commit d0dba79a90
Author: Gerald Combs <gerald@wireshark.org>
Date:   Wed Feb 25 09:20:43 2026 -0800

    Build: 4.4.14 [skip ci]

commit 16fdb087a6
Author: Giovanni Musto <giovanni.musto@italdesign.it>
Date:   Wed Feb 25 13:51:23 2026 +0100

    BLF: Avoid error in case of truncated log container

    (cherry picked from commit 84fbadc5229ee785a1610ccb349a90f7e8fd854e)

    Co-authored-by: Giovanni Musto <giovanni.musto@italdesign.it>

commit 6647382977
Author: Gerald Combs <gerald@wireshark.org>
Date:   Tue Feb 24 15:46:08 2026 -0800

    Windows: Update GnuTLS to 3.8.12

    AI-Assisted: no
    (cherry picked from commit d3fba774c1a00b17d3d0c6f964ad92285ee59f85)

commit a2afba4f62
Author: Gerald Combs <gerald@wireshark.org>
Date:   Mon Feb 23 13:15:54 2026 -0800

    extcap: Update our SSH configuration location on Windows

    "%d" expands to the user's home directory, so we need to append
    ".ssh\config" instead of "config".

    AI-Assisted: no
    (cherry picked from commit 81b24a356cd6e487d8f975c8c93be5817ddfb9c1)

commit bc76b56480
Author: John Thacker <johnthacker@gmail.com>
Date:   Fri Oct 17 16:27:18 2025 -0400

    extcap: Do not try to read c:\etc\ssh\ssh_config on Windows

    libssh by default tries to read /etc/ssh/ssh_config on all OSes.
    OpenSSH Portable for Windows by default doesn't create a systemwide
    ssh_config, and if it does, it puts it in %ProgramData%\ssh and
    has to check permissions because it might be owned and writable
    by a non-admin user. Since C:\etc does not exist by default on
    Windows, that's a possible security issue.

    https://github.com/PowerShell/openssh-portable/blob/latestw_all/ssh.c#L591-L597
    https://github.com/PowerShell/openssh-portable/blob/latestw_all/contrib/win32/openssh/config.h.vs#L1723

    Only read from the OpenSSH default configuration file in the user
    home directory on Windows.

    Fix #20806.

    (cherry picked from commit 845c005ab8549b4a84a11d1d2505effad2c00528)

commit 926a14ba0e
Author: Gerald Combs <gerald@wireshark.org>
Date:   Tue Feb 24 11:32:08 2026 -0800

    GitLab CI: Disable the macOS Build merge request job for now

commit 855ed48688
Author: Gerald Combs <gerald@wireshark.org>
Date:   Mon Feb 23 14:57:39 2026 -0800

    Prep for 4.4.14

commit 955f647114
Author: Gerald Combs <gerald@wireshark.org>
Date:   Sun Feb 22 10:17:45 2026 +0000

    [Automatic update for 2026-02-22]

    Update manuf, services enterprise numbers, translations, and other items.

commit ad6a3c83c8
Author: Gerald Combs <gerald@wireshark.org>
Date:   Fri Feb 20 15:34:33 2026 -0800

    Windows: Update libssh to 0.11.4

commit 5e76149c8c
Author: Guy Harris <gharris@sonic.net>
Date:   Tue Feb 17 17:21:26 2026 -0800

    tshark: fix exit status.

    Exit with WS_EXIT_INVALID_FILTER, not WS_EXIT_INVALID_INTERFACE, if the
    read filter isn't valid.

    (cherry picked from commit e53cc35bc7fa3740ead1b3afa6c210f59c12e54b)

commit 2d7abdc17a
Author: Gerald Combs <gerald@wireshark.org>
Date:   Sun Feb 15 10:22:35 2026 +0000

    [Automatic update for 2026-02-15]

    Update manuf, services enterprise numbers, translations, and other items.

commit 2aad4e27dc
Author: Jaap Keuter <jaap.keuter@xs4all.nl>
Date:   Thu Feb 12 18:24:13 2026 +0000

    Fix const pointer usage for gcc.

