== PostgreSQL Weekly News - November 19 2017 ==

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

== PostgreSQL Weekly News - November 19 2017 ==

David Fetter
== PostgreSQL Weekly News - November 19 2017 ==

== PostgreSQL Product News ==

pgMail 1.4 released.

pg_prefaulter, a tool used to mitigate the effects of PostgreSQL replication
lag, released.

== PostgreSQL Jobs for November ==


== PostgreSQL Local ==

PGDAY Brasilia 2017 will take place December 2, 2017.  The CfP
is open at

PGConf Local: Austin will be held December 4 - 5, 2017. Program available
and registration is now open at

PGConf.ASIA 2017 will take place on December 4-6 2017 in Akihabara, Tokyo,

FOSDEM PGDay 2018, a one day conference held before the main FOSDEM event will
be held in Brussels, Belgium, on Feb 2nd, 2018.

Prague PostgreSQL Developer Day 2018 (P2D2 2018) is a two-day
conference that will be held on February 14-15 2018 in Prague, Czech Republic.
The CfP is open until January 5, 2018 at https://p2d2.cz/callforpapers

PGConf India 2018 will be on February 22-23, 2018 in Bengaluru, Karnataka.
Proposals are due via https://goo.gl/forms/F9hRjOIsaNasVOAz2 by November 15th, 2017.

PostgreSQL@SCaLE is a two day, two track event which takes place on
March 8-9, 2018, at Pasadena Convention Center, as part of SCaLE 16X.
The CfP is open through October 31, 2017 at

Nordic PGDay 2018 will be held in Oslo, Norway, at the Radisson Blu Hotel
Nydalen, on March 13, 2018.  The CfP is open through December 31, 2017 at

pgDay Paris 2018 will be held in Paris, France at the Espace Saint-Martin, on
March 15 2018.  The CfP is open until December 31, 2017.

PGConf APAC 2018 will be held in Singapore March 22-24, 2018.  The CfP is open
at http://2018.pgconfapac.org/cfp through December 4, 2017.

The German-speaking PostgreSQL Conference 2018 will take place on April 13th,
2018 in Berlin.  The CfP is open until January 09, 2018 at
http://2018.pgconf.de/de/callforpapers.html and the conference site is at

PGCon 2018 will take place in Ottawa on May 29 - June 2018.  The CfP goes out

PGConf.Brazil 2018 will take place in São Paulo, Brazil on August 3-4 2018. The
CfP will open soon.

== PostgreSQL in the News ==

Planet PostgreSQL: http://planet.postgresql.org/

PostgreSQL Weekly News is brought to you this week by David Fetter

Submit news and announcements by Sunday at 3:00pm EST5EDT.  Please send English
language ones to [hidden email], German language to [hidden email], Italian
language to [hidden email].

== Applied Patches ==

Noah Misch pushed:

- Don't call pgwin32_message_to_UTF16() without CurrentMemoryContext.
  PostgreSQL running as a Windows service crashed upon calling write_stderr()
  before MemoryContextInit().  This fix completes work started in
  5735efee15540765315aa8c1a230575e756037f7.  Messages this early contain only
  ASCII bytes; if we removed the CurrentMemoryContext requirement, the ensuing
  conversions would have no effect.  Back-patch to 9.3 (all supported versions).
  Takayuki Tsunakawa, reviewed by Michael Paquier.  Discussion:

- Install Windows crash dump handler before all else.  Apart from calling
  write_stderr() on failure, the handler depends on no PostgreSQL facilities.
  We have experienced crashes before reaching the former call site.  Given such
  an early crash, this change cannot hurt and may produce a helpful dump.
  Absent an early crash, this change has no effect.  Back-patch to 9.3 (all
  supported versions).  Takayuki Tsunakawa Discussion:

- MSVC: Rebuild spiexceptions.h when out of date.  Also, add a warning to catch
  future instances of naming a nonexistent file as a prerequisite.  Back-patch
  to 9.3 (all supported versions).

