Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 18:27:42 2026 +0100

    Release 3.8.14
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 18:32:54 2026 +0100

    devel/nix: update flake.lock, change revdeps, more tools
    
    tlspool is marked broken in nixpkgs and looks dead upstream,
    so I'm replacing it with some other revdeps.
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 17:54:55 2026 +0100

    po: sync from translationproject.org
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 21:24:38 2026 +0100

    NEWS: add an entry for CVE-2026-127231
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Xuqing Yang <rigelyoung@icloud.com>
Date:   Fri Aug 21 22:43:29 2026 +0800

    x509: make name constraints deinitialization idempotent
    
    Clear list pointers after freeing them
    so that an allocation failure while reinitializing name constraints
    cannot leave dangling pointers for a later cleanup.
    
    Fixes: #1906
    Fixes: CVE-2026-127231
    Fixes: GNUTLS-SA-2026-11-18-3
    Severity: Low
    Introduced-in: 22cba2f35742cef9137de8f1bcbde6ef15e2e979 3.8.13
    Reported-by: Xuqing Yang <rigelyoung@icloud.com>
    Signed-off-by: Xuqing Yang <rigelyoung@icloud.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 21:16:31 2026 +0100

    NEWS: add an entry for #1875
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 9 19:56:11 2026 +0100

    cfg.mk: consolidate sc_error_message_uppercase exceptions
    
    Suggested-by: Daiki Ueno <ueno@gnu.org>
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Aug 14 18:31:48 2026 +0200

    tests/x509-upnconstraint: cover three more cases, including #1875
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Thu Aug 13 16:35:19 2026 +0200

    x509/verify: check all SAN types
    
    Ensure we're calling check_crt for every name constraint type we've defined.
    Previously UPN Subject Alternative names could be accepted
    even in the presence of name constraints excluding them.
    UPN name constraint matching remains unsupported,
    but the generic logic now correctly fails validation.
    
    Fixes: #1875
    Reported-by: Yu-Fu Fu <yufu@yfu.tw>
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Aug 14 18:24:54 2026 +0200

    tests/x509-upnconstraint: extend with full chain verification
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Aug 14 16:27:30 2026 +0200

    tests/x509-upnconstraint: refactor to abstract verify_leaf
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Aug 14 15:54:28 2026 +0200

    tests/x509-upnconstraint: generate own test certificates
    
    Regenerate the test certificates for UPN name constraint tests with
    openssl. Loses the authenticity of an old Windows-generated certificate,
    but lets me issue more leaf certs to cover extra testing scenarios.
    
    The test itself was looking only on name constraints of the intermediate
    and the SANs of the leaf, and these were preserved identical.
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Aug 14 16:40:03 2026 +0200

    tests/x509-upnconstraint: shorten and simplify
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Aug 3 18:40:38 2026 +0200

    tests/name-constraints: add tests for unsupported constraints
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Nov 11 19:39:39 2026 +0100

    tests/x509-dnconstraint: extend for permitted constraints
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 21:16:42 2026 +0100

    NEWS: add an entry for #1890
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 21:14:25 2026 +0100

    NEWS: add an entry for #1836 / #1851 / #1894 / #1895
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Aug 3 17:58:47 2026 +0200

    x509/name_constraints: reject on unsupported permitted constraints
    
    Since we don't match SAN against unsupported permitted name constraints,
    fail closed and reject if any permitted constraints are present.
    
    If a CA certificate contained nameConstraints with
    permittedSubtrees of URI or directoryName types,
    GnuTLS would silently ignore them and accept any leaf SAN of that type.
    For types without SAN-constraint matching implemented, it fails closed now:
    in the presence of constraints for an unsupported SAN type,
    any certificate featuring a SAN of that type is rejected.
    
    One consequence is that certs with a valid subject DN are now rejected,
    not just the invalid ones.
    
