Detect fclose() failures; given "ln -s /dev/full $PGDATA/devfull",
"pg_file_write('devfull', 'x', true)" now fails as it should. Don't
leak a stream when fwrite() fails. Remove a born-ineffective test that
aimed to skip zero-length writes. Back-patch to 9.2 (all supported
versions).
Commits 89e0bac86 et al replaced newlines with spaces in object names
printed in SQL comments, but we neglected to consider that the same
names are also printed by "pg_restore -l", and a newline would render
the output unparseable by "pg_restore -L". Apply the same replacement
in "-l" output. Since "pg_restore -L" doesn't actually examine any
object names, only the dump ID field that starts each line, this is
enough to fix things for its purposes.
The previous fix was treated as a security issue, and we might have
done that here as well, except that the issue was reported publicly
to start with. Anyway it's hard to see how this could be exploited
for SQL injection; "pg_restore -L" doesn't do much with the file
except parse it for leading integers.
Per bug #14587 from Milos Urbanek. Back-patch to all supported versions.
Discussion: https://postgr.es/m/20170310155318.1425.30483@wrigleys.postgresql.org
The IANA timezone crew continues to chip away at their project of removing
timezone abbreviations that have no real-world currency from their
database. The tzdata2017a update removes all such abbreviations for
South American zones, as well as much of the Pacific. This breaks some
test cases in timestamptz.sql that were expecting America/Santiago and
America/Caracas to have non-numeric abbreviations.
The test cases involving America/Santiago seem to have selected that
zone more or less at random, so just replace it with America/New_York,
which is of similar longitude. The cases involving America/Caracas are
harder since they were chosen to test a time-varying zone abbreviation
around a point where it changed meaning in the backwards direction.
Fortunately, Europe/Moscow has a similar case in 2014, and the MSK/MSD
abbreviations are well enough attested that IANA seems unlikely to
decide to remove them from the database in future.
With these changes, this regression test should pass when using any IANA
zone database from 2015 or later. One could wish that there were a few
years more daylight on how out-of-date your zone database can be ... but
really the --with-system-tzdata option is only meant for use on platforms
where the zone database is kept up-to-date pretty faithfully, so I do not
think this is a big objection.
Discussion: https://postgr.es/m/6749.1489087470@sss.pgh.pa.us
Large chunks (those too large for any palloc freelist) are managed as
separate blocks. Formerly, realloc'ing or pfree'ing such a chunk required
O(N) time in a context with N blocks, since we had to traipse down the
singly-linked block list to locate the block's predecessor before we could
fix the list links. This can result in O(N^2) runtime in situations where
large numbers of such chunks are manipulated within one context. Cases
like that were not foreseen in the original design of aset.c, and indeed
didn't arise until fairly recently. But such problems can now occur in
reorderbuffer.c and in hash joining, both of which make repeated large
requests without scaling up their request size as they do so, and which
will free their requests in not-necessarily-LIFO order.
To fix, change the block list from singly-linked to doubly-linked.
This adds another 4 or 8 bytes to ALLOC_BLOCKHDRSZ, but that doesn't
seem like unacceptable overhead, since aset.c's blocks are normally
8K or more, and never less than 1K in current practice.
In passing, get rid of some redundant AllocChunkGetPointer() calls in
AllocSetRealloc (the compiler might be smart enough to optimize these
away anyway, but no need to assume that) and improve AllocSetCheck's
checking of block header fields.
Back-patch to 9.4 where reorderbuffer.c appeared. We could take this
further back, but currently there's no evidence that it would be useful.
Discussion: https://postgr.es/m/CAMkU=1x1hvue1XYrZoWk_omG0Ja5nBvTdvgrOeVkkeqs71CV8g@mail.gmail.com
Not only did it not accept --builtin as a synonym for -b, but what it did
accept as a synonym was --tpc-b (huh?), which it got even further wrong
by marking as no_argument, so that if you did try that you got a core
dump. I suppose this is leftover from some early design for the new
switches added by commit 8bea3d221, but it's still pretty sloppy work.
