Package "postgresql-16"

  postgresql-16 (16.9-0ubuntu0.24.10.1) oracular-security; urgency=medium

  * New upstream version (LP: #2110377).

    + A dump/restore is not required for those running 16.X.

    + However, if you have any self-referential foreign key constraints on
      partitioned tables, it may be necessary to recreate those constraints to
      ensure that they are being enforced correctly. Follow the steps below to
      do so.

    + Also, if you have any BRIN bloom indexes, it may be advisable to reindex
      them after updating. Follow the steps below to do so.

    + Also, if you are upgrading from a version earlier than 16.5, see those
      release notes as well please.

    + Avoid one-byte buffer overread when examining invalidly-encoded strings
      that are claimed to be in GB18030 encoding (Noah Misch, Andres Freund)

      While unlikely, a SIGSEGV crash could occur if an incomplete multibyte
      character appeared at the end of memory. This was possible both in the
      server and in libpq-using applications. (CVE-2025-4207)

    + Handle self-referential foreign keys on partitioned tables correctly
      (Álvaro Herrera)

      Creating or attaching partitions failed to make the required catalog
      entries for a foreign-key constraint, if the table referenced by the
      constraint was the same partitioned table. This resulted in failure to
      enforce the constraint fully.

      To fix this, you should drop and recreate any self-referential foreign
      keys on partitioned tables, if partitions have been created or attached
      since the constraint was created. Bear in mind that violating rows might
      already be present, in which case recreating the constraint will fail,
      and you'll need to fix up those rows before trying again.

    + Avoid data loss when merging compressed BRIN summaries in
      brin_bloom_union() (Tomas Vondra)

      The code failed to account for decompression results not being identical
      to the input objects, which would result in failure to add some of the
      data to the merged summary, leading to missed rows in index searches.

      This mistake was present back to v14 where BRIN bloom indexes were
      introduced, but this code path was only rarely reached then. It's
      substantially more likely to be hit in v17 because parallel index builds
      now use the code.

    + Details about these and many further changes can be found at:
      https://www.postgresql.org/docs/16/release-16-9.html.

  * d/postgresql-16.NEWS: Update.

 -- Athos Ribeiro <email address hidden> Sun, 11 May 2025 06:15:48 -0300

  postgresql-16 (16.8-0ubuntu0.24.10.1) oracular-security; urgency=medium

  * New upstream version (LP: #2099900).

    + This release encompasses changes from upstream's 16.7 and 16.8
      releases. The former contains fixes for CVE-2025-1094 (among other
      things), and the latter was a hotfix for a problem caused by the CVE
      fix from 16.7.

    + A dump/restore is not required for those running 16.X.

    + However, if you are upgrading from a version earlier than 16.5, see
      those release notes as well please.

    + Harden PQescapeString and allied functions against invalidly-encoded
      input strings (Andres Freund, Noah Misch)

      Data-quoting functions supplied by libpq now fully check the encoding
      validity of their input. If invalid characters are detected, they report
      an error if possible. For the ones that lack an error return convention,
      the output string is adjusted to ensure that the server will report
      invalid encoding and no intervening processing will be fooled by bytes
      that might happen to match single quote, backslash, etc.

      The purpose of this change is to guard against SQL-injection attacks
      that are possible if one of these functions is used to quote crafted
      input. There is no hazard when the resulting string is sent directly to
      a PostgreSQL server (which would check its encoding anyway), but there
      is a risk when it is passed through psql or other client-side code.
      Historically such code has not carefully vetted encoding, and in many
      cases it's not clear what it should do if it did detect such a problem.

      This fix is effective only if the data-quoting function, the server, and
      any intermediate processing agree on the character encoding that's being
      used. Applications that insert untrusted input into SQL commands should
      take special care to ensure that that's true.

      Applications and drivers that quote untrusted input without using these
      libpq functions may be at risk of similar problems. They should first
      confirm the data is valid in the encoding expected by the server.

      The PostgreSQL Project thanks Stephen Fewer for reporting this problem.
      (CVE-2025-1094)

    + Improve behavior of libpq's quoting functions (Andres Freund, Tom Lane)

      The changes made for CVE-2025-1094 had one serious oversight:
      PQescapeLiteral() and PQescapeIdentifier() failed to honor their string
      length parameter, instead always reading to the input string's trailing
      null. This resulted in including unwanted text in the output, if the
      caller intended to truncate the string via the length parameter. With
      very bad luck it could cause a crash due to reading off the end of
      memory.