    Fixes: #1836
    Fixes: #1851
    Fixes: #1890
    Fixes: #1894
    Fixes: #1895
    Reported-by: Zou Dikai
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Reported-by: Jack Lloyd <jack.lloyd@gmail.com>
    Reported-by: Mitchell Benjamin of Revamp Studio <mitchell@revampstudio.com.au>
    Reported-by: Corban Villa of UC Berkeley <corban.villa@berkeley.edu>
    Suggested-by: Daiki Ueno <ueno@gnu.org>
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Thu Aug 20 14:22:36 2026 +0200

    x509/name_constraints: simplify has_constraints_of_type
    
    get_excluded returns either 0 or a negative error
    
    Suggested-by: Daiki Ueno <ueno@gnu.org>
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Aug 3 17:56:18 2026 +0200

    x509/name_constraints: treat unsupported constraints equally
    
    Don't special-case URI/DN SAN
    and reject all kinds of unsupported constraints equally
    if they're present in the certificate.
    
    This commit is required for the follow-up #1875 fix as well.
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Nov 11 17:56:00 2026 +0100

    x509/name_constraints: count subject as name for DN constraint
    
    DN constraints are special and look beyond just SAN.
    Previously we rejected everything with a DN constraint.
    Here we slightly relax that check only rejecting if DN is present:
    in the SAN, or as a non-empty subject.
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Nov 11 23:06:19 2026 +0100

    tests/cert-tests/name-constraints.sh: add a generated fixture
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Nov 11 18:38:10 2026 +0100

    tests/x509-dnconstraint: add
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Aug 3 16:05:21 2026 +0200

    x509/name_constraints: rename check_unsupported_constraint*...
    
    ... functions to lower confusion:
    * check_unsupported_constraint -> has_constraints_of_type
      since all it does is a SAN type presence check on nc
    * check_unsupported_constraint2 -> check_unsupported_constraint
      since it does the generic handling of unsupported constraint type
      for a given certificate
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Aug 10 20:03:04 2026 +0200

    x509/name_constraints: document IP address support in check_crt
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 16 14:05:55 2026 +0100

    NEWS: add an entry for #1920
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Jul 29 19:02:36 2026 +0200

    tests/name-constraints-ip: add tests for #1920
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Jul 29 18:32:03 2026 +0200

    x509: fix cross-family IP constraints handling
    
    An IPv6 permitted name constraint did not preclude matching IPv4 SAN
    and vice versa.
    The addresses from the address family not in the permitted constraints
    are now rejected.
    
    Fixes: #1920
    Reported-by: Huzaifa Sidhpurwala, Red Hat Product Security <huzaifas@redhat.com>
    Introduced-in: 6c166c70a30a483a81ecf3b1ee4ff83dbf480bf5 3.5.4
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Nov 13 13:46:29 2026 +0100

    NEWS: add an entry for #1925
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Wed May 13 16:33:59 2026 +0200

    Initialize reasons in CRL DistributionPoint parsing
    
    Fixes: #1925
    Introduced-in: 2bd323f728d75c44a2d7398503178b75e5b63263 3.3.0
    Reported-by: Huzaifa Sidhpurwala, Red Hat Product Security <huzaifas@redhat.com>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 9 17:35:06 2026 +0100

    NEWS: add an entry for #1857
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Thu May 7 12:49:32 2026 +0900

    nettle: reject DSA private keys with q > 512 in deterministic mode
    
    _gnutls_dsa_compute_k assumes that the maximum subgroup size of a DSA
    private key is 521 bits, to be in line with ECDSA with P-521. However,
    the DSA buffer in _wrap_nettle_pk_sign in lib/nettle/pk.c only holds
    512 bits, and there was no check to ensure that. This adds the check
    and exercises it in the sign-verify-deterministic test.
    
    Fixes: #1857
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Nov 9 17:13:34 2026 +0100

    NEWS: add an entry for #1866
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Tue May 5 16:30:09 2026 +0200

    x509/output.c: fix leak in print_crldist
    
    Fixes: #1866
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Nov 6 22:51:05 2026 +0100

    NEWS: add an entry for #1867
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Tue May 5 15:18:23 2026 +0200

    keys-win.c: fix HCRYPTHASH leak on OOM
    
    Fixes: #1867
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Thu Nov 5 19:20:17 2026 +0100

