1
1
<!-- doc/src/sgml/release-9.0.sgml -->
2
2
<!-- See header comment in release.sgml about typical markup -->
4
<sect1 id="release-9-0-7">
5
<title>Release 9.0.7</title>
8
<title>Release Date</title>
9
<simpara>2012-02-27</simpara>
13
This release contains a variety of fixes from 9.0.6.
14
For information about new features in the 9.0 major release, see
15
<xref linkend="release-9-0">.
19
<title>Migration to Version 9.0.7</title>
22
A dump/restore is not required for those running 9.0.X.
26
However, if you are upgrading from a version earlier than 9.0.6,
27
see the release notes for 9.0.6.
33
<title>Changes</title>
39
Require execute permission on the trigger function for
40
<command>CREATE TRIGGER</> (Robert Haas)
44
This missing check could allow another user to execute a trigger
45
function with forged input data, by installing it on a table he owns.
46
This is only of significance for trigger functions marked
47
<literal>SECURITY DEFINER</>, since otherwise trigger functions run
48
as the table owner anyway. (CVE-2012-0866)
54
Remove arbitrary limitation on length of common name in SSL
55
certificates (Heikki Linnakangas)
59
Both <application>libpq</> and the server truncated the common name
60
extracted from an SSL certificate at 32 bytes. Normally this would
61
cause nothing worse than an unexpected verification failure, but there
62
are some rather-implausible scenarios in which it might allow one
63
certificate holder to impersonate another. The victim would have to
64
have a common name exactly 32 bytes long, and the attacker would have
65
to persuade a trusted CA to issue a certificate in which the common
66
name has that string as a prefix. Impersonating a server would also
67
require some additional exploit to redirect client connections.
74
Convert newlines to spaces in names written in <application>pg_dump</>
75
comments (Robert Haas)
79
<application>pg_dump</> was incautious about sanitizing object names
80
that are emitted within SQL comments in its output script. A name
81
containing a newline would at least render the script syntactically
82
incorrect. Maliciously crafted object names could present a SQL
83
injection risk when the script is reloaded. (CVE-2012-0868)
89
Fix btree index corruption from insertions concurrent with vacuuming
94
An index page split caused by an insertion could sometimes cause a
95
concurrently-running <command>VACUUM</> to miss removing index entries
96
that it should remove. After the corresponding table rows are removed,
97
the dangling index entries would cause errors (such as <quote>could not
98
read block N in file ...</>) or worse, silently wrong query results
99
after unrelated rows are re-inserted at the now-free table locations.
100
This bug has been present since release 8.2, but occurs so infrequently
101
that it was not diagnosed until now. If you have reason to suspect
102
that it has happened in your database, reindexing the affected index
109
Fix transient zeroing of shared buffers during WAL replay (Tom Lane)
113
The replay logic would sometimes zero and refill a shared buffer, so
114
that the contents were transiently invalid. In hot standby mode this
115
can result in a query that's executing in parallel seeing garbage data.
116
Various symptoms could result from that, but the most common one seems
117
to be <quote>invalid memory alloc request size</>.
123
Fix postmaster to attempt restart after a hot-standby crash (Tom Lane)
127
A logic error caused the postmaster to terminate, rather than attempt
128
to restart the cluster, if any backend process crashed while operating
135
Fix <command>CLUSTER</>/<command>VACUUM FULL</> handling of toast
136
values owned by recently-updated rows (Tom Lane)
140
This oversight could lead to <quote>duplicate key value violates unique
141
constraint</> errors being reported against the toast table's index
142
during one of these commands.
148
Update per-column permissions, not only per-table permissions, when
149
changing table owner (Tom Lane)
153
Failure to do this meant that any previously granted column permissions
154
were still shown as having been granted by the old owner. This meant
155
that neither the new owner nor a superuser could revoke the
156
now-untraceable-to-table-owner permissions.
162
Support foreign data wrappers and foreign servers in
163
<command>REASSIGN OWNED</> (Alvaro Herrera)
167
This command failed with <quote>unexpected classid</> errors if
168
it needed to change the ownership of any such objects.
174
Allow non-existent values for some settings in <command>ALTER
175
USER/DATABASE SET</> (Heikki Linnakangas)
179
Allow <varname>default_text_search_config</>,
180
<varname>default_tablespace</>, and <varname>temp_tablespaces</> to be
181
set to names that are not known. This is because they might be known
182
in another database where the setting is intended to be used, or for the
183
tablespace cases because the tablespace might not be created yet. The
184
same issue was previously recognized for <varname>search_path</>, and
185
these settings now act like that one.
191
Avoid crashing when we have problems deleting table files post-commit
196
Dropping a table should lead to deleting the underlying disk files only
197
after the transaction commits. In event of failure then (for instance,
198
because of wrong file permissions) the code is supposed to just emit a
199
warning message and go on, since it's too late to abort the
200
transaction. This logic got broken as of release 8.4, causing such
201
situations to result in a PANIC and an unrestartable database.
207
Recover from errors occurring during WAL replay of <command>DROP
208
TABLESPACE</> (Tom Lane)
212
Replay will attempt to remove the tablespace's directories, but there
213
are various reasons why this might fail (for example, incorrect
214
ownership or permissions on those directories). Formerly the replay
215
code would panic, rendering the database unrestartable without manual
216
intervention. It seems better to log the problem and continue, since
217
the only consequence of failure to remove the directories is some
224
Fix race condition in logging AccessExclusiveLocks for hot standby
229
Sometimes a lock would be logged as being held by <quote>transaction
230
zero</>. This is at least known to produce assertion failures on
231
slave servers, and might be the cause of more serious problems.
