]> granicus.if.org Git - postgresql/log
postgresql
13 years agoPL/Python custom SPI exceptions
Peter Eisentraut [Mon, 28 Feb 2011 16:41:10 +0000 (18:41 +0200)]
PL/Python custom SPI exceptions

This provides a separate exception class for each error code that the
backend defines, as well as the ability to get the SQLSTATE from the
exception object.

Jan Urbański, reviewed by Steve Singer

13 years agoAdd documentation for data-modifying statements in WITH clauses.
Tom Lane [Mon, 28 Feb 2011 06:17:18 +0000 (01:17 -0500)]
Add documentation for data-modifying statements in WITH clauses.

Marko Tiikkaja, somewhat reworked by Tom

13 years agoDocument that last vacuum statistics and counts are for non-FULL vacuums.
Bruce Momjian [Sun, 27 Feb 2011 23:03:49 +0000 (18:03 -0500)]
Document that last vacuum statistics and counts are for non-FULL vacuums.

13 years agoPL/Python explicit subtransactions
Peter Eisentraut [Sun, 27 Feb 2011 15:09:56 +0000 (17:09 +0200)]
PL/Python explicit subtransactions

Adds a context manager, obtainable by plpy.subtransaction(), to run a
group of statements in a subtransaction.

Jan Urbański, reviewed by Steve Singer, additional scribbling by me

13 years agoRemove remaining expected file for Python 2.2
Peter Eisentraut [Sun, 27 Feb 2011 12:00:19 +0000 (14:00 +0200)]
Remove remaining expected file for Python 2.2

We don't have complete expected coverage for Python 2.2 anyway, so it
doesn't seem worth keeping this one around that no one appears to be
updating anyway.  Visual inspection of the differences ought to be
good enough for those few who care about this obsolete Python version.

13 years agoRefactor the executor's API to support data-modifying CTEs better.
Tom Lane [Sun, 27 Feb 2011 18:43:29 +0000 (13:43 -0500)]
Refactor the executor's API to support data-modifying CTEs better.

The originally committed patch for modifying CTEs didn't interact well
with EXPLAIN, as noted by myself, and also had corner-case problems with
triggers, as noted by Dean Rasheed.  Those problems show it is really not
practical for ExecutorEnd to call any user-defined code; so split the
cleanup duties out into a new function ExecutorFinish, which must be called
between the last ExecutorRun call and ExecutorEnd.  Some Asserts have been
added to these functions to help verify correct usage.

It is no longer necessary for callers of the executor to call
AfterTriggerBeginQuery/AfterTriggerEndQuery for themselves, as this is now
done by ExecutorStart/ExecutorFinish respectively.  If you really need to
suppress that and do it for yourself, pass EXEC_FLAG_SKIP_TRIGGERS to
ExecutorStart.

Also, refactor portal commit processing to allow for the possibility that
PortalDrop will invoke user-defined code.  I think this is not actually
necessary just yet, since the portal-execution-strategy logic forces any
non-pure-SELECT query to be run to completion before we will consider
committing.  But it seems like good future-proofing.

13 years agoBe less detailed about reporting shared memory failure by avoiding the
Bruce Momjian [Sun, 27 Feb 2011 17:21:25 +0000 (12:21 -0500)]
Be less detailed about reporting shared memory failure by avoiding the
output of actual Postgres parameter _values_ related to shared memory,
and suggesting that these are only possible parameters to reduce.

13 years agoFix verbose display of REPLICATION role attribute
Magnus Hagander [Sun, 27 Feb 2011 11:35:31 +0000 (12:35 +0100)]
Fix verbose display of REPLICATION role attribute

Josh Kupershmidt

13 years agoFix markup for pg_options_to_table() to report the return column names,
Bruce Momjian [Sun, 27 Feb 2011 08:50:12 +0000 (03:50 -0500)]
Fix markup for pg_options_to_table() to report the return column names,
per suggestion from Andrew.

13 years agoIncrease the default for wal_sender_delay from 200ms to 1s. Now that WAL
Heikki Linnakangas [Sat, 26 Feb 2011 21:36:24 +0000 (23:36 +0200)]
Increase the default for wal_sender_delay from 200ms to 1s. Now that WAL
sender is immediately woken up by transaction commit, there's no need to
wake up so aggressively.

13 years agoAdd a configure check for Python.h if building with python.
Andrew Dunstan [Sat, 26 Feb 2011 19:17:57 +0000 (14:17 -0500)]
Add a configure check for Python.h if building with python.

13 years agoDocument that pg_options_to_table() also works for
Bruce Momjian [Sat, 26 Feb 2011 17:47:29 +0000 (12:47 -0500)]
Document that pg_options_to_table() also works for
pg_attribute.attoptions.

13 years agoFix doc patch --- pg_options_to_table() returns "setof record".
Bruce Momjian [Sat, 26 Feb 2011 17:26:25 +0000 (12:26 -0500)]
Fix doc patch --- pg_options_to_table() returns "setof record".

