When detaching partition in concurrent mode, it's possible for partition
descriptors to not match the set that was recently seen when the plan
was made, causing an assertion failure or (in production builds) failure
to construct a working plan. The case that was reported involves
prepared statements, but I think it may be possible to hit this bug
without that too.
The problem is that CreatePartitionPruneState is constructing a
PartitionPruneState under the assumption that new partitions can be
added, but never removed, but it turns out that this isn't true: a
prepared statement gets replanned when the DETACH CONCURRENTLY session
sends out its invalidation message, but if the invalidation message
arrives after ExecInitAppend started, we would build a partition
descriptor without the partition, and then CreatePartitionPruneState
would refuse to work with it.
CreatePartitionPruneState already contains code to deal with the new
descriptor having more partitions than before (and behaving for the
extra partitions as if they had been pruned), but doesn't have code to
deal with less partitions than before, and it is naïve about the case
where the number of partitions is the same. We could simply add that a
new stanza for less partitions than before, and in simple testing it
works to do that; but it's possible to press the test scripts even
further and hit the case where one partition is added and a partition is
removed quickly enough that we see the same number of partitions, but
they don't actually match, causing hangs during execution.
To cope with both these problems, we now memcmp() the arrays of
partition OIDs, and do a more elaborate mapping (relying on the fact
that both OID arrays are in partition-bounds order) if they're not
identical.
Backpatch to 14, where DETACH CONCURRENTLY appeared.
Reported-by: yajun Hu <1026592243@qq.com>
Reviewed-by: Tender Wang <tndrwang@gmail.com>
Discussion: https://postgr.es/m/18377-e0324601cfebdfe5@postgresql.org
In commit af0e7deb4a, I removed a call to RelationCloseSmgr(), because
the dangling SMgrRelation was no longer an issue. However, we still
need the call when the relation's relfilelocator changes, so that the
new relfilelocator takes effect immediately.
Reported-by: Alexander Lakhin <exclusion@gmail.com>
Discussion: https://www.postgresql.org/message-id/987b1c8c-8c91-4847-ca0e-879f421680ff%40gmail.com
0452b461bc adds alternative orderings of group-by keys during the query
optimization. This new feature is controlled by the new GUC parameter
enable_group_by_reordering, which accidentally came without the documentation.
This commit adds the missing documentation for that GUC.
Reported-by: Bruce Momjian
Discussion: https://postgr.es/m/ZnDx2FYlba_OafQd%40momjian.us
Author: Andrei Lepikhov
Reviewed-by: Pavel Borisov, Alexander Korotkov
RT_NODE_16_SEARCH_EQ() performs comparisions using vector registers
on x64-64 and aarch64. We apply a mask to the resulting bitfield
to eliminate irrelevant bits that may be set. This ensures correct
behavior, but Valgrind complains of the partially-uninitialised
values. So far the warnings have only occurred on aarch64, which
explains why this hasn't been seen earlier.
To fix this warning, initialize the whole fixed-sized part of the nodes
upon allocation, rather than just do the minimum initialization to
function correctly. The initialization for node48 is a bit different
in that the 256-byte slot index array must be populated with "invalid
index" rather than zero. Experimentation has shown that compilers
tend to emit code that uselessly memsets that array twice. To avoid
pessimizing this path, swap the order of the slot_idxs[] and isset[]
arrays so we can initialize with two non-overlapping memset calls.
Reported by Tomas Vondra
Analysis and patch by Tom Lane, reviewed by Masahiko Sawada. I
investigated the behavior of memset calls to overlapping regions,
leading to the above tweaks to node48 as discussed in the thread.
Discussion: https://postgr.es/m/120c63ad-3d12-415f-a7bf-3da451c31bf6%40enterprisedb.com
Apply const qualifiers to char * arguments and fields throughout the
jsonapi. This allows the top-level APIs such as
pg_parse_json_incremental() to declare their input argument as const.
