Skip to content

Wöchentlicher PostgreSQL Newsletter - 26. Juni 2011

Der Originalartikel befindet sich unter:

== Wöchentlicher PostgreSQL Newsletter - 26. Juni 2011 ==

== PostgreSQL Produkt Neuigkeiten ==

HTSQL 2.1-RC1, eine High-Level Abfragesprache für relationale
Datenbanken, ist erschienenen.

MicroOLAP Database Designer 1.8.0 für PostgreSQL ist erschienen.

psycopg2 2.4.2, ein in Python geschriebener Connector für
PostgreSQL, ist erschienen.

== PostgreSQL Jobs im Juni ==

== PostgreSQL Lokal ==

CHAR(11), die PostgreSQL Konferenz für Clustering, Hochverfügbarkeit
und Replikation hat jetzt die Registrierung eröffnet.
Die Konferenz findet am 11. und 12. Juli 2011 in Cambridge, UK statt.

PostgreSQL Conference China 2011 findet in Guangzhou am
15. und 16. Juli 2011 statt.

PDXPUG hostet PgDay am Sonntag, dem 24. Juli 2011, einen Tag
vor der OSCON in Portland, Oregon, USA. Details hier:

Postgres Open 2011, eine Konferenz die sich auf den Umbruch der
Datenbankindustrie durch PostgreSQL konzentriert, wird vom 14. bis 16.
September 2011 in Chicago, Illinois im Westin Michigan Avenue
Hotel stattfinden.

PG-Day Denver 2011 findet am Samstag, dem 17. September 2011 auf dem
Auraria Campus in der Nähe von Downtown Denver, Colorado statt.

PostgreSQL Conference West (#PgWest) findet vom 27. bis 30. September
2011 im San Jose Convention Center in Jan Jose, Kalifornen, USA statt.

PostgreSQL Conference Europe 2011 findet vom 18. bis
21. Oktober in Amsterdam statt.

pgbr findet in Sao Paulo, Brazilien, am 3. und 4. November 2011 statt.

PGConf.DE 2011 ist die Deutschsprachige PostgreSQL Konferenz
und wird am 11. November 2011 im Rheinischen Industriemuseum
in Oberhausen, Deutschland, stattfinden. Der Call for Papers ist offen.

== PostgreSQL in den News ==

Planet PostgreSQL:

Dieser wöchentliche PostgreSQL Newsletter wurde erstellt von David Fetter.

Sende Neuigkeiten und Ankündigungen bis Sonntag, 15 Uhr Pazifischer
Zeit. Bitte sende englische Beiträge an, deutsche an, italienische an, spanische an

== Reviews ==

Noah Misch reviewed the patch to fix XML handling for the bug which
produced NULL where empty string was appropriate.

== Applied Patches ==

Tom Lane pushed:

- Fix thinko in previous patch to always update
  pg_class.reltuples/relpages.  I mis-simplified the test where
  ANALYZE decided if it could get away without doing anything: under
  the new regime, that's never allowed.  Per bug #6068 from Jeff
  Janes.  Back-patch to 8.4, just like previous patch.

- Add REL9_1_STABLE to the set of branches tracked by git_changelog.

- Fix thinko in previous patch for optimizing EXISTS-within-EXISTS.
  When recursing after an optimization in
  pull_up_sublinks_qual_recurse, the available_rels value passed down
  must include only the relations that are in the righthand side of
  the new SEMI or ANTI join; it's incorrect to pull up a sub-select
  that refers to other relations, as seen in the added test case.  Per
  report from BangarRaju Vadapalli.  While at it, rethink the idea of
  recursing below a NOT EXISTS.  That is essentially the same
  situation as pulling up ANY/EXISTS sub-selects that are in the ON
  clause of an outer join, and it has the same disadvantage: we'd
  force the two joins to be evaluated according to the syntactic
  nesting order, because the lower join will most likely not be able
  to commute with the ANTI join.  That could result in having to form
  a rather large join product, whereas the handling of a correlated
  subselect is not quite that dumb.  So until we can handle those
  cases better, #ifdef NOT_USED that case.  (I think it's okay to pull
  up in the EXISTS/ANY cases, because SEMI joins aren't so inflexible
  about ordering.) Back-patch to 8.4, same as for previous patch in
  this area.  Fortunately that patch hadn't made it into any shipped
  releases yet.

- Fix missed use of "cp -i" in an example, per Fujii Masao.  Also be
  more careful about markup: use & not just &.

- Apply upstream fix for blowfish signed-character bug
  (CVE-2011-2483).  A password containing a character with the high
  bit set was misprocessed on machines where char is signed (which is
  most).  This could cause the preceding one to three characters to
  fail to affect the hashed result, thus weakening the password.  The
  result was also unportable, and failed to match some other blowfish
  implementations such as OpenBSD's.  Since the fix changes the output
  for such passwords, upstream chose to provide a compatibility hack:
  password salts beginning with $2x$ (instead of the usual $2a$ for
  blowfish) are intentionally processed "wrong" to give the same hash
  as before.  Stored password hashes can thus be modified if necessary
  to still match, though it'd be better to change any affected
  passwords.  In passing, sync a couple other upstream changes that
  marginally improve performance and/or tighten error checking.
  Back-patch to all supported branches.  Since this issue is already
  public, no reason not to commit the fix ASAP.