Stephen Frost pushed:

- Fix typo.  Determinisitcally -> Deterministically Author: Michael Paquier
  <[hidden email]> Discussion:

Álvaro Herrera pushed:

- Mention CREATE/DROP STATISTICS in event triggers docs.  The new commands are
  reported by event triggers, but they weren't documented as such.  Repair.
  Author: David Rowley Discussion:

- Simplify index_[constraint_]create API.  Instead of passing large swaths of
  boolean arguments, define some flags that can be used in a bitmask.  This
  makes it easier not only to figure out what each call site is doing, but also
  to add some new flags.  The flags are split in two -- one set for index_create
  directly and another for constraints.  index_create() itself receives both,
  and then passes down the latter to index_constraint_create(), which can also
  be called standalone.  Discussion:
  Reviewed-by: Simon Riggs

Robert Haas pushed:

- Track in the plan the types associated with PARAM_EXEC parameters.  Up until
  now, we only tracked the number of parameters, which was sufficient to
  allocate an array of Datums of the appropriate size, but not sufficient to,
  for example, know how to serialize a Datum stored in one of those slots.  An
  upcoming patch wants to do that, so add this tracking to make it possible.
  Patch by me, reviewed by Tom Lane and Amit Kapila.  Discussion:

- Push target list evaluation through Gather Merge.  We already do this for
  Gather, but it got overlooked for Gather Merge.  Amit Kapila, with review and
  minor revisions by Rushabh Lathia and by me.  Discussion:

- Add parallel_leader_participation GUC.  Sometimes, for testing, it's useful to
  have the leader do nothing but read tuples from workers; and it's possible
  that could work out better even in production.  Thomas Munro, reviewed by Amit
  Kapila and by me.  A few final tweaks by me.  Discussion:

- Fix typo.  Jesper Pedersen Discussion:

- Centralize executor-related partitioning code.  Some code is moved from
  partition.c, which has grown very quickly lately; splitting the executor parts
  out might help to keep it from getting totally out of control.  Other code is
  moved from execMain.c.  All is moved to a new file execPartition.c.
  get_partition_for_tuple now has a new interface that more clearly separates
  executor concerns from generic concerns.  Amit Langote.  A slight comment
  tweak by me.  Discussion:

- Pass InitPlan values to workers via Gather (Merge).  If a PARAM_EXEC parameter
  is used below a Gather (Merge) but the InitPlan that computes it is attached
  to or above the Gather (Merge), force the value to be computed before starting
  parallelism and pass it down to all workers.  This allows us to use
  parallelism in cases where it previously would have had to be rejected as
  unsafe.  We do - in this case - lose the optimization that the value is only
  computed if it's actually used.  An alternative strategy would be to have the
  first worker that needs the value compute it, but one downside of that
  approach is that we'd then need to select a parallel-safe path to compute the
  parameter value; it couldn't for example contain a Gather (Merge) node.  At
  some point in the future, we might want to consider both approaches.
  Independent of that consideration, there is a great deal more work that could
  be done to make more kinds of PARAM_EXEC parameters parallel-safe.  This
  infrastructure could be used to allow a Gather (Merge) on the inner side of a
  nested loop (although that's not a very appealing plan) and cases where the
  InitPlan is attached below the Gather (Merge) could be addressed as well using
  various techniques.  But this is a good start.  Amit Kapila, reviewed and
  revised by me.  Reviewing and testing from Kuntal Ghosh, Haribabu Kommi, and
  Tushar Ahuja.  Discussion:

- Update postgresql.conf.sample to match pg_settings classificaitons.  A handful
  of settings, most notably shared_preload_libraries, were just plain the wrong
  place compared to their assigned config_group value in guc.c (and thus
  pg_settings).  In other cases the names of the sections in
  postgresql.conf.sample were mildly different from the corresponding entries in
  config_group_names[].  Make it all consistent.  Adrián Escoms, reviewed by me.

- Fix typo in comment.  Etsuro Fujita Discussion:

- Fix broken cleanup interlock for GIN pending list.  The pending list must (for
  correctness) always be cleaned up by vacuum, and should (for the avoidance of
  surprising behavior) always be cleaned up by an explicit call to
  gin_clean_pending_list, but cleanup is optional when inserting.  The old logic
  got this backward: cleanup was forced if (stats == NULL), but that's going to
  be *false* when vacuuming and *true* for inserts.  Masahiko Sawada, reviewed
  by me.  Discussion:

- Remove redundant line from Makefile.  Masahiko Sawada, reviewed by Michael
  Paquier Discussion:

- Set proargmodes for satisfies_hash_partition.  It appears that proargmodes
  should always be set for variadic functions, but satifies_hash_partition had
  it as NULL.  In addition to fixing the problem, add a regression test to guard
  against future mistakes of this type.