237
Track the OID counter correctly during WAL replay, even when it wraps
242
Previously the OID counter would remain stuck at a high value until the
243
system exited replay mode. The practical consequences of that are
244
usually nil, but there are scenarios wherein a standby server that's
245
been promoted to master might take a long time to advance the OID
246
counter to a reasonable value once values are needed.
252
Prevent emitting misleading <quote>consistent recovery state reached</>
253
log message at the beginning of crash recovery (Heikki Linnakangas)
260
<structname>pg_stat_replication</>.<structfield>replay_location</>
265
Previously, the value shown would be wrong until at least one WAL
266
record had been replayed.
272
Fix regular expression back-references with <literal>*</> attached
277
Rather than enforcing an exact string match, the code would effectively
278
accept any string that satisfies the pattern sub-expression referenced
279
by the back-reference symbol.
283
A similar problem still afflicts back-references that are embedded in a
284
larger quantified expression, rather than being the immediate subject
285
of the quantifier. This will be addressed in a future
286
<productname>PostgreSQL</> release.
292
Fix recently-introduced memory leak in processing of
293
<type>inet</>/<type>cidr</> values (Heikki Linnakangas)
297
A patch in the December 2011 releases of <productname>PostgreSQL</>
298
caused memory leakage in these operations, which could be significant
299
in scenarios such as building a btree index on such a column.
305
Fix dangling pointer after <command>CREATE TABLE AS</>/<command>SELECT
306
INTO</> in a SQL-language function (Tom Lane)
310
In most cases this only led to an assertion failure in assert-enabled
311
builds, but worse consequences seem possible.
317
Avoid double close of file handle in syslogger on Windows (MauMau)
321
Ordinarily this error was invisible, but it would cause an exception
322
when running on a debug version of Windows.
328
Fix I/O-conversion-related memory leaks in plpgsql
329
(Andres Freund, Jan Urbanski, Tom Lane)
333
Certain operations would leak memory until the end of the current
340
Improve <application>pg_dump</>'s handling of inherited table columns
345
<application>pg_dump</> mishandled situations where a child column has
346
a different default expression than its parent column. If the default
347
is textually identical to the parent's default, but not actually the
348
same (for instance, because of schema search path differences) it would
349
not be recognized as different, so that after dump and restore the
350
child would be allowed to inherit the parent's default. Child columns
351
that are <literal>NOT NULL</> where their parent is not could also be
352
restored subtly incorrectly.
358
Fix <application>pg_restore</>'s direct-to-database mode for
359
INSERT-style table data (Tom Lane)
363
Direct-to-database restores from archive files made with
364
<option>--inserts</> or <option>--column-inserts</> options fail when
365
using <application>pg_restore</> from a release dated September or
366
December 2011, as a result of an oversight in a fix for another
367
problem. The archive file itself is not at fault, and text-mode
374
Allow <application>pg_upgrade</> to process tables containing
375
<type>regclass</> columns (Bruce Momjian)
379
Since <application>pg_upgrade</> now takes care to preserve
380
<structname>pg_class</> OIDs, there was no longer any reason for this
387
Make <application>libpq</> ignore <literal>ENOTDIR</> errors
388
when looking for an SSL client certificate file
393
This allows SSL connections to be established, though without a
394
certificate, even when the user's home directory is set to something
395
like <literal>/dev/null</>.
401
Fix some more field alignment issues in <application>ecpg</>'s SQLDA area
408
Allow <literal>AT</> option in <application>ecpg</>
409
<literal>DEALLOCATE</> statements (Michael Meskes)
413
The infrastructure to support this has been there for awhile, but
414
through an oversight there was still an error check rejecting the case.
420
Do not use the variable name when defining a varchar structure in ecpg
427
Fix <filename>contrib/auto_explain</>'s JSON output mode to produce
428
valid JSON (Andrew Dunstan)
432
The output used brackets at the top level, when it should have used
439
Fix error in <filename>contrib/intarray</>'s <literal>int[] &
440
int[]</> operator (Guillaume Lelarge)
444
If the smallest integer the two input arrays have in common is 1,
445
and there are smaller values in either array, then 1 would be
446
incorrectly omitted from the result.
452
Fix error detection in <filename>contrib/pgcrypto</>'s
453
<function>encrypt_iv()</> and <function>decrypt_iv()</>
458
These functions failed to report certain types of invalid-input errors,
459
and would instead return random garbage values for incorrect input.
465
Fix one-byte buffer overrun in <filename>contrib/test_parser</>
470
The code would try to read one more byte than it should, which would
471
crash in corner cases.
472
Since <filename>contrib/test_parser</> is only example code, this is
473
not a security issue in itself, but bad example code is still bad.
479
Use <function>__sync_lock_test_and_set()</> for spinlocks on ARM, if
480
available (Martin Pitt)
484
This function replaces our previous use of the <literal>SWPB</>
485
instruction, which is deprecated and not available on ARMv6 and later.
486
Reports suggest that the old code doesn't fail in an obvious way on
487
recent ARM boards, but simply doesn't interlock concurrent accesses,
488
leading to bizarre failures in multiprocess operation.
494
Use <option>-fexcess-precision=standard</> option when building with
495
gcc versions that accept it (Andrew Dunstan)
499
This prevents assorted scenarios wherein recent versions of gcc will
500
produce creative results.
506
Allow use of threaded Python on FreeBSD (Chris Rees)
510
Our configure script previously believed that this combination wouldn't
511
work; but FreeBSD fixed the problem, so remove that error check.
4
520
<sect1 id="release-9-0-6">
5
521
<title>Release 9.0.6</title>