Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2015-02-06 13:18:20 UTC

* Add 15-to_char_buffer_overflow.patch and 16-to_char_buffer_overflow_time.patch:
  Fix buffer overruns in to_char() [CVE-2015-0241]
* Add 17-pgcrypto_pullf_read_max_overflow.patch and 18-pgcrypto_imath_fixes.patch:
  Fix buffer overruns in contrib/pgcrypto [CVE-2015-0243]
* Add 19-ensure_frontend_backend_sync.patch:
  Fix possible loss of frontend/backend protocol synchronization after an
  error [CVE-2015-0244]
* Add 20-column_privilege_leak.patch:
  Fix information leak via constraint-violation error messages
  [CVE-2014-8161]
* Note: CVE-2015-0242 does not affect Ubuntu packages as we use glibc's
  snprintf().

Author: Martin Pitt
Revision Date: 2015-02-06 13:18:20 UTC

* Add 15-to_char_buffer_overflow.patch and 16-to_char_buffer_overflow_time.patch:
  Fix buffer overruns in to_char() [CVE-2015-0241]
* Add 17-pgcrypto_pullf_read_max_overflow.patch and 18-pgcrypto_imath_fixes.patch:
  Fix buffer overruns in contrib/pgcrypto [CVE-2015-0243]
* Add 19-ensure_frontend_backend_sync.patch:
  Fix possible loss of frontend/backend protocol synchronization after an
  error [CVE-2015-0244]
* Add 20-column_privilege_leak.patch:
  Fix information leak via constraint-violation error messages
  [CVE-2014-8161]
* Note: CVE-2015-0242 does not affect Ubuntu packages as we use glibc's
  snprintf().

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2014-07-29 14:47:30 UTC

* New upstream bug fix release: (LP: #1348176)
  - Various data integrity and other bug fixes.
  - Secure Unix-domain sockets of temporary postmasters started during make
     check.
     Any local user able to access the socket file could connect as the
     server's bootstrap superuser, then proceed to execute arbitrary code as
     the operating-system user running the test, as we previously noted in
     CVE-2014-0067. This change defends against that risk by placing the
     server's socket in a temporary, mode 0700 subdirectory of /tmp.
  - See release notes for details:
    http://www.postgresql.org/docs/current/static/release-8-4-22.html
* Drop pg_regress patch to run tests with socket in /tmp, obsolete with
  above upstream changes and not applicable any more.

Author: Martin Pitt
Revision Date: 2014-07-24 18:17:34 UTC

* New upstream bug fix release: (LP: #1348176)
  - Various data integrity and other bug fixes.
  - Secure Unix-domain sockets of temporary postmasters started during make
     check.
     Any local user able to access the socket file could connect as the
     server's bootstrap superuser, then proceed to execute arbitrary code as
     the operating-system user running the test, as we previously noted in
     CVE-2014-0067. This change defends against that risk by placing the
     server's socket in a temporary, mode 0700 subdirectory of /tmp.
  - See release notes for details:
    http://www.postgresql.org/docs/current/static/release-8-4-22.html
* Drop pg_regress patch to run tests with socket in /tmp, obsolete with
  above upstream changes and not applicable any more.
* Add debian/postgresql-8.4.NEWS to point out that upstream support ends
  now.

Author: Martin Pitt
Revision Date: 2014-07-29 14:47:30 UTC

* New upstream bug fix release: (LP: #1348176)
  - Various data integrity and other bug fixes.
  - Secure Unix-domain sockets of temporary postmasters started during make
     check.
     Any local user able to access the socket file could connect as the
     server's bootstrap superuser, then proceed to execute arbitrary code as
     the operating-system user running the test, as we previously noted in
     CVE-2014-0067. This change defends against that risk by placing the
     server's socket in a temporary, mode 0700 subdirectory of /tmp.
  - See release notes for details:
    http://www.postgresql.org/docs/current/static/release-8-4-22.html
* Drop pg_regress patch to run tests with socket in /tmp, obsolete with
  above upstream changes and not applicable any more.