13 years agoTable function support for PL/Python
Peter Eisentraut [Sat, 26 Feb 2011 14:53:11 +0000 (16:53 +0200)]
Table function support for PL/Python

This allows functions with multiple OUT parameters returning both one
or multiple records (RECORD or SETOF RECORD).

Jan Urbański, reviewed by Hitoshi Harada

13 years agoDocument pg_options_to_table() (not previously documented)
Bruce Momjian [Sat, 26 Feb 2011 06:55:26 +0000 (01:55 -0500)]
Document pg_options_to_table() (not previously documented)

13 years agoFix order of shutdown processing when CTEs contain inter-references.
Tom Lane [Sat, 26 Feb 2011 04:53:34 +0000 (23:53 -0500)]
Fix order of shutdown processing when CTEs contain inter-references.

We need ExecutorEnd to run the ModifyTable nodes to completion in
reverse order of initialization, not forward order.  Easily done
by constructing the list back-to-front.

13 years agoSupport data-modifying commands (INSERT/UPDATE/DELETE) in WITH.
Tom Lane [Fri, 25 Feb 2011 23:56:23 +0000 (18:56 -0500)]
Support data-modifying commands (INSERT/UPDATE/DELETE) in WITH.

This patch implements data-modifying WITH queries according to the
semantics that the updates all happen with the same command counter value,
and in an unspecified order.  Therefore one WITH clause can't see the
effects of another, nor can the outer query see the effects other than
through the RETURNING values.  And attempts to do conflicting updates will
have unpredictable results.  We'll need to document all that.

This commit just fixes the code; documentation updates are waiting on
author.

Marko Tiikkaja and Hitoshi Harada

13 years agoUpdate pageinspect--1.0.sql to match the upgrade script
Alvaro Herrera [Fri, 25 Feb 2011 22:39:02 +0000 (19:39 -0300)]
Update pageinspect--1.0.sql to match the upgrade script

Per comment from Tom

13 years agoFix pageinspect's heap_page_item to return infomasks as 32 bit values
Alvaro Herrera [Fri, 25 Feb 2011 22:04:25 +0000 (19:04 -0300)]
Fix pageinspect's heap_page_item to return infomasks as 32 bit values

HeapTupleHeader's t_infomask and t_infomask2 are defined as 16-bit
unsigned integers, so when the 16th bit was set, heap_page_item was
returning them as negative values, which was ugly.

The change to pageinspect--unpackaged--1.0.sql allows a module upgraded
from 9.0 to be cleanly updated from the previous definition.

13 years agoNamed restore point improvements.
Robert Haas [Thu, 24 Feb 2011 23:58:38 +0000 (18:58 -0500)]
Named restore point improvements.

Emit a log message when creating a named restore point, and improve
documentation for pg_create_restore_point().

Euler Taveira de Oliveira,  per suggestions from Thom Brown, with some
additional wordsmithing by me.

13 years agoMore psql tab-completion for new commands.
Itagaki Takahiro [Thu, 24 Feb 2011 12:05:40 +0000 (21:05 +0900)]
More psql tab-completion for new commands.

- ALTER FOREIGN DATA WRAPPER with HANDLER
- ALTER TABLE VALIDATE CONSTRAINT
- ALTER TYPE ADD VALUE
- COPY with ENCODING and FORCE NOT NULL
- CREATE FOREIGN DATA WRAPPER with HANDLER
- CREATE TRIGGER ... INSTEAD OF

13 years agoAdd tab-completion for CREATE UNLOGGED TABLE in psql,
Itagaki Takahiro [Thu, 24 Feb 2011 01:13:27 +0000 (10:13 +0900)]
Add tab-completion for CREATE UNLOGGED TABLE in psql,
and fix unexpected completion for DROP TEMP and UNIQUE.

13 years agoUpdate wording about information schema and name which views potentially
Bruce Momjian [Wed, 23 Feb 2011 04:31:45 +0000 (23:31 -0500)]
Update wording about information schema and name which views potentially
can have duplicates, per request from Tom.

13 years agoMake the second words lowercase in psql's \d titles for unlogged tables.
Itagaki Takahiro [Wed, 23 Feb 2011 00:54:32 +0000 (09:54 +0900)]
Make the second words lowercase in psql's \d titles for unlogged tables.

13 years agoAdd a relkind field to RangeTblEntry to avoid some syscache lookups.
Tom Lane [Wed, 23 Feb 2011 00:23:23 +0000 (19:23 -0500)]
Add a relkind field to RangeTblEntry to avoid some syscache lookups.

The recent additions for FDW support required checking foreign-table-ness
in several places in the parse/plan chain.  While it's not clear whether
that would really result in a noticeable slowdown, it seems best to avoid
any performance risk by keeping a copy of the relation's relkind in
RangeTblEntry.  That might have some other uses later, anyway.
Per discussion.

