If ANALYZE found no repeated non-null entries in its sample, it set the
column's stadistinct value to -1.0, intending to indicate that the entries
are all distinct. But what this value actually means is that the number
of distinct values is 100% of the table's rowcount, and thus it was
overestimating the number of distinct values by however many nulls there
are. This could lead to very poor selectivity estimates, as for example
in a recent report from Andreas Joseph Krogh. We should discount the
stadistinct value by whatever we've estimated the nulls fraction to be.
(That is what will happen if we choose to use a negative stadistinct for
a column that does have repeated entries, so this code path was just
inconsistent.)
In addition to fixing the stadistinct entries stored by several different
ANALYZE code paths, adjust the logic where get_variable_numdistinct()
forces an "all distinct" estimate on the basis of finding a relevant unique
index. Unique indexes don't reject nulls, so there's no reason to assume
that the null fraction doesn't apply.
Back-patch to all supported branches. Back-patching is a bit of a judgment
call, but this problem seems to affect only a few users (else we'd have
identified it long ago), and it's bad enough when it does happen that
destabilizing plan choices in a worse direction seems unlikely.
Patch by me, with documentation wording suggested by Dean Rasheed
Report: <VisenaEmail.26.df42f82acae38a58.156463942b8@tc7-visena>
Discussion: <16143.1470350371@sss.pgh.pa.us>
This oversight could cause logical decoding to fail to decode an outer
transaction containing changes, if a subtransaction had an XID but no
actual changes. Per bug #14279 from Marko Tiikkaja. Patch by Marko
based on analysis by Andrew Gierth.
Discussion: <20160804191757.1430.39011@wrigleys.postgresql.org>
In _bt_unlink_halfdead_page(), we might fail to find an immediate left
sibling of the target page, perhaps because of corruption of the page
sibling links. The code intends to cope with this by just abandoning
the deletion attempt; but what actually happens is that it fails outright
due to releasing the same buffer lock twice. (And error recovery masks
a second problem, which is possible leakage of a pin on another page.)
Seems to have been introduced by careless refactoring in commit efada2b8e.
Since there are multiple cases to consider, let's make releasing the buffer
lock in the failure case the responsibility of _bt_unlink_halfdead_page()
not its caller.
Also, avoid fetching the leaf page's left-link again after we've dropped
lock on the page. This is probably harmless, but it's not exactly good
coding practice.
Per report from Kyotaro Horiguchi. Back-patch to 9.4 where the faulty code
was introduced.
Discussion: <20160803.173116.111915228.horiguchi.kyotaro@lab.ntt.co.jp>
Beginning with the next development cycle, PG servers will report two-part
not three-part version numbers. Fix libpq so that it will compute the
correct numeric representation of such server versions for reporting by
PQserverVersion(). It's desirable to get this into the field and
back-patched ASAP, so that older clients are more likely to understand the
new server version numbering by the time any such servers are in the wild.
(The results with an old client would probably not be catastrophic anyway
for a released server; for example "10.1" would be interpreted as 100100
which would be wrong in detail but would not likely cause an old client to
misbehave badly. But "10devel" or "10beta1" would result in sversion==0
which at best would result in disabling all use of modern features.)
Extracted from a patch by Peter Eisentraut; comments added by me
Patch: <802ec140-635d-ad86-5fdf-d3af0e260c22@2ndquadrant.com>
In 9.5, two arguments were introduced into pg_replication_origin_xact_reset()
by mistake while they are actually not used at all. We cannot fix this issue
for 9.5 anymore because it needs a catalog version bump. Instead, we add
a note about those unused arguments into the document.
Reviewed-By: Andres Freund
DST law changes in Kemerovo and Novosibirsk. Historical corrections for
Azerbaijan, Belarus, and Morocco. Asia/Novokuznetsk and Asia/Novosibirsk
now use numeric time zone abbreviations instead of invented ones. Zones
for Antarctic bases and other locations that have been uninhabited for
portions of the time span known to the tzdata database now report "-00"
rather than "zzz" as the zone abbreviation for those time spans.
Also, I decided to remove some of the timezone/data/ files that we don't
use. At one time that subdirectory was a complete copy of what IANA
distributes in the tzdata tarballs, but that hasn't been true for a long
time. There seems no good reason to keep shipping those specific files
but not others; they're just bloating our tarballs.