Per bug #14580 from Stepan Pesternikov. Back-patch to 9.6 where the
error was introduced.
Report: https://postgr.es/m/20170307123347.25054.73207@wrigleys.postgresql.org
pg_dump has always handled the public schema in a special way when it
comes to the "--clean" option. To wit, we do not drop or recreate the
public schema in "normal" mode, but when we are run in "--clean" mode
then we do drop and recreate the public schema.
When running in "--clean" mode, the public schema is dropped and then
recreated and it is recreated with the normal schema-default privileges
of "nothing". This is unlike how the public schema starts life, which
is to have CREATE and USAGE GRANT'd to the PUBLIC role, and that is what
is recorded in pg_init_privs.
Due to this, in "--clean" mode, pg_dump would mistakenly only dump out
the set of privileges required to go from the initdb-time privileges on
the public schema to whatever the current-state privileges are. If the
privileges were not changed from initdb time, then no privileges would
be dumped out for the public schema, but with the schema being dropped
and recreated, the result was that the public schema would have no ACLs
on it instead of what it should have, which is the initdb-time
privileges.
Practically speaking, this meant that pg_dump with --clean mode dumping
a database where the ACLs on the public schema were not changed from the
default would, upon restore, result in a public schema with *no*
privileges GRANT'd, not matching the state of the existing database
(where the initdb-time privileges would have been CREATE and USAGE to
the PUBLIC role for the public schema).
To fix, adjust the query in getNamespaces() to ignore the pg_init_privs
entry for the public schema when running in "--clean" mode, meaning that
the privileges for the public schema would be dumped, correctly, as if
it was going from a newly-created schema to the current state (which is,
indeed, what will happen during the restore thanks to the DROP/CREATE).
Only the public schema is handled in this special way by pg_dump, no
other initdb-time objects are dropped/recreated in --clean mode.
Back-patch to 9.6 where the bug was introduced.
Discussion: https://postgr.es/m/3534542.o3cNaKiDID%40techfox
We attached no schema label to comments for procedural languages, casts,
transforms, operator classes, operator families, or text search objects.
The first three categories of objects don't really have schemas, but
pg_dump treats them as if they do, and it seems like the TocEntry fields
for their comments had better match the TocEntry fields for the parent
objects. (As an example of a possible hazard, the type names in a CAST
will be formatted with the assumption of a particular search_path, so
failing to ensure that this same path is active for the COMMENT ON command
could lead to an error or to attaching the comment to the wrong cast.)
In the last six cases, this was a flat-out error --- possibly mine to
begin with, but it was a long time ago.
The security label for a procedural language was likewise not correctly
labeled as to schema, and both the comment and security label for a
procedural language were not correctly labeled as to owner.
In simple cases the restore would accidentally work correctly anyway, since
these comments and security labels would normally get emitted right after
the owning object, and so the search path and active user would be correct
anyhow. But it could fail in corner cases; for example a schema-selective
restore would omit comments it should include.
Giuseppe Broccolo noted the oversight, and proposed the correct fix, for
text search dictionary objects; I found the rest by cross-checking other
dumpComment() calls. These oversights are ancient, so back-patch all
the way.
Discussion: https://postgr.es/m/CAFzmHiWwwzLjzwM4x5ki5s_PDMR6NrkipZkjNnO3B0xEpBgJaA@mail.gmail.com
When performing a pg_upgrade, we copy the files behind pg_largeobject
and pg_largeobject_metadata, allowing us to avoid having to dump out and
reload the actual data for large objects and their ACLs.
Unfortunately, that isn't all of the information which can be associated
with large objects. Currently, we also support COMMENTs and SECURITY
LABELs with large objects and these were being silently dropped during a
pg_upgrade as pg_dump would skip everything having to do with a large
object and pg_upgrade only copied the tables mentioned to the new
cluster.
