- Thu Aug 17 23:08:08 UTC 2000

[libfaim.git] / CHANGES

No release numbers
------------------
 - Thu Aug 17 23:08:08 UTC 2000
   - More leak fixes (eric, eric, eric...)

 - Thu Aug 17 22:53:03 UTC 2000
   - Change the SNAC cache to a hash instead of list
   - Fixed a few leaks relating to aim_gettlv_str()

 - Thu Aug 17 14:04:18 UTC 2000
   - Fix the fd closing bug introduced last week

 - Tue Aug  8 04:25:35 UTC 2000
   - Preserve subtype and priv after aim_conn_close

 - Tue Aug  8 04:15:47 UTC 2000
   - Fixed double-calling of faim_mutex_init
   - conn->type preserved after aim_conn_close

 - Mon Jul 17 01:56:31 UTC 2000
   - Added 0004/000c callback (for message acknowledgments)
      - This goes with the AIM_IMFLAGS_ACK option that has been there
        for ages now.  When you send a message with that option, you'll
        get a message back containing the screenname, telling you it was sent.
   - Moved the aim_conn_kill's out of aim_get_command() and into
      the client.  They're only closed in there now.  This is so that you 
      can detect what type of connection died.  You'll want to handle
      BOS and chat connections in special ways most likely.  You'll also
      need to call aim_conn_kill when you're done with it.
   - Added/fixed the Rate Change handlers.  Now passes up a long that
      should contain a number.  I'm currently in pattern-searching mode.

 - Sun Jul 16 22:59:39 GMT 2000
   - Fixed the locking in the rendezvous code.  Hopefully is safe now.

 - Sun Jul 16 11:03:28 GMT 2000
   - Fixed 0001/000a chat printf.  (Actually a one liner a few days ago.)
   - Started integrating josh's 3k rendezvous patch.  Currently only
      directim is working.  And its got bugs.  But he's not online
      right now, so I can't bitch.  I'd love to bitch here, but I'm 
      tired. 
     - Can definitly accept directim connections and send/recieve.
     - See faimtest for the new callbacks.  (Urg.)
     - Here weeee gooooo.  (Not sure if this will break everyone or not.)

 - Fri Jun 30 00:04:47 UTC 2000
   - Removed the forceful ICQ-related warnings, just in case they're wrong
   - Added three new fields to the client info.  
      - **REQUIRES CLIENT CHANGES -- see faimtest for new initializer
   - Cleaned up aim_send_login()
   - Put in a nice hidious fix for AOL's latest bit of the drama.  See
      aim_readtlvchain() for the details.

 - Mon Jun 26 07:53:02 UTC 2000
   - Added utils/aimdebugd for playing with things -- see the README
   - Added aim_im.c::aim_parse_outgoing_im().  Probably not useful
      unless you're writing a server or something (or hacking aimdebugd).
   - aim_send_im() now sends the same data as AIM4 does (three more bytes)
   - Added aim_debugconn_sendconnect() for aimdebugd-enablement.

 - Sat Jun 24 02:14:07 UTC 2000
   - Added fakelocks for platforms without pthreads or that have no
      use for them.
   - Added workaround for MSG_WAITALL for people with the very broken
      libc that comes with earlier Red Hat's.

 - Sat Jun 24 00:44:24 UTC 2000
   - Support zero-type TLVs in aim_extractinfo()
     - AOL now ocassionally throws extra zeros in so in order to break
         clients.  Aparently zero-type TLVs are a special case -- if the
         type is zero, then _there is no length_ field.  AOL are geniuses.
     - Note that you only get these if you send a bad client string, such
         as the top one in faimtest.  If you #if 0 that one out and use the
         second one, you'll be fine.  (Unless you just want to see the message
         for yourself.  Its nifty. (FREE!))
   - ICBM parser now uses aim_extractuserinfo() so that it can benefit
       from the above changes, as well as clean up the whole "is it the 
       first or the second one of this type?" issue
     - This also fixes the changes AOL made in away messages.
   = Fixed a possible buffer overflow when AOL changes the max sn length
   - Increased MAXSNLEN to 32 (they appear to be valid)
     - If you start getting messages from someone called "AOL Instant
        Messenger", see comment one, paragraph two.
   - aim_select() now returns a -1 if there are no connections open
   - Connections are now killed completely if there is a major error on them
   - faimtest now calls aim_logoff() on connection error