Some conditions resulted in "return" directly out of a PG_TRY block,
which left the exception stack dangling, and to add insult to injury
failed to restore the state of set_latch_on_sigusr1.
This is a bug only in 9.5; in HEAD it was accidentally fixed by commit
db0f6cad4, which removed the surrounding PG_TRY block. However, I (tgl)
chose to apply the patch to HEAD as well, because the old coding was
gratuitously different from WaitForBackgroundWorkerStartup(), and there
would indeed have been no bug if it were done like that to start with.
Dmitry Ivanov
Discussion: <1637882.WfYN5gPf1A@abook>
This is apparently harmless on Windows, but on Unix it results in an
assertion failure. We'd not noticed because this code doesn't get
used on Unix unless you build with -DEXEC_BACKEND. Bug was evidently
introduced by sloppy refactoring in commit 31c453165.
Thomas Munro
Discussion: <CAEepm=1VOnbVx4wsgQFvj94hu9jVt2nVabCr7QiooUSvPJXkgQ@mail.gmail.com>
As noted by Alvaro, there are CHECK_FOR_INTERRUPTS() calls in the shm_mq.c
functions called by HandleParallelMessages(). I believe they're all
unreachable since we always pass nowait = true, but it doesn't seem like
a great idea to assume that no such call will ever be reachable from
HandleParallelMessages(). If that did happen, there would be a risk of a
recursive call to HandleParallelMessages(), which it does not appear to be
designed for --- for example, there's nothing that would prevent
out-of-order processing of received messages. And certainly such cases
cannot easily be tested. So let's prevent it by holding off interrupts for
the duration of the function. Back-patch to 9.5 which contains identical
code.
Discussion: <14869.1470083848@sss.pgh.pa.us>
This back-patches commit a5fe473ad (notably, marking ParallelMessagePending
as volatile, which is not particularly optional). I also back-patched some
previous cosmetic changes to remove unnecessary diffs between the two
branches. I'm unsure how much of this code is actually reachable in 9.5,
but to the extent that it is reachable, it needs to be maintained, and
minimizing cross-branch diffs will make that easier.
Since -c plus -C requests dropping and recreating the target database
as a whole, not dropping individual objects in it, we should assume that
the public schema already exists and need not be created. The previous
coding considered only the state of the -c option, so it would emit
"CREATE SCHEMA public" anyway, leading to an unexpected error in restore.
Back-patch to 9.2. Older versions did not accept -c with -C so the
issue doesn't arise there. (The logic being patched here dates to 8.0,
cf commit 2193121fa, so it's not really wrong that it didn't consider
the case at the time.)
Note that versions before 9.6 will still attempt to emit REVOKE/GRANT
on the public schema; but that happens without -c/-C too, and doesn't
seem to be the focus of this complaint. I considered extending this
stanza to also skip the public schema's ACL, but that would be a
misfeature, as it'd break cases where users intentionally changed that
ACL. The real fix for this aspect is Stephen Frost's work to not dump
built-in ACLs, and that's not going to get back-ported.
Per bugs #13804 and #14271. Solution found by David Johnston and later
rediscovered by me.
Report: <20151207163520.2628.95990@wrigleys.postgresql.org>
Report: <20160801021955.1430.47434@wrigleys.postgresql.org>
This coding pattern creates a race condition, because if an interesting
interrupt happens after we've checked InterruptPending but before we reset
our latch, the latch-setting done by the signal handler would get lost,
and then we might block at WaitLatch in the next iteration without ever
noticing the interrupt condition. You can put the CHECK_FOR_INTERRUPTS
before WaitLatch or after ResetLatch, but not between them.
Aside from fixing the bugs, add some explanatory comments to latch.h
to perhaps forestall the next person from making the same mistake.
In HEAD, also replace gather_readnext's direct call of
HandleParallelMessages with CHECK_FOR_INTERRUPTS. It does not seem clean
or useful for this one caller to bypass ProcessInterrupts and go straight
to HandleParallelMessages; not least because that fails to consider the
InterruptPending flag, resulting in useless work both here
(if InterruptPending isn't set) and in the next CHECK_FOR_INTERRUPTS call
(if it is).