As the file copies happen after the catalog dump and reload, we can't
simply include the COMMENTs and SECURITY LABELs in pg_dump's binary-mode
output but we also have to include the actual large object definition as
well. With the definition, comments, and security labels in the pg_dump
output and the file copies performed by pg_upgrade, all of the data and
metadata associated with large objects is able to be successfully pulled
forward across a pg_upgrade.
In 9.6 and master, we can simply adjust the dump bitmask to indicate
which components we don't want. In 9.5 and earlier, we have to put
explciit checks in in dumpBlob() and dumpBlobs() to not include the ACL
or the data when in binary-upgrade mode.
Adjustments made to the privileges regression test to allow another test
(large_object.sql) to be added which explicitly leaves a large object
with a comment in place to provide coverage of that case with
pg_upgrade.
Back-patch to all supported branches.
Discussion: https://postgr.es/m/20170221162655.GE9812@tamriel.snowman.net
With RLS active, "COPY tab TO ..." failed under -DRELCACHE_FORCE_RELEASE,
and would sometimes fail without that, because it used the relation name
directly from the relcache as part of the parsetree it's building. That
becomes a potentially-dangling pointer as soon as the relcache entry is
closed, a bit further down. Typical symptom if the relcache entry chanced
to get cleared would be "relation does not exist" error with a garbage
relation name, or possibly a core dump; but if you were really truly
unlucky, the COPY might copy from the wrong table.
Per report from Andrew Dunstan that regression tests fail with
-DRELCACHE_FORCE_RELEASE. The core tests now pass for me (but have
not tried "make check-world" yet).
Discussion: https://postgr.es/m/7b52f900-0579-cda9-ae2e-de5da17090e6@2ndQuadrant.com
This reliably fails with -DRELCACHE_FORCE_RELEASE, as reported by
Andrew Dunstan, and could sometimes fail in normal operation, resulting
in a wrong persistence value being used for the transient table.
It's not immediately clear to me what effects that might have beyond
the risk of a crash while accessing OldHeap->rd_rel->relpersistence,
but it's probably not good.
Bug introduced by commit f41872d0c, and made substantially worse by
commit 85b506bbf, which added a second such access significantly
later than the heap_close. I doubt the first reference could fail
in a production scenario, but the second one definitely could.
Discussion: https://postgr.es/m/7b52f900-0579-cda9-ae2e-de5da17090e6@2ndQuadrant.com
Likewise in RestoreSnapshot(). Do so by copying between the user buffer
and a stack buffer of known alignment. Back-patch to 9.6, where this
last applies cleanly. In master, the select_parallel test dies with
SIGBUS on "Oracle Solaris 10 1/13 s10s_u11wos_24a SPARC", building
32-bit with gcc 4.9.2. In 9.6 and 9.5, the buffers in question happen
to be sufficiently-aligned, and this change is mere insurance against
future 9.6 changes or extension code compromising that.
Newer Perl or IPC::Run versions default to appending the filename to string
exceptions, e.g. the exception
psql timed out
is thrown as
psql timed out at /usr/share/perl5/vendor_perl/IPC/Run.pm line 2961.
To handle this, match exceptions with !~ rather than ne.
From: Craig Ringer <craig@2ndquadrant.com>
Reviewed-by: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Previously the pg_upgrade standby upgrade instructions said not to
execute pgcrypto.sql, but it should have referenced the extension
command "CREATE EXTENSION pgcrypto". This patch makes that doc change.
Reported-by: a private bug report
Backpatch-through: 9.4, where standby instructions were added
We have a portable way of writing uint64 constants, but whoever wrote
this macro didn't know about it.
While at it, fix unsafe under-parenthesization of arguments. That might
be moot, because there are already good reasons not to use the macro on
anything more complicated than a simple variable, but it's still poor
practice.
Per buildfarm warnings.
The logic for removing excess trigrams from the result was faulty.
It intends to avoid merging the initial and final states of the NFA,
which is necessary, but in testing whether removal of a specific trigram
would cause that, it failed to consider the combined effects of all the
state merges that that trigram's removal would cause. This could result
in a broken final graph that would never match anything, leading to GIN
or GiST indexscans not finding anything.