    NEWS: add an entry for CVE-2026-127040
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Tue May 5 14:51:00 2026 +0200

    rsa_psk.c: add missing return
    
    Fixes: #1868
    Fixes: CVE-2026-127040
    Fixes: GNUTLS-SA-2026-11-18-2
    Severity: Low
    Introduced-in: f032324af70b8420f7b550f3277efd4a9cd2b2dd 3.7.7
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Jul 27 15:19:30 2026 +0200

    NEWS: add an entry for #1865
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Tue May 5 14:38:58 2026 +0200

    ktls.c: free allocated memory in _gnutls_ktls_set_keys
    
    Fixes: #1865
    Introduced-in: 22fddddc53cfb2b2d9f73a27bc1ea147f53cd746 3.8.2
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Tue Jul 14 19:08:39 2026 +0200

    NEWS: add an entry for CVE-2026-127039
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Mon May 4 11:33:15 2026 +0200

    _gnutls_pcert_to_auth_info: always zeroize cert list after cleaning it up
    
    Fixes: #1863
    Fixes: CVE-2026-127039
    Fixes: GNUTLS-SA-2026-11-18-1
    Severity: Low
    Introduced-in: bb3bc56370e3b291219695bec5d4e1cbfc3241be 2.5.9
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Jul 10 15:25:52 2026 +0200

    NEWS: add an entry for #1856 / #1872
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Thu Apr 30 14:00:01 2026 +0200

    p11_ciphers, p11_macs: add missing error checks
    
    Fixes: #1856
    Fixes: #1872
    Related: #1821
    Introduced-in: a0e381827ca10b001cb0fc82e6ee79f10554f829 3.8.10
    Introduced-in: ce47488add59abbd0a061ceccf27182a66c34a36 3.8.10
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Nov 4 19:29:33 2026 +0100

    NEWS: reformat, add proper 3.8.14 section
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Jul 10 10:44:30 2026 +0200

    lib/x509: validate asn1_der_decoding_startEnd boundaries
    
    The current implementation of libtasn1
    does not let any of these conditions happen,
    so the check is purely for the case a bug creeps into it later.
    
    Reported-by: Kamil Frankowicz <kamil.frankowicz@cert.pl>
    Related: #1818
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Zoltan Fridrich <zfridric@redhat.com>
Date:   Tue May 5 17:06:07 2026 +0200

    priority.c: add missing return in gnutls_priority_certificate_type_list2
    
    Fixes: #1870
    Reported-by: Joshua Rogers of AISLE Research Team <joshua@joshua.hu>
    Signed-off-by: Zoltan Fridrich <zfridric@redhat.com>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Thu Aug 20 11:44:03 2026 +0900

    priority: refactor @KEYWORD expansion logic
    
    This rewrites _gnutls_resolve_priorities to be more readable by:
    
    - using memchr and memeq for non-NUL terminated substrings, instead of
      strchr and strncmp
    - using stpcpy for NUL terminated strings, instead of memcpy
    - ensuring that a pointer to the first occurrence of ':' should never
      be NULL (if it is not found, it will point to the end of the input)
    - making it clear that the loop breaks after the first @KEYWORD is
      expanded, by moving the post-processing out of the loop
    - renaming/removing cryptic local variables: ss, ss_len, ss_next,
      ss_next_len, ...
    - avoiding reuse of local variable p for different purposes
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Oct 14 14:48:04 2026 +0200

    secrets: check whether MAC algorithms are FIPS approved
    
    This should not realistically happen with what FIPS currently approves,
    so this is rather future-proofing.
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Thu Sep 24 17:19:15 2026 +0200

    secrets: switch FIPS state in _tls13_init_secret2
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Thu Sep 24 17:15:58 2026 +0200

    secrets: use _gnutls_hkdf_extract in _tls13_init_secret2
    
    Suggested-by: Daiki Ueno <ueno@gnu.org>
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Tue Nov 10 16:51:56 2026 +0100

    .mailmap: use personal email for Alexander Sosedkin
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Tue Nov 10 16:46:01 2026 +0100

    Revert "gl: override gnulib gl_array_list.c to initialize count earlier"
    
