1. 05 Apr, 2011 1 commit
  2. 17 Feb, 2011 1 commit
  3. 06 Feb, 2011 3 commits
  4. 04 Feb, 2011 1 commit
  5. 02 Feb, 2011 2 commits
  6. 01 Feb, 2011 1 commit
  7. 30 Jan, 2011 1 commit
  8. 21 Jan, 2011 1 commit
  9. 20 Jan, 2011 1 commit
    • Jonathan-Christofer Demay's avatar
      Fix build issue when not using Kerberos. · 4e028028
      Jonathan-Christofer Demay authored
      Apparently when building without Kerberos support the entire contents of
      camel-sasl-gssapi.c is omitted, which breaks generating API docs because
      camel-scan.c expects a camel_sasl_gssapi_get_type() regardless.
      
      With this commit, when building with Kerberos support, CamelSaslGssapi
      is still a registered CamelSasl subclass but it does absolutely nothing
      and is never used by Camel's SASL support.
      4e028028
  10. 19 Jan, 2011 1 commit
    • Matthew Barnes's avatar
      Treat empty cache files as nonexistent. · ee21a86b
      Matthew Barnes authored
      For some reason, cached mail message files are sometimes winding up as
      zero-length files.  Still need to figure out how this is happening, but
      in the meantime teach Camel to disregard zero-length cache files.
      ee21a86b
  11. 17 Jan, 2011 2 commits
  12. 11 Jan, 2011 1 commit
  13. 10 Jan, 2011 1 commit
  14. 27 Dec, 2010 1 commit
  15. 20 Dec, 2010 1 commit
  16. 16 Dec, 2010 1 commit
  17. 15 Dec, 2010 1 commit
  18. 10 Dec, 2010 1 commit
  19. 09 Dec, 2010 1 commit
  20. 02 Dec, 2010 1 commit
  21. 26 Nov, 2010 1 commit
  22. 24 Nov, 2010 2 commits
  23. 23 Nov, 2010 1 commit
  24. 22 Nov, 2010 1 commit
  25. 19 Nov, 2010 2 commits
  26. 16 Nov, 2010 1 commit
  27. 15 Nov, 2010 2 commits
  28. 13 Nov, 2010 1 commit
    • David Woodhouse's avatar
      Fix memory leak recently introduced in imapx QRESYNC · 7d697801
      David Woodhouse authored
      Commit e02cb958 ("Fix QRESYNC protocol violation (RH #648082))
      introduced a potential leak of the 'lastuid' and 'firstuid' strings.
      
      It would almost never happen, because it would require that we're
      selecting a folder on a QRESYNC-capable server where we *have*
      previously seen messages in that folder, but for some reason we don't
      have a record of the uidvalidity and previous modification sequence
      (modseq). The only time that's ever likely to happen is perhaps the
      first time we connect to the server after QRESYNC is enabled. But that's
      no excuse.
      
      Fix this by simplifying the logic. There's no point in using QRESYNC if
      our local cache knows of no messages in the folder *anyway*, so remove
      that separate check and merge the 'if (total)' part into the existing
      condition for actually issuing QRESYNC.
      
      Spotted by Bojan Smojver.
      7d697801
  29. 12 Nov, 2010 1 commit
  30. 11 Nov, 2010 1 commit
    • David Woodhouse's avatar
      Fix QRESYNC protocol violation (RH #648082) · e02cb958
      David Woodhouse authored
      The sequence-set '1:*' isn't valid in QRESYNC, and newer versions of
      Dovecot (>=2.0) have started to object to it.
      
      Ideally we'd omit the parameter completely, since it's supposed to be
      optional — but *older* versions of Dovecot (<=1.2.13) require it.
      
      We only really care about messages which we already knew about. For any
      *new* messages, we're going to fetch the headers and the flags in an
      explicit FETCH command anyway, and we'll ignore anything that QRESYNC
      tells us about them. So just use the first and last UID from our existing
      summary cache as the range for the QRESYNC command.
      
      Fix a shadow declaration of the 'uid' variable in an inner block, while
      we're at it.
      e02cb958
  31. 10 Nov, 2010 1 commit
  32. 09 Nov, 2010 1 commit
  33. 08 Nov, 2010 1 commit