 - Fri Jun 23 22:38:47 UTC 2000
   - Adds start of adverts support
   - Adds aim_bos_nop().  You may want to send this occassionally
   - Fixes aim_rxhandlers.c to work for people with three-letter
       screen names

 - The week prior to Fri Jun 16 19:37:09 UTC 2000
   - Rethought the rendezvous stuff.  DEFINIT WORK IN PROGRESS
     - Do not bother trying to use this.  This will probably break
         your clients.  You may or may not want to fix them, since 
         I may change it again soon.  
     - I REPEAT: This really is a developer-only version.  It barely
         compiles.  I'm just committing it so I don't have so many
         trees floating around.
   - Added stuff to support OFT as well as OSCAR protocols
   - Added aim_send_im_direct() for directim ("IM Image")
   - Added caps getting/putting routines 
   - Added more aim_login fixes to avoid more AOL stupidity
     - This is the same as went to libfaim-devel yesterday, so don't upgrade
       just to get these changes.
   - Removed aimicq_encode_password -- now use the same table
       for both ICQ and AIM
   - Added AUTHORS

 - Tue Jun  6 01:36:48 UTC 2000
   - Inverted gethostbyname2() check.  Not sure how that ended up 
       in that state.
   - Added some of the initial framework to support file transfers
   - Added in a few checks for that
     - Can currently parse and send a denial message.
   - Added cookie caching.  (Sorry, I just like saying that.)
   - Rearranged channel 2 ICBM parsing (detects the different
       rendezvous types better and hopefully accuratly).
   - Killed the connection array.  Now a list.  (Suits the
       upcoming features better.)  --- REQUIRES CLIENT CHANGES...
     - In most cases, you need to change aim_conn_close() calls
         to aim_conn_kill().  This will free them as well as 
         close them.  If you don't do this, you'll pollute the
         connection list and leak memory.
   - Possibly a few other minor things (some cleanups to faimtest for one)

 - Fri Jun  2 23:27:28 UTC 2000
   - Cleaned up aim_send_login slightly
   - Added aimicq_encode_password to support the new hash
       values required for logging in with ICQ
   - Removed crash when email and/or reg status aren't
       in the auth response (they're not in ICQ responses)
   - So now as long as you specify a client version of 
       4.30.3141 in the client info, and your ICQ number
       as your screen name, you can login to ICQ via libfaim
       and use it just like an AIM account.
   - Added AIM_CB_SPECIAL_CONNERR callback for recieving 
       messages such as disconnect reasonses ("you logged
       on from a different machine", etc).
   - Added faimtest's userinfo handler to support away messages

 - Fri Jun  2 15:57:27 UTC 2000
   - Add creation time, maxmsglen, and various unknowns to chat info
     update callback.
   - Remove 'nothing works yet' comment from top of aim_chat.c (it works!)

 - Wed May 31 01:17:24 UTC 2000
   - Added constants for message sizes (MAXMSGLEN, MAXCHATMSGLEN)
     - Note that the new max message size is defined as 7988bytes,
         drastically larger than WinAIM lets you put out.  This is a
         libfaim-exclusive feature :)
   - Reduced base buffer size on aim_send_im()  
 
 - Tue May 30 22:32:31 UTC 2000
   - Added lock around tx seqnum to prevent that race when transmitting
       from diverse threads. (actually committed earlier today)
   - Added aim_meta.c and a few constants and things to keep
       track of "releases".  
   - Added mkbuildinfo.sh to automatically update build string
       "constants" on every invocation of 'make all'.

 - Mon May 29 12:08:28 GMT 2000
   - Rearranged aim_tx_flushqueue(); moved write operation 
       to aim_tx_sendframe()
   - Turned aim_tx_enqueue() into a macro that calls sess->tx_enqueue,
       a function pointer to whatever you want to use to enqueue
       things for transmition (or not)
   - Old aim_tx_enqueue becomes aim_tx_enqueue__queuebased.  Added
       aim_tx_enqueue__immediate for doing immediate writes.  Default
       is to use queue-based procedure.
   - Cleaned up parts of aim_conn.c
   - Added locking around the sockets themselves.  Should allow
       for full cross-thread usage.
     - Unfortunatly, only pthreads are supported at this time.
       If you don't have pthreads on your arch, implement the
       macros your arch needs and send me a patch.  (A SysV
       semaphore implementation would be trivial, as would a 
       simple integer-based lock.)