- Update postgresql.conf.sample comment for bgwriter_lru_maxpages.  Commit
  14ca9abfbe4643408ad6ed3279f2f6366cafb3f1 should have done this, but did not.
  Jeff Janes Discussion:

Tom Lane pushed:

- Allow running just selected steps of pgbench's initialization sequence.  This
  feature caters to specialized use-cases such as running the normal pgbench
  scenario with nonstandard indexes, or inserting other actions between steps of
  the initialization sequence.  The normal sequence of initialization actions is
  broken down into half a dozen steps which can be executed in a user-specified
  order, to the extent to which that's sensible.  The actions themselves aren't
  changed, except to make them more robust against nonstandard uses: * all four
  tables are now dropped in one DROP command, to reduce assumptions about what
  foreign key relationships exist; * all four tables are now truncated at the
  start of the data load step, for consistency; * the foreign key creation
  commands now specify constraint names, to prevent accidentally creating
  duplicate constraints by executing the 'f' step twice.  Make some cosmetic
  adjustments in the messages emitted by pgbench so that it's clear which steps
  are getting run, and so that the messages agree with the documented names of
  the steps.  In passing, fix failure to enforce that the -v option is used only
  in benchmarking mode.  Masahiko Sawada, reviewed by Fabien Coelho,
  editorialized a bit by me Discussion:

- Document changes in large-object privilege checking.  Commit 5ecc0d738 removed
  the hard-wired superuser checks in lo_import and lo_export in favor of
  protecting them with SQL permissions, but failed to adjust the documentation
  to match.  Fix that, and add a <caution> paragraph pointing out the nontrivial
  security hazards involved with actually granting such permissions.  (It's
  still better than ALLOW_DANGEROUS_LO_FUNCTIONS, though.) Also, commit
  ae20b23a9 caused large object read/write privilege to be checked during
  lo_open() rather than in the actual read or write calls.  Document that.

- Rearrange c.h to create a "compiler characteristics" section.  Generalize
  section 1 to handle stuff that is principally about the compiler (not
  libraries), such as attributes, and collect stuff there that had been dropped
  into various other parts of c.h.  Also, push all the gettext macros into
  section 8, so that section 0 is really just inclusions rather than inclusions
  and random other stuff.  The primary goal here is to get
  pg_attribute_aligned() defined before section 3, so that we can use it with
  int128.  But this seems like good cleanup anyway.  This patch just moves macro
  definitions around, and shouldn't result in any changes in generated code.
  But I'll push it out separately to see if the buildfarm agrees.  Discussion:

- Prevent int128 from requiring more than MAXALIGN alignment.  Our initial work
  with int128 neglected alignment considerations, an oversight that came back to
  bite us in bug #14897 from Vincent Lachenal.  It is unsurprising that int128
  might have a 16-byte alignment requirement; what's slightly more surprising is
  that even notoriously lax Intel chips sometimes enforce that.  Raising
  MAXALIGN seems out of the question: the costs in wasted disk and memory space
  would be significant, and there would also be an on-disk compatibility break.
  Nor does it seem very practical to try to allow some data structures to have
  more-than-MAXALIGN alignment requirement, as we'd have to push knowledge of
  that throughout various code that copies data structures around.  The only way
  out of the box is to make type int128 conform to the system's alignment
  assumptions.  Fortunately, gcc supports that via its __attribute__(aligned())
  pragma; and since we don't currently support int128 on non-gcc-workalike
  compilers, we shouldn't be losing any platform support this way.  Although we
  could have just done pg_attribute_aligned(MAXIMUM_ALIGNOF) and called it a
  day, I did a little bit of extra work to make the code more portable than
  that: it will also support int128 on compilers without
  __attribute__(aligned()), if the native alignment of their 128-bit-int type is
  no more than that of int64.  Add a regression test case that exercises the one
  known instance of the problem, in parallel aggregation over a bigint column.
  This will need to be back-patched, along with the preparatory commit
  91aec93e6.  But let's see what the buildfarm makes of it first.  Discussion:

