Commit e09996ff8d was one brick shy of a load: it didn't insist
that the detected JSON number be the whole of the supplied string.
This allowed inputs such as "2016-01-01" to be misdetected as valid JSON
numbers. Per bug #13906 from Dmitry Ryabov.
In passing, be more wary of zero-length input (I'm not sure this can
happen given current callers, but better safe than sorry), and do some
minor cosmetic cleanup.
listForeignTables' invocation of processSQLNamePattern did not match up
with the other ones that handle potentially-schema-qualified names; it
failed to make use of pg_table_is_visible() and also passed the name
arguments in the wrong order. Bug seems to have been aboriginal in commit
0d692a0dc9. It accidentally sort of worked as long as you didn't
inquire too closely into the behavior, although the silliness was later
exposed by inconsistencies in the test queries added by 59efda3e50
(which I probably should have questioned at the time, but didn't).
Per bug #13899 from Reece Hart. Patch by Reece Hart and Tom Lane.
Back-patch to all affected branches.
We entirely randomly chose to initialize port->remote_host just after
printing the log_connections message, when we could perfectly well do it
just before, allowing %h and %r to work for that message. Per gripe from
Artem Tomyuk.
This will enable PL/Java to be cleanly compiled, as dynloader.h is a
requirement.
Report by Chapman Flack
Patch by Michael Paquier
Backpatch through 9.1
We can never leak more than one token, but we shouldn't do that. We
don't bother closing it in the error paths since the process will
exit shortly anyway.
Christian Ullrich
pg_dump's original approach to handling extension member objects was to
run around and clear (or set) their dump flags rather late in its data
collection process. Unfortunately, quite a lot of code expects those flags
to be valid before that; which was an entirely reasonable expectation
before we added extensions. In particular, this explains Karsten Hilbert's
recent report of pg_upgrade failing on a database in which an extension
has been installed into the pg_catalog schema. Its objects are initially
marked as not-to-be-dumped on the strength of their schema, and later we
change them to must-dump because we're doing a binary upgrade of their
extension; but we've already skipped essential tasks like making associated
DO_SHELL_TYPE objects.
To fix, collect extension membership data first, and incorporate it in the
initial setting of the dump flags, so that those are once again correct
from the get-go. This has the undesirable side effect of slightly
lengthening the time taken before pg_dump acquires table locks, but testing
suggests that the increase in that window is not very much.
Along the way, get rid of ugly special-case logic for deciding whether
to dump procedural languages, FDWs, and foreign servers; dump decisions
for those are now correct up-front, too.
In 9.3 and up, this also fixes erroneous logic about when to dump event
triggers (basically, they were *always* dumped before). In 9.5 and up,
transform objects had that problem too.
Since this problem came in with extensions, back-patch to all supported
versions.
Commit 803716013d installed a safeguard against loading plpython2
and plpython3 at the same time, but asserted that both could still be
used in the same database, just not in the same session. However, that's
not actually all that practical because dumping and reloading will fail
(since both libraries necessarily get loaded into the restoring session).
pg_upgrade is even worse, because it checks for missing libraries by
loading every .so library mentioned in the entire installation into one
session, so that you can have only one across the whole cluster.
We can improve matters by not throwing the error immediately in _PG_init,
but only when and if we're asked to do something that requires calling
into libpython. This ameliorates both of the above situations, since
while execution of CREATE LANGUAGE, CREATE FUNCTION, etc will result in
loading plpython, it isn't asked to do anything interesting (at least
not if check_function_bodies is off, as it will be during a restore).
It's possible that this opens some corner-case holes in which a crash
could be provoked with sufficient effort. However, since plpython
only exists as an untrusted language, any such crash would require
superuser privileges, making it "don't do that" not a security issue.
To reduce the hazards in this area, the error is still FATAL when it
does get thrown.
Per a report from Paul Jones. Back-patch to 9.2, which is as far back
as the patch applies without work. (It could be made to work in 9.1,
but given the lack of previous complaints, I'm disinclined to expend
effort so far back. We've been pretty desultory about support for
Python 3 in 9.1 anyway.)
A scan for missed proisstrict markings in the core code turned up
these functions:
brin_summarize_new_values
pg_stat_reset_single_table_counters
pg_stat_reset_single_function_counters
pg_create_logical_replication_slot
pg_create_physical_replication_slot
pg_drop_replication_slot
The first three of these take OID, so a null argument will normally look
like a zero to them, resulting in "ERROR: could not open relation with OID
0" for brin_summarize_new_values, and no action for the pg_stat_reset_XXX
functions. The other three will dump core on a null argument, though this
is mitigated by the fact that they won't do so until after checking that
the caller is superuser or has rolreplication privilege.