It also reduces the number of unconstify() calls.
Reviewed-by: Andrew Dunstan <andrew@dunslane.net>
Discussion: https://www.postgresql.org/message-id/flat/f732b014-f614-4600-a437-dba5a2c3738b%40eisentraut.org
afterTriggerInvokeEvents and AfterTriggerExecute have always
treated it as an error if the trigger OID mentioned in a queued
after-trigger event can't be found. However, that fails to
account for the edge case where the trigger's been dropped in
the current transaction since queueing the event. There seems
no very good reason to disallow that case, so instead silently
do nothing if the trigger OID can't be found.
This does give up a little bit of bug-detection ability, but I don't
recall that these error messages have ever actually revealed a bug,
so it seems mostly theoretical. Alternatives such as marking
pending events DONE at the time of dropping a trigger would be
complicated and perhaps introduce bugs of their own.
Per bug #18517 from Alexander Lakhin. Back-patch to all
supported branches.
Discussion: https://postgr.es/m/18517-af2d19882240902c@postgresql.org
This changes the import library name from 'postgres.exe.lib' to
'postgres.lib', which is what it was with the old MSVC build system.
Extension builds use that name.
Bug: #18513
Reported-by: Muralikrishna Bandaru <muralikrishna.bandaru@enterprisedb.com>
Contrary to what the comment for the "check" struct member claims,
'pg_upgrade --check' performs only the checks and does not ask the
user for permission to make changes.
Reviewed-by: Daniel Gustafsson
Discussion: https://postgr.es/m/ZnHk7ci5IuTWVc_c%40nathan
Currently, when the ON EMPTY clause is not present, the ON ERROR
clause (implicit or explicit) dictates the behavior when jsonpath
evaluation in ExecEvalJsonExprPath() results in an empty sequence.
That is an oversight in the commit 6185c9737c.
This commit fixes things so that a NULL is returned instead in that
case which is the default behavior when the ON EMPTY clause is not
present.
Reported-by: Markus Winand
Discussion: https://postgr.es/m/F7DD1442-265C-4220-A603-CB0DEB77E91D%40winand.at
Previously, GetJsonPathVar() allowed a jsonpath expression to
reference any prefix of a PASSING variable's name. For example, the
following query would incorrectly work:
SELECT JSON_QUERY(context_item, jsonpath '$xy' PASSING val AS xyz);
The fix ensures that the length of the variable name mentioned in a
jsonpath expression matches exactly with the name of the PASSING
variable before comparing the strings using strncmp().
Reported-by: Alvaro Herrera (off-list)
Discussion: https://postgr.es/m/CA+HiwqFGkLWMvELBH6E4SQ45qUHthgcRH6gCJL20OsYDRtFx_w@mail.gmail.com
In cost_memoize_rescan(), when calculating the hit_ratio using the calls
and ndistinct estimations, if the value that was set in
MemoizePath.calls had not been processed through clamp_row_est(), then it
was possible that it was set to some non-integer value which could result
in ndistinct being 1 higher than calls due to estimate_num_groups()
performing clamp_row_est() on its input_rows. This could result in
hit_ratio values slightly below 0.0, which would cause an Assert failure.
The value of MemoizePath.calls comes from the final parameter in the
create_memoize_path() function, of which we only have one true caller of.
That caller passes outer_path->rows. All the core code I looked at
always seems to call clamp_row_est() on the Path.rows, so there might
have been no issues with any core Paths causing troubles here. The bug
report was about a CustomPath with a non-clamped row estimated.
The misbehavior as a result of this seems to be mostly limited to the
Assert() failing. Aside from that, it seems the Memoize costs would
just come out slightly higher than they should have, which is likely
fairly harmless.