    This reverts commit ed1dcde92aa40f11effadd6678e9439ecbaa21c2.
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Tue Nov 10 16:44:39 2026 +0100

    gnulib: update to 2026-11-09 41bae5c2f3
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Oct 2 16:16:48 2026 +0200

    devel/release-steps.md: add a step for web-pages
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Thu Oct 1 18:08:17 2026 +0200

    cfg.mk: use a timezone-independent cutoff
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Sep 30 17:54:12 2026 +0200

    devel/release-steps.md: ship the release tarball built with Nix
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Mon Sep 28 15:35:59 2026 +0200

    devel/nix: add Nix flake for building release tarballs
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Oct 30 12:29:54 2026 +0100

    gl: override gnulib gl_array_list.c to initialize count earlier
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Oct 2 19:04:10 2026 +0200

    .gitlab-ci.yml: add checks for NEWS header in release-ready
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Sep 30 15:45:34 2026 +0200

    Makefile.am: move symbol/abi checks from dist to distcheck
    
    I want to generate prerelease tarballs even with abi-dump out of sync,
    so I don't want dist to be gated on it.
    The release steps instruct the maintainer to run distcheck anyway.
    An extra manual CI job is added for extra visibility/convenience.
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Sat Nov 7 09:19:11 2026 +0900

    mr, rsa-validate: prefix private symbols with _gnutls_
    
    On platforms without support for linker version script, we expose
    symbols matching '^(gnutls|_gnutls).*'. Add "_gnutls_" prefix to
    miller_rabin_enhanced and rsa_validate_sp800_56b_r2_public so they are
    exposed to the tests.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Fri Nov 6 18:37:44 2026 +0900

    tests: explicitly add nettle and gmp cflags/libs to mr/rsa-validate
    
    The commit 6ec9bc98aeeffdb7b358dbf5602879f83d5d20c6 turned out to be
    insufficient to address the CI failure on Homebrew installation of
    Nettle and GMP. While this could be resolved using *-uninstalled.pc,
    let's simply pass $(NETTLE_*) and $(GMP_*) to CFLAGS and LIBS.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Fri Nov 6 18:34:23 2026 +0900

    devel: revert back abi-dump submodule to our own location
    
    The commit c8a32e10b4ed75a9ec2b1c4bbf431add2ff8973d moves the abi-dump
    git URL to a personal fork; this is not the way we do. Update
    devel/libgnutls.abignore instead, as suggested in CONTRIBUTING.md.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Alejandro Pacheco <alepacheco99@gmail.com>
Date:   Thu Oct 15 04:59:24 2026 +0200

    str-iconv: convert the NFC-normalized UTF-16 in _gnutls_utf8_to_ucs2
    
    Since the conversion to UCS-2 started normalizing to NFC,
    _gnutls_utf8_to_ucs2() has sized its output from the normalized
    string but copied it from the unnormalized one. A password that is not
    already in NFC was therefore converted into the wrong UCS-2 string:
    decomposed input was cut short (U+0065 U+0301 became U+0065 alone),
    and input that NFC makes longer, such as U+0958, made the copy read
    past the end of the unnormalized buffer. The UCS-2 string is what
    PKCS#12 password-based key derivation uses, for PKCS#12 MACs and
    PKCS#12-encrypted PKCS#8 keys.
    
    Copy from the normalized buffer, and make change_u16_endianness()
    take a count of 16-bit code units, so that both converters pass the
    length of the buffer they hand it. _gnutls_ucs2_to_utf8() now rejects
    input of odd length instead of reading one byte past it while
    swapping the last, partial code unit; a BMPString of odd length is
    malformed, so the DN printer shows it in hex and a certificate policy
    with such an explicitText fails to parse. Passwords already in NFC
    convert exactly as before.
    
    tests/conv-utf8.c checks that decomposed input is composed, that a
    character NFC expands is converted in full, and that odd-length UCS-2
    is rejected.
    
    Signed-off-by: Alejandro Pacheco <alepacheco99@gmail.com>

Author: Mohammadreza Khellat <mkhellat@gmail.com>
Date:   Thu Nov 5 15:08:59 2026 +0330

    abstract: address review on ML-DSA raw pubkey APIs
    
    Retarget the new symbols from GNUTLS_3_8_15 to the existing unreleased
    GNUTLS_3_8_14 section (3.8.14 is not tagged yet), set key->bits on
    import like the other raw helpers, and add the test file copyright
    holder. Bump devel/abi-dump to match.
    