In addition, the pg_logical_slot_get/peek[_binary]_changes family was
intentionally marked nonstrict, but failed to make nullness checks on all
the arguments; so again a null-pointer-dereference crash is possible but
only for superusers and rolreplication users.
Add the missing ARGISNULL checks to the latter functions, and mark the
former functions as strict in pg_proc. Make that change in the back
branches too, even though we can't force initdb there, just so that
installations initdb'd in future won't have the issue. Since none of these
bugs rise to the level of security issues (and indeed the pg_stat_reset_XXX
functions hardly misbehave at all), it seems sufficient to do this.
In addition, fix some order-of-operations oddities in the slot_get_changes
family, mostly cosmetic, but not the part that moves the function's last
few operations into the PG_TRY block. As it stood, there was significant
risk for an error to exit without clearing historical information from
the system caches.
The slot_get_changes bugs go back to 9.4 where that code was introduced.
Back-patch appropriate subsets of the pg_proc changes into all active
branches, as well.
Given syntactically wrong input, widget_in() could call atof() with an
indeterminate pointer argument, typically leading to a crash; or if it
didn't do that, it might return a NULL pointer, which again would lead
to a crash since old-style C functions aren't supposed to do things
that way. Fix that by correcting the off-by-one syntax test and
throwing a proper error rather than just returning NULL.
Also, since widget_in and widget_out have been marked STRICT for a
long time, their tests for null inputs are just dead code; remove 'em.
In the oldest branches, also improve widget_out to use snprintf not
sprintf, just to be sure.
In passing, get rid of a long-since-useless sprintf into a local buffer
that nothing further is done with, and make some other minor coding
style cleanups.
In the intended regression-testing usage of these functions, none of
this is very significant; but if the regression test database were
left around in a production installation, these bugs could amount
to a minor security hazard.
Piotr Stefaniak, Michael Paquier, and Tom Lane
These functions readily crash when passed a NULL input value. The tests
themselves do not pass NULL values to them; but when the regression
database is used as a basis for fuzz testing, they cause a lot of noise.
Also, if someone were to leave a regression database lying about in a
production installation, these would create a minor security hazard.
Andreas Seltenreich
Turns out the only reason initdb -X worked is that pg_mkdir_p won't
whine if you point it at something that's a symlink to a directory.
Otherwise, the attempt to create pg_xlog/ just like all the other
subdirectories would have failed. Let's be a little more explicit
about what's happening. Oversight in my patch for bug #13853
(mea culpa for not testing -X ...)
When we're creating subdirectories of PGDATA during initdb, we know darn
well that the parent directory exists (or should exist) and that the new
subdirectory doesn't (or shouldn't). There is therefore no need to use
anything more complicated than mkdir(). Using pg_mkdir_p() just opens us
up to unexpected failure modes, such as the one exhibited in bug #13853
from Nuri Boardman. It's not very clear why pg_mkdir_p() went wrong there,
but it is clear that we didn't need to be trying to create parent
directories in the first place. We're not even saving any code, as proven
by the fact that this patch nets out at minus five lines.
Since this is a response to a field bug report, back-patch to all branches.
pg_ctl is using isatty() to verify whether the process is running in a
terminal, and if not it sends its output to Windows' Event Log ... which
does the wrong thing when the output has been redirected to a pipe, as
reported in bug #13592.
To fix, make pg_ctl use the code we already have to detect service-ness:
in the master branch, move src/backend/port/win32/security.c to src/port
(with suitable tweaks so that it runs properly in backend and frontend
environments); pg_ctl already has access to pgport so it Just Works. In
older branches, that's likely to cause trouble, so instead duplicate the
required code in pg_ctl.c.
Author: Michael Paquier
Bug report and diagnosis: Egon Kocjan
Backpatch: all supported branches
The order of inclusion of .o files makes a difference in linker output;
not a functional difference, but still a bitwise difference, which annoys
some packagers who would like reproducible builds.
