After pg_upgrade, it is possible that some tuples' Xmax have multixacts
corresponding to the old installation; such multixacts cannot have
running members anymore. In many code sites we already know not to read
them and clobber them silently, but at least when VACUUM tries to freeze
a multixact or determine whether one needs freezing, there's an attempt
to resolve it to its member transactions by calling GetMultiXactIdMembers,
and if the multixact value is "in the future" with regards to the
current valid multixact range, an error like this is raised:
ERROR: MultiXactId 123 has not been created yet -- apparent wraparound
and vacuuming fails. Per discussion with Andrew Gierth, it is completely
bogus to try to resolve multixacts coming from before a pg_upgrade,
regardless of where they stand with regards to the current valid
multixact range.
It's possible to get from under this problem by doing SELECT FOR UPDATE
of the problem tuples, but if tables are large, this is slow and
tedious, so a more thorough solution is desirable.
To fix, we realize that multixacts in xmax created in 9.2 and previous
have a specific bit pattern that is never used in 9.3 and later (we
already knew this, per comments and infomask tests sprinkled in various
places, but we weren't leveraging this knowledge appropriately).
Whenever the infomask of the tuple matches that bit pattern, we just
ignore the multixact completely as if Xmax wasn't set; or, in the case
of tuple freezing, we act as if an unwanted value is set and clobber it
without decoding. This guarantees that no errors will be raised, and
that the values will be progressively removed until all tables are
clean. Most callers of GetMultiXactIdMembers are patched to recognize
directly that the value is a removable "empty" multixact and avoid
calling GetMultiXactIdMembers altogether.
To avoid changing the signature of GetMultiXactIdMembers() in back
branches, we keep the "allow_old" boolean flag but rename it to
"from_pgupgrade"; if the flag is true, we always return an empty set
instead of looking up the multixact. (I suppose we could remove the
argument in the master branch, but I chose not to do so in this commit).
This was broken all along, but the error-facing message appeared first
because of commit 8e9a16ab8f and was partially fixed in a25c2b7c4d.
This fix, backpatched all the way back to 9.3, goes approximately in the
same direction as a25c2b7c4d but should cover all cases.
Bug analysis by Andrew Gierth and Álvaro Herrera.
A number of public reports match this bug:
https://www.postgresql.org/message-id/20140330040029.GY4582@tamriel.snowman.nethttps://www.postgresql.org/message-id/538F3D70.6080902@publicrelay.comhttps://www.postgresql.org/message-id/556439CF.7070109@pscs.co.ukhttps://www.postgresql.org/message-id/SG2PR06MB0760098A111C88E31BD4D96FB3540@SG2PR06MB0760.apcprd06.prod.outlook.comhttps://www.postgresql.org/message-id/20160615203829.5798.4594@wrigleys.postgresql.org
something.* IS NOT NULL means that every attribute of the row is not
NULL, not that the row itself is non-NULL (e.g. because it's coming
from below an outer join. Use (somevar.*)::pg_catalog.text IS NOT
NULL instead.
Ashutosh Bapat, per a report by Rushabh Lathia. Reviewed by
Amit Langote and Etsuro Fujita. Schema-qualification added by me.
There was some very strange code here, dating to commit b525bf77, that
purported to work around an ancient gcc bug by forcing a float4 comparison
to be done as int instead. Commit 5871b8848 broke that when it changed
one side of the comparison to "double" but left the comparison code alone.
Commit f576b17cd doubled down on the weirdness by introducing a "volatile"
marker, which had nothing to do with the actual problem.
Guess that the gcc bug, even if it's still present in the wild, was
triggered by comparison of float4's and can be avoided if we store the
result of cnt_sml() into a double before comparing to the double "nlimit".
This will at least work correctly on non-broken compilers, and it's way
more readable.
Per bug #14202 from Greg Navis. Add a regression test based on his
example.
Report: <20160620115321.5792.10766@wrigleys.postgresql.org>
As shown by buildfarm reports, dblink_build_sql_insert and
dblink_build_sql_update are *not* parallel safe, because they
may attempt to access temporary tables of the local session.
Although dblink_build_sql_delete doesn't actually touch the
contents of the referenced table, it seems consistent and prudent
to mark it PARALLEL RESTRICTED too.
This requires some core changes as well so that we can properly
WAL-log the truncation. Specifically, it changes the format of the
XLOG_SMGR_TRUNCATE WAL record, so bump XLOG_PAGE_MAGIC.
Patch by me, reviewed but not fully endorsed by Andres Freund.
Almost all functions provided by this extension are PARALLEL
RESTRICTED. Mostly, that's because the leader's TCP connections won't
be shared with the workers, but in some cases like dblink_get_pkey
it's because they obtain locks which might be released early if taken
within a parallel worker. dblink_fdw_validator probably can't be used
in a query anyway, but there would be no problem from the point of
view of parallel query if it were, so it's PARALLEL SAFE.
