reconstruct-changelog should ignore signed-off-by from commits

[Andreas Hasenack]

When git-ubuntu is reconstructing the debian/changelog file, it should ignore the "Signed-off-by" messages that some developers use in their commits, otherwise the reconstructed changelog looks like this:
strongswan (5.6.3-1ubuntu1) cosmic; urgency=medium

  * Merge with Debian unstable. Remaining changes:
        + Clean up d/strongswan-starter.postinst: section about runlevel changes
Signed-off-by: Christian Ehrhardt <email address hidden>
        + Clean up d/strongswan-starter.postinst: Removed entire section on opportunistic encryption disabling - this was never in strongSwan and won't be see upstream issue #2160.
Signed-off-by: Christian Ehrhardt <email address hidden>
    + d/rules: Removed patching ipsec.conf on build (not using the
      debconf-managed config.)
    + d/ipsec.secrets.proto: Removed ipsec.secrets.inc reference (was
      used for debconf-managed include of private key).
    + Mass enablement of extra plugins and features to allow a user to use
      strongswan for a variety of extra use cases without having to rebuild.
      - d/control: Add required additional build-deps
      - d/control: Mention addtionally enabled plugins
      - d/rules: Enable features at configure stage
      - d/libbstrongswan-extra-plugins.install: Add plugins (so, lib, conf)
      - d/libstrongswan.install: Add plugins (so, conf)
Signed-off-by: Christian Ehrhardt <email address hidden>
    + d/strongswan-starter.install: Install pool feature, which is useful since
      we have attr-sql plugin enabled as well using it.
Signed-off-by: Christian Ehrhardt <email address hidden>
    + Add plugin kernel-libipsec to allow the use of strongswan in containers
(...)

[Robie Basak (racb)]

Thanks! I wonder if we can make this broader to ignore any RFC822-style "headers" at the end of the commit message.

[Bryce Harrington (bryce)]

> Thanks! I wonder if we can make this broader to ignore any RFC822-style "headers" at the end of the commit message.

They certainly don't belong in the changelog entry either, however in my experience Signed-off-by seems to be the only one that pops up commonly. My guess is people set signoff in their global .gitconfig since it's required by various upstream projects, and so it just sneaks in accidentally to git-ubuntu commits.

To be honest, the presence of other RFC822-style headers would be a flag that something might be wrong with the tree (i.e. maybe someone applied a patch directly via git am rather than stored it in debian/patches).