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2014-02-20 13:15:23 UTC

* New upstream security/bugfix release. (LP: #1282677)
  - Shore up GRANT ... WITH ADMIN OPTION restrictions.
    Granting a role without ADMIN OPTION is supposed to prevent the grantee
    from adding or removing members from the granted role, but this
    restriction was easily bypassed by doing SET ROLE first. The security
    impact is mostly that a role member can revoke the access of others,
    contrary to the wishes of his grantor. Unapproved role member additions
    are a lesser concern, since an uncooperative role member could provide
    most of his rights to others anyway by creating views or SECURITY
    DEFINER functions. (CVE-2014-0060)
  - Prevent privilege escalation via manual calls to PL validator functions.
    The primary role of PL validator functions is to be called implicitly
    during CREATE FUNCTION, but they are also normal SQL functions that a
    user can call explicitly. Calling a validator on a function actually
    written in some other language was not checked for and could be
    exploited for privilege-escalation purposes. The fix involves adding a
    call to a privilege-checking function in each validator function.
    Non-core procedural languages will also need to make this change to
    their own validator functions, if any. (CVE-2014-0061)
  - Avoid multiple name lookups during table and index DDL.
    If the name lookups come to different conclusions due to concurrent
    activity, we might perform some parts of the DDL on a different table
    than other parts. At least in the case of CREATE INDEX, this can be used
    to cause the permissions checks to be performed against a different
    table than the index creation, allowing for a privilege escalation
    attack. (CVE-2014-0062)
  - Prevent buffer overrun with long datetime strings.
    The MAXDATELEN constant was too small for the longest possible value of
    type interval, allowing a buffer overrun in interval_out(). Although the
    datetime input functions were more careful about avoiding buffer
    overrun, the limit was short enough to cause them to reject some valid
    inputs, such as input containing a very long timezone name. The ecpg
    library contained these vulnerabilities along with some of its own.
    (CVE-2014-0063)
  - Prevent buffer overrun due to integer overflow in size calculations.
    Several functions, mostly type input functions, calculated an allocation
    size without checking for overflow. If overflow did occur, a too-small
    buffer would be allocated and then written past. (CVE-2014-0064)
  - Prevent overruns of fixed-size buffers.
    Use strlcpy() and related functions to provide a clear guarantee that
    fixed-size buffers are not overrun. Unlike the preceding items, it is
    unclear whether these cases really represent live issues, since in most
    cases there appear to be previous constraints on the size of the input
    string. Nonetheless it seems prudent to silence all Coverity warnings of
    this type. (CVE-2014-0065)
  - Avoid crashing if crypt() returns NULL.
    There are relatively few scenarios in which crypt() could return NULL,
    but contrib/chkpass would crash if it did. One practical case in which
    this could be an issue is if libc is configured to refuse to execute
    unapproved hashing algorithms (e.g., "FIPS mode"). (CVE-2014-0066)
  - Document risks of make check in the regression testing instructions
    Since the temporary server started by make check uses "trust"
    authentication, another user on the same machine could connect to it as
    database superuser, and then potentially exploit the privileges of the
    operating-system user who started the tests. A future release will
    probably incorporate changes in the testing procedure to prevent this
    risk, but some public discussion is needed first. So for the moment,
    just warn people against using make check when there are untrusted users
    on the same machine. (CVE-2014-0067)
* The upstream tarballs no longer contain a plain HISTORY file, but point to
  the html documentation. Add 70-history.patch to note the location of these
  files in our changelog.gz file.

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2012-09-25 07:34:06 UTC

* New upstream bug fix release: (LP: #1055944)
  - Fix planner's assignment of executor parameters, and fix executor's
    rescan logic for CTE plan nodes.
    These errors could result in wrong answers from queries that scan
    the same WITH subquery multiple times.
  - Improve page-splitting decisions in GiST indexes.
    Multi-column GiST indexes might suffer unexpected bloat due to this
    error.
  - Fix cascading privilege revoke to stop if privileges are still held.
    If we revoke a grant option from some role "X", but "X" still holds
    that option via a grant from someone else, we should not
    recursively revoke the corresponding privilege from role(s) "Y"
    that "X" had granted it to.
  - Fix handling of SIGFPE when PL/Perl is in use.
    Perl resets the process's SIGFPE handler to SIG_IGN, which could
    result in crashes later on. Restore the normal Postgres signal
    handler after initializing PL/Perl.
  - Prevent PL/Perl from crashing if a recursive PL/Perl function is
    redefined while being executed.
  - Work around possible misoptimization in PL/Perl.
    Some Linux distributions contain an incorrect version of
    "pthread.h" that results in incorrect compiled code in PL/Perl,
    leading to crashes if a PL/Perl function calls another one that
    throws an error.

Author: Martin Pitt
Revision Date: 2012-09-25 07:34:06 UTC

* New upstream bug fix release: (LP: #1055944)
  - Fix planner's assignment of executor parameters, and fix executor's
    rescan logic for CTE plan nodes.
    These errors could result in wrong answers from queries that scan
    the same WITH subquery multiple times.
  - Improve page-splitting decisions in GiST indexes.
    Multi-column GiST indexes might suffer unexpected bloat due to this
    error.
  - Fix cascading privilege revoke to stop if privileges are still held.
    If we revoke a grant option from some role "X", but "X" still holds
    that option via a grant from someone else, we should not
    recursively revoke the corresponding privilege from role(s) "Y"
    that "X" had granted it to.
  - Fix handling of SIGFPE when PL/Perl is in use.
    Perl resets the process's SIGFPE handler to SIG_IGN, which could
    result in crashes later on. Restore the normal Postgres signal
    handler after initializing PL/Perl.
  - Prevent PL/Perl from crashing if a recursive PL/Perl function is
    redefined while being executed.
  - Work around possible misoptimization in PL/Perl.
    Some Linux distributions contain an incorrect version of
    "pthread.h" that results in incorrect compiled code in PL/Perl,
    leading to crashes if a PL/Perl function calls another one that
    throws an error.