Report and patch by Christoph Berg
pgwin32_recv() has treated a non-error return of zero bytes from WSARecv()
as being a reason to block ever since the current implementation was
introduced in commit a4c40f140d. However, so far as one can tell
from Microsoft's documentation, that is just wrong: what it means is
graceful connection closure (in stream protocols) or receipt of a
zero-length message (in message protocols), and neither case should result
in blocking here. The only reason the code worked at all was that control
then fell into the retry loop, which did *not* treat zero bytes specially,
so we'd get out after only wasting some cycles. But as of 9.5 we do not
normally reach the retry loop and so the bug is exposed, as reported by
Shay Rojansky and diagnosed by Andres Freund.
Remove the unnecessary test on the byte count, and rearrange the code
in the retry loop so that it looks identical to the initial sequence.
Back-patch of commit 90e61df813. The
original plan was to apply this only to 9.5 and up, but after discussion
and buildfarm testing, it seems better to back-patch. The noblock code
path has been at risk of this problem since it was introduced (in 9.0);
if it did happen in pre-9.5 branches, the symptom would be that a walsender
would wait indefinitely rather than noticing a loss of connection. While
we lack proof that the case has been seen in the field, it seems possible
that it's happened without being reported.
Commit c7e27becd2 fixed this on the backend side, but we neglected
the fact that several code paths in pg_dump were printing reloptions
values that had not gotten massaged by ruleutils. Apply essentially the
same quoting logic in those places, too.
spg_text_inner_consistent is capable of reconstructing an empty string
to pass down to the next index level; this happens if we have an empty
string coming in, no prefix, and a dummy node label. (In practice, what
is needed to trigger that is insertion of a whole bunch of empty-string
values.) Then, we will arrive at the next level with in->level == 0
and a non-NULL (but zero length) in->reconstructedValue, which is valid
but the Assert tests weren't expecting it.
Per report from Andreas Seltenreich. This has no impact in non-Assert
builds, so should not be a problem in production, but back-patch to
all affected branches anyway.
In passing, remove a couple of useless variable initializations and
shorten the code by not duplicating DatumGetPointer() calls.
As pointed out by Michael Paquier, recovery_min_apply_delay didn't exist
in 9.0-9.3, making the release note text not very useful. Instead make it
talk about recovery_target_xid, which did exist then.
9.0 is already out of support, but we can fix the text in the newer
branches' copies of its release notes.
flatten_reloptions() supposed that it didn't really need to do anything
beyond inserting commas between reloption array elements. However, in
principle the value of a reloption could be nearly anything, since the
grammar allows a quoted string there. Any restrictions on it would come
from validity checking appropriate to the particular option, if any.
A reloption value that isn't a simple identifier or number could thus lead
to dump/reload failures due to syntax errors in CREATE statements issued
by pg_dump. We've gotten away with not worrying about this so far with
the core-supported reloptions, but extensions might allow reloption values
that cause trouble, as in bug #13840 from Kouhei Sutou.
To fix, split the reloption array elements explicitly, and then convert
any value that doesn't look like a safe identifier to a string literal.
(The details of the quoting rule could be debated, but this way is safe
and requires little code.) While we're at it, also quote reloption names
if they're not safe identifiers; that may not be a likely problem in the
field, but we might as well try to be bulletproof here.
It's been like this for a long time, so back-patch to all supported
branches.
Kouhei Sutou, adjusted some by me
A report from Andy Colson showed that gincostestimate() was not being
nearly paranoid enough about whether to believe the statistics it finds in
the index metapage. The problem is that the metapage stats (other than the
pending-pages count) are only updated by VACUUM, and in the worst case
could still reflect the index's original empty state even when it has grown
to many entries. We attempted to deal with that by scaling up the stats to
match the current index size, but if nEntries is zero then scaling it up
still gives zero. Moreover, the proportion of pages that are entry pages
vs. data pages vs. pending pages is unlikely to be estimated very well by
scaling if the index is now orders of magnitude larger than before.
We can improve matters by expanding the use of the rule-of-thumb estimates
I introduced in commit 7fb008c5ee: if the index has grown by more
than a cutoff amount (here set at 4X growth) since VACUUM, then use the
rule-of-thumb numbers instead of scaling. This might not be exactly right
but it seems much less likely to produce insane estimates.
I also improved both the scaling estimate and the rule-of-thumb estimate
to account for numPendingPages, since it's reasonable to expect that that
is accurate in any case, and certainly pages that are in the pending list
are not either entry or data pages.