    AI-Assisted: no

    (cherry picked from commit 96d27343151d90a3de642d6544f969392ccffe73)

    Co-authored-by: Dexter Gerig <dexgerig@gmail.com>

commit e70f956c3e
Author: Pascal Quantin <pascal@wireshark.org>
Date:   Tue Feb 10 19:33:18 2026 +0100

    Diameter: add a missing 3GPP RAT-Type AVP value

    AI-Assisted: no

    (cherry picked from commit 822055418e407d9cd9ca94384431fe388498aa4c)

    Co-authored-by: Pascal Quantin <pascal@wireshark.org>

commit eedc5cb41a
Author: Pascal Quantin <pascal@wireshark.org>
Date:   Mon Feb 9 23:32:31 2026 +0100

    Diameter update 3GPP RAT-Type AVP values

    Based on 3GPP 29.212 v19.1.0.
    Fixes #21012

    AI-Assisted: no

    (cherry picked from commit a4ace7abcb885463108e96171d6ef2bf96cbe659)

    Co-authored-by: Pascal Quantin <pascal@wireshark.org>

commit 502ddd2d6e
Author: Gerald Combs <gerald@wireshark.org>
Date:   Sun Feb 8 10:18:05 2026 +0000

    [Automatic update for 2026-02-08]

    Update manuf, services enterprise numbers, translations, and other items.

commit b25886fa6e
Author: John Thacker <johnthacker@gmail.com>
Date:   Sun Feb 8 00:21:08 2026 +0000

    SOCKS: Create a proxy conversation for each outer conversation

    To support multiple SOCKS over TCP connections on the same outer
    5-tuple (consecutively, of course, not concurrently), associate the
    inner proxy conversation with the outer conversation, and create
    a new inner conversation when the outer conversation changes.

    Fix #21006

    AI-Assisted: no

    (cherry picked from commit 624e46fb5545503ea42cfa26cabeb093f01502bf)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit d487e0e07b
Author: John Thacker <johnthacker@gmail.com>
Date:   Sat Feb 7 02:51:04 2026 +0000

    RF4CE: Check that the input data is long enough

    Prevent illegal memory access.

    Fix: #21009

    AI-Assisted: no

    (cherry picked from commit 17215397c1a5fbb2ef8764b3ec29ec45cde9c153)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit b571934a30
Author: John Thacker <johnthacker@gmail.com>
Date:   Fri Feb 6 05:55:03 2026 -0500

    TDS: Fix dissection of variable-length data types in RPC

    DATENTYPE does not have a column maximum length (TYPE_VARLEN) because
    the length is fixed at 3 if non-NULL.

    Retrieve the SCALE value and pass it to dissect_tds_type_varbyte so that
    the types which have a length that depends on the SCALE (TIMENTYPE,
    DATETIME2NTYPE, and DATETIMEOFFSETNTYPE) are dissected correctly.

    Fix the retrieval and display of variable length integers when 64-bit
    and of variable length floats; the incorrect offset (the offset of the
    length value) was being used instead.

    https://learn.microsoft.com/en-us/openspecs/windows_protocols/ms-tds/cbe9c510-eae6-4b1f-9893-a098944d430a

    Fix: #21001

    AI-Assisted: no
    (backported from commit 15ffb508bba4945beba8df98bc32c486e6a8ec23)

commit fee713500f
Author: Balint Reczey <balint@balintreczey.hu>
Date:   Fri Dec 5 12:30:10 2025 +0100

    wsutil: Restore removed ws_base32_decode() to fix ABI breakage

    (cherry picked from commit e502c94a5df9a0575ebc568051eb85e106e2725f)

    Conflicts:
            wsutil/CMakeLists.txt

commit 960359b754
Author: Alexis La Goutte <alexis.lagoutte@gmail.com>
Date:   Wed Feb 4 10:14:34 2026 +0100

    artnet: Fix wrong RDM Status for Output Status

    Fix: #20980

    (cherry picked from commit 7551171435761167d44f8b37ec2fef9502ca7daf)

    Co-authored-by: Alexis La Goutte <alexis.lagoutte@gmail.com>

commit ca20762cf1
Author: John Thacker <johnthacker@gmail.com>
Date:   Wed Feb 4 03:01:32 2026 +0000

    Zigbee Direct: Set necessary packet pointer to the global short_table

    The Zigbee NWK dissector requires a ieee802154_packet structure to be
    passed as dissector data. When Zigbee Direct is used to tunnel Zigbee
    NPDUs, there is no IEEE 802.15.4 MAC layer and header to fill that out.