To fix, add a "tentParent" field that is used only within this loop,
and set it to show state merges that we are tentatively going to do.
While examining a particular arc, we must chase up through tentParent
links as well as regular parent links (the former can only appear atop
the latter), and we must account for state init/fin flag merges that
haven't actually been done yet.
To simplify the latter, combine the separate init and fin bool fields
into a bitmap flags field. I also chose to get rid of the "children"
state list, which seems entirely inessential.
Per bug #14563 from Alexey Isayko, which the added test cases are based on.
Back-patch to 9.3 where this code was added.
Report: https://postgr.es/m/20170222111446.1256.67547@wrigleys.postgresql.org
Discussion: https://postgr.es/m/8816.1487787594@sss.pgh.pa.us
Walsender uses the local buffers for each outgoing and incoming message.
Previously when creating replication slot, walsender forgot to initialize
one of them and which can cause the segmentation fault error. To fix this
issue, this commit changes walsender so that it always initialize them
before it executes the requested replication command.
Back-patch to 9.4 where replication slot was introduced.
Problem report and initial patch by Stas Kelvich, modified by me.
Report: https://www.postgresql.org/message-id/A1E9CB90-1FAC-4CAD-8DBA-9AA62A6E97C5@postgrespro.ru
Several places in fd.c had badly-thought-through handling of error returns
from lseek() and close(). The fact that those would seldom fail on valid
FDs is probably the reason we've not noticed this up to now; but if they
did fail, we'd get quite confused.
LruDelete and LruInsert actually just Assert'd that lseek never fails,
which is pretty awful on its face.
In LruDelete, we indeed can't throw an error, because that's likely to get
called during error abort and so throwing an error would probably just lead
to an infinite loop. But by the same token, throwing an error from the
close() right after that was ill-advised, not to mention that it would've
left the LRU state corrupted since we'd already unlinked the VFD from the
list. I also noticed that really, most of the time, we should know the
current seek position and it shouldn't be necessary to do an lseek here at
all. As patched, if we don't have a seek position and an lseek attempt
doesn't give us one, we'll close the file but then subsequent re-open
attempts will fail (except in the somewhat-unlikely case that a
FileSeek(SEEK_SET) call comes between and allows us to re-establish a known
target seek position). This isn't great but it won't result in any state
corruption.
Meanwhile, having an Assert instead of an honest test in LruInsert is
really dangerous: if that lseek failed, a subsequent read or write would
read or write from the start of the file, not where the caller expected,
leading to data corruption.
In both LruDelete and FileClose, if close() fails, just LOG that and mark
the VFD closed anyway. Possibly leaking an FD is preferable to getting
into an infinite loop or corrupting the VFD list. Besides, as far as I can
tell from the POSIX spec, it's unspecified whether or not the file has been
closed, so treating it as still open could be the wrong thing anyhow.
I also fixed a number of other places that were being sloppy about
behaving correctly when the seekPos is unknown.
Also, I changed FileSeek to return -1 with EINVAL for the cases where it
detects a bad offset, rather than throwing a hard elog(ERROR). It seemed
pretty inconsistent that some bad-offset cases would get a failure return
while others got elog(ERROR). It was missing an offset validity check for
the SEEK_CUR case on a closed file, too.
Back-patch to all supported branches, since all this code is fundamentally
identical in all of them.
Discussion: https://postgr.es/m/2982.1487617365@sss.pgh.pa.us
These are only supported in to_char, not in the other direction, but the
documentation failed to mention that. Also, describe TZ/tz as printing the
time zone "abbreviation", not "name", because what they print is elsewhere
referred to that way. Per bug #14558.
One case in the PL/Tcl tests is observed to fail on RHEL5 with a Turkish
time zone setting. It's not clear if this is an old Tcl bug or something
odd about the zone data, but in any case that test is meant to see if the
Tcl [clock] command works at all, not what its corner-case behaviors are.