Author: Jamie Strandboge
Revision Date: 2012-08-16 17:10:53 UTC

* New upstream security/bug fix release:
 - Prevent access to external files/URLs via XML entity references
   (Noah Misch, Tom Lane)
   xml_parse() would attempt to fetch external files or URLs as needed
   to resolve DTD and entity references in an XML value, thus allowing
   unprivileged database users to attempt to fetch data with the
   privileges of the database server. While the external data wouldn't
   get returned directly to the user, portions of it could be exposed
   in error messages if the data didn't parse as valid XML; and in any
   case the mere ability to check existence of a file might be useful
   to an attacker. (CVE-2012-3489)
 - Prevent access to external files/URLs via "contrib/xml2"'s
   xslt_process() (Peter Eisentraut)
   libxslt offers the ability to read and write both files and URLs
   through stylesheet commands, thus allowing unprivileged database
   users to both read and write data with the privileges of the
   database server. Disable that through proper use of libxslt's
   security options. (CVE-2012-3488)
   Also, remove xslt_process()'s ability to fetch documents and
   stylesheets from external files/URLs. While this was a documented
   "feature", it was long regarded as a bad idea. The fix for
   CVE-2012-3489 broke that capability, and rather than expend effort
   on trying to fix it, we're just going to summarily remove it.
 - Prevent too-early recycling of btree index pages (Noah Misch)
   When we allowed read-only transactions to skip assigning XIDs, we
   introduced the possibility that a deleted btree page could be
   recycled while a read-only transaction was still in flight to it.
   This would result in incorrect index search results. The
   probability of such an error occurring in the field seems very low
   because of the timing requirements, but nonetheless it should be
   fixed.
 - Fix crash-safety bug with newly-created-or-reset sequences (Tom
   Lane)
   If "ALTER SEQUENCE" was executed on a freshly created or reset
   sequence, and then precisely one nextval() call was made on it, and
   then the server crashed, WAL replay would restore the sequence to a
   state in which it appeared that no nextval() had been done, thus
   allowing the first sequence value to be returned again by the next
   nextval() call. In particular this could manifest for serial
   columns, since creation of a serial column's sequence includes an
   "ALTER SEQUENCE OWNED BY" step.
 - Ensure the "backup_label" file is fsync'd after pg_start_backup()
   (Dave Kerr)
 - Back-patch 9.1 improvement to compress the fsync request queue
   (Robert Haas)
   This improves performance during checkpoints. The 9.1 change has
   now seen enough field testing to seem safe to back-patch.
 - Only allow autovacuum to be auto-canceled by a directly blocked
   process (Tom Lane)
   The original coding could allow inconsistent behavior in some
   cases; in particular, an autovacuum could get canceled after less
   than deadlock_timeout grace period.
 - Improve logging of autovacuum cancels (Robert Haas)
 - Fix log collector so that log_truncate_on_rotation works during the
   very first log rotation after server start (Tom Lane)
 - Fix WITH attached to a nested set operation
   (UNION/INTERSECT/EXCEPT) (Tom Lane)
 - Ensure that a whole-row reference to a subquery doesn't include any
   extra GROUP BY or ORDER BY columns (Tom Lane)
 - Disallow copying whole-row references in CHECK constraints and
   index definitions during "CREATE TABLE" (Tom Lane)
   This situation can arise in "CREATE TABLE" with LIKE or INHERITS.
   The copied whole-row variable was incorrectly labeled with the row
   type of the original table not the new one. Rejecting the case
   seems reasonable for LIKE, since the row types might well diverge
   later. For INHERITS we should ideally allow it, with an implicit
   coercion to the parent table's row type; but that will require more
   work than seems safe to back-patch.
 - Fix memory leak in ARRAY(SELECT ...) subqueries (Heikki
   Linnakangas, Tom Lane)
 - Fix extraction of common prefixes from regular expressions (Tom
   Lane)
   The code could get confused by quantified parenthesized
   subexpressions, such as ^(foo)?bar. This would lead to incorrect
   index optimization of searches for such patterns.
 - Fix bugs with parsing signed "hh":"mm" and "hh":"mm":"ss" fields in
   interval constants (Amit Kapila, Tom Lane)
 - Report errors properly in "contrib/xml2"'s xslt_process() (Tom
   Lane)
 - Update time zone data files to tzdata release 2012e for DST law
   changes in Morocco and Tokelau

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2012-02-27 15:13:58 UTC

* New upstream bug fix/security release: (LP: #941912)
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 00git_inet_cidr_unpack.patch, 04-armel-tas.patch: applied upstream.

Author: Martin Pitt
Revision Date: 2012-02-27 15:13:58 UTC

* New upstream bug fix/security release: (LP: #941912)
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 00git_inet_cidr_unpack.patch, 04-armel-tas.patch: applied upstream.