    A zeroed out packet structure has been passed in, but apparently that
    packet structure should always contain a pointer to a global lookup
    table. (The Zigbee NWK dissector might need some checking to see if
    the address is already set, etc.)

    Also set the COL_PROTOCOL fence immediately before calling the
    Zigbee NWK dissector, and insert a separator.

    Fix #20977

    AI-Assisted: no

    (cherry picked from commit 029738e8f7736fe65cda10c1995d343fc7cd2751)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit bce1b764ed
Author: John Thacker <johnthacker@gmail.com>
Date:   Tue Feb 3 21:57:37 2026 +0000

    Zigbee Direct: Make sure a loop iterates

    The loop variable i was not advanced when a test was false, resulting
    in an infinite loop. Use for loops for clarity. Also switch to using
    a GHashTableIter, which is faster than creating a list of all the values
    and also doesn't require freeing. (It was previously leaking.)

    AI-Assisted: no

    (cherry picked from commit e9183fadd2266dfd64d4984c293301cc247ca5d2)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit 21546744bc
Author: Pascal Quantin <pascal@wireshark.org>
Date:   Tue Feb 3 20:34:44 2026 +0100

    NAS-5GS: correction of Rejected NSSAI

    Use of correct offset value

    (cherry picked from commit 1f2a9840894498f9b8226a305f66b465013f21ab)

    Co-authored-by: Joakim Karlsson <oakimk@gmail.com>

commit 93a3ba774f
Author: John Thacker <johnthacker@gmail.com>
Date:   Sun Feb 1 15:30:20 2026 +0000

    rsa: Use proper free function

    gnutls_x509_privkey_export_rsa_raw() allocates buffers with
    gnutls_malloc() and they must be freed with gnutls_free().

    AI-Assisted: no

    (cherry picked from commit 5b1dfffd1869c5c08859b01c1e21bedb82ce3799)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit 947a937b68
Author: Gerald Combs <gerald@wireshark.org>
Date:   Sun Feb 1 10:17:43 2026 +0000

    [Automatic update for 2026-02-01]

    Update manuf, services enterprise numbers, translations, and other items.

commit 7c7b861a44
Author: John Thacker <johnthacker@gmail.com>
Date:   Fri Jan 30 14:47:29 2026 +0000

    secrets: Fix return type

    The GnuTLS return types are ints and negative on error. Casting it to a
    bool makes all errors into positive one, which obscures the failure.

    This probably worked ok before converting to C99 types
    ( 2562674df0d294223d49ee9dfea7cc9680b00098 ) because a gboolean is
    really a int, and casting back and forth between an int and a
    gboolean doesn't coerce the value into 1 the way it does with a
    stdbool bool.

    AI-Assisted: no

    (cherry picked from commit 806af16d88d0fc7adbaa24f984995e4779a40604)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit 5e80615ebc
Author: Michael Mann <mmann78@netscape.net>
Date:   Mon Jan 26 16:44:58 2026 +0000

    USB-HID: Bugfix resource exhaustion in parse_report_descriptor()

    Sanity range check was removed in 739666a7f5acc270204980e01b4069caf5060f30, restore it

    AI-Assisted: no
    Fixes #20972

    (cherry picked from commit 6f753c79b7c8ac382e6383dfabd7d5be6e2b722c)

    Co-authored-by: Michael Mann <mmann78@netscape.net>

commit 710b5bc1fc
Author: Gerald Combs <gerald@wireshark.org>
Date:   Sun Jan 25 10:17:20 2026 +0000

    [Automatic update for 2026-01-25]

    Update manuf, services enterprise numbers, translations, and other items.

commit 1652adaaa1
Author: Dr. Lars Völker <lars.voelker@technica-engineering.de>
Date:   Sat Jan 24 19:12:29 2026 +0100

    TECMP: Fixing FlexRay Replay format

    The Replay Format was not implemented correctly and is being fixed.

commit 820da2a82f
Author: Anders Broman <a.broman58@gmail.com>
Date:   Sat Jan 24 09:49:27 2026 +0100

    Qt: Speed up expert info proxy model

    When opening the Expert Dialog, the tap information and GUI is
    periodically updated to avoid the program looking frozen. Since
    expert infos have been added, this causes a relayout. Part of a
    relayout is querying for the existence of children (partly to
    add the indicator).

