This restores a parse error that was thrown (though only in the ORDER BY
case) by the original collation patch.  I had removed it in my recent
revisions because it was thrown at a place where collations now haven't
been computed yet; but I thought of another way to handle it.

Throwing the error at parse time, rather than leaving it to be done at
runtime, is good because a syntax error pointer is helpful for localizing
the problem.  We can reasonably assume that the comparison function for a
collatable datatype will complain if it doesn't have a collation to use.
Now the planner might choose to implement GROUP or DISTINCT via hashing,
in which case no runtime error would actually occur, but it seems better
to throw error consistently rather than let the error depend on what the
planner chooses to do.  Another possible objection is that the user might
specify a nondefault sort operator that doesn't care about collation
... but that's surely an uncommon usage, and it wouldn't hurt him to throw
in a COLLATE clause anyway.  This change also makes the ORDER BY/GROUP
BY/DISTINCT case more consistent with the UNION/INTERSECT/EXCEPT case,
which was already coded to throw this error even though the same objections
could be raised there.

Opening a catcache's index could require reading from that cache's own
catalog, which of course would acquire AccessShareLock on the catalog.
So the original coding here risks locking index before heap, which could
deadlock against another backend trying to get exclusive locks in the
normal order.  Because InitCatCachePhase2 is only called when a backend
has to start up without a relcache init file, the deadlock was seldom seen
in the field.  (And by the same token, there's no need to worry about any
performance disadvantage; so not much point in trying to distinguish
exactly which catalogs have the risk.)

Bug report, diagnosis, and patch by Nikhil Sontakke.  Additional commentary
by me.  Back-patch to all supported branches.

Instead of playing cute games with pathkeys, just build a direct
representation of the intended sub-select, and feed it through
query_planner to get a Path for the index access.  This is a bit slower
than 9.1's previous method, since we'll duplicate most of the overhead of
query_planner; but since the whole optimization only applies to rather
simple single-table queries, that probably won't be much of a problem in
practice.  The advantage is that we get to do the right thing when there's
a partial index that needs the implicit IS NOT NULL clause to be usable.
Also, although this makes planagg.c be a bit more closely tied to the
ordering of operations in grouping_planner, we can get rid of some coupling
to lower-level parts of the planner.  Per complaint from Marti Raudsepp.

ensure that they use different checkpoints as the starting point. We use
the checkpoint redo location as a unique identifier for the base backup in
the end-of-backup record, and in the backup history file name.

Bug spotted by Fujii Masao.

There's a lot we can't test very well without platform dependencies,
but the C/POSIX collations should now work the same way everywhere.

The local variable "sock" can be unused depending on compilation flags.
But there seems no particular need for it, since the kernel calls can
just as easily say port->sock instead.

Install just one instance of the "C" and "POSIX" collations into
pg_collation, rather than one per encoding.  Make these instances exist
and do something useful even in machines without locale_t support: to wit,
it's now possible to force comparisons and case-folding functions to use C
locale in an otherwise non-C database, whether or not the platform has
support for using any additional collations.

Fix up severely broken upper/lower/initcap functions, too: the C/POSIX
fastpath now does what it is supposed to, and non-default collations are
handled correctly in single-byte database encodings.

Merge the two separate collation hashtables that were being maintained in
pg_locale.c, and be more wary of the possibility that we fail partway
through filling a cache entry.

notes.

Remove excessive linking to pg_ctl manual page.

Reorder incompatibility sections.

Thom Brown

All expression nodes now have an explicit output-collation field, unless
they are known to only return a noncollatable data type (such as boolean
or record).  Also, nodes that can invoke collation-aware functions store
a separate field that is the collation value to pass to the function.
This avoids confusion that arises when a function has collatable inputs
and noncollatable output type, or vice versa.

Also, replace the parser's on-the-fly collation assignment method with
a post-pass over the completed expression tree.  This allows us to use
a more complex (and hopefully more nearly spec-compliant) assignment
rule without paying for it in extra storage in every expression node.

Fix assorted bugs in the planner's handling of collations by making
collation one of the defining properties of an EquivalenceClass and
by converting CollateExprs into discardable RelabelType nodes during
expression preprocessing.

This removes an overloading of two authentication options where
one is very secure (peer) and one is often insecure (ident). Peer
is also the name used in libpq from 9.1 to specify the same type
of authentication.

Also make initdb select peer for local connections when ident is
chosen, and ident for TCP connections when peer is chosen.

ident keyword in pg_hba.conf is still accepted and maps to peer
authentication.

Don't leave partial filename information on the last status or
dangling characters if the previous filename was longer than
the current one.

When adding an inheritance parent to a table, an AccessShareLock on the
parent isn't strong enough to prevent trouble, so take
ShareUpdateExclusiveLock instead.  Since this is a behavior change,
albeit a fairly unobtrusive one, and since we have only one report
from the field, no back-patch.

Report by Jon Nelson, analysis by Alvaro Herrera, fix by me.

This is advantageous because the BG writer is alive until much later in
the shutdown sequence than WAL writer; we want to make sure that it's
possible to shut off synchronous replication during a smart shutdown,
else it might not be possible to complete the shutdown at all.

Per very reasonable gripes from Fujii Masao and Simon Riggs.

Noted by Tom Lane.

the time on the primary that the transaction was generated.

Per a gripe from Thom Brown about my previous commit in this area,
commit 9a56dc33.

The last version in which these options were documented is now EOL, so
it's time to get rid of them for real.  We now use GNU-style long
options instead.

Without this, the startup process goes into a tight loop, consuming
100% of one CPU and failing to respond to interrupts.

Erik Rijkers and me

Report and patch from Daniel Popowich, bug #5842
(with some debugging help from Alex Hunsaker)

The maximum value of deadlock_timeout, max_standby_archive_delay,
max_standby_streaming_delay, log_min_duration_statement, and
log_autovacuum_min_duration was INT_MAX/1000 milliseconds, which is
about 35min, which is too short for some practical uses.  Raise the
maximum value to INT_MAX; the code that uses the parameters already
supports that just fine.

Fujii Masao, but with the proposed behavior change reverted, and the
rest adjusted accordingly.

1. Don't ignore query cancel interrupts.  Instead, if the user asks to
cancel the query after we've already committed it, but before it's on
the standby, just emit a warning and let the COMMIT finish.

2. Don't ignore die interrupts (pg_terminate_backend or fast shutdown).
Instead, emit a warning message and close the connection without
acknowledging the commit.  Other backends will still see the effect of
the commit, but there's no getting around that; it's too late to abort
at this point, and ignoring die interrupts altogether doesn't seem like
a good idea.

3. If synchronous_standby_names becomes empty, wake up all backends
waiting for synchronous replication to complete.  Without this, someone
attempting to shut synchronous replication off could easily wedge the
entire system instead.

4. Avoid depending on the assumption that if a walsender updates
MyProc->syncRepState, we'll see the change even if we read it without
holding the lock.  The window for this appears to be quite narrow (and
probably doesn't exist at all on machines with strong memory ordering)
but protecting against it is practically free, so do that.

5. Remove useless state SYNC_REP_MUST_DISCONNECT, which isn't needed and
doesn't actually do anything.

There's still some further work needed here to make the behavior of fast
shutdown plausible, but that looks complex, so I'm leaving it for a
separate commit.  Review by Fujii Masao.

Tom Lane

This has been broken for years, and I'm not sure why it has not been
noticed before, but now a very modern Cygwin breaks on it, and the fix
is clearly correct. Backpatching to all live branches.

Pavel Stehule