Reported-by: Kohei KaiGai <kaigai@heterodb.com>
Discussion: https://postgr.es/m/CAOP8fzZnTU+N64UYJYogb1hN-5hFP+PwTb3m_cnGAD7EsQwrKw@mail.gmail.com
Reviewed-by: Richard Guo
Backpatch-through: 14, where Memoize was introduced
Reconstruction of an SP-GiST index by REINDEX CONCURRENTLY may
insert some REDIRECT tuples. This will typically happen in
a transaction that lacks an XID, which leads either to assertion
failure in spgFormDeadTuple or to insertion of a REDIRECT tuple
with zero xid. The latter's not good either, since eventually
VACUUM will apply GlobalVisTestIsRemovableXid() to the zero xid,
resulting in either an assertion failure or a garbage answer.
In practice, since REINDEX CONCURRENTLY locks out index scans
till it's done, it doesn't matter whether it inserts REDIRECTs
or PLACEHOLDERs; and likewise it doesn't matter how soon VACUUM
reduces such a REDIRECT to a PLACEHOLDER. So in non-assert builds
there's no observable problem here, other than perhaps a little
index bloat. But it's not behaving as intended.
To fix, remove the failing Assert in spgFormDeadTuple, acknowledging
that we might sometimes insert a zero XID; and guard VACUUM's
GlobalVisTestIsRemovableXid() call with a test for valid XID,
ensuring that we'll reduce such a REDIRECT the first time VACUUM
sees it. (Versions before v14 use TransactionIdPrecedes here,
which won't fail on zero xid, so they really have no bug at all
in non-assert builds.)
Another solution could be to not create REDIRECTs at all during
REINDEX CONCURRENTLY, making the relevant code paths treat that
case like index build (which likewise knows that no concurrent
index scans can be happening). That would allow restoring the
Assert in spgFormDeadTuple, but we'd still need the VACUUM change
because redirection tuples with zero xid may be out there already.
But there doesn't seem to be a nice way for spginsert() to tell that
it's being called in REINDEX CONCURRENTLY without some API changes,
so we'll leave that as a possible future improvement.
In HEAD, also rename the SpGistState.myXid field to redirectXid,
which seems less misleading (since it might not in fact be our
transaction's XID) and is certainly less uninformatively generic.
Per bug #18499 from Alexander Lakhin. Back-patch to all supported
branches.
Discussion: https://postgr.es/m/18499-8a519c280f956480@postgresql.org
In the wake of the previous commit, we're not doing anything
with that argument. Hence, revert the portions of 534287403
that added that argument and taught the callers to pass it.
Passing the ownerId requires additional syscache lookups in
some code paths, which'd be fine if we were doing anything
useful with the info, but it seems inadvisable if we're not.
Committed separately since there's some thought that we might
want to un-revert this in future, in case it's decided that
storing the original owner ID explicitly in pg_init_privs
is worth doing.
Discussion: https://postgr.es/m/CAMT0RQSVgv48G5GArUvOVhottWqZLrvC5wBzBa4HrUdXe9VRXw@mail.gmail.com
Commit 534287403 invented SHARED_DEPENDENCY_INITACL entries in
pg_shdepend, but installed them only for non-owner roles mentioned
in a pg_init_privs entry. This turns out to be the wrong thing,
because there is nothing to cue REASSIGN OWNED to go and update
pg_init_privs entries when the object's ownership is reassigned.
That leads to leaving dangling entries in pg_init_privs, as
reported by Hannu Krosing. Instead, install INITACL entries for
all roles mentioned in pg_init_privs entries (except pinned roles),
and change ALTER OWNER to not touch them, just as it doesn't
touch pg_init_privs entries.
REASSIGN OWNED will now substitute the new owner OID for the old
in pg_init_privs entries. This feels like perhaps not quite the
right thing, since pg_init_privs ought to be a historical record
of the state of affairs just after CREATE EXTENSION. However,
it's hard to see what else to do, if we don't want to disallow
dropping the object's original owner. In any case this is
better than the previous do-nothing behavior, and we're unlikely
to come up with a superior solution in time for v17.