    The normal process of querying for children in a QSortFilterProxyModel
    calculates, if necessary, the mapping from the source model to the
    proxy model. This mapping is cached so it can be used later, including
    in other member functions. However, the mapping has to be recalculated
    each time new items have been added and the model reset in TapDraw.
    The mapping involves sorting the rows added to the model. It is expensiv
     to repeatedly periodically sort as items are added.

    Determining whether or not there are children does not require
    sorting, nor counting all the children, as it can stop as soon as
    one visible child is found. Override the hasChildren function to
    calculate whether or not there are children in a much faster way,
    without calculating and caching the mapping. This reduces the time
    to open the Expert Dialog dramatically, and also makes it O(N) in
    average time instead of something like O(N^2 log N).

    Fixes #20970

    AI-Assisted: no

    (cherry picked from commit f2791659ed855ca7e8699e00666f003561201b47)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit a148b08f21
Author: Anders Broman <a.broman58@gmail.com>
Date:   Sat Jan 24 08:39:43 2026 +0100

    Qt: Speed up expert info item row calculation

    The childItems_ list is only appended to, the items are not sorted
    or removed (until destruction). An item is only added as the child
    of a single parent, which determines its row from that point onwards.
    So we can cache the row upon insertion instead of using indexOf_ and
    searching for the child in the parent's list of children. That
    reduces a frequently called O(N) operation to O(1).

    Ping #20970

    AI-Assisted: no

    (cherry picked from commit be66fffbaae3915d25462fd82b96999b774cffdc)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit 6416718af6
Author: John Thacker <johnthacker@gmail.com>
Date:   Wed Jan 21 23:28:27 2026 +0000

    TLS Utils: For JA4, test if ALPN characters are alphanumeric

    The JA4 technical details indicate that the hex representation is
    used when the first and last characters are non (ASCII) alphanumeric,
    not non (ASCII) printable.

    https://github.com/FoxIO-LLC/ja4/blob/main/technical_details/JA4.md#alpn-extension-value

    Fix #20966

    AI-Assisted: no

    (cherry picked from commit c074ae7e74ee59b408bf99c2a25419ecc4fcdda2)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit 983aadac63
Author: Gerald Combs <gerald@wireshark.org>
Date:   Sun Jan 18 10:17:18 2026 +0000

    [Automatic update for 2026-01-18]

    Update manuf, services enterprise numbers, translations, and other items.

commit f9fc043271
Author: Guy Harris <gharris@sonic.net>
Date:   Wed Jan 14 23:47:29 2026 -0800

    dumpcap man page: fix glitches.

    Remove stray > after option names. Remove an extra space.

    AI-Assisted: no

    (cherry picked from commit 7a08b32b59e2164d5fe9ee5ba0e9f5f7a97189ea)

commit 03932a881b
Author: John Thacker <johnthacker@gmail.com>
Date:   Thu Jan 15 21:58:12 2026 +0000

    tls-utils: Update TLS 1.3 SignatureScheme list

    Update the value string for the TLS SignatureScheme for recent
    registrations, primarily post quantum cryptography.

    https://www.iana.org/assignments/tls-parameters/tls-parameters.xhtml#tls-signaturescheme

    Note that the various private use range used by OpenQuantumSafe could be
    updated, but this commit does not do so (but does update the URL to
    point to the current location.)

    Fix #20953

    (cherry picked from commit c5bb3f71d6e8efeaf97fbbd20bd92fc8681e3bbe)

    Co-authored-by: John Thacker <johnthacker@gmail.com>

commit 73f1cf9c56
Author: Gerald Combs <gerald@wireshark.org>
Date:   Wed Jan 14 12:08:46 2026 -0800

    Version: 4.4.13 → 4.4.14 [skip ci]