Therefore we have no need to test exactly which week a Sunday midnight is
considered to fall into. Probe the following Tuesday instead.
Discussion: https://postgr.es/m/797.1487517822@sss.pgh.pa.us
The recovery.conf file that's generated is specifically for replication,
and not needed (or wanted) for regular backup restore, so indicate that
in the message.
Also add to the existing rather half-baked description of PROFILE,
which does exactly the same thing, but I think people use it differently.
Discussion: https://postgr.es/m/16461.1487361849@sss.pgh.pa.us
This causes a warning with the old html-docs toolchain, though not with the
new. I had originally supposed that we needed both <indexterm> entries to
get both a primary index entry and a see-also link; but evidently not,
as pointed out by Fabien Coelho.
Discussion: https://postgr.es/m/alpine.DEB.2.20.1702161616060.5445@lancre
Make the typedefs for output plugins consistent with project style;
they were previously not even consistent with each other as to layout
or inclusion of parameter names. Make the documentation look the same,
and fix errors therein (missing and misdescribed parameters).
Back-patch because of the documentation bugs.
The loops in ExecHashJoinNewBatch(), ExecHashIncreaseNumBatches(), and
ExecHashRemoveNextSkewBucket() are all capable of iterating over many
tuples without ever doing a CHECK_FOR_INTERRUPTS, so that the backend
might fail to respond to SIGINT or SIGTERM for an unreasonably long time.
Fix that. In the case of ExecHashJoinNewBatch(), it seems useful to put
the added CHECK_FOR_INTERRUPTS into ExecHashJoinGetSavedTuple() rather
than directly in the loop, because that will also ensure that both
principal code paths through ExecHashJoinOuterGetTuple() will do a
CHECK_FOR_INTERRUPTS, which seems like a good idea to avoid surprises.
Back-patch to all supported branches.
Tom Lane and Thomas Munro
Discussion: https://postgr.es/m/6044.1487121720@sss.pgh.pa.us
It wouldn't complete "TO" after the variable name, which is certainly
minor enough. But since we do complete "TO" after "SET variable ...",
and since this case used to work pre-9.6, I think this is a bug.
Also, fix the query used to collect the variable names; whoever last
touched it evidently didn't understand how the pieces are supposed
to fit together. It accidentally worked anyway, because readline
ignores irrelevant completions, but it was randomly unlike the ones
around it, and could be a source of actual bugs if someone copied
it as a prototype for another query.
Jeff Janes noted that the error cursor position shown for some errors
would vary when operator_precedence_warning is turned on. We'd prefer
that option to have no undocumented effects, so this isn't desirable.
To fix, make sure that an AEXPR_PAREN node has the same exprLocation
as its child node.
(Note: it would be a little cheaper to use @2 here instead of an
exprLocation call, but there are cases where that wouldn't produce
the identical answer, so don't do it like that.)
Back-patch to 9.5 where this feature was introduced.
Discussion: https://postgr.es/m/CAMkU=1ykK+VhhcQ4Ky8KBo9FoaUJH3f3rDQB8TkTXi-ZsBRUkQ@mail.gmail.com
The ALTER TABLE ALTER TYPE implementation can issue DROP INDEX and
CREATE INDEX to refit existing indexes for the new column type. Since
this CREATE INDEX is an implementation detail of an index alteration,
the ensuing DefineIndex() should skip ACL checks specific to index
creation. It already skips the namespace ACL check. Make it skip the
tablespace ACL check, too. Back-patch to 9.2 (all supported versions).
Reviewed by Tom Lane.
The S/390 members of the buildfarm are showing failures indicating
that they're having trouble with the rint() calls I added yesterday.
There's no good reason for that, and I wonder if it is a compiler bug
similar to the one we worked around in d9476b838. Try to fix it using
the same method as before, namely to store the result of rint() back
into a "double" variable rather than immediately converting to int64.
(This isn't entirely waving a dead chicken, since on machines with
wider-than-double float registers, the extra store forces a width
conversion. I don't know if S/390 is like that, but it seems worth
trying.)