13 years agoAdd PL/Python functions for quoting strings
Peter Eisentraut [Tue, 22 Feb 2011 21:33:44 +0000 (23:33 +0200)]
Add PL/Python functions for quoting strings

Add functions plpy.quote_ident, plpy.quote_literal,
plpy.quote_nullable, which wrap the equivalent SQL functions.

To be able to propagate char * constness properly, make the argument
of quote_literal_cstr() const char *.  This also makes it more
consistent with quote_identifier().

Jan Urbański, reviewed by Hitoshi Harada, some refinements by Peter
Eisentraut

13 years agoFix a couple of unlogged tables goofs.
Robert Haas [Tue, 22 Feb 2011 19:42:45 +0000 (14:42 -0500)]
Fix a couple of unlogged tables goofs.

"SELECT ... INTO UNLOGGED tabname" works, but wasn't documented; CREATE
UNLOGGED SEQUENCE and CREATE UNLOGGED VIEW failed an assertion, instead
of throwing a sensible error.

Latter issue reported by Itagaki Takahiro; patch review by Tom Lane.

13 years agoAllow binary I/O of type "void".
Tom Lane [Tue, 22 Feb 2011 18:08:22 +0000 (13:08 -0500)]
Allow binary I/O of type "void".

void_send is useful for the same reason that void_out doesn't throw error,
namely that someone might do "select void_returning_func(...)"  from a
client that prefers to operate in binary mode.  The void_recv function may
or may not have any practical use, but we provide it for symmetry.

Radosław Smogura

13 years agoReword information_schema duplicate constraint warning text.
Bruce Momjian [Tue, 22 Feb 2011 14:06:18 +0000 (09:06 -0500)]
Reword information_schema duplicate constraint warning text.

13 years agoRemove ExecRemoveJunk(), which is no longer used anywhere.
Tom Lane [Tue, 22 Feb 2011 02:41:08 +0000 (21:41 -0500)]
Remove ExecRemoveJunk(), which is no longer used anywhere.

This was a leftover from the pre-8.1 design of junkfilters.  It doesn't
seem to have any reason to live, since it's merely a combination of two
easy function calls, and not a well-designed combination at that (it
encourages callers to leak the result tuple).

13 years agoFix dangling-pointer problem in before-row update trigger processing.
Tom Lane [Tue, 22 Feb 2011 02:18:04 +0000 (21:18 -0500)]
Fix dangling-pointer problem in before-row update trigger processing.

ExecUpdate checked for whether ExecBRUpdateTriggers had returned a new
tuple value by seeing if the returned tuple was pointer-equal to the old
one.  But the "old one" was in estate->es_junkFilter's result slot, which
would be scribbled on if we had done an EvalPlanQual update in response to
a concurrent update of the target tuple; therefore we were comparing a
dangling pointer to a live one.  Given the right set of circumstances we
could get a false match, resulting in not forcing the tuple to be stored in
the slot we thought it was stored in.  In the case reported by Maxim Boguk
in bug #5798, this led to "cannot extract system attribute from virtual
tuple" failures when trying to do "RETURNING ctid".  I believe there is a
very-low-probability chance of more serious errors, such as generating
incorrect index entries based on the original rather than the
trigger-modified version of the row.

In HEAD, change all of ExecBRInsertTriggers, ExecIRInsertTriggers,
ExecBRUpdateTriggers, and ExecIRUpdateTriggers so that they continue to
have similar APIs.  In the back branches I just changed
ExecBRUpdateTriggers, since there is no bug in the ExecBRInsertTriggers
case.

13 years agoMove information_schema duplicate constraint note to the top of the
Bruce Momjian [Tue, 22 Feb 2011 01:14:02 +0000 (20:14 -0500)]
Move information_schema duplicate constraint note to the top of the
information schema documentation because it affects several tables.

13 years agoUpdate docs about new role privileges reference, per Alvaro.
Bruce Momjian [Mon, 21 Feb 2011 14:15:09 +0000 (09:15 -0500)]
Update docs about new role privileges reference, per Alvaro.

13 years agoFix pg_server_to_client, that was broken in the previous commit.
Itagaki Takahiro [Mon, 21 Feb 2011 07:26:58 +0000 (16:26 +0900)]
Fix pg_server_to_client, that was broken in the previous commit.

13 years agoAdd ENCODING option to COPY TO/FROM and file_fdw.
Itagaki Takahiro [Mon, 21 Feb 2011 05:08:04 +0000 (14:08 +0900)]
Add ENCODING option to COPY TO/FROM and file_fdw.
File encodings can be specified separately from client encoding.
If not specified, client encoding is used for backward compatibility.

Cases when the encoding doesn't match client encoding are slower
than matched cases because we don't have conversion procs for other
encodings. Performance improvement would be be a future work.

Original patch by Hitoshi Harada, and modified by me.

