changes

oscar message protocol stack

No release numbers
------------------
 - Fri Oct 26 15:18:10 PDT 2001
  - Fix buddy icons. (the checksum is only 16bits. hum.)

 - Sun Oct 14 19:45:54 PDT 2001
  - Fix aim_icq_setstatus
  - Minor touchups
  - Parse 1/f (selfinfo)
  - Add ICQ status defines

 - Wed Oct  3 11:07:22 PDT 2001
  - Fix awful bug in ft listenestablish.
  - struct aim_userinfo_s becomes aim_userinfo_t.
      - I haven't decided what to do with this.  I like it in its current
          form to an extent, but it is also not flexible enough.  But it would
          become too complex and cumbersome if I were to do it more flexibly.
      - Also added a few accessor functions that should be used instead of
          directly accessing the struct, particularly for the capabilities.

 - Wed Sep 19 18:50:34 PDT 2001
  - Remove non-standard default features from aim_send_im_ext.
    - Sadly, AOL 5.0 ignores messages with non-standard features.  Once
        AOL releases AOL 6 for the Macintosh, there will no longer be any
        reason for anyone to use AOL 5, and this will be a non-issue.
  - Check for 0x2711 TLV in buddy icons.  This is weird.
  - Add AIM_IMFLAG_ACTIVEBUDDY.
  - Fix ICQ login. (Oops.)

 - Tue Sep 11 16:41:11 PDT 2001
  - Fix chat invites.
  - Keep faimtest from passing va_list's. That scares me.

 - Tue Sep 11 04:42:28 PDT 2001
  - Fix faim_internal/export confusion for the TLV functions
  - Add aim_addtlvtochain_userinfo() (used by AP)
  - Clean up ICBM Channel 1 parsing ever so slightly
  - Add multipart message support in both directions, and some 
      fun examples of it in faimtest.
  - Yes, I'm insane.

 - Mon Sep 10 06:15:43 PDT 2001
  - Remove $Id line from md5.h. I really really hate those things.
  - Remove all that locking crap. It never worked right anyway.
  - Add conn->inside for things that should never ever be seen by clients
  - Store supported groups list for each connection
  - Support migrations.
   - faimtest now theoretically supports doing this.  It happens so rarely
       that it makes it difficult to test.  For more complicated clients, a
       flag will have to be kept so that some things that happen during login
       don't get triggered the second time around.  But faimtest doesn't have
       such problems.
  - Added some random commentary.

 - Sun Sep  9 03:01:28 PDT 2001
  - Three fixes in directim for gaim.  It works now. Yay.

 - Sat Sep  8 21:26:27 PDT 2001
  - Minor bug directim_connect

 - Sat Sep  8 20:18:34 PDT 2001
  - Split up the ICBM Channel 2 handlers a bit more
  - Add a stub parser for recieving buddy lists.  It will stay that way until
      someone decides on a good API for passing lists like this to the client.
      In case someone does, the same standard will be used for setbuddylist and
      the as yet nonexistant support for server-side buddy lists.
  - Make infochange work again. I still don't like it, but I don't want to 
      fix it.

 - Sat Sep  8 19:05:05 PDT 2001
  - Redo ICQ login
  - Fix search by email (oops).

 - Sat Sep  8 17:07:09 PDT 2001
  - Fix directim so you can do it more than once without crashing.
   - This removes the connect/disconnect callbacks.  They were pointless, as
       you can get the same information at a better spot by looking for when
       aim_getcommand returns -1, just like is done for FLAP connections.
   - (This was causing aim_conn_kill to be called twice for the same connect,
       once in the client callback, and once more when aim_getcommand returned
       the -1).
  - Add aim_conn_(close|kill)_rend().  This is for destroying the cookie and
     the internal data sections in one spot.
  - Fix a bug in aim_connrst (and hence aim_logout()) that caused intdata/priv
     to not be freed in those cases.  Evil.

 - Sat Sep  8 07:32:27 PDT 2001
  - Clean up ft.c slightly.  Direct IM now works.  In both directions.
   - This could still use a lot more help.  It should use bstreams more.

 - Sat Sep  8 00:55:46 PDT 2001
  - Chatnav loop fix. (Err. I shouldn't code when I'm that tired.)
  - Remove some things from BUGS
  - Remove USE_MACROS from faimconfig.h.  Not used, and frivelous when it was.

