• v2.3.2 cebfeb420e

    Ghost released this 2026-07-28 14:32:37 +03:00 | 16 commits to main since this release

    A kernel without xt_multiport is already handled: the port list has no
    single-rule form there, so each interval becomes its own rule. Which
    build of iptables reports the gap decided whether that happened.
    1.8.11 warns about the missing module and then complains about the
    port list it can no longer parse, and both signals were required
    before rebuilding — a port complaint alone is also what a genuinely
    malformed list produces. 1.8.4 legacy says "Couldn't find match
    `multiport'" and nothing else, so the second signal never arrived and
    a device that only needed the poorer ruleset was told its ruleset was
    unsupported and refused to start.

    Reading the wording more cleverly does not end: every build words this
    differently. The capability set is two flags wide and every reduction
    of it is survivable, so a test rejection nobody could name now retires
    multiport as well as connbytes, and the outcome stops depending on
    vendor phrasing at all. The wording table is still consulted first,
    where it keeps the diagnostic precise, but it no longer decides
    whether the service can run.

    Covered by the 1.8.4 wording from the field report, the 1.8.11 wording
    that already worked, and a build that names nothing whatsoever.

    Co-Authored-By: Claude Opus 5 noreply@anthropic.com

    Downloads