 - Sun May 21 14:59:20 GMT 2000
   - Added infotype parameter to aim_getinfo() for requesting
       different types of messages.  AIM_GETINFO_GENERALINFO
       will get you the normal stuff, AIM_GETINFO_AWAYMESSAGE
       will get the normal stuff, with the profile replaced with
       the user's away message.  Also, the callback adds an 
       unsigned short to the end to give you the value that
       was originally passed to the request.
   - There were a few server-pertinent changes that didn't get 
       documented from last month.  I have no idea what they were.

 - Sun Apr  2 07:29:11 UTC 2000
   - Redid much of the tx path to remove the superfluous copy.  But
       it touches every function that transmits.  Its been a long 
       time in coming.
   - Added caps parameter to aim_bos_setprofile().  You can now
       control what capabilties you show up as others to
   - Added ->capabilities to userinfo.  It should be nonzero when
       its coming off buddy oncoming information, but NOT requested
       user info.

 - Sun Apr  2 01:45:15 UTC 2000
   - Increased AIM_CONN_MAX from 5 to 7.  Precausionary only
   - Removed deprecated TIS_TELNET_PROXY from faim/faimconfig.h
   - Added macros for the simple integer functions in aim_util.c
   - Removed usage of aim_failsaferead() 
      - Still not sure why I had it to begin with.  I think it 
        was a remament from when we used non-blocking sockets
        (which I don't know why I did either)
   - Removed some of the oddities in aim_get_command()
      - gprof showed this taking 20kns on my k6-300.  Should be faster now.
      - Added conn parameter.  This should prevent aim_select() from
        getting called too many times in a row.

 - Thu Mar 23 08:45:40 UTC 2000
   - Removed aim_countconn() > 0 check in aim_select(), its logically redundent
   - Added aim_putuserinfo() (inverse of aim_extractuserinfo())
   - Added aim_sendbuddyoncoming/offgoing() 
   - Rearranged loop in rxdispatch()
   - Remove aim_conn_close() if connections dead in aim_get_command()

 - Thu Mar 23 00:44:32 UTC 2000
   - Added a check to purge_rxqueue to skip handled commands

 - Mon Mar 20 05:30:59 UTC 2000
   - Added some server-only functions for login
   - Added aim_counttlvchain()
   - Added aim_sncmp() and aim_snlen()

 - Sun Mar 19 06:07:52 UTC 2000
   - Added a parameter to aim_select to return event type 
     - REQUIRES CLIENT CHANGES.
   - For the most part rewrote the tx and rx queuing code 
     - Should fix many, many outstanding problems in and related
         to that code, including one that keeps memory from freeing 
   - Fixed several bugs in various places
   - Reformated a lot of code, and did general cleanups
   - Should have a generally more robust lib now.

 - Sun Mar 12 00:07:40 UTC 2000
   - Fixed a robustness problem in aim_handleredirect_middle()
   - Added TLV chain creation routines (yes, aimd is progressing)

 - Mon Jan  3 04:07:55 UTC 2000
   - Fixed bug in aim_snac.c
   - Fixed condition where commands read from connections that have 
       been closed were still left in the queue.  Now cancelled.
   - Added some printfs to aim_info to get more informative crahes
   - Fixed a bug in aim_rxqueue::purge

 - Sun Jan  2 10:31:19 UTC 2000
   - Cleanups in aim_info.c
   - Can compile with -Ddebug=100 again
   - Implemented chat: Joining/Leaving, info parsing, IM parsing/sending
   - Implemented some chatnav: rights req/parsing, room creation

 - Thu Dec 30 10:08:42 UTC 1999
   - Fixed bug in aim_im.c when (encoding == NULL) || (lang == NULL)
   - Added detection of voice chat requests
   - Added AIM_CLASS_* defines, including new Away flag
   - Added awaymsg parameter to bos_setprofile.
     - If awaymsg is nonnull, you will be advertised as being away (your
         class will be ORed with AIM_CLASS_AWAY), otherwise you'll show 
         up normal.

