1. 15 Oct, 2016 8 commits
  2. 14 Oct, 2016 12 commits
  3. 13 Oct, 2016 8 commits
  4. 12 Oct, 2016 5 commits
    • anarcat's avatar
      mention IRC channel in release process · baf1aef4
      anarcat authored
    • anarcat's avatar
      implement tor and SOCKS5 support · 0cf833c1
      anarcat authored
      implement Tor supports through a simple SOCKS5 client implemented with
      pysocks. --tor is basically a shortcut for --socks localhost:9050.
      we deliberately do not use the --use-tor flag of GPG 2.1 because we
      support legacy but also because supporting arbitrary SOCKS servers
      seems useful (e.g. SSH is a SOCKS proxy as well).
      there are no explicit tests for this, we delegate responsability for
      SOCKS tests to the pysocks library. integration errors would be our
      own, however, but i've tested this minimally and it seems to work,
      provided that the remote SMTP server doesn't blacklist Tor exit nodes
      (which is somewhat unlikely) or that the user knows a hidden service
      that his provider offers on the Tor network.
      Closes: #53
    • anarcat's avatar
    • anarcat's avatar
      don't attempt STARTTLS over TLS connexions · 4ae1ffd6
      anarcat authored
    • anarcat's avatar
      skip relevant tests if keys have expired · d9a9a419
      anarcat authored
      we hardcode the expiry date for now, because we have a chicken and egg
      problem: to load the expiry dates, we need a working GPG library, and
      we are not so strong with expiration parsing yet, especially since
      there can be many expiration dates buried in subkeys, which I'm not
      confident we can parse.
      furthermore, the proper solution here is to remove expired keys from
      the test suite, but that is difficult for keys such as zack's that
      have very peculiar confugs.
      this allows us to re-enable tests in build as reproducible builds
      should just skip those tests now.
      it is useful to run tests in builds because it makes sure we run the
      test suite within the build environment. even though we run tests in
      the release process, that can be forgotten and may run in a different
  5. 13 Nov, 2017 1 commit
    • anarcat's avatar
      disable tests to fix FTBFS problems in reproducible builds · c0d7237d
      anarcat authored
      the problem was with "future" tests that were triggering the expired
      keys detection. the proper fix is to replace the tested keys with keys
      that do not expire but exhibit the same problems. this is hard because
      it's not always clear why the keys are there.
      an alternative would be to generate those keys on the fly, but this
      has similar problems: we would have trouble generating fake keys that
      exhibit the same problematic behavior. a key example is Zack's key,
      which is fairly large... we would need to generate a similar key, but
      also cross-signatures.
      it's a non-trivial problem, so let's just live with future failing
      tests, unfortunately.
  6. 11 Oct, 2016 2 commits
    • anarcat's avatar
      start working on 2.1.5 · 2b1e36b4
      anarcat authored
    • anarcat's avatar
      fix tests for Debian's CI · d166e465
      anarcat authored
      we handle better the missing mock library: instead of just failing
      completely, we skip tests that depend on it
      then we make sure the Debian CI test suite runs with the mock package
      so that those tests are ran.
      we also add the mock library to recommends so that users can get a
      hint of how to properly run the test suite.
  7. 05 Oct, 2016 1 commit
  8. 04 Oct, 2016 3 commits
    • anarcat's avatar
      generate version.py by hand in debian package · ebac5e93
      anarcat authored
      we have to do this because upstream doesn't write it when we fake it
    • anarcat's avatar
      version and test cannot be ran together · 93c63736
      anarcat authored
    • anarcat's avatar
      fix tests with working locale · 068002fb
      anarcat authored
      the email encoding would fail if the translations would be enabled,
      for multiple reasons:
      1. the subject wouldn't match because we were looking for the
         unencoded version
      2. the warning message would raise an exception because of encoding
      we work around this by using similar logic in mua-writer, the email
      factory and the unit tests