Author: Martin Pitt
Revision Date: 2012-02-25 11:17:15 UTC

* Urgency medium due to security fixes.
* New upstream bug fix/security release:
  - Require execute permission on the trigger function for "CREATE
    TRIGGER".
    This missing check could allow another user to execute a trigger
    function with forged input data, by installing it on a table he
    owns. This is only of significance for trigger functions marked
    SECURITY DEFINER, since otherwise trigger functions run as the
    table owner anyway. (CVE-2012-0866)
  - Remove arbitrary limitation on length of common name in SSL
    certificates.
    Both libpq and the server truncated the common name extracted from
    an SSL certificate at 32 bytes. Normally this would cause nothing
    worse than an unexpected verification failure, but there are some
    rather-implausible scenarios in which it might allow one
    certificate holder to impersonate another. The victim would have to
    have a common name exactly 32 bytes long, and the attacker would
    have to persuade a trusted CA to issue a certificate in which the
    common name has that string as a prefix. Impersonating a server
    would also require some additional exploit to redirect client
    connections. (CVE-2012-0867)
  - Convert newlines to spaces in names written in pg_dump comments.
    pg_dump was incautious about sanitizing object names that are
    emitted within SQL comments in its output script. A name containing
    a newline would at least render the script syntactically incorrect.
    Maliciously crafted object names could present a SQL injection risk
    when the script is reloaded. (CVE-2012-0868)
  - Fix btree index corruption from insertions concurrent with
    vacuuming.
    An index page split caused by an insertion could sometimes cause a
    concurrently-running "VACUUM" to miss removing index entries that
    it should remove. After the corresponding table rows are removed,
    the dangling index entries would cause errors (such as "could not
    read block N in file ...") or worse, silently wrong query results
    after unrelated rows are re-inserted at the now-free table
    locations. This bug has been present since release 8.2, but occurs
    so infrequently that it was not diagnosed until now. If you have
    reason to suspect that it has happened in your database, reindexing
    the affected index will fix things.
  - Update per-column permissions, not only per-table permissions, when
    changing table owner.
    Failure to do this meant that any previously granted column
    permissions were still shown as having been granted by the old
    owner. This meant that neither the new owner nor a superuser could
    revoke the now-untraceable-to-table-owner permissions.
  - Allow non-existent values for some settings in "ALTER USER/DATABASE
    SET".
    Allow default_text_search_config, default_tablespace, and
    temp_tablespaces to be set to names that are not known. This is
    because they might be known in another database where the setting
    is intended to be used, or for the tablespace cases because the
    tablespace might not be created yet. The same issue was previously
    recognized for search_path, and these settings now act like that
    one.
  - Avoid crashing when we have problems deleting table files
    post-commit.
    Dropping a table should lead to deleting the underlying disk files
    only after the transaction commits. In event of failure then (for
    instance, because of wrong file permissions) the code is supposed
    to just emit a warning message and go on, since it's too late to
    abort the transaction. This logic got broken as of release 8.4,
    causing such situations to result in a PANIC and an unrestartable
    database.
  - Track the OID counter correctly during WAL replay, even when it
    wraps around.
    Previously the OID counter would remain stuck at a high value until
    the system exited replay mode. The practical consequences of that
    are usually nil, but there are scenarios wherein a standby server
    that's been promoted to master might take a long time to advance
    the OID counter to a reasonable value once values are needed.
  - Fix regular expression back-references with - attached.
    Rather than enforcing an exact string match, the code would
    effectively accept any string that satisfies the pattern
    sub-expression referenced by the back-reference symbol.
    A similar problem still afflicts back-references that are embedded
    in a larger quantified expression, rather than being the immediate
    subject of the quantifier. This will be addressed in a future
    PostgreSQL release.
  - Fix recently-introduced memory leak in processing of inet/cidr
    values.
  - Fix dangling pointer after "CREATE TABLE AS"/"SELECT INTO" in a
    SQL-language function.
    In most cases this only led to an assertion failure in
    assert-enabled builds, but worse consequences seem possible.
  - Fix I/O-conversion-related memory leaks in plpgsql.
  - Improve pg_dump's handling of inherited table columns.
    pg_dump mishandled situations where a child column has a different
    default expression than its parent column. If the default is
    textually identical to the parent's default, but not actually the
    same (for instance, because of schema search path differences) it
    would not be recognized as different, so that after dump and
    restore the child would be allowed to inherit the parent's default.
    Child columns that are NOT NULL where their parent is not could
    also be restored subtly incorrectly.
  - Fix pg_restore's direct-to-database mode for INSERT-style table
    data.
    Direct-to-database restores from archive files made with
    "--inserts" or "--column-inserts" options fail when using
    pg_restore from a release dated September or December 2011, as a
    result of an oversight in a fix for another problem. The archive
    file itself is not at fault, and text-mode output is okay.
  - Allow AT option in ecpg DEALLOCATE statements.
    The infrastructure to support this has been there for awhile, but
    through an oversight there was still an error check rejecting the
    case.
  - Fix error in "contrib/intarray"'s int[] & int[] operator.
    If the smallest integer the two input arrays have in common is 1,
    and there are smaller values in either array, then 1 would be
    incorrectly omitted from the result.
  - Fix error detection in "contrib/pgcrypto"'s encrypt_iv() and
    decrypt_iv().
    These functions failed to report certain types of invalid-input
    errors, and would instead return random garbage values for
    incorrect input.
  - Fix one-byte buffer overrun in "contrib/test_parser".
    The code would try to read one more byte than it should, which
    would crash in corner cases. Since "contrib/test_parser" is only
    example code, this is not a security issue in itself, but bad
    example code is still bad.
  - Use __sync_lock_test_and_set() for spinlocks on ARM, if available.
    This function replaces our previous use of the SWPB instruction,
    which is deprecated and not available on ARMv6 and later. Reports
    suggest that the old code doesn't fail in an obvious way on recent
    ARM boards, but simply doesn't interlock concurrent accesses,
    leading to bizarre failures in multiprocess operation.
  - Use "-fexcess-precision=standard" option when building with gcc
    versions that accept it.
    This prevents assorted scenarios wherein recent versions of gcc
    will produce creative results.
  - Allow use of threaded Python on FreeBSD.
    Our configure script previously believed that this combination
    wouldn't work; but FreeBSD fixed the problem, so remove that error
    check.
* Drop 04-armel-tas.patch, applied upstream.