 - Wed Dec 29 10:06:35 UTC 1999
   - Fixed small bug in IM parser
   - Added stubs for the capability TLVs in userinfo.

 - Wed Dec 29 09:14:45 UTC 1999
   - Added a capability block to aim_bos_setprofile.  Can now get chat 
       invites again.
   - Extended ICBM parser to support channel 2 messages (chat invites)
     - A channel parameter has been prepended to the varargs -- REQUIRES 
         CLIENT CHANGES.
     - Extended faimtest to support chat invites.
   - Changed faimtest to get sn/password from environment

 - Wed Dec 29 04:17:03 UTC 1999
   - Added -g to CFLAGS
   - Added aim_sendconnack() to aim_login.c (needed for newer login)
   - Added code for the new SNAC-based login/auth procedure. (see SNACLOGIN
       in faim/faimconfig.h for why its not enabled)
   - Reimplemented aim_authparse(), aim_handleredirect() using TLVlists
       - The old auth_failed callback is now integrated into the
           success one.  If there was an error, logininfo->errorcode is nonzero
   - Fiddled with version information.  Added aim_setversions()
   - Added table of SNAC names for showing unknown snacs (jbm)
   - Added a middle handler for MOTD
   - Added new authorization SNACs to faim/aim_cbtypes.h   

 - Sun Dec 26 22:59:10 UTC 1999
   - Renamed login_phase1_struct to aim_login_struct
   - Changed cookie and sn to be static arrays in aim_login_struct
   - Integrated the Jabber-faim changes.  (temas)  [BIG CLIENT CHANGES]
     - Added aim_session_t, removed all global variables
     - Changed all functions to accept a session pointer
     - Removed aim_global.c
     - Updated faimtest to use an aim_session_t.
     - Removed all cases where logininfo was passed as vararg to client
   - Fixed small bug in aim_newconn's 'fixing' of host:port addresses
   - Added an install rule to the makefile (installs headers+so only!)
   - Enabled USE_SNAC_FOR_IMS by default, assuming it got fixed
       by n's new aim_snac.c from ages ago
   - Implemented a middle handler for 0004/0001 message errors, added
       snacid lookup to get illfated destination SN and pass to client
   - Implemented a short middle handler for offgoing buddy.

 - Fri Dec 24 21:30:06 UTC 1999
   - Added an error-counting Read() that has been sitting in my inbox
   - Cleaned up header files, created aim_cbtypes.h.
   - Added void * to aim_conn_t for private client use if they want. (Orb)
   - Removed all stderr output.  All output is important, and goes to stdout.
   - Renamed isautoresponse in IM parser to icbmflags.
   - Added Orb's fix for the new login code, deleted old (see Orb, I do read
       your mail....eventually).
   - Added mailing lists to README.

 - Fri Dec 24 11:12:34 UTC 1999
   - Cleaned up both outgoing and incoming ICBM handling.  Anything
       that crashes around there is no longer libfaims fault!
       - The encoding flags are now passed up to the client.
   - Added several TLV routines to parse large blocks of continuous
       TLV triplets.  Not terribly effecient, but quite elegent in usage.
   - Added icbm_setparams() back in from way-back-long-ago.  It hasn't
       been implemented in a long time, but I think we should still send it.

 - Fri Dec 24 01:23:06 UTC 1999
   - Fixed a very minor bug in aim_newconn().
   - Added aimutil_get{16,32}()
   - Added aim_extractuserinfo() for extracting  user data
       blocks and putting them into struct aim_userinfo_s's.
         - Added a loop to print out info on extraneous TLVs.
         - Put in lots of comments.
   - Added parse_oncoming_middle() to parse the user data
       block of that packet.  Now passes struct aim_userinfo_s
       to client.
   - Rearranged parse_userinfo_middle().  Now passes an entire
       userinfo struct to client instead of individual variables.
   - Convered the version of parse_im_middle() thats actually getting
       used to pass up a userinfo struct.  
   - Updated faimtest to accept new structs.