    Signed-off-by: Mohammadreza Khellat <mkhellat@gmail.com>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Thu Nov 5 18:05:33 2026 +0900

    tests: use HOGWEED_{CFLAGS,LIBS} to build mr and rsa-validate
    
    These tests rely on bignums and thus need to link to libhogweed in
    addition to libnettle, which is already a dependency of libhogweed.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Thu Nov 5 16:52:50 2026 +0900

    global: declare ASN.1 schema objects as asn1_node_const
    
    Outside of global.c, where those ASN.1 object is
    initialized/deinitialized, they are only treated as const. This adds
    wrapper functions which returns asn1_node_const instead of asn1_node
    to enforce this. This will also suppress -Wuseless-cast warning.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Sat Aug 22 18:28:02 2026 +0900

    priority: replace name_val_array.h with a simpler implementation
    
    The previously existed name_val_array.h implemented an array of
    key-value pairs in a tricky way: as a linked list, each node has a
    contiguous region allocated to hold both key and value, which may have
    slightly improved performance by lowering the number of malloc calls,
    though this is not on a performance critical path. This rewrites it
    with gl_linkedhash_list.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Fri Sep 18 10:26:16 2026 +0900

    bootstrap.conf: import hashcode-mem instead of hash-pjw-bare
    
    The hash-pjw-bare module is deprecated as well as the
    "hash-pjw-bare.h" in favor of the hashcode-mem module. Switch to it
    and also use the suggested #include "hashcode-mem.h".
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Mohammadreza Khellat <mkhellat@gmail.com>
Date:   Fri Oct 30 11:27:26 2026 +0330

    devel: bump abi-dump for ML-DSA raw pubkey APIs
    
    distcheck abi-check-latest failed because the new
    gnutls_pubkey_{export,import}_mldsa_raw symbols were absent from
    libgnutls-latest-x86_64.abi.
    
    Point devel/abi-dump at mkhellat/abi-dump@c163bc7 (draft MR
    gnutls/abi-dump!12) so CI can fetch the regenerated dump. After that
    abi-dump MR merges, retarget .gitmodules to gnutls/abi-dump and bump
    to the upstream SHA.
    
    Signed-off-by: Mohammadreza Khellat <mkhellat@gmail.com>

Author: Mohammadreza Khellat <mkhellat@gmail.com>
Date:   Fri Oct 30 10:44:04 2026 +0330

    abstract: update symbols.last for ML-DSA raw pubkey APIs
    
    fedora-docdist distcheck runs symbol-check and rejected the branch
    because gnutls_pubkey_{export,import}_mldsa_raw were in libgnutls.map
    (GNUTLS_3_8_15) but missing from devel/symbols.last.
    
    Refresh the sorted symbol list (equivalent to make files-update for
    the symbol-check half) so dist-hook can proceed.
    
    Signed-off-by: Mohammadreza Khellat <mkhellat@gmail.com>

Author: Mohammadreza Khellat <mkhellat@gmail.com>
Date:   Fri Oct 30 09:06:39 2026 +0330

    abstract: add ML-DSA raw pubkey export/import APIs
    
    ML-DSA is supported experimentally, but applications such as DNSSEC
    need the FIPS 204 public key octet string rather than only
    SubjectPublicKeyInfo. Add gnutls_pubkey_export_mldsa_raw() and
    gnutls_pubkey_import_mldsa_raw() mirroring the other *_raw helpers.
    
    Includes a round-trip test against the existing ML-DSA-44/65/87
    fixtures, NEWS note, symbol map (GNUTLS_3_8_15), and manpage wiring.
    
    Closes: #1917
    Signed-off-by: Mohammadreza Khellat <mkhellat@gmail.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Tue Oct 27 18:17:16 2026 +0100

    po: bundle translations
    
    Bundle translations instead of fetching them during bootstrap,
    since there's no source of them that's stable over time.
    This is required for the release tarballs to be reproducible.
    