      In addition, modify all these quoting functions so that when invalid
      encoding is detected, an invalid sequence is substituted for just the
      first byte of the presumed character, not all of it. This reduces the
      risk of problems if a calling application performs additional processing
      on the quoted string.

    + Details about these and many further changes can be found at:
      https://www.postgresql.org/docs/16/release-16-7.html and
      https://www.postgresql.org/docs/16/release-16-8.html.

  * d/postgresql-16.NEWS: Update.

 -- Athos Ribeiro <email address hidden> Mon, 24 Feb 2025 12:03:16 -0300

  postgresql-16 (16.6-0ubuntu0.24.10.1) oracular-security; urgency=medium

  * New upstream version (LP: #2085196).

    + This release encompasses changes from upstream's 16.5 and 16.6
      releases. The former contains fixes for 4 CVEs (among other
      things), and the latter was hotfix for a problem caused by one of
      the CVE fixes from 16.5.

    + A dump/restore is not required for those running 16.X.

    + However, if you are upgrading from a version earlier than 16.3, see
      those release notes as well please.

    + Ensure cached plans are marked as dependent on the calling role when
      RLS applies to a non-top-level table reference (Nathan Bossart)

      If a CTE, subquery, sublink, security invoker view, or coercion
      projection in a query references a table with row-level security
      policies, we neglected to mark the resulting plan as potentially
      dependent on which role is executing it. This could lead to later query
      executions in the same session using the wrong plan, and then returning
      or hiding rows that should have been hidden or returned instead.

      The PostgreSQL Project thanks Wolfgang Walther for reporting this
      problem.
      (CVE-2024-10976)

    + Make libpq discard error messages
      received during SSL or GSS protocol negotiation (Jacob Champion)

      An error message received before encryption negotiation is completed
      might have been injected by a man-in-the-middle, rather than being real
      server output. Reporting it opens the door to various security hazards;
      for example, the message might spoof a query result that a careless user
      could mistake for correct output. The best answer seems to be to
      discard such data and rely only on libpq's own report of the connection
      failure.

      The PostgreSQL Project thanks Jacob Champion for reporting this
      problem.
      (CVE-2024-10977)

    + Fix unintended interactions between SET SESSION AUTHORIZATION
      and SET ROLE (Tom Lane)

      The SQL standard mandates that SET SESSION AUTHORIZATION have a
      side-effect of doing SET ROLE NONE. Our implementation of that was
      flawed, creating more interaction between the two settings than
      intended. Notably, rolling back a transaction that had done SET SESSION
      AUTHORIZATION would revert ROLE to NONE even if that had not been the
      previous state, so that the effective user ID might now be different
      from what it had been before the transaction. Transiently setting
      session_authorization in a function SET clause had a similar effect. A
      related bug was that if a parallel worker inspected
      current_setting('role'), it saw none even when it should see something
      else.

      The PostgreSQL Project thanks Tom Lane for reporting this problem.
      (CVE-2024-10978)

    + Prevent trusted PL/Perl code from changing environment variables
      (Andrew Dunstan, Noah Misch)

      The ability to manipulate process environment variables such as PATH
      gives an attacker opportunities to execute arbitrary code. Therefore,
      trustedPLs must not offer the ability to do that. To fix plperl,
      replace %ENV with a tied hash that rejects any modification attempt with
      a warning. Untrusted plperlu retains the ability to change the
      environment.

      The PostgreSQL Project thanks Coby Abrams for reporting this problem.
      (CVE-2024-10979)

    + Restore functionality of ALTER {ROLE|DATABASE} SET
      role (Tom Lane, Noah Misch)

      The fix for CVE-2024-10978 accidentally caused settings for role to
      not be applied if they come from non-interactive sources, including
      previous ALTER {ROLE|DATABASE} commands and the PGOPTIONS environment
      variable.

    + Details about these and many further changes can be found at:
      https://www.postgresql.org/docs/16/release-16-5.html and
      https://www.postgresql.org/docs/16/release-16-6.html.

  * d/postgresql-16.NEWS: Create.

 -- Sergio Durigan Junior <email address hidden> Mon, 25 Nov 2024 16:05:41 -0500