13 years agoMerge two documentation permission chapters into a single chapter.
Bruce Momjian [Mon, 21 Feb 2011 03:20:57 +0000 (22:20 -0500)]
Merge two documentation permission chapters into a single chapter.

13 years agoMinor logic fix for new levenshtein implementation.
Tom Lane [Sun, 20 Feb 2011 19:55:07 +0000 (14:55 -0500)]
Minor logic fix for new levenshtein implementation.

Alexander Korotkov

13 years agoAdd contrib/file_fdw foreign-data wrapper for reading files via COPY.
Tom Lane [Sun, 20 Feb 2011 19:06:59 +0000 (14:06 -0500)]
Add contrib/file_fdw foreign-data wrapper for reading files via COPY.

This is both very useful in its own right, and an important test case
for the core FDW support.

This commit includes a small refactoring of copy.c to expose its option
checking code as a separately callable function.  The original patch
submission duplicated hundreds of lines of that code, which seemed pretty
unmaintainable.

Shigeru Hanada, reviewed by Itagaki Takahiro and Tom Lane

13 years agoImplement an API to let foreign-data wrappers actually be functional.
Tom Lane [Sun, 20 Feb 2011 05:17:18 +0000 (00:17 -0500)]
Implement an API to let foreign-data wrappers actually be functional.

This commit provides the core code and documentation needed.  A contrib
module test case will follow shortly.

Shigeru Hanada, Jan Urbanski, Heikki Linnakangas

13 years agoAdjust documentation wording of window function ORDER BY to not mention
Bruce Momjian [Sun, 20 Feb 2011 05:01:08 +0000 (00:01 -0500)]
Adjust documentation wording of window function ORDER BY to not mention
avg() because it was inaccurate.

13 years agoDocumentation update for UTF16 surrogate pairs in UTF8 encoding.
Bruce Momjian [Sun, 20 Feb 2011 00:00:19 +0000 (19:00 -0500)]
Documentation update for UTF16 surrogate pairs in UTF8 encoding.

Marko Kreen

13 years agoDocument that an SQL-standard-compliant query that expects a unique row
Bruce Momjian [Sat, 19 Feb 2011 23:21:03 +0000 (18:21 -0500)]
Document that an SQL-standard-compliant query that expects a unique row
from information_schema.referential_constraints might return several in
Postgres because we allow duplicate-named constraints in the same
schema.

13 years agoInvalidate PL/Python functions with composite type argument when the
Peter Eisentraut [Sat, 19 Feb 2011 14:52:24 +0000 (16:52 +0200)]
Invalidate PL/Python functions with composite type argument when the
type changes.

The invalidation will cause the type information to be refetched, and
everything will work.

Jan Urbański, reviewed by Alex Hunsaker

13 years agoInitialize variable to quiet compiler.
Bruce Momjian [Sat, 19 Feb 2011 13:14:32 +0000 (08:14 -0500)]
Initialize variable to quiet compiler.

13 years agoSet psql client encoding from locale by default
Peter Eisentraut [Sat, 19 Feb 2011 06:54:58 +0000 (08:54 +0200)]
Set psql client encoding from locale by default

Add a new libpq connection option client_encoding (which includes the
existing PGCLIENTENCODING environment variable), which besides an
encoding name accepts a special value "auto" that tries to determine
the encoding from the locale in the client's environment, using the
mechanisms that have been in use in initdb.

psql sets this new connection option to "auto" when running from a
terminal and not overridden by setting PGCLIENTENCODING.

original code by Heikki Linnakangas, with subsequent contributions by
Jaime Casanova, Peter Eisentraut, Stephen Frost, Ibrar Ahmed

13 years agoCreate the catalog infrastructure for foreign-data-wrapper handlers.
Tom Lane [Sat, 19 Feb 2011 05:06:18 +0000 (00:06 -0500)]
Create the catalog infrastructure for foreign-data-wrapper handlers.

Add a fdwhandler column to pg_foreign_data_wrapper, plus HANDLER options
in the CREATE FOREIGN DATA WRAPPER and ALTER FOREIGN DATA WRAPPER commands,
plus pg_dump support for same.  Also invent a new pseudotype fdw_handler
with properties similar to language_handler.

This is split out of the "FDW API" patch for ease of review; it's all stuff
we will certainly need, regardless of any other details of the FDW API.
FDW handler functions will not actually get called yet.

In passing, fix some omissions and infelicities in foreigncmds.c.

Shigeru Hanada, Jan Urbanski, Heikki Linnakangas

13 years agoImprove text search prefix doc addition.
Bruce Momjian [Sat, 19 Feb 2011 04:23:52 +0000 (23:23 -0500)]
Improve text search prefix doc addition.

13 years agoDocument that text search comparisons happen after text search
Bruce Momjian [Sat, 19 Feb 2011 03:39:17 +0000 (22:39 -0500)]
Document that text search comparisons happen after text search
configuration rules.