This thinko seems to have been introduced in the initial coding of
storage/ipc/shm_mq.c (commit ec9037df2), and then blindly copied into all
the subsequent parallel-query support logic. Back-patch relevant hunks
to 9.4 to extirpate the error everywhere.
Discussion: <1661.1469996911@sss.pgh.pa.us>
The help message for pg_basebackup specifies that the numbers 0 through 9
are accepted as valid values of -Z option. But, previously -Z 0 was rejected
as an invalid compression level.
Per discussion, it's better to make pg_basebackup treat 0 as valid
compression level meaning no compression, like pg_dump.
Back-patch to all supported versions.
Reported-By: Jeff Janes
Reviewed-By: Amit Kapila
Discussion: CAMkU=1x+GwjSayc57v6w87ij6iRGFWt=hVfM0B64b1_bPVKRqg@mail.gmail.com
This text was added by commit ff213239c, and not long thereafter obsoleted
by commit 4adc2f72a (which made the test depend on NBuffers instead); but
nobody noticed the need for an update. Commit 9563d5b5e adds some further
dependency on maintenance_work_mem, but the existing verbiage seems to
cover that with about as much precision as we really want here. Let's
just take it all out rather than leaving ourselves open to more errors of
omission in future. (That solution makes this change back-patchable, too.)
Noted by Peter Geoghegan.
Discussion: <CAM3SWZRVANbj9GA9j40fAwheQCZQtSwqTN1GBTVwRrRbmSf7cg@mail.gmail.com>
An evident copy-and-pasteo in commit 2bd9e412f broke the non-blocking
aspect of pq_putmessage_noblock(), causing it to behave identically to
pq_putmessage(). That function is nowadays used only in walsender.c,
so that the net effect was to cause walsenders to hang up waiting for
the receiver in situations where they should not.
Kyotaro Horiguchi
Patch: <20160728.185228.58375982.horiguchi.kyotaro@lab.ntt.co.jp>
Per the fgets() specification, it cannot return without reading some data
unless it reports EOF or error. So the code here assumed that the data
buffer would necessarily be nonempty when we go to check for a newline
having been read. However, Agostino Sarubbo noticed that this could fail
to be true if the first byte of the data is a NUL (\0). The fgets() API
doesn't really work for embedded NULs, which is something I don't feel
any great need for us to worry about since we generally don't allow NULs
in SQL strings anyway. But we should not access off the end of our own
buffer if the case occurs. Normally this would just be a harmless read,
but if you were unlucky the byte before the buffer would contain '\n'
and we'd overwrite it with '\0', and if you were really unlucky that
might be valuable data and psql would crash.
Agostino reported this to pgsql-security, but after discussion we concluded
that it isn't worth treating as a security bug; if you can control the
input to psql you can do far more interesting things than just maybe-crash
it. Nonetheless, it is a bug, so back-patch to all supported versions.
Commits 4452000f3 et al established semantics for NullTest.argisrow that
are a bit different from its initial conception: rather than being merely
a cache of whether we've determined the input to have composite type,
the flag now has the further meaning that we should apply field-by-field
testing as per the standard's definition of IS [NOT] NULL. If argisrow
is false and yet the input has composite type, the construct instead has
the semantics of IS [NOT] DISTINCT FROM NULL. Update the comments in
primnodes.h to clarify this, and fix ruleutils.c and deparse.c to print
such cases correctly. In the case of ruleutils.c, this merely results in
cosmetic changes in EXPLAIN output, since the case can't currently arise
in stored rules. However, it represents a live bug for deparse.c, which
would formerly have sent a remote query that had semantics different
from the local behavior. (From the user's standpoint, this means that
testing a remote nested-composite column for null-ness could have had
unexpected recursive behavior much like that fixed in 4452000f3.)
In a related but somewhat independent fix, make plancat.c set argisrow
to false in all NullTest expressions constructed to represent "attnotnull"
constructs. Since attnotnull is actually enforced as a simple null-value
check, this is a more accurate representation of the semantics; we were
previously overpromising what it meant for composite columns, which might
possibly lead to incorrect planner optimizations. (It seems that what the
SQL spec expects a NOT NULL constraint to mean is an IS NOT NULL test, so
arguably we are violating the spec and should fix attnotnull to do the
other thing. If we ever do, this part should get reverted.)