- Further refactoring of c.h and nearby files.  This continues the work of
  commit 91aec93e6 by getting rid of a lot of Windows-specific funny business in
  "section 0".  Instead of including pg_config_os.h in different places
  depending on platform, let's standardize on putting it before the system
  headers, and in consequence reduce win32.h to just what has to appear before
  the system headers or the body of c.h (the latter category seems to include
  only PGDLLIMPORT and PGDLLEXPORT).  The rest of what was in win32.h is moved
  to a new sub-include of port.h, win32_port.h.  Some of what was in port.h
  seems to better belong there too.  It's possible that I missed some
  declaration ordering dependency that needs to be preserved, but hopefully the
  buildfarm will find that out in short order.  Unlike the previous commit, no
  back-patch, since this is just cleanup not a prerequisite for a bug fix.
  Discussion: https://postgr.es/m/29650.1510761080@...

- Fix bogus logic for checking data dirs' versions within pg_upgrade.  Commit
  9be95ef15 failed to cure all of the redundancy here: we were actually calling
  get_major_server_version() three times for each of the old and new data
  directories.  While that's not enormously expensive, it's still sloppy.  A.
  Akenteva Discussion:

- Define _WINSOCK_DEPRECATED_NO_WARNINGS in all MSVC builds.  Commit 0fb54de9a
  thought that this was only needed in VS2015 and later, but buildfarm member
  woodlouse shows that at least VS2013 whines as well.  Let's just define it
  regardless of MSVC version; it should be harmless enough in older releases.
  Also, in the wake of ed9b3606d, it seems better to put it in win32_port.h
  where <winsock2.h> is included.  Since this is only suppressing a pedantic
  compiler warning, I don't feel a need for a back-patch.  Discussion:

- Make PL/Python handle domain-type conversions correctly.  Fix PL/Python so
  that it can handle domains over composite, and so that it enforces domain
  constraints correctly in other cases that were not always done properly
  before.  Notably, it didn't do arrays of domains right (oversight in commit
  c12d570fa), and it failed to enforce domain constraints when returning a
  composite type containing a domain field, and if a transform function is being
  used for a domain's base type then it failed to enforce domain constraints on
  the result.  Also, in many places it missed checking domain constraints on
  null values, because the plpy_typeio code simply wasn't called for Py_None.
  Rather than try to band-aid these problems, I made a significant refactoring
  of the plpy_typeio logic.  The existing design of recursing for array and
  composite members is extended to also treat domains as containers requiring
  recursion, and the APIs for the module are cleaned up and simplified.  The
  patch also modifies plpy_typeio to rely on the typcache more than it did
  before (which was pretty much not at all).  This reduces the need for
  repetitive lookups, and lets us get rid of an ad-hoc scheme for detecting
  changes in composite types.  I added a couple of small features to typcache to
  help with that.  Although some of this is fixing bugs that long predate v11, I
  don't think we should risk a back-patch: it's a significant amount of code
  churn, and there've been no complaints from the field about the bugs.  Tom
  Lane, reviewed by Anthony Bykov Discussion:

- Clean up warnings in MinGW builds.  Experimentation with modern MinGW
  (specifically the 5.0.2 version packaged for Fedora 26) shows that its version
  of sys/stat.h *does* provide S_IRGRP and friends, contrary to the expectation
  of win32_port.h.  This results in an astonishing number of compiler warnings,
  and perhaps in incorrect code --- I'm not sure if the nonzero values supplied
  by MinGW's header actually do anything.  Hence, adjust win32_port.h to only
  define these macros if <sys/stat.h> doesn't.  This might be worth
  back-patching, but given the lack of complaints so far, I'm not too excited
  about it.