Erik Rijkers

13 years agoUn-break building with BTREE_BUILD_STATS.
Tom Lane [Fri, 18 Feb 2011 19:05:42 +0000 (14:05 -0500)]
Un-break building with BTREE_BUILD_STATS.

This has been broken for awhile, but not clear it's worth back-patching.

Euler Taveira de Oliveira

13 years agoFix parallel pg_restore to handle comments on POST_DATA items correctly.
Tom Lane [Fri, 18 Feb 2011 18:11:45 +0000 (13:11 -0500)]
Fix parallel pg_restore to handle comments on POST_DATA items correctly.

The previous coding would try to process all SECTION_NONE items in the
initial sequential-restore pass, which failed if they were dependencies of
not-yet-restored items.  Fix by postponing such items into the parallel
processing pass once we have skipped any non-PRE_DATA item.

Back-patch into 9.0; the original parallel-restore coding in 8.4 did not
have this bug, so no need to change it.

Report and diagnosis by Arnd Hannemann.

13 years agoOne more hack to make contrib upgrades from 9.0 match fresh 9.1 installs.
Tom Lane [Fri, 18 Feb 2011 16:55:57 +0000 (11:55 -0500)]
One more hack to make contrib upgrades from 9.0 match fresh 9.1 installs.

intarray and tsearch2 both reference core support functions in their GIN
opclasses, and the signatures of those functions changed for 9.1.  We added
backwards-compatible pg_proc entries for the functions in order to allow
9.0 dump files to be restored at all, but that hack leaves the opclasses
pointing at pg_proc entries different from what they'd point to if the
contrib modules were installed fresh in 9.1.  To forestall any possibility
of future problems, fix the opclasses to match fresh installs via the
expedient of direct UPDATEs on pg_amproc in the update-from-unpackaged
scripts.  (Yech ... but the alternatives are worse, or require far more
effort than seems justified right now.)

Note: updating pg_amproc is sufficient because there will be no pg_depend
entries corresponding to these dependencies, since the referenced functions
are all pinned.

13 years agoUse $INDENT instead of `which` to find the indent binary
Alvaro Herrera [Fri, 18 Feb 2011 15:47:48 +0000 (12:47 -0300)]
Use $INDENT instead of `which` to find the indent binary

Per discussion after my commit o yesterday.

13 years agoMake a hard state change from catchup to streaming mode.
Simon Riggs [Fri, 18 Feb 2011 15:07:26 +0000 (15:07 +0000)]
Make a hard state change from catchup to streaming mode.
More useful state change for monitoring purposes, plus a
required change for synchronous replication patch.

13 years agoSeparate messages for standby replies and hot standby feedback.
Simon Riggs [Fri, 18 Feb 2011 11:31:49 +0000 (11:31 +0000)]
Separate messages for standby replies and hot standby feedback.
Allow messages to be sent at different times, and greatly reduce
the frequency of hot standby feedback. Refactor to allow additional
message types.

13 years agoProperly initialize variables
Magnus Hagander [Fri, 18 Feb 2011 10:59:57 +0000 (11:59 +0100)]
Properly initialize variables

Kevin Grittner

13 years agoApplied a patch by Zoltán Böszörményi that makes ecpg's parser accept dynamic cursorn...
Michael Meskes [Fri, 18 Feb 2011 10:16:16 +0000 (11:16 +0100)]
Applied a patch by Zoltán Böszörményi that makes ecpg's parser accept dynamic cursornames even in WHERE CURRENT OF clauses.

13 years agoFix an uninitialized field in DR_copy.
Itagaki Takahiro [Fri, 18 Feb 2011 05:29:40 +0000 (14:29 +0900)]
Fix an uninitialized field in DR_copy.

Shigeru HANADA

13 years agoAdd transaction-level advisory locks.
Itagaki Takahiro [Fri, 18 Feb 2011 05:04:34 +0000 (14:04 +0900)]
Add transaction-level advisory locks.

They share the same locking namespace with the existing session-level
advisory locks, but they are automatically released at the end of the
current transaction and cannot be released explicitly via unlock
functions.

Marko Tiikkaja, reviewed by me.

13 years agoConvert Postgres arrays to Perl arrays on PL/perl input arguments
Alvaro Herrera [Fri, 18 Feb 2011 01:11:50 +0000 (22:11 -0300)]
Convert Postgres arrays to Perl arrays on PL/perl input arguments

More generally, arrays are turned in Perl array references, and row and
composite types are turned into Perl hash references.  This is done
recursively, in a way that's natural to every Perl programmer.

To avoid a backwards compatibility hit, the string representation of
each structure is also available if the function requests it.

Authors: Alexey Klyukin and Alex Hunsaker.
Some code cleanups by me.