 - Fri Sep  7 21:18:51 PDT 2001
  - Make icon field names uniform
  - Add AIM_IMFLAGS_CUSTOMFEATURES.  This allows the client to send/recieve
     the field in IMs that show client information.
    - This can be used to identify other open source OSCAR clients, if any
        one is interested.
  - Increase the size of args->icbmflags to 32bits
  - Make sure that extended-only flags are not set for aim_send_im()

 - Fri Sep  7 19:59:43 PDT 2001
  - Fix buddyicons. Yay! (checksums are 32bits all the time. duh.)

 - Mon Sep  3 18:48:26 PDT 2001
  - Reformat everything to use real tabs (and to my latest coding style)
  - Abstract out the numerical data types to fu8/16/32_t for portability.
  - AIM_FRAMETYPE_OSCAR -> AIM_FRAMETYPE_FLAP.  This makes more sense.
  - aim_conn_t's FLAP sequence number was a signed int.  Oops.
  - Get rid of the 'struct' on all types suffixed with _t.  That's been 
     annoying me for a while.  They're all real typedefs now.
     - Yes, client people are going to be rather annoyed by this.  But it 
         doesn't stop there.  Keep reading.
  - Make the 'struct aim_rxcblist_t' type local to rxhandlers.c.
  - Combine the command_tx_struct and command_rx_struct into aim_frame_t.
     - Mostly aim_frame_t is command_rx_struct.  For command_tx_struct, the
         same structure is used, except where ->sent was, ->handled is now 
         used.
     - This makes things a lot easier, everywhere.
     - Get rid of ->lock.  Pointless.  If libfaim is ever preemptible, it 
         has much more important problems than that.
  - Welcome to aim_bstream_t.  No more direct buffer accesses.  Anywhere.
     - In fact I plan on getting rid of the aimutil macros completly.
     - This isn't complete yet.  It will be better later.  Believe me.  Maybe
         even make more sense.  
     - More advanced and configurable boundschecking is coming, too.
  - Clean up lots of stuff, everywhere.
  - Rearrange the implementation of the TLV functions, saving lots of code.
      - I'm rather pleased with the way some things got implemented with this,
          particularly in places where TLVs contain TLVs.
  - Get rid of aim_puttlv_() functions.  Those were gross. Use tlvchains.
  - XOR login is temporarily broken.  I'll fix it someday. ("Someone" needs it.)
  - Fix the return values of most everything -- OSCAR functions should all 
      return 0 on sucess, negative errno on failure.
  - There are several things braced with #ifdef MID_REWROTE_ALL_THE_CRAP.
      Consider that code broken and nonfunctional for now.
  - I think I may have broken buddy icons.  Remind me to fix that.
  - Renovate faimtest substantially.  Reformat, split up, update to new types.

 - Wed Aug 29 16:59:24 PDT 2001
  - Pass up entire icon triplet (checksum/length/timestamp) in all cases
     that it is recieved.
  - Eric was paying more attention than I am.

 - Sat Aug 25 19:46:38 PDT 2001
  - Pass "Maximum visible message length" to client in chat info update
    - This is the real maximum message length now.  There is still a
      "complexity" limit that will prevent you from sending messages 
      even if they are less than this limit (102 right now), but I don't 
      know how that is calculated.

 - Sat Aug 25 19:19:16 PDT 2001
  - Keep parsing missed calls for entire packet
    - This is really easy to test now.  Just talk to someone with
      their minmsginterval set high (like 10seconds or something)

 - Sat Aug 25 18:35:13 PDT 2001
  - aim_sendconnack -> aim_sendflapver
  - Remove addicbmparam.  This was identical to aim_seticbmparam
  - aim_seticbmparam now takes a struct with the parameters in it.  See
      faimtest.  !!! REQUIRES CLIENT CHANGES if you use aim_seticbmparam, which
      you really really should -- the defaults are insanely limited
  - Fix rendezvous IM parsing that I broke last time.
  - ICBM parameter info callback gives same struct as above.  !!! REQUIRES
      CLIENT CHANGES
  - Reformat login.c and part of faimtest.c
  - !!! NOTE: don't do anything in your motd handler.  If you do, it will
      get run twice, sine for some reason AOL is sending two MOTDs now.