- Prevent to_number() from losing data when template doesn't match exactly.
  Non-data template patterns would consume characters whether or not those
  characters were what the pattern expected, for example SELECT
  TO_NUMBER('1234', '9,999'); produced 134 because the '2' got eaten by the
  comma pattern.  This seems undesirable, not least because it doesn't happen in
  Oracle.  For the ',' and 'G' template patterns, we can fix this by consuming
  characters only if they match what the pattern would output.  For non-data
  patterns such as 'L' and 'TH', it seems impractical to tighten things up to
  the point of consuming only exact matches to what the pattern would output;
  but we can improve matters quite a lot by redefining the behavior as "consume
  only characters that aren't digits, signs, decimal point, or comma".  Also,
  fix it so that the behavior is to consume the number of *characters* the
  pattern would output, not the number of *bytes*.  The old coding would do
  surprising things with non-ASCII currency symbols, for example.  (It would be
  good to apply that rule for literal text as well, but this commit only fixes
  it for non-data patterns.) Oliver Ford, reviewed by Thomas Munro and Nathan
  Wagner, and whacked around a bit more by me Discussion:

- Provide modern examples of how to auto-start Postgres on macOS.  The scripts
  in contrib/start-scripts/osx don't work at all on macOS 10.10 (Yosemite) or
  later, because they depend on SystemStarter which Apple deprecated long ago
  and removed in 10.10.  Add a new subdirectory contrib/start-scripts/macos with
  scripts that use the newer launchd infrastructure.  Since this problem is
  independent of which Postgres version you're using, back-patch to all
  supported branches.  Discussion:

- Remove contrib/start-scripts/osx/.  Since those scripts haven't worked at all
  in macOS releases of 2014 and later, and aren't the recommended way to do it
  on any release since 2005, there seems little point carrying them into the
  future.  It's very unlikely that anyone would be installing PG >= 11 on a
  macOS release where they couldn't use contrib/start-scripts/macos/.
  Discussion: https://postgr.es/m/31338.1510763554@...

- Fix quoted-substring handling in format parsing for to_char/to_number/etc.
  This code evidently intended to treat backslash as an escape character within
  double-quoted substrings, but it was sufficiently confused that cases like
  ..."foo\\"... did not work right: the second backslash managed to quote the
  double-quote after it, despite being quoted itself.  Rewrite to get that
  right, while preserving the existing behavior outside double-quoted
  substrings, which is that backslash isn't special except in the combination
  \".  Comparing to Oracle, it seems that their version of to_char() for
  timestamps allows literal alphanumerics only within double quotes, while
  non-alphanumerics are allowed outside quotes; backslashes aren't special
  anywhere; there is no way at all to emit a literal double quote.  (Bizarrely,
  their to_char() for numbers is different; it doesn't allow literal text at all
  AFAICT.)  The fact that they don't treat backslash as special justifies our
  existing behavior for backslash outside double quotes.  I considered making
  backslash inside double quotes act the same way (ie, special only if before
  "), which in a green field would be a more consistent behavior.  But that
  would likely break more existing SQL code than what this patch does.  Add some
  test cases illustrating this behavior.  (Only the last new case actually
  changes behavior in this commit.) Little of this behavior was documented,
  either, so fix that.  Discussion:

- Improve to_date/to_number/to_timestamp behavior with multibyte characters.
  The documentation says that these functions skip one input character per
  literal (non-pattern) format character.  Actually, though, they skipped one
  input *byte* per literal *byte*, which could be hugely confusing if either
  data or format contained multibyte characters.  To fix, adjust the FormatNode
  representation and parse_format() so that multibyte format characters are
  stored as one FormatNode not several, and adjust the data-skipping bits to
  advance by pg_mblen() not necessarily one byte.  There's no user-visible
  behavior change on the to_char() side, although the internal representation
  changes.  Commit e87d4965b had already fixed most places where we skip
  characters on the basis of non-literal format patterns to advance by
  characters not bytes, but this gets one more place, the SKIP_THth macro.  I
  think everything in formatting.c gets that right now.  It'd be nice to have
  some regression test cases covering this behavior; but of course there's no
  way to do so in an encoding-agnostic way, and many of the interesting aspects
  would also require unportable locale selections.  So I've not bothered here.
  Discussion: https://postgr.es/m/28186.1510957703@...