As a somewhat separate issue, adjust the estimation equations that are
concerned with extra fetches for partial-match searches. These equations
suppose that a fraction partialEntries / numEntries of the entry and data
pages will be visited as a consequence of a partial-match search. Now,
it's physically impossible for that fraction to exceed one, but our
estimate of partialEntries is mostly bunk, and our estimate of numEntries
isn't exactly gospel either, so we could arrive at a silly value. In the
example presented by Andy we were coming out with a value of 100, leading
to insane cost estimates. Clamp the fraction to one to avoid that.
Like the previous patch, back-patch to all supported branches; this
problem can be demonstrated in one form or another in all of them.
Common mathematical convention is that exponentiation associates right to
left. We aren't going to change the parser for this, but we could note
it in the operator's description. (It's already noted in the operator
precedence/associativity table, but users might not look there.)
Per bug #13829 from Henrik Pauli.
Tone down an overly strong statement about which pseudo-types PLs are
likely to allow. Add "event_trigger" to the list, as well as
"pg_ddl_command" in 9.5/HEAD. Back-patch to 9.3 where event_trigger
was added.
For some reason, we've been overlooking the fact that pg_receivexlog
and pg_recvlogical are using wrong translation domains all along,
so their output hasn't ever been translated. The right domain is
pg_basebackup, not their own executable names.
Noticed by Ioseph Kim, who's been working on the Korean translation.
Backpatch pg_receivexlog to 9.2 and pg_recvlogical to 9.4.
Both Blowfish and DES implementations of crypt() can take arbitrarily
long time, depending on the number of rounds specified by the caller;
make sure they can be interrupted.
Author: Andreas Karlsson
Reviewer: Jeff Janes
Backpatch to 9.1.
When pg_dump prompts the user for a password, it remembers the password
for possible re-use by parallel worker processes. However, libpq might
have extracted the password from a connection string originally passed
as "dbname". Since we don't record the original form of dbname but
break it down to host/port/etc, the password gets lost. Fix that by
retrieving the actual password from the PGconn.
(It strikes me that this whole approach is rather broken, as it will also
lose other information such as options that might have been present in
the connection string. But we'll leave that problem for another day.)
In passing, get rid of rather silly use of malloc() for small fixed-size
arrays.
Back-patch to 9.3 where parallel pg_dump was introduced.
Report and fix by Zeus Kronion, adjusted a bit by Michael Paquier and me
This is necessary so that REASSIGN OWNED does the right thing with
composite types, to wit, that it also alters ownership of the type's
pg_class entry -- previously, the pg_class entry remained owned by the
original user, which caused later other failures such as the new owner's
inability to use ALTER TYPE to rename an attribute of the affected
composite. Also, if the original owner is later dropped, the pg_class
entry becomes owned by a non-existant user which is bogus.
To fix, create a new routine AlterTypeOwner_oid which knows whether to
pass the request to ATExecChangeOwner or deal with it directly, and use
that in shdepReassignOwner rather than calling AlterTypeOwnerInternal
directly. AlterTypeOwnerInternal is now simpler in that it only
modifies the pg_type entry and recurses to handle a possible array type;
higher-level tasks are handled by either AlterTypeOwner directly or
AlterTypeOwner_oid.
I took the opportunity to add a few more objects to the test rig for
REASSIGN OWNED, so that more cases are exercised. Additional ones could
be added for superuser-only-ownable objects (such as FDWs and event
triggers) but I didn't want to push my luck by adding a new superuser to
the tests on a backpatchable bug fix.
Per bug #13666 reported by Chris Pacejo.
This is a backpatch of commit 756e7b4c9d to branches 9.1 -- 9.4.
When REASSIGN and ALTER OWNER TO are used, both the object owner and ACL
list should be changed from the old owner to the new owner. This patch
fixes types, foreign data wrappers, and foreign servers to change their
ACL list properly; they already changed owners properly.
Report by Alexey Bashtanov
This is a backpatch of commit 59367fdf97 (for bug #9923) by Bruce
Momjian to branches 9.1 - 9.4; it wasn't backpatched originally out of
concerns that it would create a backwards compatibility problem, but per
discussion related to bug #13666 that turns out to have been misguided.
(Therefore, the entry in the 9.5 release notes should be removed.)
Note that 9.1 didn't have privileges on types (which were introduced by
commit 729205571e), so this commit only changes foreign-data related
objects in that branch.