 - Tue Dec 21 06:18:50 UTC 1999
   - Fixed a Win32 header problem
 - Tue Dec 21 03:44:13 UTC 1999
   - Latency timers now update on rx as well as tx.  Gets rid of even more
       of the rate problems that most clients are having.  
   - Renamed lasttx and settxlatency to lastactivity and setlatency, respec.
   - Integrated fixes needed for Win32 -- should compile cleanly now (DMP)
   - Cleaned up some places, use aim_putsnac everywhere now.

 - Sun Sep 26 20:04:20 MST 1999
   - Reduced the IM parsing to look for 3 zeros instead of 4 -- NEEDS WORK
       - This was needed to work with MacAIM3 and some WinAIM3s.
   - Added aim_conn_settxlatency() for throttling outgoing frames -- NEEDS WORK
   - Added an int to the userinfo and incoming IM user callbacks for new
       TLV that AOL put it in -- its the number of seconds elapsed since
       the user logged in
   - Worked more on the callbacks (more internal rearrangements)
   - Fixed bug in aim_select() (returning negative fds)
   - Clear out logininfo struct before putting more data into it
   - Other bugfixes that I can't particularly remember.

 - Tue Aug 24 03:13:12 UTC 1999   --- TRANSITION RELEASE!!
   - Added jbm's new aim_rxqueue.c, which should crash less
   - Started the overhaul on the callback system.  No where near complete yet.

 - Sun Aug  1 03:02:17 UTC 1999
   - Added aimutil_*()s in aim_util.c for raw byte placement
   - Cleaned up aim_im.c, aim_auth.c, and aim_login.c using aimutil_*
   - Added AIM_IMFLAGS, flags option to aim_send_im(), removed
       aim_send_im_away()
   - Added client_info parameter to aim_send_login()
   - Tweaked aim_send_im() (AOL changed a few things, there's some more
       changes left to do)
   - Some Chat stuff changed, still no where near functional
   - Finally remembered to switch the license to LGPL (from GPL)
   - Permit/Deny (blocking) list support added
   - Released a snapshot

 - Sat Jul 31 05:28:38 UTC 1999
   - Changed aim_bos_setdeny() to aim_bos_changevisibility() and actually
       did the implementation.  

 - Fri Jul 23 17:45:22 UTC 1999
   - Work around for the AOL change in IP syntax (thanks to Eric Peyton)
   - Released snapshot

 - Sun Apr 25 23:31:44 UTC 1999
   - Fixed stupid off-by-one bug in aim_logoff()

 - Sun Apr 25 22:43:31 UTC 1999
   - Renamed/reclassified missed IM errors (now all callbacks -- no backend handling)
   - Killed aim_login() -- all connections now opened in frontend
   - aim_conn_close() will not close fd's < 3
   - Released snapshot.

 - Sat Apr 10 22:44:07 UTC 1999
   - Changed how rx_command->handled works; now uses return value from callback
   - Changed the if's in the dispatcher (aim_rxhandlers.c) to switch()es

 - Sat Apr 10 03:51:21 UTC 1999
   - Started CHANGES
   - Put in n's aim_snac.c.
   - Updated aim_{add,remove}_buddy() to use aim_conn
   - Updated aim_usersearch_address() to use aim_conn
   - Cleaned up aim.h a bit (removed old *phase* protos)
   - Changed gethostbyname() to gethostbyname2() (for MacOS X)
   - Changed aim_newconn() to pass connection errors in connstruct->status
     - Updated faimtest.c to detect connection errors
   - Fixed a small typo in aim_im::aim_parse_incoming_im_middle()
   - Added free()s in that function and to userinfo counterpart (fewer leaks)
   - Released snapshot

 - Sat Apr  3 06:43:14 UTC 1999
   - Released snapshot

 - Wed Dec 30 01:20:59 MST 1998
   - Released snapshot

 - Mon Dec  7 16:57:57 MST 1998
   - Released snapshot

 - Sat Dec  5 00:01:53 MST 1998
   - Released snapshot
 
 - Thu Nov 26 17:50:02 MST 1998
   -  Released snapshot