Author: Martin Pitt
Revision Date: 2012-01-05 12:41:27 UTC

* Add 00git_inet_cidr_unpack.patch: Revert the behavior of inet/cidr
  functions to not unpack the arguments. This fixes the memory leak when
  sorting inet values. Patch taken from upstream git HEAD. Spotted during
  testing in LP #904631.
* 01-armel-tas.patch: Turn slock_t datatype into an int, and define
  S_UNLOCK() to call __sync_lock_release() instead of using the default
  implementation. This complies to the gcc built-in atomic operations
  specifiction more strictly and now also works on the Panda boards.
  (LP: #904828)

Author: Ubuntu Archive Auto-Sync
Revision Date: 2011-12-15 10:14:04 UTC

Automated backport upload; no source changes.

Author: Martin Pitt
Revision Date: 2011-04-20 16:10:45 UTC

* New upstream bug fix release: (LP: #767165)
  - If your installation was upgraded from a previous major release by
    running pg_upgrade, you should take action to prevent possible data loss
    due to a now-fixed bug in pg_upgrade. The recommended solution is to run
    "VACUUM FREEZE" on all TOAST tables. More information is available at
    http://wiki.postgresql.org/wiki/20110408pg_upgrade_fix.
  - Fix pg_upgrade's handling of TOAST tables.
    This error poses a significant risk of data loss for installations
    that have been upgraded with pg_upgrade. This patch corrects the
    problem for future uses of pg_upgrade, but does not in itself cure
    the issue in installations that have been processed with a buggy
    version of pg_upgrade.
  - Suppress incorrect "PD_ALL_VISIBLE flag was incorrectly set"
    warning.
  - Disallow including a composite type in itself.
  - Avoid potential deadlock during catalog cache initialization.
  - Fix dangling-pointer problem in BEFORE ROW UPDATE trigger handling
    when there was a concurrent update to the target tuple.
  - Disallow "DROP TABLE" when there are pending deferred trigger
    events for the table.
    Formerly the "DROP" would go through, leading to "could not open
    relation with OID nnn" errors when the triggers were eventually
    fired.
  - Prevent crash triggered by constant-false WHERE conditions during
    GEQO optimization.
  - Improve planner's handling of semi-join and anti-join cases.
  - Fix selectivity estimation for text search to account for NULLs.
  - Improve PL/pgSQL's ability to handle row types with dropped columns.
  - Fix PL/Python memory leak involving array slices.
  - Fix pg_restore to cope with long lines (over 1KB) in TOC files.
  - Put in more safeguards against crashing due to division-by-zero
    with overly enthusiastic compiler optimization. (Closes: #616180)

Author: Martin Pitt
Revision Date: 2011-02-01 22:30:52 UTC

* New upstream security/bug fix release: (LP: #711318)
  - Fix buffer overrun in "contrib/intarray"'s input function for the
    query_int type.
    This bug is a security risk since the function's return address
    could be overwritten. Thanks to Apple Inc's security team for
    reporting this issue and supplying the fix. (CVE-2010-4015)
  - Avoid failures when "EXPLAIN" tries to display a simple-form CASE
    expression.
    If the CASE's test expression was a constant, the planner could
    simplify the CASE into a form that confused the expression-display
    code, resulting in "unexpected CASE WHEN clause" errors.
  - Fix assignment to an array slice that is before the existing range
    of subscripts.
    If there was a gap between the newly added subscripts and the first
    pre-existing subscript, the code miscalculated how many entries
    needed to be copied from the old array's null bitmap, potentially
    leading to data corruption or crash.
  - Avoid unexpected conversion overflow in planner for very distant
    date values.
    The date type supports a wider range of dates than can be
    represented by the timestamp types, but the planner assumed it
    could always convert a date to timestamp with impunity.
  - Fix pg_restore's text output for large objects (BLOBs) when
    standard_conforming_strings is on.
    Although restoring directly to a database worked correctly, string
    escaping was incorrect if pg_restore was asked for SQL text output
    and standard_conforming_strings had been enabled in the source
    database.
  - Fix erroneous parsing of tsquery values containing ... &
    !(subexpression) | ... .
    Queries containing this combination of operators were not executed
    correctly. The same error existed in "contrib/intarray"'s query_int
    type and "contrib/ltree"'s ltxtquery type.
  - Fix bug in "contrib/seg"'s GiST picksplit algorithm.
    This could result in considerable inefficiency, though not actually
    incorrect answers, in a GiST index on a seg column. If you have
    such an index, consider "REINDEX"ing it after installing this
    update. (This is identical to the bug that was fixed in
    "contrib/cube" in the previous update.)

Author: Martin Pitt
Revision Date: 2011-05-22 15:41:07 UTC

* debian/postgresql-8.4.postrm: Clean up pg_ctl.conf on purge.
* debian/control, debian/rules: Drop usage of pycentral. We don't ship any
  Python extension/module, so we don't need a python helper at all.
  (Closes: #616948)
* Add 16-cmsgcred-size.patch: Fix size of struct cmsgcred to fix ident
  authentication on kFreeBSD 64 bit. Thanks to Petr Salinger for the patch!
  (Closes: #612888)

Author: Peter Pearse
Revision Date: 2011-04-07 12:52:19 UTC

Use multiarch paths when cross building.