Back-patch, same as the previous commit.
Discussion: <10682.1469566308@sss.pgh.pa.us>
The docs failed to explain that LIKE INCLUDING INDEXES would not preserve
the names of indexes and associated constraints. Also, it wasn't mentioned
that EXCLUDE constraints would be copied by this option. The latter
oversight seems enough of a documentation bug to justify back-patching.
In passing, do some minor copy-editing in the same area, and add an entry
for LIKE under "Compatibility", since it's not exactly a faithful
implementation of the standard's feature.
Discussion: <20160728151154.AABE64016B@smtp.hushmail.com>
start_postmaster() registered stop_postmaster_atexit as an atexit(3)
callback each time through, although the obvious intention was to do
so only once per program run. The extra registrations were harmless,
so long as we didn't exceed ATEXIT_MAX, but still it's a bug.
Artur Zakirov, with bikeshedding by Kyotaro Horiguchi and me
Discussion: <d279e817-02b5-caa6-215f-cfb05dce109a@postgrespro.ru>
The description of udt_privileges view contained an incorrect copy-pasted word.
Back-patch to 9.2 where udt_privileges view was added.
Author: Alexander Law
The SQL standard appears to specify that IS [NOT] NULL's tests of field
nullness are non-recursive, ie, we shouldn't consider that a composite
field with value ROW(NULL,NULL) is null for this purpose.
ExecEvalNullTest got this right, but eval_const_expressions did not,
leading to weird inconsistencies depending on whether the expression
was such that the planner could apply constant folding.
Also, adjust the docs to mention that IS [NOT] DISTINCT FROM NULL can be
used as a substitute test if a simple null check is wanted for a rowtype
argument. That motivated reordering things so that IS [NOT] DISTINCT FROM
is described before IS [NOT] NULL. In HEAD, I went a bit further and added
a table showing all the comparison-related predicates.
Per bug #14235. Back-patch to all supported branches, since it's certainly
undesirable that constant-folding should change the semantics.
Report and patch by Andrew Gierth; assorted wordsmithing and revised
regression test cases by me.
Report: <20160708024746.1410.57282@wrigleys.postgresql.org>
Welsh (cy_GB) apparently sorts 'dd' after 'f', creating problems
analogous to the odd sorting of 'aa' in Danish. Adjust regression
test case to not use data that provokes that.
Jeff Janes
Patch: <CAMkU=1zx-pqcfSApL2pYDQurPOCfcYU0wJorsmY1OrYPiXRbLw@mail.gmail.com>
In btree_gin and citext, avoid some not-particularly-interesting
dependencies on the sorting of 'aa'. In tsearch2, use COLLATE "C" to
remove an uninteresting dependency on locale sort order (and thereby
allow removal of a variant expected-file).
Also, in citext, avoid assuming that lower('I') = 'i'. This isn't relevant
to Danish but it does fail in Turkish.
Another peculiarity of Danish locale is that it has an unusual idea
of how to sort upper vs. lower case. One of the pltcl test cases has
an issue with that. Now that COLLATE works in all supported branches,
we can just change the test to be locale-independent, and get rid of
the variant expected file that used to support non-C locales.
Some tests added in 9.5 depended on 'aa' sorting before 'bb', which
doesn't hold true in Danish. Use slightly different test data to
avoid the problem.
Jeff Janes
Report: <CAMkU=1w-cEDbA+XHdNb=YS_4wvZbs66Ni9KeSJKAJGNJyOsgQw@mail.gmail.com>
runtime.sgml used to contain a table of estimated shared memory consumption
rates for max_connections and some other GUCs. Commit 390bfc643 removed
that on the well-founded grounds that (a) we weren't maintaining the
entries well and (b) it no longer mattered so much once we got out from
under SysV shmem limits. But it missed that there were even-more-obsolete
versions of some of those numbers in comments in postgresql.conf.sample.
Remove those too. Back-patch to 9.3 where the aforesaid commit went in.