Discussion: http://www.postgresql.org/message-id/20151216224004.GL2618@alvherre.pgsqlhttp://www.postgresql.org/message-id/10227.1450373793@sss.pgh.pa.us
Rather than expect the Query returned by get_view_query() to be
read-only and then copy bits and pieces of it out, simply copy the
entire structure when we get it. This addresses an issue where
AcquireRewriteLocks, which is called by acquireLocksOnSubLinks(),
scribbles on the parsetree passed in, which was actually an entry
in relcache, leading to segfaults with certain view definitions.
This also future-proofs us a bit for anyone adding more code to this
path.
The acquireLocksOnSubLinks() was added in commit c3e0ddd40.
Back-patch to 9.3 as that commit was.
psql offered USING, WHERE, and SET in this context, but SET is not a valid
possibility here. Seems to have been a thinko in commit f5ab0a14ea
which added DELETE's USING option.
Turns out we must set rl_basic_word_break_characters *before* we call
rl_initialize() the first time, because it will quietly copy that value
elsewhere --- but only on the first call. (Love these undocumented
dependencies.) I broke this yesterday in commit 2ec477dc8108339d;
like that commit, back-patch to all active branches. Per report from
Pavel Stehule.
It emerges that libreadline doesn't notice terminal window size change
events unless they occur while collecting input. This is easy to stumble
over if you resize the window while using a pager to look at query output,
but it can be demonstrated without any pager involvement. The symptom is
that queries exceeding one line are misdisplayed during subsequent input
cycles, because libreadline has the wrong idea of the screen dimensions.
The safest, simplest way to fix this is to call rl_reset_screen_size()
just before calling readline(). That causes an extra ioctl(TIOCGWINSZ)
for every command; but since it only happens when reading from a tty, the
performance impact should be negligible. A more valid objection is that
this still leaves a tiny window during entry to readline() wherein delivery
of SIGWINCH will be missed; but the practical consequences of that are
probably negligible. In any case, there doesn't seem to be any good way to
avoid the race, since readline exposes no functions that seem safe to call
from a generic signal handler --- rl_reset_screen_size() certainly isn't.
It turns out that we also need an explicit rl_initialize() call, else
rl_reset_screen_size() dumps core when called before the first readline()
call.
rl_reset_screen_size() is not present in old versions of libreadline,
so we need a configure test for that. (rl_initialize() is present at
least back to readline 4.0, so we won't bother with a test for it.)
We would need a configure test anyway since libedit's emulation of
libreadline doesn't currently include such a function. Fortunately,
libedit seems not to have any corresponding bug.
Merlin Moncure, adjusted a bit by me
Apparently, there are bugs in this code that cause it to loop endlessly.
That bug still needs more research, but in the meantime it's clear that
the loop is missing a check for interrupts so that it can be cancelled
timely.
Backpatch to 9.1 -- this has been missing since 49475aab8d.
If libpq ran out of memory while constructing the result set, it would hang,
waiting for more data from the server, which might never arrive. To fix,
distinguish between out-of-memory error and not-enough-data cases, and give
a proper error message back to the client on OOM.
There are still similar issues in handling COPY start messages, but let's
handle that as a separate patch.
Michael Paquier, Amit Kapila and me. Backpatch to all supported versions.
e3f4cfc7 introduced a LWLockHeldByMe() call, without the corresponding
Assert() surrounding it.
Spotted by Coverity.
Backpatch: 9.1+, like the previous commit
This has worked that way for a long time, maybe always, but you would
not have known it from the documentation. Also back-patch the notes
I added to HEAD earlier today about behavior of the "-f -" switch,
which likewise have been valid for many releases.
These would leak random xlog positions if a walsender used for backup would
a walsender slot previously used by a replication walsender.
In passing also fix a couple of cases where the xlog pointer is directly
compared to zero instead of using XLogRecPtrIsInvalid, noted by
Michael Paquier.
Changing the tablespace of an unlogged relation did not WAL log the
creation and content of the init fork. Thus, after a standby is
promoted, unlogged relation cannot be accessed anymore, with errors
like:
ERROR: 58P01: could not open file "pg_tblspc/...": No such file or directory
Additionally the init fork was not synced to disk, independent of the
configured wal_level, a relatively small durability risk.
Investigation of that problem also brought to light that, even for
permanent relations, the creation of !main forks was not WAL logged,
i.e. no XLOG_SMGR_CREATE record were emitted. That mostly turns out not
to be a problem, because these files were created when the actual
relation data is copied; nonexistent files are not treated as an error
condition during replay. But that doesn't work for empty files, and
generally feels a bit haphazard. Luckily, outside init and main forks,
empty forks don't occur often or are not a problem.