Author: Martin Pitt
Revision Date: 2011-02-01 22:30:52 UTC

* New upstream security/bug fix release: (LP: #711318)
  - Fix buffer overrun in "contrib/intarray"'s input function for the
    query_int type.
    This bug is a security risk since the function's return address
    could be overwritten. Thanks to Apple Inc's security team for
    reporting this issue and supplying the fix. (CVE-2010-4015)
  - Avoid failures when "EXPLAIN" tries to display a simple-form CASE
    expression.
    If the CASE's test expression was a constant, the planner could
    simplify the CASE into a form that confused the expression-display
    code, resulting in "unexpected CASE WHEN clause" errors.
  - Fix assignment to an array slice that is before the existing range
    of subscripts.
    If there was a gap between the newly added subscripts and the first
    pre-existing subscript, the code miscalculated how many entries
    needed to be copied from the old array's null bitmap, potentially
    leading to data corruption or crash.
  - Avoid unexpected conversion overflow in planner for very distant
    date values.
    The date type supports a wider range of dates than can be
    represented by the timestamp types, but the planner assumed it
    could always convert a date to timestamp with impunity.
  - Fix pg_restore's text output for large objects (BLOBs) when
    standard_conforming_strings is on.
    Although restoring directly to a database worked correctly, string
    escaping was incorrect if pg_restore was asked for SQL text output
    and standard_conforming_strings had been enabled in the source
    database.
  - Fix erroneous parsing of tsquery values containing ... &
    !(subexpression) | ... .
    Queries containing this combination of operators were not executed
    correctly. The same error existed in "contrib/intarray"'s query_int
    type and "contrib/ltree"'s ltxtquery type.
  - Fix bug in "contrib/seg"'s GiST picksplit algorithm.
    This could result in considerable inefficiency, though not actually
    incorrect answers, in a GiST index on a seg column. If you have
    such an index, consider "REINDEX"ing it after installing this
    update. (This is identical to the bug that was fixed in
    "contrib/cube" in the previous update.)

Author: Martin Pitt
Revision Date: 2010-12-21 21:17:08 UTC

* New upstream bug fix release: (LP: #693157)
  - Force the default wal_sync_method to be fdatasync on Linux.
    The default on Linux has actually been fdatasync for many years,
    but recent kernel changes caused PostgreSQL to choose open_datasync
    instead. This choice did not result in any performance improvement,
    and caused outright failures on certain filesystems, notably ext4
    with the data=journal mount option.
  - Fix assorted bugs in WAL replay logic for GIN indexes.
    This could result in "bad buffer id: 0" failures or corruption of
    index contents during replication.
  - Fix recovery from base backup when the starting checkpoint WAL
    record is not in the same WAL segment as its redo point.
  - Fix persistent slowdown of autovacuum workers when multiple workers
    remain active for a long time.
    The effective vacuum_cost_limit for an autovacuum worker could drop
    to nearly zero if it processed enough tables, causing it to run
    extremely slowly.
  - Add support for detecting register-stack overrun on IA64.
    The IA64 architecture has two hardware stacks. Full prevention of
    stack-overrun failures requires checking both.
  - Add a check for stack overflow in copyObject().
    Certain code paths could crash due to stack overflow given a
    sufficiently complex query.
  - Fix detection of page splits in temporary GiST indexes.
    It is possible to have a "concurrent" page split in a temporary
    index, if for example there is an open cursor scanning the index
    when an insertion is done. GiST failed to detect this case and
    hence could deliver wrong results when execution of the cursor
    continued.
  - Fix error checking during early connection processing.
    The check for too many child processes was skipped in some cases,
    possibly leading to postmaster crash when attempting to add the new
    child process to fixed-size arrays.
  - Improve efficiency of window functions.
    Certain cases where a large number of tuples needed to be read in
    advance, but work_mem was large enough to allow them all to be held
    in memory, were unexpectedly slow. percent_rank(), cume_dist() and
    ntile() in particular were subject to this problem.
  - Avoid memory leakage while "ANALYZE"'ing complex index expressions.
  - Ensure an index that uses a whole-row Var still depends on its
    table.
    An index declared like create index i on t (foo(t.-)) would not
    automatically get dropped when its table was dropped.
  - Do not "inline" a SQL function with multiple OUT parameters.
    This avoids a possible crash due to loss of information about the
    expected result rowtype.
  - Behave correctly if ORDER BY, LIMIT, FOR UPDATE, or WITH is
    attached to the VALUES part of INSERT ... VALUES.
  - Fix constant-folding of COALESCE() expressions.
    The planner would sometimes attempt to evaluate sub-expressions
    that in fact could never be reached, possibly leading to unexpected
    errors.
  - Fix postmaster crash when connection acceptance (accept() or one of
    the calls made immediately after it) fails, and the postmaster was
    compiled with GSSAPI support.
  - Fix missed unlink of temporary files when log_temp_files is active.
    If an error occurred while attempting to emit the log message, the
    unlink was not done, resulting in accumulation of temp files.
  - Add print functionality for InhRelation nodes.
    This avoids a failure when debug_print_parse is enabled and certain
    types of query are executed.
  - Fix incorrect calculation of distance from a point to a horizontal
    line segment.
    This bug affected several different geometric distance-measurement
    operators.
  - Fix incorrect calculation of transaction status in ecpg.
  - Fix PL/pgSQL's handling of "simple" expressions to not fail in
    recursion or error-recovery cases.
  - Fix PL/Python's handling of set-returning functions.
    Attempts to call SPI functions within the iterator generating a set
    result would fail.
  - Fix bug in "contrib/cube"'s GiST picksplit algorithm.
    This could result in considerable inefficiency, though not actually
    incorrect answers, in a GiST index on a cube column. If you have
    such an index, consider "REINDEX"ing it after installing this
    update.
  - Don't emit "identifier will be truncated" notices in
    "contrib/dblink" except when creating new connections.
  - Fix potential coredump on missing public key in "contrib/pgcrypto".
  - Fix memory leak in "contrib/xml2"'s XPath query functions.