- Minor editing for README-SSI.  Fix some grammatical issues, try to
  clarify a couple of proofs, make the terminology more consistent.

- Fix symlink for errcodes.h so it works in VPATH builds from
  tarballs.  backend/Makefile was treating errcodes.h as a header
  always generated during build, but actually it's a header provided
  in tarballs.  Hence, must use the absolute-symlink recipe, not the
  relative-symlink one.  Per bug #6072 from Hartmut Raschick.

- Add missing -I switch for VPATH builds.  Per bug #6073 from Hartmut

- Update alternative expected file for recent sequence test changes.

- Revert "Don't select log_cnt in sequence regression tests." This
  reverts commit addf11f9a264417aa467d4e135b9a8afc59f172a.  The right
  fix for the problem is to update the alternative expected file, not
  to lobotomize the test case.

- Undo overly enthusiastic de-const-ification.  s/const//g wasn't
  exactly what I was suggesting here ... parameter declarations of the
  form "const structtype *param" are good and useful, so put those
  occurrences back.  Likewise, avoid casting away the const in a
  "const void *" parameter.

Peter Eisentraut pushed:

- Produce HISTORY file consistently as ASCII.  The release notes may
  contain non-ASCII characters (for contributor names), which lynx
  converts to the encoding determined by the current locale.  The get
  output that is deterministic and easily readable by everyone, we
  make lynx produce LATIN1 and then convert that to ASCII with
  transliteration for the non-ASCII characters.

- Message style and spelling improvements

- Add a bit of markup.

- Add information schema views role_udt_grants, udt_privileges,

- Replace tabs by spaces in makefile comment.

- Add the possibility to pass --flag arguments to xgettext calls.  The
  --flag argument can be used to tell xgettext the arguments of which
  functions should be flagged with c-format in the PO files, instead
  of guessing based on the presence of format specifiers, which fails
  if no format specifiers are present but the translation accidentally
  introduces one.  Appropriate flag settings have been added for each
  message catalog.  based on a patch by Christoph Berg for bug #6066

- Make _ be automatically included in GETTEXT_TRIGGERS.  Since it's
  globally defined in c.h, it should be treated as a gettext trigger

- Replace := by = in files.  It currently doesn't make a
  difference, but it's inconsistent with most other usage, and it
  might interfere with a future patch, so I'll change it all in a
  separate commit.  Also, replace tabs with spaces for alignment.

- Refactor common gettext triggers.  Put gettext trigger words that
  are common to the backend and backend modules into a makefile
  variable to include everywhere, to avoid error-prone repetitions.

- Remove some cruft from files

- Add missing file to GETTEXT_FILES.  This doesn't actually change the
  resulting set of strings, but better be correct.

Michael Meskes pushed:

- Fixed string in German translation that causes segfault.  Applied
  patch by Christoph Berg <> to replace placeholder "%s" by
  correct string.

Alvaro Herrera pushed:

- Remove extra copying of TupleDescs for heap_create_with_catalog.
  Some callers were creating copies of tuple descriptors to pass to
  that function, stating in code comments that it was necessary
  because it modified the passed descriptor.  Code inspection reveals
  this not to be true, and indeed not all callers are passing copies
  in the first place.  So remove the extra ones and the misleading
  comments about this behavior as well.

Heikki Linnakangas pushed:

- Fix bug introduced by recent SSI patch to merge ROLLED_BACK and
  MARKED_FOR_DEATH flags into one.  We still need the ROLLED_BACK flag
  to mark transactions that are in the process of being rolled back.
  To be precise, ROLLED_BACK now means that a transaction has already
  been discounted from the count of transactions with the oldest xmin,
  but not yet removed from the list of active transactions.  Dan Ports

- Fix bug in PreCommit_CheckForSerializationFailure.  A transaction
  that has already been marked as PREPARED cannot be killed.  Kill the
  current transaction instead.  One of the prepared_xacts regression
  tests actually hits this bug.  I removed the anomaly from the
  duplicate-gids test so that it fails in the intended way, and added
  a new test to check serialization failures with a prepared
  transaction.  Dan Ports

- Adjust the alternative expected output file for prepared_xacts test
  case, used when max_prepared_transactions=0, for the recent changes
  in the test case.

- Remove pointless const qualifiers from function arguments in the SSI
  code.  As Tom Lane pointed out, "const Relation foo" doesn't
  guarantee that you can't modify the data the "foo" pointer points
  to. It just means that you can't change the pointer to point to
  something else within the function, which is not very useful.

Robert Haas pushed:

- Performance tuning advice for SSI.  Kevin Grittner, with additional
  wordsmithing by me.