    Conceptually reverts 493c15b49908de420a8bf9e473eaf0a120203fa6
    
    Signed-off-by: Alexander Sosedkin <monk@unboiled.info>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Wed Oct 21 19:50:01 2026 +0900

    nettle: check validity of RSA public keys before use
    
    This adds a public key plausibility check based on SP800-56B rev2,
    before RSA-OAEP encryption and RSA-PSS signature verification.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Thu Oct 22 08:52:38 2026 +0900

    fips: allow stack-based indicator context for internal use
    
    This adds a way of allocating struct gnutls_fips140_context_st on
    stack, so that a service indicator context can be temporarily put
    aside.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Wed Oct 21 17:10:53 2026 +0900

    nettle: implement partial public key validation for RSA
    
    This implements explicit partial public key validation as defined in
    SP800-56Brev2 6.4.2.2, based on SP800-89 5.3.3.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Wed Oct 21 07:37:46 2026 +0900

    nettle: implement Enhanced Miller-Rabin primarity testing
    
    This implements Enhanced Miller-Rabin probablistic primarity testing
    as defined in FIPS 186-5 B.3.2.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Ghadi Elie Rahme <ghadi.rahme@canonical.com>
Date:   Mon Oct 19 15:32:33 2026 +0000

    tests: Add sensitive privkey DSA test
    
    Signed-off-by: Ghadi Elie Rahme <ghadi.rahme@canonical.com>

Author: Ghadi Elie Rahme <ghadi.rahme@canonical.com>
Date:   Mon Oct 12 16:14:38 2026 +0000

    lib/pkcs11: Fix DSA private key type fetching
    
    CKA_VALUE cannot be fetched when it is protected by
    CKA_SENSITIVE being set to CK_TRUE. This causes DSA
    private key information to fail to retrieved. Simplify
    pkcs11_read_privkey_type to only retrieve CKA_PRIME and
    simplify gnutls_pkcs11_obj_get_pk_algorithm to only use
    CKA_PRIME (DSA_P) since we can retrieve the key length from
    CKA_PRIME bit length alone.
    
    Signed-off-by: Ghadi Elie Rahme <ghadi.rahme@canonical.com>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Fri Oct 23 12:39:09 2026 +0900

    devel: update gnulib submodule
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Fri Oct 23 10:17:40 2026 +0900

    devel: update abi-dump submodule
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Sahana Prasad <sahana@redhat.com>
Date:   Mon Sep 7 18:32:05 2026 +0200

    Add --enable-tls11-support configure option (RFC 8996)
    
    Add a compile-time toggle for TLS 1.0, TLS 1.1, and DTLS 1.0
    support. The option defaults to enabled; a future release will
    disable it by default.
    
    Signed-off-by: Sahana Prasad <sahana@redhat.com>

Author: Stanislav Zidek <szidek@redhat.com>
Date:   Fri Oct 2 18:19:04 2026 +0200

    interop tests: update argument for allowing local provisioning
    
    Signed-off-by: Stanislav Zidek <szidek@redhat.com>

Author: Brad Smith <brad@comstyle.com>
Date:   Thu Oct 1 13:54:19 2026 -0400

    aarch64: check elf_aux_info return value
    
    elf_aux_info() returns an errno value (0 on success) rather than -1
    with errno set. Ignoring the return value leaves "c" uninitialized
    when the call fails, and the capability bits are then derived from
    garbage. Treat a failure as "no capabilities reported" so that none
    of the accelerated implementations are registered and the generic
    code is used.
    
    Signed-off-by: Brad Smith <brad@comstyle.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Fri Sep 11 16:14:22 2026 +0200

    secrets: pass actual zeroes in _tls13_init_secret2
    
    https://datatracker.ietf.org/doc/html/rfc8446#section-7.1
    states that key schedule should start with "0", defined as
    > a string of Hash.length bytes set to zero.
    Passing "" to gnutls_hmac_fast() is computationally equivalent,
    but trips a FIPS indicator because this API has a length check.
    
    A third option would've been to call _gnutls_hkdf_extract()
    and handle the indicator state like the other functions do,
    but this feels a different kind of quirky
    and starts hitting a different set of approved algorithm checks,
    so I went for a minimal change instead.
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: Daniel P. Berrangé <berrange@redhat.com>
Date:   Wed Sep 9 12:54:50 2026 +0100

    lib: document retries explicitly for gnutls_record_uncork
    
    Document explicitly that when gnutls_record_uncork() is unable
    to send all pending data, the session remains in a corked state
    and the API call must be retried to flush data.
    
    Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>

Author: Alexander Sosedkin <monk@unboiled.info>
Date:   Wed Aug 26 16:35:54 2026 +0200

    nettle/pk: mark ML-KEM as FIPS-unapproved
    
    ML-KEM needs a self-test to be approved in FIPS mode.
    
    Signed-off-by: Alexander Sosedkin <asosedkin@redhat.com>

Author: d-Dudas <vid.dudas03@gmail.com>
Date:   Wed Aug 26 19:58:08 2026 +0300

    Remove fedora-nettle-minigmb CI jobs.
    
    Cleaned up the already commented fedora-nettle-minigmb
    CI jobs from .gitlab-ci.yml.
    
    Signed-off-by: David Dudas <vid.dudas03@gmail.com>

Author: d-Dudas <vid.dudas03@gmail.com>
Date:   Wed Aug 26 10:14:30 2026 +0300

    [#1946] Remove --with-nettle-mini configure option
    
    Signed-off-by: David Dudas <vid.dudas03@gmail.com>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Fri Aug 14 08:12:38 2026 +0900

    ext/cookie: switch to buffer based interface to parse extension data
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Daiki Ueno <ueno@gnu.org>
Date:   Thu Aug 20 16:15:45 2026 +0900

    hello_ext: add buffer-based extension parsing function
    
    This adds parse_func in hello_ext_entry_st, which takes a
    gnutls_buffer_st, unlike the existing recv_func which takes raw data
    and the size. If set, parse_func is preferred over recv_func, and it
    is checked that there are no trailing bytes in the extension data
    after parsing.
    
    Signed-off-by: Daiki Ueno <ueno@gnu.org>

Author: Tim Rühsen <tim.ruehsen@gmx.de>
Date:   Wed Aug 19 22:52:35 2026 +0200

    _gnutls_alt_name_assign_virt_type: Fix memory leak
    
    Co-authored-by: Daiki Ueno <ueno@gnu.org>
    Signed-off-by: Tim Rühsen <tim.ruehsen@gmx.de>

Author: Simon Josefsson <jas@josefsson.org>
Date:   Tue Aug 18 16:45:52 2026 +0200

    Remove minitasn1
    
    Closes #1938
    
    Signed-off-by: Simon Josefsson <simon@josefsson.org>

Author: Tim Rühsen <tim.ruehsen@gmx.de>
Date:   Tue Aug 18 18:04:17 2026 +0200

    gnutls_certificate_set_x509_crl: Use gnutls_free instead of free
    
    Signed-off-by: Tim Rühsen <tim.ruehsen@gmx.de>

Author: Tim Rühsen <tim.ruehsen@gmx.de>
Date:   Tue Aug 18 11:15:19 2026 +0200

    pkcs11: Fix macro REPEAT_ON_INVALID_HANDLE
    
    The macro REPEAT_ON_INVALID_HANDLE contained a return, which did not do cleanups
    like freeing memory or unlocking mutexes. This leads to memory leaks and even
    deadlocks under certain error conditions.
    
    Signed-off-by: Tim Rühsen <tim.ruehsen@gmx.de>

Author: Tim Rühsen <tim.ruehsen@gmx.de>
Date:   Tue Aug 18 11:25:35 2026 +0200

    x509_privkey_verify_seed: Fix memory leak
    
    The cleanup was missing for the case where seed is NULL or has 0 length.
    
    Signed-off-by: Tim Rühsen <tim.ruehsen@gmx.de>

Author: Tim Rühsen <tim.ruehsen@gmx.de>
Date:   Tue Aug 18 10:49:54 2026 +0200

    session_ticket: Use gnutls_free instead of free
    
    Signed-off-by: Tim Rühsen <tim.ruehsen@gmx.de>

Author: Christian Häggström <kalvdans@gmail.com>
Date:   Sun Aug 16 17:31:54 2026 +0200

    doc: document GNUTLS_PIN and GNUTLS_SO_PIN environment variables