Back-patch commit 1c1a7cbd6a, along with subsequent portability
fixes, into all active branches. Also, back-patch commits 696027727 and
596857043 (addition of zic -P option) into 9.1 and 9.2, just to reduce
differences between the branches. src/timezone/ is now largely identical
in all active branches, except that in 9.1, pgtz.c retains the
initial-timezone-selection code that was moved over to initdb in 9.2.
Ordinarily we wouldn't risk this much code churn in back branches, but it
seems necessary in this case, because among the changes are two feature
additions in the "zic" zone data file compiler (a larger limit on the
number of allowed DST transitions, and addition of a "%z" escape in zone
abbreviations). IANA have not yet started to use those features in their
tzdata files, but presumably they will before too long. If we don't update
then we'll be unable to adopt new timezone data. Also, installations built
with --with-system-tzdata (which includes most distro-supplied builds, I
believe) might fail even if we don't update our copies of the data files.
There are assorted bug fixes too, mostly affecting obscure timezones or
post-2037 dates.
Discussion: <13601.1468868947@sss.pgh.pa.us>
9.4 added a second description of GET DIAGNOSTICS that was totally
independent of the existing one, resulting in each description lying to the
extent that it claimed the set of status items it described was complete.
Fix that, and do some minor markup improvement.
Also some other small fixes per bug #14258 from Dilian Palauzov.
Discussion: <20160718181437.1414.40802@wrigleys.postgresql.org>
The old code used SEQ_MINVALUE to get the smallest int64 value. This
was done as a convenience to avoid having to deal with INT64_IS_BUSTED,
but that is obsolete now. Also, it is incorrect because the smallest
int64 value is actually SEQ_MINVALUE-1. Fix by using PG_INT64_MIN.
The Assert() here seems unreasonably optimistic. Andreas Seltenreich
found that it could fail with NaNs in the input geometries, and it
seems likely to me that it might fail in corner cases due to roundoff
error, even for ordinary input values. As a band-aid, make the function
return SQL NULL instead of crashing.
Report: <87d1md1xji.fsf@credativ.de>
When heap_update needs to look for a page for the new tuple version,
because the current one doesn't have sufficient free space, or when
columns have to be processed by the tuple toaster, it has to release the
lock on the old page during that. Otherwise there'd be lock ordering and
lock nesting issues.
To avoid concurrent sessions from trying to update / delete / lock the
tuple while the page's content lock is released, the tuple's xmax is set
to the current session's xid.
That unfortunately was done without any WAL logging, thereby violating
the rule that no XIDs may appear on disk, without an according WAL
record. If the database were to crash / fail over when the page level
lock is released, and some activity lead to the page being written out
to disk, the xid could end up being reused; potentially leading to the
row becoming invisible.
There might be additional risks by not having t_ctid point at the tuple
itself, without having set the appropriate lock infomask fields.
To fix, compute the appropriate xmax/infomask combination for locking
the tuple, and perform WAL logging using the existing XLOG_HEAP_LOCK
record. That allows the fix to be backpatched.
This issue has existed for a long time. There appears to have been
partial attempts at preventing dangers, but these never have fully been
implemented, and were removed a long time ago, in
11919160 (cf. HEAP_XMAX_UNLOGGED).
In master / 9.6, there's an additional issue, namely that the
visibilitymap's freeze bit isn't reset at that point yet. Since that's a
new issue, introduced only in a892234f83, that'll be fixed in a
separate commit.
Author: Masahiko Sawada and Andres Freund
Reported-By: Different aspects by Thomas Munro, Noah Misch, and others
Discussion: CAEepm=3fWAbWryVW9swHyLTY4sXVf0xbLvXqOwUoDiNCx9mBjQ@mail.gmail.com
Backpatch: 9.1/all supported versions
0ac5ad5 started to compress infomask bits in WAL records. Unfortunately
the replay routines for XLOG_HEAP_LOCK/XLOG_HEAP2_LOCK_UPDATED forgot to
reset the HEAP_XMAX_INVALID (and some other) hint bits.
Luckily that's not problematic in the majority of cases, because after a
crash/on a standby row locks aren't meaningful. Unfortunately that does
not hold true in the presence of prepared transactions. This means that
after a crash, or after promotion, row level locks held by a prepared,
but not yet committed, prepared transaction might not be enforced.
