Bugs
----
    * clearsig: keep lineendings as they are. Remember that trailings
      blanks are not hashed.  Funny: pgp263in works fine even with
      a source file with CR,LF but GnuPG and pgp263in has problems
      if the clearsign has been created by pgp263ia.
      Needs more investigation - anyone?

Important
----------
    * Check revocation and expire stuff.  PLEASE: THIS MUST BE TESTED!

    * Check calculation of key validity. PLEASE: IT IS IMPORTED THAT
      THIS GET TESTED.

    * It has been reported that lockfiles are not removed in all cases.
      cleanup is done with atexit() and all signals trigger exit() -
      anything wrong with this?  - ah yes: a signal while still in
      dotlock_make

    * See why we always get this "Hmmm public key lost"

    * print a warning when a revoked/expired secret key is used.

    * Allow the use of a the faked RNG only for keys which are
      flagged as INSECURE.

> 0.9.1: I get some occasional segfaults during 'make check' unless I use
> --with-included-zlib. It's strange, I have zlib-1.1.2 on one machine, and
> zlib-1.1.3 on another, and both of them fail on occasion (maybe half of the


Needed
------
    * remove more "Fixmes"

    * Replace Blowfish by Twofish and add the new encrypted packet typ
      which has a MACing option (append SHA1 hash to the plaintext and
      encrypt this all) - We need an identifier for Twofish to put this
      one into the cipher preferences.

    * The -export-dynamic flag to ld works only for FreeBSD 3.0.  It does
      not exist on FreeBSD's 2.2.x version of ld.
      Also, on my FreeBSD 2.2-stable box, i simply removed the
      -Wl,-export-dynamic flag from my Makefile and it linked and seems to
      be working OK so far.

Minor Bugs
----------

Nice to have
------------
    * preferences of hash algorithms are not yet used.
    * new menu to delete signatures and list signature in menu
    * Replace the SIGUSR1 stuff by semaphores to avoid loss of a signal.
      or use POSIX.4 realtime signals.
    * add test cases for invalid data (scrambled armor or other random data)
    * add checking of armor trailers
    * Burn the buffers used by fopen(), or use read(2). Does this
      really make sense?
    * change the fake_data stuff to mpi_set_opaque