Andreas Karlsson
The new pg_check_visible() and pg_check_frozen() functions can be used to
verify that the visibility map bits for a relation's data pages match the
actual state of the tuples on those pages.
Amit Kapila and Robert Haas, reviewed (in earlier versions) by Andres
Freund. Additional testing help by Thomas Munro.
All functions provided by this extension are PARALLEL RESTRICTED,
because they provide information about the connection state. Parallel
workers don't have this information and therefore these functions
can't be executed in a worker (but they can be present in a query some
other part of which uses parallelism).
Andreas Karlsson
Commit 749a787c5b bumped the extension
version on all of these extensions already, and we haven't had a
release since then, so we can make further changes without bumping the
extension version again. Take this opportunity to mark all of the
functions exported by these modules PARALLEL SAFE -- except for
pg_trgm's set_limit(). Mark that one PARALLEL RESTRICTED, because it
makes a persistent change to a GUC value.
Note that some of the markings added by this commit don't have any
effect; for example, gseg_picksplit() isn't likely to be mentioned
explicitly in a query and therefore it's parallel-safety marking will
never be consulted. But this commit just marks everything for
consistency: if it were somehow used in a query, that would be fine as
far as parallel query is concerned, since it does not consult any
backend-private state, attempt to write data, etc.
Andreas Karlsson, with a few revisions by me.
As discovered by Andreas Seltenreich via sqlsmith, it's possible for a
remote join to need to generate a target list which contains a
PlaceHolderVar which would need to be evaluated on the remote server.
This happens when we try to push down a join tree which contains outer
joins and the nullable side of the join contains a subquery which
evauates some expression which can go to NULL above the level of the
join. Since the deparsing logic can't build a remote query that
involves subqueries, it fails while trying to produce an SQL query
that can be sent to the remote side. Detect such cases and don't try
to push down the join at all.
It's actually fine to push down the join if the PlaceHolderVar needs
to be evaluated at the current join level. This patch makes a small
change to build_tlist_to_deparse so that this case will work.
Amit Langote, Ashutosh Bapat, and me.
Extension scripts should never use CREATE OR REPLACE for initial object
creation. If there is a collision with a pre-existing (probably
user-created) object, we want extension installation to fail, not silently
overwrite the user's object. Bloom and sslinfo both violated this precept.
Also fix a number of scripts that had no standard header (the file name
comment and the \echo...\quit guard). Probably the \echo...\quit hack
is less important now than it was in 9.1 days, but that doesn't mean
that individual extensions get to choose whether to use it or not.
And fix a couple of evident copy-and-pasteos in file name comments.
No need for back-patch: the REPLACE bugs are both new in 9.6, and the
rest of this is pretty much cosmetic.
Andreas Karlsson and Tom Lane
Andreas Seltenreich reports that it is possible for a PlaceHolderVar
to creep into this tlist, and I fear that even after that's fixed we
might have other, similar bugs in this area either now or in the
future. There's a lot of action-at-a-distance here, because the
validity of this assertion depends on core planner behavior; so, let's
use elog() to make sure we catch this even in non-assert builds,
rather than just crashing.
All functions provided by this extension are PARALLEL SAFE. Given the
general prohibition against write operations in parallel queries, it is
perhaps a bit surprising that pg_stat_statements_reset() is parallel safe.
But since it only modifies shared memory, not the database, it's OK.
Andreas Karlsson
In commits 9ff60273e3 and dbe2328959 I (tgl) fixed the
signatures of a bunch of contrib's GIN and GIST support functions so that
they would pass validation by the recently-added amvalidate functions.
The backend does not actually consult or check those signatures otherwise,
so I figured this was basically cosmetic and did not require an extension
version bump. However, Alexander Korotkov pointed out that that would
leave us in a pretty messy situation if we ever wanted to redefine those
functions later, because there wouldn't be a unique way to name them.
Since we're going to be bumping these extensions' versions anyway for
parallel-query cleanups, let's take care of this now.
Andreas Karlsson, adjusted for more search-path-safety by me
Fix still another bug in commit 35fcb1b3d: it failed to fully initialize
the SortSupport states it introduced to allow the executor to re-check
ORDER BY expressions containing distance operators. That led to a null
pointer dereference if the sortsupport code tried to use ssup_cxt. The
problem only manifests in narrow cases, explaining the lack of previous
field reports. It requires a GiST-indexable distance operator that lacks
SortSupport and is on a pass-by-ref data type, which among core+contrib
seems to be only btree_gist's interval opclass; and it requires the scan
to be done as an IndexScan not an IndexOnlyScan, which explains how
btree_gist's regression test didn't catch it. Per bug #14134 from
Jihyun Yu.