Discussion: 20160715192319.ubfuzim4zv3rqnxv@alap3.anarazel.de
Backpatch: 9.3, the oldest branch on which 0ac5ad5 is present.
When key-share locking a tuple that has been not-key-updated, and the
update is a committed transaction, in some cases we raised
serializability errors:
ERROR: could not serialize access due to concurrent update
Because the key-share doesn't conflict with the update, the error is
unnecessary and inconsistent with the case that the update hasn't
committed yet. This causes problems for some usage patterns, even if it
can be claimed that it's sufficient to retry the aborted transaction:
given a steady stream of updating transactions and a long locking
transaction, the long transaction can be starved indefinitely despite
multiple retries.
To fix, we recognize that HeapTupleSatisfiesUpdate can return
HeapTupleUpdated when an updating transaction has committed, and that we
need to deal with that case exactly as if it were a non-committed
update: verify whether the two operations conflict, and if not, carry on
normally. If they do conflict, however, there is a difference: in the
HeapTupleBeingUpdated case we can just sleep until the concurrent
transaction is gone, while in the HeapTupleUpdated case this is not
possible and we must raise an error instead.
Per trouble report from Olivier Dony.
In addition to a couple of test cases that verify the changed behavior,
I added a test case to verify the behavior that remains unchanged,
namely that errors are raised when a update that modifies the key is
used. That must still generate serializability errors. One
pre-existing test case changes behavior; per discussion, the new
behavior is actually the desired one.
Discussion: https://www.postgresql.org/message-id/560AA479.4080807@odoo.comhttps://www.postgresql.org/message-id/20151014164844.3019.25750@wrigleys.postgresql.org
Backpatch to 9.3, where the problem appeared.
GiST index build could go into an infinite loop when presented with boxes
(or points, circles or polygons) containing NaN component values. This
happened essentially because the code assumed that x == x is true for any
"double" value x; but it's not true for NaNs. The looping behavior was not
the only problem though: we also attempted to sort the items using simple
double comparisons. Since NaNs violate the trichotomy law, qsort could
(in principle at least) get arbitrarily confused and mess up the sorting of
ordinary values as well as NaNs. And we based splitting choices on box size
calculations that could produce NaNs, again resulting in undesirable
behavior.
To fix, replace all comparisons of doubles in this logic with
float8_cmp_internal, which is NaN-aware and is careful to sort NaNs
consistently, higher than any non-NaN. Also rearrange the box size
calculation to not produce NaNs; instead it should produce an infinity
for a box with NaN on one side and not-NaN on the other.
I don't by any means claim that this solves all problems with NaNs in
geometric values, but it should at least make GiST index insertion work
reliably with such data. It's likely that the index search side of things
still needs some work, and probably regular geometric operations too.
But with this patch we're laying down a convention for how such cases
ought to behave.
Per bug #14238 from Guang-Dih Lei. Back-patch to 9.2; the code used before
commit 7f3bd86843 is quite different and doesn't lock up on my simple
test case, nor on the submitter's dataset.
Report: <20160708151747.1426.60150@wrigleys.postgresql.org>
Discussion: <28685.1468246504@sss.pgh.pa.us>
Since IMPORT FOREIGN SCHEMA has an INTO clause, pl/pgsql needs to be
aware of that and avoid capturing the INTO as an INTO-variables clause.
This isn't hard, though it's annoying to have to make IMPORT a plpgsql
keyword just for this. (Fortunately, we have the infrastructure now
to make it an unreserved keyword, so at least this shouldn't break any
existing pl/pgsql code.)
Per report from Merlin Moncure. Back-patch to 9.5 where IMPORT FOREIGN
SCHEMA was introduced.
Report: <CAHyXU0wpHf2bbtKGL1gtUEFATCY86r=VKxfcACVcTMQ70mCyig@mail.gmail.com>
Change assorted places in our Perl code that did things like
system("prog $path/file");
to do it more like
system('prog', "$path/file");
which is safe against spaces and other special characters in the path
variable. The latter was already the prevailing style, but a few bits
of code hadn't gotten this memo. Back-patch to 9.4 as relevant.
Michael Paquier, Kyotaro Horiguchi
Discussion: <20160704.160213.111134711.horiguchi.kyotaro@lab.ntt.co.jp>