13 years agopgindent run on plperl.c
Alvaro Herrera [Thu, 17 Feb 2011 19:40:13 +0000 (16:40 -0300)]
pgindent run on plperl.c

13 years agoUse $INDENT rather than indent throughout the pgindent code
Alvaro Herrera [Thu, 17 Feb 2011 19:37:47 +0000 (16:37 -0300)]
Use $INDENT rather than indent throughout the pgindent code

This allows the user to change the path to be used more easily.
Also, change URL in README.

13 years agoFix tsmatchsel() to account properly for null rows.
Tom Lane [Fri, 18 Feb 2011 00:00:49 +0000 (19:00 -0500)]
Fix tsmatchsel() to account properly for null rows.

ts_typanalyze.c computes MCE statistics as fractions of the non-null rows,
which seems fairly reasonable, and anyway changing it in released versions
wouldn't be a good idea.  But then ts_selfuncs.c has to account for that.
Failure to do so results in overestimates in columns with a significant
fraction of null documents.  Back-patch to 8.4 where this stuff was
introduced.

Jesper Krogh

13 years agoFix upgrade of contrib/xml2 from 9.0.
Tom Lane [Thu, 17 Feb 2011 23:11:28 +0000 (18:11 -0500)]
Fix upgrade of contrib/xml2 from 9.0.

Update script was being sloppy about two functions that have been changed
since 9.0.

13 years agoFix upgrade of contrib/intarray and contrib/unaccent from 9.0.
Tom Lane [Thu, 17 Feb 2011 22:45:09 +0000 (17:45 -0500)]
Fix upgrade of contrib/intarray and contrib/unaccent from 9.0.

Take care of a couple of discrepancies between what you get from a fresh
install and what the first-draft update-from-unpackaged scripts produced.

13 years agoFix upgrade of contrib/btree_gist from 9.0.
Tom Lane [Thu, 17 Feb 2011 21:37:34 +0000 (16:37 -0500)]
Fix upgrade of contrib/btree_gist from 9.0.

The initial version of the update-from-unpackaged script neglected to
include the <> operators that were added to the opclasses during 9.1.
To make this script produce the same final state as the regular install
script, use the same ALTER OPERATOR FAMILY trick as in pg_trgm.

13 years agoAdd client_hostname field to pg_stat_activity.
Robert Haas [Thu, 17 Feb 2011 21:03:28 +0000 (16:03 -0500)]
Add client_hostname field to pg_stat_activity.

Peter Eisentraut, reviewed by Steve Singer, Alvaro Herrera, and me.

13 years agoPrevent possible compiler warnings.
Robert Haas [Thu, 17 Feb 2011 21:00:23 +0000 (16:00 -0500)]
Prevent possible compiler warnings.

Simon Riggs reports that rnode.dbNode and rnode.spcNode were
generating unused variable warnings on gcc 4.4.3 with CFLAGS=-O1

13 years agoTypo fix
Bruce Momjian [Thu, 17 Feb 2011 20:57:03 +0000 (15:57 -0500)]
Typo fix

Erik Rijkers

13 years agoFix contrib/pg_trgm to have smoother updates from 9.0.
Tom Lane [Thu, 17 Feb 2011 20:03:30 +0000 (15:03 -0500)]
Fix contrib/pg_trgm to have smoother updates from 9.0.

Take care of some loose ends in the update-from-unpackaged script, and
apply some ugly hacks to ensure that it produces the same catalog state
as the fresh-install script.  Per discussion, this seems like a safer
plan than having two different catalog states that both call themselves
"pg_trgm 1.0", even if it's not immediately clear that the subtle
differences would ever matter.

Also, fix the stub function gin_extract_trgm() so that it works instead
of just bleating.  Needed because this function will get called during a
regular dump and reload, if there are any indexes using its opclass.
The user won't have an opportunity to update the extension till later,
so telling him to do so is unhelpful.

13 years agoRemove doc mention about read committed in upsert example.
Bruce Momjian [Thu, 17 Feb 2011 19:24:14 +0000 (14:24 -0500)]
Remove doc mention about read committed in upsert example.

13 years agoDocument that plpgsql upsert example can fail if insert triggers return
Bruce Momjian [Thu, 17 Feb 2011 18:36:36 +0000 (13:36 -0500)]
Document that plpgsql upsert example can fail if insert triggers return
unique exception, or if not used in read committed transaction mode.

13 years agoIn documentation example, use concat_values() instead of concat()
Bruce Momjian [Thu, 17 Feb 2011 17:37:13 +0000 (12:37 -0500)]
In documentation example, use concat_values() instead of concat()
because concat() is a built-in function.

Erik Rijkers

13 years agoAdd some words of caution to elog.c.
Robert Haas [Thu, 17 Feb 2011 15:29:42 +0000 (10:29 -0500)]
Add some words of caution to elog.c.

Stephen Frost, somewhat rewritten by me

13 years agoTypo fixes.
Robert Haas [Thu, 17 Feb 2011 13:07:31 +0000 (08:07 -0500)]
Typo fixes.