Add the required WAL logging and syncing to disk.
Reported-By: Michael Paquier
Author: Michael Paquier and Andres Freund
Discussion: 20151210163230.GA11331@alap3.anarazel.de
Backpatch: 9.1, where unlogged relations were introduced
Recent releases of libxml2 do not provide error context reports for errors
detected at the very end of the input string. This appears to be a bug, or
at least an infelicity, introduced by the fix for libxml2's CVE-2015-7499.
We can hope that this behavioral change will get undone before too long;
but the security patch is likely to spread a lot faster/further than any
follow-on cleanup, which means this behavior is likely to be present in the
wild for some time to come. As a stopgap, add a variant regression test
expected-file that matches what you get with a libxml2 that acts this way.
As reported in bug #13809 by Alexander Ashurkov, the code for REASSIGN
OWNED hadn't gotten word about user mappings. Deal with them in the
same way default ACLs do, which is to ignore them altogether; they are
handled just fine by DROP OWNED. The other foreign object cases are
already handled correctly by both commands.
Also add a REASSIGN OWNED statement to foreign_data test to exercise the
foreign data objects. (The changes are just before the "cleanup" phase,
so it shouldn't remove any existing live test.)
Reported by Alexander Ashurkov, then independently by Jaime Casanova.
More fuzz testing by Andreas Seltenreich exposed that the planner did not
cope well with chains of lateral references. If relation X references Y
laterally, and Y references Z laterally, then we will have to scan X on the
inside of a nestloop with Z, so for all intents and purposes X is laterally
dependent on Z too. The planner did not understand this and would generate
intermediate joins that could not be used. While that was usually harmless
except for wasting some planning cycles, under the right circumstances it
would lead to "failed to build any N-way joins" or "could not devise a
query plan" planner failures.
To fix that, convert the existing per-relation lateral_relids and
lateral_referencers relid sets into their transitive closures; that is,
they now show all relations on which a rel is directly or indirectly
laterally dependent. This not only fixes the chained-reference problem
but allows some of the relevant tests to be made substantially simpler
and faster, since they can be reduced to simple bitmap manipulations
instead of searches of the LateralJoinInfo list.
Also, when a PlaceHolderVar that is due to be evaluated at a join contains
lateral references, we should treat those references as indirect lateral
dependencies of each of the join's base relations. This prevents us from
trying to join any individual base relations to the lateral reference
source before the join is formed, which again cannot work.
Andreas' testing also exposed another oversight in the "dangerous
PlaceHolderVar" test added in commit 85e5e222b1. Simply rejecting
unsafe join paths in joinpath.c is insufficient, because in some cases
we will end up rejecting *all* possible paths for a particular join, again
leading to "could not devise a query plan" failures. The restriction has
to be known also to join_is_legal and its cohort functions, so that they
will not select a join for which that will happen. I chose to move the
supporting logic into joinrels.c where the latter functions are.
Back-patch to 9.3 where LATERAL support was introduced.
At the end of crash recovery, unlogged relations are reset to the empty
state, using their init fork as the template. The init fork is copied to
the main fork without going through shared buffers. Unfortunately WAL
replay so far has not necessarily flushed writes from shared buffers to
disk at that point. In normal crash recovery, and before the
introduction of 'fast promotions' in fd4ced523 / 9.3, the
END_OF_RECOVERY checkpoint flushes the buffers out in time. But with
fast promotions that's not the case anymore.
To fix, force WAL writes targeting the init fork to be flushed
immediately (using the new FlushOneBuffer() function). In 9.5+ that
flush can centrally be triggered from the code dealing with restoring
full page writes (XLogReadBufferForRedoExtended), in earlier releases
that responsibility is in the hands of XLOG_HEAP_NEWPAGE's replay
function.
Backpatch to 9.1, even if this currently is only known to trigger in
9.3+. Flushing earlier is more robust, and it is advantageous to keep
the branches similar.
Typical symptoms of this bug are errors like
'ERROR: index "..." contains unexpected zero page at block 0'
shortly after promoting a node.
Reported-By: Thom Brown
Author: Andres Freund and Michael Paquier
Discussion: 20150326175024.GJ451@alap3.anarazel.de
Backpatch: 9.1-