- Merge near-duplicate code in RI triggers.  Merge ri_restrict_del and
  ri_restrict_upd into one function ri_restrict.  Create a function ri_setnull
  that is the common implementation of RI_FKey_setnull_del and
  RI_FKey_setnull_upd.  Likewise create a function ri_setdefault that is the
  common implementation of RI_FKey_setdefault_del and RI_FKey_setdefault_upd.
  All of these pairs of functions were identical except for needing to check for
  no-actual-key-change in the UPDATE cases; the one extra if-test is a small
  price to pay for saving so much code.  Aside from removing about 400 lines of
  essentially duplicate code, this allows us to recognize that we were uselessly
  caching two identical plans whenever there were pairs of triggers using these
  duplicated functions (which is likely very common).  Ildar Musin, reviewed by
  Ildus Kurbangaliev Discussion:

- Fix compiler warning in rangetypes_spgist.c.  On gcc 7.2.0, comparing pointer
  to (Datum) 0 produces a warning.  Treat it as a simple pointer to avoid that;
  this is more consistent with comparable code elsewhere, anyway.  Tomas Vondra

Peter Eisentraut pushed:

- Remove TRUE and FALSE.  Code should be using true and false.  Existing code
  can be changed to those in a backward compatible way.  The definitions in the
  ecpg header files are left around to avoid upsetting those users
  unnecessarily.  Reviewed-by: Michael Paquier <[hidden email]>

- Refactor routine to test connection to SSL server.  Move the sub-routines
  wrappers to check if a connection to a server is fine or not into the test
  main module. This is useful for other tests willing to check connectivity into
  a server.  Author: Michael Paquier <[hidden email]>

- Support channel binding 'tls-unique' in SCRAM.  This is the basic feature set
  using OpenSSL to support the feature.  In order to allow the frontend and the
  backend to fetch the sent and expected TLS Finished messages, a PG-like API is
  added to be able to make the interface pluggable for other SSL
  implementations.  This commit also adds a infrastructure to facilitate the
  addition of future channel binding types as well as libpq parameters to
  control the SASL mechanism names and channel binding names.  Those will be
  added by upcoming commits.  Some tests are added to the SSL test suite to test
  SCRAM authentication with channel binding.  Author: Michael Paquier
  <[hidden email]> Reviewed-by: Peter Eisentraut
  <[hidden email]>

- Consistently catch errors from Python _New() functions.  Python Py*_New()
  functions can fail and return NULL in out-of-memory conditions.  The previous
  code handled that inconsistently or not at all.  This change organizes that
  better.  If we are in a function that is called from Python, we just check for
  failure and return NULL ourselves, which will cause any exception information
  to be passed up.  If we are called from PostgreSQL, we consistently create an
  "out of memory" error.  Reviewed-by: Tom Lane <[hidden email]>

Andrew Dunstan pushed:

- Disable test_session_hooks test module until buildfarm issues are sorted out.

- Disable installcheck tests for test_session_hooks.  The module requires a
  preloaded library and the defect can't be cured by a LOAD instruction in the
  test script. To achieve this we override the installcheck target in the
  module's Makefile, and exclude ithe module in vcregress.pl.  Along the way,
  revert commit 9989f92aabd.

- Add hooks for session start and session end.  These hooks can be used in
  loadable modules. A simple test module is included.  Discussion:
  https://postgr.es/m/20170720204733.40f2b7eb.nagata@... Fabrízio de
  Royes Mello  and Yugo Nagata Reviewed by Michael Paquier and Aleksandr

- Back out the session_start and session_end hooks feature.  It's become
  apparent during testing that there are problems with at least the testing
  regime. I don't think we should have it without a working test regime, and the
  difficulties might indicate implementation problems anyway, so I'm backing out
  the whole thing until that's sorted out.  This reverts commits 7459484 9989f92

Andres Freund pushed:

- Provide DSM segment to ExecXXXInitializeWorker functions.  Previously,
  executor nodes running in parallel worker processes didn't have access to the
  dsm_segment object used for parallel execution.  In order to support resource
  management based on DSM segment lifetime, they need that.  So create a
  ParallelWorkerContext object to hold it and pass it to all InitializeWorker
  functions.  Author: Thomas Munro Reviewed-By: Andres Freund Discussion:

- Remove BufFile's isTemp flag.  The isTemp flag controls whether buffile.c
  chops BufFile data up into 1GB segments on disk.  Since it was badly named and
  always true, get rid of it.  Author: Thomas Munro (based on suggestion by
  Peter Geoghegan) Reviewed-By: Peter Geoghegan, Andres Freund Discussion:

== Pending Patches ==

Thomas Munro sent in another revision of a patch to implement ldaps.