 - Wed Aug 22 22:11:18 PDT 2001
  - Reformat the rest of im.c
  - Fix a few returns, particularly in error cases

 - Thu Aug 16 06:17:57 PDT 2001
  - Reformat a bit of im.c and aim.h
  - Rearrange encoding/subencoding checks in IM parsing
 
 - Fri Aug  3 14:19:37 EDT 2001
  - Naje aun_sebd_un retyrn sabe vakyes

 - Thu Aug  2 13:28:37 EDT 2001
  - Uhm. Why does arcanejill think its in EDT?
  - Cache the username on warnings so errors work right.
  - Someday, I'm going to convert all of libfaim to that formatting style, too.

 -  Wed Jun 27 00:13:20 EDT 2001
  - Some comments about UNICODE in faimtest.

 - Fri Jun 15 17:51:27 EDT 2001
  - Make faimtest compile on RH7.1 (new gcc).
  - Add aim_conn_getsess()
  - Add aim_getconn_type_all() and aim_getconn_fd().
  - Make aim_connrst() and aim_conn_getnext() static to conn.c.
  - Why does arcanejill think she is in EDT? Oh well.

 - Thu Jun 14 17:37:47 PDT 2001
  - Rearrange aim_setprofile().  It will now let you _not_ send a profile,
      if you really want to (pass NULL).  Note that this is quite different
      than sending a _blank_ profile.  Also fixes the "asci" bug.

 - Mon Jun  4 12:57:46 PDT 2001
  - Fix chatnav.  Whoops.

 - Mon Jun  4 12:45:18 PDT 2001
  - Don't use snacid's for generic requests (gets around recent blocks)
  - If they start doing snacid-based blocks again, an audit will have
      to be done to see what WinAIM requests uses snacids for and what 
      it leaves it blank for.  That will be tedious and I'm not going to 
      do it. 

 - Thu May 31 23:00:50 PDT 2001
  - Store two hardcoded hashes in login.c.  Sigh.
  - Add prototype for aim_callhandler_noparam in aim_internal.h

 - Sat May 26 03:58:27 PDT 2001
  - Aha! This is why I could never get icons to work before.
  - Luckily the checksum is painfully simple.

 - Sat May 26 01:37:39 PDT 2001
  - So this is my big merge with the ActiveBuddy version...
    - Remove use of tlvchains from IM parsing for better performance
    - Add in buddy icon support
    - Add support for sending UNICODE messages.
  - Split up the more complicated options to aim_send_im into a structure
      that is passed to aim_send_im_ext().  This makes it easier to send
      simple messages, and only slightly harder to send complex ones.
     - aim_send_im_ext() must be used if you're using buddy icons and/or you
         are sending UNICODE messages.
  - UNICODE is fully explained in commentary above aim_send_im_ext() in im.c.
  - The ISO-8859-1 option is also added.  Use when necessary, to avoid making
      WinAIM barf.
  - This removes the msglen parameter to aim_send_im because that was only
      needed to do UNICODE, which can now only be done by aim_send_im_ext.

 - Wed May 23 18:05:46 PDT 2001
  - I upgraded libtool.  Which broke the faimtest makefile.
  - Parse locate rights
  - You shouldn't ask so many questions, Eric.

 - Wed May 16 12:48:13 PDT 2001
  - Thanks Rob and Eric

 - Thu May 10 17:37:23 PDT 2001
  - I haven't had lunch yet.  What should I have?

 - Wed May  9 15:21:51 PDT 2001
  - I had pizza for lunch.

 - Mon May  7 15:19:55 PDT 2001
  - Rename rxhandler_t to aim_rxhandler_t.  This has been 
     bugging me for a while.

 - Thu May  3 15:47:23 PDT 2001
  - Clean up getcap/putcaps
  - Add GAMES2 and SENDBUDDYLIST caps (the latter is new in winaim 4.7beta)
  - Update hash to latest value

 - Wed Apr 25 11:51:39 PDT 2001
  - I think this was a bug.

 - Tue Apr 24 17:36:03 PDT 2001
  - Let Eric generate the hash if he wants.

 - Mon Apr 23 15:58:11 PDT 2001
  - On a clear day, you can see forever.