Thom Brown

13 years agoRemove rule to build sepgsql-regtest.pp.
Robert Haas [Thu, 17 Feb 2011 11:39:13 +0000 (06:39 -0500)]
Remove rule to build sepgsql-regtest.pp.

Instead, document how to build it manually.

Per discussion.  KaiGai Kohei, with some wordsmithing by me.

13 years agoFix blatantly uninitialized variable in recent commit.
Tom Lane [Thu, 17 Feb 2011 00:53:20 +0000 (19:53 -0500)]
Fix blatantly uninitialized variable in recent commit.

Doesn't anybody around here pay attention to compiler warnings?

13 years agoFix bogus test for hypothetical indexes in get_actual_variable_range().
Tom Lane [Thu, 17 Feb 2011 00:24:45 +0000 (19:24 -0500)]
Fix bogus test for hypothetical indexes in get_actual_variable_range().

That function was supposing that indexoid == 0 for a hypothetical index,
but that is not likely to be true in any non-toy implementation of an index
adviser, since assigning a fake OID is the only way to know at EXPLAIN time
which hypothetical index got selected.  Fix by adding a flag to
IndexOptInfo to mark hypothetical indexes.  Back-patch to 9.0 where
get_actual_variable_range() was added.

Gurjeet Singh

13 years agoAdd backwards-compatible declarations of some core GIN support functions.
Tom Lane [Wed, 16 Feb 2011 22:24:06 +0000 (17:24 -0500)]
Add backwards-compatible declarations of some core GIN support functions.

These are needed to support reloading dumps of 9.0 installations containing
contrib/intarray or contrib/tsearch2.  Since not only regular dump/reload
but binary upgrade would fail, it seems worth the trouble to carry these
stubs for awhile.  Note that the contrib opclasses referencing these
functions will still work fine, since GIN doesn't actually pay any
attention to the declared signature of a support function.

13 years agoAlso process psqlrc when running psql -l
Peter Eisentraut [Wed, 16 Feb 2011 21:14:38 +0000 (23:14 +0200)]
Also process psqlrc when running psql -l

This was previously not very useful, but with many people customizing
the linestyle, it is nice for a consistent appearance.

13 years agoBetter support for thread-support flag detection with clang
Peter Eisentraut [Wed, 16 Feb 2011 21:06:36 +0000 (23:06 +0200)]
Better support for thread-support flag detection with clang

When testing the stderr produced by various thread-support flags, also
run a compilation in addition to a link, because clang warns on
certain flags when compiling but not when linking.

13 years agoFix for warnings-free compilation with Python 3.2
Peter Eisentraut [Wed, 16 Feb 2011 20:19:29 +0000 (22:19 +0200)]
Fix for warnings-free compilation with Python 3.2

The first argument of PyEval_EvalCode() was changed from PyCodeObject*
to PyObject* because of PEP 384.

13 years agoIn pg_upgrade, no need to initialize global struct values; they are
Bruce Momjian [Wed, 16 Feb 2011 20:16:57 +0000 (15:16 -0500)]
In pg_upgrade, no need to initialize global struct values;  they are
always zeros.  Also no need to free memory before we exit.

13 years agoHot Standby feedback for avoidance of cleanup conflicts on standby.
Simon Riggs [Wed, 16 Feb 2011 19:29:37 +0000 (19:29 +0000)]
Hot Standby feedback for avoidance of cleanup conflicts on standby.
Standby optionally sends back information about oldestXmin of queries
which is then checked and applied to the WALSender's proc->xmin.
GetOldestXmin() is modified slightly to agree with GetSnapshotData(),
so that all backends on primary include WALSender within their snapshots.
Note this does nothing to change the snapshot xmin on either master or
standby. Feedback piggybacks on the standby reply message.
vacuum_defer_cleanup_age is no longer used on standby, though parameter
still exists on primary, since some use cases still exist.

Simon Riggs, review comments from Fujii Masao, Heikki Linnakangas, Robert Haas

13 years agoMake a no-op ALTER EXTENSION UPDATE give just a NOTICE, not ERROR.
Tom Lane [Wed, 16 Feb 2011 17:39:45 +0000 (12:39 -0500)]
Make a no-op ALTER EXTENSION UPDATE give just a NOTICE, not ERROR.

This seems a bit more user-friendly.

13 years agoCleanup ClusterInfo initialization in pg_upgrade
Alvaro Herrera [Wed, 16 Feb 2011 17:06:17 +0000 (14:06 -0300)]
Cleanup ClusterInfo initialization in pg_upgrade

13 years agoWAL receiver shouldn't try to send a reply when dying.
Robert Haas [Wed, 16 Feb 2011 15:24:50 +0000 (10:24 -0500)]
WAL receiver shouldn't try to send a reply when dying.

Per report from, and discussion with, Fujii Masao.