Author: Martin Pitt
Revision Date: 2010-07-19 23:29:03 UTC

* Migrate to a common init script for all server versions, to avoid
  providing the "postgresql" service in multiple packages (which causes
  insserv to complain bitterly):
  - Drop debian/postgresql-8.4.init.
  - debian/control: Bump dependency to postgresql-common to ensure we have a
    common /etc/init.d/postgresql init script.
  - debian/postgresql-8.4.preinst: Remove/rename our init script on upgrade.
  - debian/postgresql-8.4.prerm: Call stop_version on upgrade.
  - debian/rules: Drop dh_installinit arguments.
  - (Closes: #585890)

Author: Martin Pitt
Revision Date: 2010-05-19 09:30:34 UTC

Automated backport upload; no source changes.

Author: Martin Pitt
Revision Date: 2010-03-13 16:44:46 UTC

* New upstream bug fix release:
  - Add new configuration parameter ssl_renegotiation_limit to control
    how often we do session key renegotiation for an SSL connection.
    This can be set to zero to disable renegotiation completely, which
    may be required if a broken SSL library is used. In particular,
    some vendors are shipping stopgap patches for CVE-2009-3555 that
    cause renegotiation attempts to fail.
  - Fix possible deadlock during backend startup.
  - Fix possible crashes due to not handling errors during relcache
    reload cleanly.
  - Fix possible crash due to use of dangling pointer to a cached plan.
  - Fix possible crash due to overenthusiastic invalidation of cached
    plan for "ROLLBACK".
  - Fix possible crashes when trying to recover from a failure in
    subtransaction start.
  - Fix server memory leak associated with use of savepoints and a
    client encoding different from server's encoding.
  - Fix incorrect WAL data emitted during end-of-recovery cleanup of a
    GIST index page split.
  - Fix bug in WAL redo cleanup method for GIN indexes.
  - Fix incorrect comparison of scan key in GIN index search.
  - Make substring() for bit types treat any negative length as meaning
    "all the rest of the string". The previous coding treated only -1 that
    way, and would produce an invalid result value for other negative
    values, possibly leading to a crash (CVE-2010-0442).
  - Fix integer-to-bit-string conversions to handle the first
    fractional byte correctly when the output bit width is wider than
    the given integer by something other than a multiple of 8 bits.
  - Fix some cases of pathologically slow regular expression matching.
  - Fix bug occurring when trying to inline a SQL function that returns
    a set of a composite type that contains dropped columns.
  - Fix bug with trying to update a field of an element of a
    composite-type array column.
  - Avoid failure when "EXPLAIN" has to print a FieldStore or
    assignment ArrayRef expression.
    These cases can arise now that "EXPLAIN VERBOSE" tries to print
    plan node target lists.
  - Avoid an unnecessary coercion failure in some cases where an
    undecorated literal string appears in a subquery within
    "UNION"/"INTERSECT"/"EXCEPT".
    This fixes a regression for some cases that worked before 8.4.
  - Avoid undesirable rowtype compatibility check failures in some
    cases where a whole-row Var has a rowtype that contains dropped
    columns.
  - Fix the STOP WAL LOCATION entry in backup history files to report
    the next WAL segment's name when the end location is exactly at a
    segment boundary.
  - Always pass the catalog ID to an option validator function
    specified in "CREATE FOREIGN DATA WRAPPER".
  - Fix some more cases of temporary-file leakage.
    This corrects a problem introduced in the previous minor release.
    One case that failed is when a plpgsql function returning set is
    called within another function's exception handler.
  - Add support for doing FULL JOIN ON FALSE.
    This prevents a regression from pre-8.4 releases for some queries
    that can now be simplified to a constant-false join condition.
  - Improve constraint exclusion processing of boolean-variable cases,
    in particular make it possible to exclude a partition that has a
    "bool_column = false" constraint.
  - Prevent treating an INOUT cast as representing binary compatibility.
  - Include column name in the message when warning about inability to
    grant or revoke column-level privileges.
    This is more useful than before and helps to prevent confusion when
    a "REVOKE" generates multiple messages, which formerly appeared to
    be duplicates.
  - When reading "pg_hba.conf" and related files, do not treat
    @something as a file inclusion request if the @ appears inside
    quote marks; also, never treat @ by itself as a file inclusion
    request.
    This prevents erratic behavior if a role or database name starts
    with @. If you need to include a file whose path name contains
    spaces, you can still do so, but you must write @"/path to/file"
    rather than putting the quotes around the whole construct.
  - Prevent infinite loop on some platforms if a directory is named as
    an inclusion target in "pg_hba.conf" and related files.
  - Fix possible infinite loop if SSL_read or SSL_write fails without
    setting errno.
    This is reportedly possible with some Windows versions of openssl.
  - Disallow GSSAPI authentication on local connections, since it
    requires a hostname to function correctly.
  - Protect ecpg against applications freeing strings unexpectedly.
  - Make ecpg report the proper SQLSTATE if the connection disappears.
  - Fix translation of cell contents in psql \d output.
  - Fix psql's numericlocale option to not format strings it shouldn't
    in latex and troff output formats.
  - Fix a small per-query memory leak in psql.
  - Make psql return the correct exit status (3) when ON_ERROR_STOP and
    --single-transaction are both specified and an error occurs during
    the implied "COMMIT".
  - Fix pg_dump's output of permissions for foreign servers.
  - Fix possible crash in parallel pg_restore due to out-of-range
    dependency IDs.
  - Fix plpgsql failure in one case where a composite column is set to
    NULL.
  - Fix possible failure when calling PL/Perl functions from PL/PerlU
    or vice versa.
  - Add volatile markings in PL/Python to avoid possible
    compiler-specific misbehavior>
  - Ensure PL/Tcl initializes the Tcl interpreter.
    The only known symptom of this oversight is that the Tcl clock
    command misbehaves if using Tcl 8.5 or later.
  - Prevent ExecutorEnd from being run on portals created within a
    failed transaction or subtransaction.
    This is known to cause issues when using "contrib/auto_explain".
  - Prevent crash in "contrib/dblink" when too many key columns are
    specified to a dblink_build_sql_- function.
  - Allow zero-dimensional arrays in "contrib/ltree" operations.
    This case was formerly rejected as an error, but it's more
    convenient to treat it the same as a zero-element array. In
    particular this avoids unnecessary failures when an ltree operation
    is applied to the result of ARRAY(SELECT ...) and the sub-select
    returns no rows.
  - Fix assorted crashes in "contrib/xml2" caused by sloppy memory
    management.
* Rebuild against libossp-uuid16. (Closes: #570108, LP: #538284)