- Add notion of a "transform function" that can simplify function
  calls.  Initially, we use this only to eliminate calls to the
  varchar() function in cases where the length is not being reduced
  and, therefore, the function call is equivalent to a RelabelType
  operation.  The most significant effect of this is that we can avoid
  a table rewrite when changing a varchar(X) column to a varchar(Y)
  column, where Y > X.  Noah Misch, reviewed by me and Alexey Klyukin

- Make deadlock_timeout PGC_SUSET rather than PGC_SIGHUP.  This allows
  deadlock_timeout to be reduced for transactions that are
  particularly likely to be involved in a deadlock, thus detecting it
  more quickly.  It is also potentially useful as a poor-man's
  deadlock priority mechanism: a transaction with a high
  deadlock_timeout is less likely to be chosen as the victim than one
  with a low deadlock_timeout.  Since that could be used to game the
  system, we make this PGC_SUSET rather than PGC_USERSET.  At some
  point, it might be worth thinking about a more explicit priority
  mechanism, since using this is far from fool-proof.  But let's see
  whether there's enough use case to justify the additional work
  before we go down that route.  Noah Misch, reviewed by Shigeru

- Add smallserial pseudotype.  This is just like serial and bigserial,
  except it generates an int2 column rather than int4 or int8.  Mike
  Pultz, reviewed by Brar Piening and Josh Kupershmidt

- Add some regression tests for serial pseudotypes.  Josh Kupershmidt

- Make the visibility map crash-safe.  This involves two main changes
  from the previous behavior.  First, when we set a bit in the
  visibility map, emit a new WAL record of type XLOG_HEAP2_VISIBLE.
  Replay sets the page-level PD_ALL_VISIBLE bit and the visibility map
  bit.  Second, when inserting, updating, or deleting a tuple, we can
  no longer get away with clearing the visibility map bit after
  releasing the lock on the corresponding heap page, because an
  intervening crash might leave the visibility map bit set and the
  page-level bit clear.  Making this work requires a bit of interface
  refactoring.  In passing, a few minor but related cleanups: change
  the test in visibilitymap_set and visibilitymap_clear to throw an
  error if the wrong page (or no page) is pinned, rather than silently
  doing nothing; this case should never occur.  Also, remove duplicate
  definitions of InvalidXLogRecPtr.  Patch by me, review by Noah

- Don't select log_cnt in sequence regression tests.  It's not
  entirely stable.  Per suggestion from Josh Kupershmidt.

- Documentation improvements for pg_locks with respect to SSI.
  Explain that querying pg_locks does not simultaneously lock both the
  normal lock manager and the predicate lock manager.  Per discussion
  with Kevin Grittner.

Bruce Momjian pushed:

- Fix pg_upgrade status message capitalization mistake.  Backpatch to
  9.1 and 9.0.  Dan McGee

- Adjust pg_upgrade check for pg_upgrade_support to happen after the
  binary directory has been validated.  Backpatch to 9.1.  Dan McGee

- Rename pg_upgrade directory validation function, for clarity.

- In pg_upgrade, fix odd function parameter wrapping.

- In pg_upgrade, check that the binary and data directories are the
  same major version.  Backpatch to 9.1.  Dan McGee

- In pg_upgrade docs, clarify that link mode uses "hard" links.
  Backpatch to 9.1 and 9.0.

Joe Conway pushed:

- Async dblink functions require a named connection, and therefore
  should use DBLINK_GET_NAMED_CONN rather than DBLINK_GET_CONN.
  Problem found by Peter Eisentraut and patch by Fujii Masao.

== Abgelehnte Patches (bis jetzt) ==

No one was disappointed this week :-)

== Eingesandte Patches ==

Alexander Korotkov and Heikki Linnakangas traded revisions of a patch
to build GiST indexes more quickly.

Cedric Villemain sent in another revision, along with a review, of the
patch to constrain backend temp file space.

Alexey Klyukin sent in another revision of the patch to allow
validating configuration files without starting the postmaster.

Pavel Stehule sent in a patch for PL/pgsql to remove an unnecessary
ccache search when a array variable is updated.

Alvaro Herrera sent in two more revisions of the patch to make
keywords in pg_hba.conf field-specific.

Per review from Steve Singer, Pavel Stehule sent in another revision
of the patch to enhance error reporting in PL/pgsql.

Peter Geoghegan sent in another revision of the patch which wakes on
postmaster death.

Simon Riggs sent in two more revisions of a patch to fix an issue in a
privious patch, which was intended to take weaker locks on some kinds
of DDL.

Robert Haas and KaiGai Kohei traded patches to rework DROP.

Andrew Dunstan sent in a WIP patch to replace pgindent.

Peter Geoghegan sent in two more revisions of a patch to wake on
postmaster death.

Magnus Hagander sent in another revision of a patch to fix some
infelicities in how pg_dump uses malloc.

Florian Pflug sent in two more revisions of a patch to fix xml

Dan Ports and Kevin Grittner traded patches to skip some calls to

Bruce Momjian sent in two revisions of a patch to change the port for

Robert Haas sent in a patch intended to reduce spinlock contention.

Mark Kirkwood reviewed and revised the patch to control backend temp
file space.


No Trackbacks


Display comments as Linear | Threaded

No comments

Add Comment

Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
E-Mail addresses will not be displayed and will only be used for E-Mail notifications.
To leave a comment you must approve it via e-mail, which will be sent to your address after submission.
Form options