New bug fix releases: 8.4.6, 8.3.13, 8.1.23

8.4.6 has been synced into Natty.

Accepted postgresql-8.4 into maverick-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Accepted postgresql-8.4 into lucid-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Accepted postgresql-8.4 into karmic-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Accepted postgresql-8.3 into hardy-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Accepted postgresql-8.1 into dapper-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

Since the packages built correctly, the upstream test suite succeeded on all architectures for all Ubuntu releases.

I ran the postgresql-common integration tests for dapper, hardy, and lucid, which showed no regressions. I'll do karmic and maverick next.

I ran the postgresql-common integration tests for karmic and maverick as well now, all pass.

This bug was fixed in the package postgresql-8.1 - 8.1.23-0ubuntu0.6.06

---------------
postgresql-8.1 (8.1.23-0ubuntu0.6.06) dapper-proposed; urgency=low

  * 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 recovery from base backup when the starting checkpoint WAL
      record is not in the same WAL segment as its redo point.
    - 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.
    - 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.
    - 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.
    - 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 PL/pgSQL's handling of "simple" expressions to not fail in
      recursion or error-recovery cases.
    - 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.
 -- Martin Pitt <email address hidden> Tue, 21 Dec 2010 21:40:48 +0100

This bug was fixed in the package postgresql-8.3 - 8.3.13-0ubuntu8.04

---------------
postgresql-8.3 (8.3.13-0ubuntu8.04) hardy-proposed; urgency=low

  * 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.
    - 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...

This bug was fixed in the package postgresql-8.4 - 8.4.6-0ubuntu9.10

---------------
postgresql-8.4 (8.4.6-0ubuntu9.10) karmic-proposed; urgency=low

  * 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...

We also need to update the existing hardy backport. Shamefully we already forgot the 8.4.5 update (which had a security fix), sorry about that.

This bug was fixed in the package postgresql-8.4 - 8.4.6-0ubuntu10.04

---------------
postgresql-8.4 (8.4.6-0ubuntu10.04) lucid-proposed; urgency=low

  * 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 reache...

This bug was fixed in the package postgresql-8.4 - 8.4.6-0ubuntu10.10

---------------
postgresql-8.4 (8.4.6-0ubuntu10.10) maverick-proposed; urgency=low

  * 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 rea...

I: Extracting postgresql-8.4_8.4.6-1.dsc ... done.
I: Building backport of postgresql-8.4 as 8.4.6-1~hardy1 ... done.

The new postgresql-8.4 package needs a newer postgresql-common as well, I'll backport that too.