While here, tighten up some coding rules about how ACLs in
pg_init_privs should never be null or empty. There's not any
obvious reason to allow that, and perhaps asserting that it's
not so will catch some bugs. (We were previously inconsistent
on the point, with some code paths taking care not to store
empty ACLs and others not.)
This leaves recordExtensionInitPrivWorker not doing anything
with its ownerId argument, but we'll deal with that separately.
catversion bump forced because of change of expected contents
of pg_shdepend when pg_init_privs entries exist.
Discussion: https://postgr.es/m/CAMT0RQSVgv48G5GArUvOVhottWqZLrvC5wBzBa4HrUdXe9VRXw@mail.gmail.com
It used to check if the replication slot exists and is active on
primary. This check might fail on slow hosts because the replication
slot might not be active at the time of this check.
The current code obtains the replication slot name from the
primary_slot_name on standby and assumes the replication slot exists
and is active on primary. If it doesn't exist, this tool will log an
error and continue.
Author: Euler Taveira <euler.taveira@enterprisedb.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Discussion: https://www.postgresql.org/message-id/776c5cac-5ef5-4001-b1bc-5b698bc0c62a%40app.fastmail.com
It used to check if the target server is connected to the primary
server (send required WAL) to rapidly react when the process won't
succeed. This code is not enough to guarantee that the recovery
process will complete. There is a window between the walreceiver
shutdown and the pg_is_in_recovery() returns false that can reach
NUM_CONN_ATTEMPTS attempts and fails.
Instead, rely only on the --recovery-timeout option to give up the
process after the specified number of seconds.
This should help with buildfarm failures on slow machines.
Author: Euler Taveira <euler.taveira@enterprisedb.com>
Reviewed-by: Hayato Kuroda <kuroda.hayato@fujitsu.com>
Discussion: https://www.postgresql.org/message-id/776c5cac-5ef5-4001-b1bc-5b698bc0c62a%40app.fastmail.com
The function has been introduced in 3682025015 to test at a low level
the new kinds of external toast datums, and would fail on OOM when
dealing with a plain storage attribute. The existing tests of
indirect_toast do not test this case, still the error generated was
confusing.
Author: Alexander Lakhin
Discussion: https://postgr.es/m/250a21e5-d677-6b2a-2692-cd4233785e37@gmail.com
DeleteInitPrivs did not get the memo about how, when dropping a
whole object (with subid == 0), you should drop entries relating
to its sub-objects too. This is visible in the test_pg_dump test
case if one drops the extension at the end: the entry for
GRANT SELECT(col1) ON regress_pg_dump_table TO public;
was still present in pg_init_privs afterwards, although it was
pointing to a dangling table OID.
Noted while fooling with a fix for REASSIGN OWNED for pg_init_privs
entries. This bug is aboriginal in the pg_init_privs feature
though, and there seems no reason not to back-patch the fix.
Oversight in 534287403. We missed this up to now because the
core regression tests create no such entries (at least up to
this test), so the only way to see the failure is to do
"make installcheck" in an installation where some other DB
has such entries. I happened to do that just now ...
Commit 667e65aac3 changed both num_dead_tuples and max_dead_tuples
columns to dead_tuple_bytes and max_dead_tuple_bytes columns,
respectively. But as per discussion, the number of dead tuples
collected still provides meaningful insights for users.
This commit reintroduces the column for the count of dead tuples,
renamed as num_dead_item_ids. It avoids confusion with the number of
dead tuples removed by VACUUM, which includes dead heap-only tuples
but excludes any pre-existing LP_DEAD items left behind by
opportunistic pruning.
Bump catalog version.
Reviewed-by: Peter Geoghegan, Álvaro Herrera, Andrey Borodin
Discussion: https://postgr.es/m/CAD21AoBL5sJE9TRWPyv%2Bw7k5Ee5QAJqDJEDJBUdAaCzGWAdvZw%40mail.gmail.com
Alvaro Herrera