Author: Ubuntu Archive Auto-Sync
Revision Date: 2010-01-19 09:11:30 UTC

Automated backport upload; no source changes.

Author: Martin Pitt
Revision Date: 2009-09-06 14:11:13 UTC

* Urgency medium due to security fix.
* New upstream security/bug fix release:
  - Disallow "RESET ROLE" and "RESET SESSION AUTHORIZATION" inside
    security-definer functions. This covers a case that was missed in the
    previous patch that disallowed "SET ROLE" and "SET SESSION
    AUTHORIZATION" inside security-definer functions. [CVE-2007-6600]
  - Fix WAL page header initialization at the end of archive recovery.
    This could lead to failure to process the WAL in a subsequent archive
    recovery.
  - Fix "cannot make new WAL entries during recovery" error.
  - Fix problem that could make expired rows visible after a crash.
    This bug involved a page status bit potentially not being set
    correctly after a server crash.
  - Make "LOAD" of an already-loaded loadable module into a no-op.
    Formerly, "LOAD" would attempt to unload and re-load the module,
    but this is unsafe and not all that useful.
  - Make window function PARTITION BY and ORDER BY items always be
    interpreted as simple expressions.
    In 8.4.0 these lists were parsed following the rules used for
    top-level GROUP BY and ORDER BY lists. But this was not correct per
    the SQL standard, and it led to possible circularity.
  - Fix several errors in planning of semi-joins. These led to wrong query
    results in some cases where IN or EXISTS was used together with another
    join.
  - Fix handling of whole-row references to subqueries that are within
    an outer join. An example is SELECT COUNT(ss.-) FROM ... LEFT JOIN
    (SELECT ...) ss ON .... Here, ss.- would be treated as
    ROW(NULL,NULL,...) for null-extended join rows, which is not the same as
    a simple NULL. Now it is treated as a simple NULL.
  - Fix locale handling with plperl. This bug could cause the server's
    locale setting to change when a plperl function is called, leading to
    data corruption.
  - Fix handling of reloptions to ensure setting one option doesn't
    force default values for others.
  - Ensure that a "fast shutdown" request will forcibly terminate open
    sessions, even if a "smart shutdown" was already in progress.
  - Avoid memory leak for array_agg() in GROUP BY queries.
  - Treat to_char(..., 'TH') as an uppercase ordinal suffix with
    'HH'/'HH12'. It was previously handled as 'th'.
  - Include the fractional part in the result of EXTRACT(second) and
    EXTRACT(milliseconds) for time and time with time zone inputs.
    This has always worked for floating-point datetime configurations,
    but was broken in the integer datetime code.
  - Fix overflow for INTERVAL 'x ms' when "x" is more than 2 million
    and integer datetimes are in use.
  - Improve performance when processing toasted values in index scans.
    This is particularly useful for PostGIS.
  - Fix a typo that disabled commit_delay.
  - Output early-startup messages to "postmaster.log" if the server is
    started in silent mode. Previously such error messages were discarded,
    leading to difficulty in debugging.
  - Remove translated FAQs. They are now on the wiki. The main FAQ was moved
    to the wiki some time ago.
  - Fix pg_ctl to not go into an infinite loop if "postgresql.conf" is
    empty.
  - Fix several errors in pg_dump's --binary-upgrade mode. pg_dump
    --binary-upgrade is used by pg_migrator.
  - Fix "contrib/xml2"'s xslt_process() to properly handle the maximum
    number of parameters (twenty).
  - Improve robustness of libpq's code to recover from errors during
    "COPY FROM STDIN".
  - Avoid including conflicting readline and editline header files when
    both libraries are installed.
  - Work around gcc bug that causes "floating-point exception" instead
    of "division by zero" on some platforms.
* debian/control: Bump Standards-Version to 3.8.3 (no changes necessary).