Peter Geoghegan
Report: <20160511154904.2603.43889@wrigleys.postgresql.org>
Per discussion, this is a more understandable and future-proof way of
exposing the setting to users. On-disk, we can still store it in words,
so as to not break on-disk compatibility with beta1.
Along the way, clean up the code associated with Bloom reloptions.
Provide explicit macros for default and maximum lengths rather than
having magic numbers buried in multiple places in the code. Drop
the adjustBloomOptions() code altogether: it was useless in view of
the fact that reloptions.c already performed default-substitution and
range checking for the options. Rename a couple of macros and types
for more clarity.
Discussion: <23767.1464926580@sss.pgh.pa.us>
blbuildempty did not do even approximately the right thing: it tried
to add a metapage to the relation's regular data fork, which already
has one at that point. It should look like the ambuildempty methods
for all the standard index types, ie, initialize a metapage image in
some transient storage and then write it directly to the init fork.
To support that, refactor BloomInitMetapage into two functions.
In passing, fix BloomInitMetapage so it doesn't leave the rd_options
field of the index's relcache entry pointing at transient storage.
I'm not sure this had any visible consequence, since nothing much
else is likely to look at a bloom index's rd_options, but it's
certainly poor practice.
Per bug #14155 from Zhou Digoal.
Report: <20160524144146.22598.42558@wrigleys.postgresql.org>
It's possible to begin and end an indexscan without ever calling
amrescan. contrib/bloom, unlike every other index AM, allocated
its "scan->opaque" storage at amrescan time, and thus would crash
in amendscan if amrescan hadn't been called. We could fix this
by putting in a null-pointer check in blendscan, but I see no very
good reason why contrib/bloom should march to its own drummer in
this respect. Let's move that initialization to blbeginscan
instead. Per report from Jeff Janes.
Commit 3151f16e18 was intended to be
a commit of a patch from Ashutosh Bapat, but instead I mistakenly
committed an earlier version from Michael Paquier (because both
patches were submitted with the same filename, and I confused them).
Michael's patch fixes the crash but doesn't actually implement the
correct test.
Repair the incorrect logic, and also expand the comments considerably
so that this is all more clear.
Ashutosh Bapat and Robert Haas
First, even if we cancel a query, we still have to roll back the
containing transaction; otherwise, the session will be left in a
failed transaction state.
Second, we need to support canceling queries whe aborting a
subtransaction as well as when aborting a toplevel transaction.
Etsuro Fujita, reviewed by Michael Paquier
Buildfarm member skink failed with symptoms suggesting that an
auto-analyze had happened and changed the plan displayed for a
test query. Although this is evidently of low probability,
regression tests that sometimes fail are no fun, so add commands
to force a bitmap scan to be chosen.
These adjustments adjust code and comments in minor ways to prevent
pgindent from mangling them. Among other things, I tried to avoid
situations where pgindent would emit "a +b" instead of "a + b", and I
tried to avoid having it break up inline comments across multiple
lines.
CHECK_PAGE_OFFSET_RANGE() has been unused forever.
CHECK_RELATION_BLOCK_RANGE() has been unused in pgstatindex.c ever since
bt_page_stats() and bt_page_items() functions were moved from pgstattuple
to pageinspect module. It still exists in pageinspect/btreefuncs.c.
Daniel Gustafsson
This reverts commit c8e81afc60.
That turns out to have been based on a faulty diagnosis of why the
VS2015 build was misbehaving. Instead, we need to fix DatumGetBool().
It appears that we can no longer get away with using V0 call convention
for bool-returning functions in newer versions of MSVC. The compiler
seems to generate code that doesn't clear the higher-order bits of the
result register, causing the bool result Datum to often read as "true"
when "false" was intended. This is not very surprising, since the
function thinks it's returning a bool-width result but fmgr_oldstyle
assumes that V0 functions return "char *"; what's surprising is that
that hack worked for so long on so many platforms.
The only functions of this description in core+contrib are in contrib/seg,
which we'd intentionally left mostly in V0 style to serve as a warning
canary if V0 call convention breaks. We could imagine hacking things
so that they're still V0 (we'd have to redeclare the bool-returning
functions as returning some suitably wide integer type, like size_t,
at the C level). But on the whole it seems better to convert 'em to V1.
We can still leave the pointer- and int-returning functions in V0 style,
so that the test coverage isn't gone entirely.
Back-patch to 9.5, since our intention is to support VS2015 in 9.5
and later. There's no SQL-level change in the functions' behavior
so back-patching should be safe enough.
Discussion: <22094.1461273324@sss.pgh.pa.us>
Michael Paquier, adjusted some by me