In passing, merge duplicate ereport() calls in float8_timestamptz().
Per buildfarm.
When converting a float value to integer microseconds, we should be careful
to round the value to the nearest integer, typically with rint(); simply
assigning to an int64 variable will truncate, causing apparently off-by-one
values in cases that should work. Most places in the datetime code got
this right, but not these two.
float8_timestamptz() is new as of commit e511d878f (9.6). Previous
versions effectively depended on interval_mul() to do roundoff correctly,
which it does, so this fixes an accuracy regression in 9.6.
The problem in make_interval() dates to its introduction in 9.4. Aside
from being careful to round not truncate, let's incorporate the hours and
minutes inputs into the result with exact integer arithmetic, rather than
risk introducing roundoff error where there need not have been any.
float8_timestamptz() problem reported by Erik Nordström, though this is
not his proposed patch. make_interval() problem found by me.
Discussion: https://postgr.es/m/CAHuQZDS76jTYk3LydPbKpNfw9KbACmD=49dC4BrzHcfPv6yA1A@mail.gmail.com
The CREATE INDEX CONCURRENTLY bug can only be triggered by row updates,
not inserts, since the problem would arise from an update incorrectly
being made HOT. Noted by Alvaro.
The problem with the original coding here is that we might receive (and
clear) a relcache invalidation signal for the target relation down inside
one of the index_open calls we're doing. Since the target is open, we
would not drop the relcache entry, just reset its rd_indexvalid and
rd_indexlist fields. But RelationGetIndexAttrBitmap() kept going, and
would eventually cache and return potentially-obsolete attribute bitmaps.
The case where this matters is where the inval signal was from a CREATE
INDEX CONCURRENTLY telling us about a new index on a formerly-unindexed
column. (In all other cases, the lock we hold on the target rel should
prevent any concurrent change in index state.) Even just returning the
stale attribute bitmap is not such a problem, because it shouldn't matter
during the transaction in which we receive the signal. What hurts is
caching the stale data, because it can survive into later transactions,
breaking CREATE INDEX CONCURRENTLY's expectation that later transactions
will not create new broken HOT chains. The upshot is that there's a window
for building corrupted indexes during CREATE INDEX CONCURRENTLY.
This patch fixes the problem by rechecking that the set of index OIDs
is still the same at the end of RelationGetIndexAttrBitmap() as it was
at the start. If not, we loop back and try again. That's a little
more than is strictly necessary to fix the bug --- in principle, we
could return the stale data but not cache it --- but it seems like a
bad idea on general principles for relcache to return data it knows
is stale.
There might be more hazards of the same ilk, or there might be a better
way to fix this one, but this patch definitely improves matters and seems
unlikely to make anything worse. So let's push it into today's releases
even as we continue to study the problem.
Pavan Deolasee and myself
Discussion: https://postgr.es/m/CABOikdM2MUq9cyZJi1KyLmmkCereyGp5JQ4fuwKoyKEde_mzkQ@mail.gmail.com
If we forcibly place a Material node atop a finished subplan, we need
to move any initPlans attached to the subplan up to the Material node,
in order to keep SS_finalize_plan() happy. I'd figured this out in
commit 7b67a0a49 for the case of materializing a cursor plan, but out of
an abundance of caution, I put the initPlan movement hack at the call
site for that case, rather than inside materialize_finished_plan().
That was the wrong thing, because it turns out to also be necessary for
the only other caller of materialize_finished_plan(), ie subselect.c.
We lacked any test cases that exposed the mistake, but bug#14524 from
Wei Congrui shows that it's possible to get an initPlan reference into
the top tlist in that case too, and then SS_finalize_plan() complains.
Hence, move the hack into materialize_finished_plan().
In HEAD, also relocate some recently-added tests in subselect.sql, which
I'd unthinkingly dropped into the middle of a sequence of related tests.
Report: https://postgr.es/m/20170202060020.1400.89021@wrigleys.postgresql.org