13 years agoAdd FOREACH IN ARRAY looping to plpgsql.
Tom Lane [Wed, 16 Feb 2011 06:52:04 +0000 (01:52 -0500)]
Add FOREACH IN ARRAY looping to plpgsql.

(I'm not entirely sure that we've finished bikeshedding the syntax details,
but the functionality seems OK.)

Pavel Stehule, reviewed by Stephen Frost and Tom Lane

13 years agopg_ctl promote
Robert Haas [Wed, 16 Feb 2011 02:28:48 +0000 (21:28 -0500)]
pg_ctl promote

Fujii Masao, reviewed by Robert Haas, Stephen Frost, and Magnus Hagander.

13 years agoExport the external file reader used in COPY FROM as APIs.
Itagaki Takahiro [Wed, 16 Feb 2011 02:19:11 +0000 (11:19 +0900)]
Export the external file reader used in COPY FROM as APIs.
They are expected to be used by extension modules like file_fdw.
There are no user-visible changes.

Itagaki Takahiro
Reviewed and tested by Kevin Grittner and Noah Misch.

13 years agoFix bug in 9.1 pg_upgrade processing of old/new relations; adjust debug
Bruce Momjian [Wed, 16 Feb 2011 00:01:33 +0000 (19:01 -0500)]
Fix bug in 9.1 pg_upgrade processing of old/new relations;  adjust debug
output.

13 years agoFix corner case for binary upgrade: extension functions in pg_catalog.
Tom Lane [Tue, 15 Feb 2011 23:09:41 +0000 (18:09 -0500)]
Fix corner case for binary upgrade: extension functions in pg_catalog.

Normally, pg_dump summarily excludes functions in pg_catalog from
consideration.  However, some extensions may create functions in pg_catalog
(adminpack already does that, and extensions for procedural languages will
likely do it too).  In binary-upgrade mode, we have to dump such functions,
or the extension will be incomplete after upgrading.  Per experimentation
with adminpack.

13 years agoMake pg_upgrade compile again
Alvaro Herrera [Tue, 15 Feb 2011 22:52:32 +0000 (19:52 -0300)]
Make pg_upgrade compile again

13 years agoAdd CheckTableNotInUse calls in DROP TABLE and DROP INDEX.
Tom Lane [Tue, 15 Feb 2011 20:49:54 +0000 (15:49 -0500)]
Add CheckTableNotInUse calls in DROP TABLE and DROP INDEX.

Recent releases had a check on rel->rd_refcnt in heap_drop_with_catalog,
but failed to cover the possibility of pending trigger events at DROP time.
(Before 8.4 we didn't even check the refcnt.)  When the trigger events were
eventually fired, you'd get "could not open relation with OID nnn" errors,
as in recent report from strk.  Better to throw a suitable error when the
DROP is attempted.

Also add a similar check in DROP INDEX.

Back-patch to all supported branches.

13 years agoAdjust pg_upgrade error message, array freeing, and add error check.
Bruce Momjian [Tue, 15 Feb 2011 20:00:07 +0000 (15:00 -0500)]
Adjust pg_upgrade error message, array freeing, and add error check.

13 years agoBring hstore's comment into line with style of other contrib comments.
Tom Lane [Tue, 15 Feb 2011 19:21:45 +0000 (14:21 -0500)]
Bring hstore's comment into line with style of other contrib comments.

All the other ones that are primarily a new datatype say "data type for
<purpose>", so make this one similar.

13 years agoFix obsolete comment.
Tom Lane [Tue, 15 Feb 2011 18:27:03 +0000 (13:27 -0500)]
Fix obsolete comment.

Comment about MaxAllocSize was not updated when the TOAST-header macros
were replaced in 8.3 "varvarlena" changes.  Per report from Frederik Ramm.

13 years agoAssorted corrections to the patch to add WAL receiver replies.
Robert Haas [Tue, 15 Feb 2011 17:02:53 +0000 (12:02 -0500)]
Assorted corrections to the patch to add WAL receiver replies.

Per reports from Fujii Masao.

13 years agoRename max_predicate_locks_per_transaction.
Robert Haas [Tue, 15 Feb 2011 13:00:04 +0000 (08:00 -0500)]
Rename max_predicate_locks_per_transaction.

The new name, max_pred_locks_per_transaction, is shorter.

Kevin Grittner, per discussion.

13 years agoAllow make check in PL directories
Peter Eisentraut [Mon, 14 Feb 2011 19:52:32 +0000 (21:52 +0200)]
Allow make check in PL directories

Also add make check-world target, and refactor pg_regress invocation
code in makefiles a bit.

13 years agoAvoid a few more SET DATA TYPE table rewrites.
Robert Haas [Tue, 15 Feb 2011 04:17:35 +0000 (23:17 -0500)]
Avoid a few more SET DATA TYPE table rewrites.

When the new type is an unconstrained domain over the old type, we don't
need to rewrite the table.

Noah Misch and Robert Haas