Noah Misch sent in a patch to fix building PL/Perl with ActiveState Perl 5.22
and MSVC.

Mark Rofail sent in another revision of a patch to implement foriegn key arrays.

Anthony Bykov sent in another revision of a patch to add a PL/PythonU TRANSFORM
for JSONB.

Dean Rasheed sent in another revision of a patch to add a RLS summary table to
the documentation.

Fabien COELHO sent in another revision of a patch to fix a pgbench regression
test failure.

Konstantin Knizhnik sent in a patch to the PostgrSQL FDW which enables
concurrent execution of aggregates by all remote servers.

Álvaro Herrera sent in three more revisions of a patch to implement indexes for
partitioned tables.

Oliver Ford sent in another revision of a patch to fix number skipping in

Thomas Munro sent in another revision of a patch to implement parallel hash.

Masahiko Sawada sent in two revisions of a patch to fix a bug where
ginInsertCleanup called from vacuum could still miss tuples to be deleted.

Andres Freund sent in two more revisions of a patch to fix freezing of a dead
HOT-updated tuple.

Moon Insung sent in a patch to add a buffer state column to pg_buffercache and
add a function to decode buffer state.

Andrew Dunstan sent in another revision of a patch to add a toast_tuple_target
parameter to tables.

Peter Eisentraut sent in another revision of a patch to implement SQL

Anthony Bykov and Pavel Stěhule traded patches to add PL/Perl TRANSFORMs for

Alexander Korotkov sent in another revision of a patch to implement incremental

Rushabh Lathia sent in another revision of a patch to implement parallel B-tree
index build sorting.

Beena Emerson sent in another revision of a patch to implement runtime partition

Ildus Kurbangaliev sent in another revision of a patch to implement custom
compression methods.

David Rowley sent in another revision of a patch to implement bms_add_range().

Nikita Glukhov sent in another revision of a patch to implement SQL/JSON.

David Rowley sent in another revision of a patch to remove [Merge]Append nodes
which contain a single subpath.

Robert Haas sent in another revision of a patch to implement wire protocol
version negotiation.

Masahiko Sawada sent in a patch to fix a bug that manifested as a failure with
user-defined data types in logical replication.

Michaël Paquier and Masahiko Sawada traded patches to fix an assertion failure
when the non-exclusive pg_stop_backup aborted.

Jeevan Chalke sent in another revision of a patch to implement partition-wise

Peter Eisentraut sent in another revision of a patch to implement transaction
control in PL procedures.

Michaël Paquier sent in another revision of a patch to add timeline ID in
backup_label file.

Stas Kelvich sent in a patch to create a flag in pgproc indicating that hw lock
is acquired and include in RunningXacts only transaction with this flag set.

Brian Cloutier sent in a patch to add PGDLLIMPORT lines to some variables for
use on Windows.

Kyotaro HORIGUCHI sent in another revision of a patch to add columns
pg_stat_all_tables.last_vacuum_index_scans, pg_stat_all_tables.vacuum_required,
and pg_stat_all_tables.last_vacuum_status/autovacuum_fail_count.

Kyotaro HORIGUCHI sent in another revision of a patch to fix walsender timeouts
when decoding a large transaction.

Masahiko Sawada sent in a doc patch to add a missing wal_receiver_status_interval
to the subscribers section.

Fabrízio de Royes Mello sent in another revision of a patch to add hooks for
session start and session end.

Amit Langote sent in another revision of a patch to prune partitions faster.

Robert Haas sent in a patch to fix multiple problems with

Amit Langote sent in a patch to fix a bug in default range partition and
constraint exclusion.

Amit Kapila sent in another revision of a patch to pass eflags to workers.

Tomas Vondra sent in another revision of a patch to implement multivariate
histograms and MCV lists.

Tomas Vondra sent in another revision of a patch to implement BRIN multi-range

GaoZengqi (高增琦) sent in another revision of a patch to add a dependency
between client executes and static libraries.

Alik Khilazhev sent in another revision of a patch to add a zipfian distribution
in pgbench.

Sent via pgsql-announce mailing list ([hidden email])
To make changes to your subscription: