing with relations excluded by constraints.  Without a defense,
estimate_num_groups() can return zero, which leads to divisions by zero
inside the planner as well as assertion failures in the executor.

An alternative fix would be to change set_dummy_rel_pathlist() to make
the size estimate for a dummy relation 1 row instead of 0, but that seemed
pretty ugly; and probably someday we'll want to drop the convention that
the minimum rowcount estimate is 1 row.

Back-patch to 8.4, as the problem can be demonstrated that far back.

Per report from Keith Fiske.

Marko Kreen

Commit d22a09dc introduced official support
for GiST consistentFns that want to cache data using the FmgrInfo fn_extra
pointer: the idea was to preserve the cached values across gistrescan(),
whereas formerly they'd been leaked.  However, there was an oversight in
that, namely that multiple scan keys might reference the same column's
consistentFn; the code would result in propagating the same cache value
into multiple scan keys, resulting in crashes or wrong answers.  Use a
separate array instead to ensure that each scan key keeps its own state.

Per bug #8143 from Joel Roller.  Back-patch to 9.2 where the bug was
introduced.

It is not used anymore.

This helps guard against changes in the set of reserved keywords from
one version to another.  In theory it should only be an issue if we
de-reserve a keyword in a newer release, since that can create the type
of problem shown in bug #8128.

Back-patch to 9.1 where the --quote-all-identifiers option was added.

Document that post-upgrade steps are likely to be the same for all
clusters with the same DDL/schemas;  this should help automated
upgrades.

Missed in commit 62e66640.

This reverts the code changes in 50c13748,
which turned out to induce crashes and not completely fix the problem
anyway.  That commit only considered single subqueries that were excluded
by constraint-exclusion logic, but actually the problem also exists for
subqueries that are appendrel members (ie part of a UNION ALL list).  In
such cases we can't add a dummy subpath to the appendrel's AppendPath list
without defeating the logic that recognizes when an appendrel is completely
excluded.  Instead, fix the problem by having setrefs.c scan the rangetable
an extra time looking for subqueries that didn't get into the plan tree.
(This approach depends on the 9.2 change that made set_subquery_pathlist
generate dummy paths for excluded single subqueries, so that the exclusion
behavior is the same for single subqueries and appendrel members.)

Note: it turns out that the appendrel form of the missed-permissions-checks
bug exists as far back as 8.4.  However, since the practical effect of that
bug seems pretty minimal, consensus is to not attempt to fix it in the back
branches, at least not yet.  Possibly we could back-port this patch once
it's gotten a reasonable amount of testing in HEAD.  For the moment I'm
just going to revert the previous patch in 9.2.

Fix the term used in variable and struct names, and comments.

Alexander Korotkov

If a standby server has a cascading standby server connected to it, it's
possible that WAL has already been sent up to the next WAL page boundary,
splitting a WAL record in the middle, when the first standby server is
promoted. Don't throw an assertion failure or error in walsender if that
happens.

Also, fix a variant of the same bug in pg_receivexlog: if it had already
received WAL on previous timeline up to a segment boundary, when the
upstream standby server is promoted so that the timeline switch record falls
on the previous segment, pg_receivexlog would miss the segment containing
the timeline switch. To fix that, have walsender send the position of the
timeline switch at end-of-streaming, in addition to the next timeline's ID.
It was previously assumed that the switch happened exactly where the
streaming stopped.

Note: this is an incompatible change in the streaming protocol. You might
get an error if you try to stream over timeline switches, if the client is
running 9.3beta1 and the server is more recent. It should be fine after a
reconnect, however.

Reported by Fujii Masao.

What we have implemented is a radix tree (or a radix trie or a patricia
trie), but the docs and code comments incorrectly called it a "suffix tree".

Alexander Korotkov

Karl O. Pinc

It is surprisingly common mistake to leave out backup_label file from a base
backup. Say more explicitly that it must be included.

Jeff Janes, with minor rewording by me.

I had time for a quick review of the notes, so here are some fixes.

Previously this state was represented by whether the view's disk file had
zero or nonzero size, which is problematic for numerous reasons, since it's
breaking a fundamental assumption about heap storage.  This was done to
allow unlogged matviews to revert to unpopulated status after a crash
despite our lack of any ability to update catalog entries post-crash.
However, this poses enough risk of future problems that it seems better to
not support unlogged matviews until we can find another way.  Accordingly,
revert that choice as well as a number of existing kluges forced by it
in favor of creating a pg_class.relispopulated flag column.

Very old versions of msgfmt choke on these specific messages, for reasons
that are unclear at the moment.  Remove them so that we can ship a beta
release and not get complaints from testers (these messages will just go
untranslated, instead, and we're hardly at 100% coverage anyway).
Peter Eisentraut will look for a better fix later.

The initial implementation of this feature was really unsupportable,
because it's relying on the physical size of an on-disk file to carry the
relation's populated/unpopulated state, which is at least a modularity
violation and could have serious long-term consequences.  We could say that
an unlogged matview goes to empty on crash, but not everybody likes that
definition, so let's just remove the feature for 9.3.  We can add it back
when we have a less klugy implementation.

I left the grammar and tab-completion support for CREATE UNLOGGED
MATERIALIZED VIEW in place, since it's harmless and allows delivering a
more specific error message about the unsupported feature.

I'm committing this separately to ease identification of what should be
reverted when/if we are able to re-enable the feature.

Andrew Dunstan

Mention this also helps in the restoring of pg_dumps.

Jeff Janes

No need to mention wal_receiver_status_interval.

Previous coding set the SQL buffer but never executed

Bug noted by me during beta testing

Removal of doc adjustment and release note mention as well.