Sun Jul 30 23:06:50 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  tagged 0.17

Sun Jul 30 23:05:41 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Remove the "prerelease" script as it's no longer needed.

Sun Jul 30 22:44:45 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Use unsigned long for the thread ids to avoid a warning.

Sun Jul 30 21:04:23 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Make libold compile in Windows.
  Change libold to build natively under Windows, at least with the mingw
  compiler.
  
  Thanks a lot to Ron Burkey (rburkey2005@earthlink.net) who provided the code
  and tested the patch.

Sat Jul 15 12:35:37 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Clarify licensing terms.

Sat Jul 15 04:52:26 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Don't force the fd to be active in net_wakeup()
  net_wakeup() forces the file descriptor it's waking up to be active, by
  setting active_fd[fd] = 1.
  
  This can cause the following race:
  
   * Lock L is held by client A
   * Client B is connected and waiting for lock L to become available
   * Client B decides to disconnect, and closes the socket
   * This triggers net_close()
   * At this point, client A releases lock L, triggering net_wakeup()
  
  Now, here's how it could go:
  
    net_close()              net_wakeup()
      active_fd[fd] = 0
                               active_fd[fd] = 1
                               net_send_cmd[fd] returns OK because
                                 the socket is not closed yet
      close()
  
  
  And that leaves fd closed, but active_fd[fd] = 1, which will cause the
  select() loop to hang up indefinitely, because it gets a "Bad file descriptor"
  error (due to a fd on the read set being closed), which causes it to rebuild
  the read set and try again. But nothing will never remove that fd from the
  set, because nothing will set active_fd[fd] = 0.
  
  The fix is not setting active_fd[fd] = 1 in net_wakeup(), it has nothing
  to do there in the first place.
  
  
  Besides, check if the fd is active before calling net_send_cmd(), because we
  know it's going to fail.
  
  Thanks to Lloyd C. Cha for the report and testing.

Sat Jul 15 04:50:42 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Don't cleanup after net_send_cmd().
  net_send_cmd() calls clean_buffer() and net_close() when it gets a write
  error, so there's no need to do cleanup when it fails. But net_wakeup() does
  just that: when net_send_cmd() fails, it also calls clean_buffer() and
  net_close().
  
  This patch fixes that by removing both calls.

Sat Jul 15 04:17:33 ART 2006  Alberto Bertogli <albertogli@telpin.com.ar>
  * Compile with -fPIC.

Sun Dec  4 19:12:47 ART 2005  Alberto Bertogli <albertogli@telpin.com.ar>
  tagged 0.16