logboris.txt

The source code of Doom legacy for windows

- improve network with supposed nettic
- added possible value for cvar
- fixed map7 bug when the 2 last babyspider die at same time (original 1.9 bug)

stop here for beta 4
here is new for beta 5

- changed menu handling mutch more usable and very easy to change
  it handle cvar automaticely and mutch more !!!
- put fix for some Fab bugs (2 sided texture)
- improve more network client send only nack and server don't send to much
  data i think it reduice dramaticaly the bandwidth needed
- fixed fraglimit

13-11-98

- fixed -left and right with the new protocol

14 15-11-98

- 2 Days to find the fucking Fab Bug ! texture cache can't be freeed without
  a total flush of the cache

20-11-98

- make the join with Fab version. Very hard task ! i have put the makecol15
  from windows version because using makecol15 from allegro add 150kb to the exe
- put maxvalue 4 to gamma.
- added vid_ticrate cvar

22-11-98

- fixed final pic shows in doom1 and bunny scroll
- 5 miniute to fix the damn bug from fab with there flatcache problem
  but i suppose this in not fixed at all is just a quik fix because i am bored
  to see damn f**ed flat !
- outch 1 line added and you can take a screenshot pressing the key that you
  want. What a wonderfull world. Ok nice added screenshot command is
  recommanded to bind it.

23-11-98

- fixed deathmatch cvar don't work in network (spwan player problem)
  the solution is : use the map command to start the game !
  it this occation i have clean lot of stuff in d_main and d_clisrv init code

24-11-98

- added load command and make it work in network
- added skill parameter in the map command

26-11-98
- added save command and cleaned the way that it work in network no more
  use ticcmd to send save game (original doom 1.9)

29-11-98

- put cv_Unsigned for lot of variable and fixed little problem with fraglimit
- fixed "got unknow netcmd 109" when change player config while game not
  started it was because the lmp execution execute the last textcmd and the
  demoversion is not the same
- fixed console display with the new system of "map" pushing using warp
  and netgame

1-12-98

- raaa i am tired to clean the code... It make me sick, i have found some bug
  using tha map command in demos, network and console. The biggest is the
  engine parameter is not set like in a single player game. secondary there
  was demoplayback=false in g_initnew (the map command execution). and thirdly
  netgame was put true when playback demos !
  My first fix is to put a g_stopdemo in the map command and in load level
  this fix to most part of problem (eatch time before calling g_initnew).
- clean the way that doom lunch the game, now g_newgame sets all parameter to
  normal and then push deathmatch and map on the command buffer.

2-12-98

- fixed little problem with console_toggle_off when playing old demos
- fixed viewz of the first frame
- try to fix te clientprediction code, i have rewrite it and found one cause
  of non function and i see it will not be so easy :(

3-12-98
- clientpreditcion work but still need more work

4-12-98
- added timedemo console command, and change timedemo no more leave program
  just print the result at the console

5-12-98
- some note here to don't forget to put in what's new because fab no more
  update there log.
    - transparent texture.
    - mutltipatch can be put in 2 sided line
    - load texture at runtime
- change the pause handling no more use the ticcmd to send it. Now the ticcmd
  is much cleaner i have 1 bits more for player command.
- Do r_loadtexture only one time (no more at eatch loadlevel). It is called
  when adding wad at runtime if it detect texture changes. If no texture
  change it only flush the texture cache. Also there was a little bug
  because R_GenerateTexture don't check if the texture is not allready loaded
  i have put a test in R_Precache level. Also i have reput the precache flat.
- fixed little bug in P_AddWadFile when searching the firstmap

6-12-98
- added sayteam command

8-12-98
- i have removed the viewz==1 at the first frame. I still don't know for what
  this hack is...
- got the client prediction running i stop here for the moment still to fix :
  - we hear all the sound twice because spirit and player spawn the sound
  - the hack bit of fab to permit the player walk in air is modified by the
    spirit, this cause some "consistancy failure". Stupidely removing this
    spirit ability will miss predict player movement !
  - remove localtic in the ticcmd_t
  - aparentely there is a "consistancy failure" sometime when the client use
    a weapon (plasma) i don't know what it is but i need more information on
    this
  other bug not related to spirit:
  - the usage of firstmap level cause reload of the firstlevel in a wad using
    map or addwad command
  - passing deathmatch 2 to deathmatch 1 don't respaw the weapons
  - due to change of map command in lmp deathmatch cvar is put to 0 to lattely

9-12-98

- respawn weapons when passing to deathmatch 2 to deathmatch 1
- fixed sound play of spirit
- fab remind my to fix the frame synchronisation TODO

18-12-98
- finely fixed the synchronisation problem in network, i was caused by a
  spupid wait a new tic before do a GetPacket() !

19-12-98
- now use explisitely the "max" for bounded cvar
  with this we can insert a other special stuff between, very usefull since
  the menu use the value 1 for the index for the maxcursor.
  All this stuff to unlimit mousesens cvar ! b鰎k
- put 1000 for fraglimit max
- put specific name to eatch deathmatch type and allow only 0,1,2,3
  i think 3 is already used to respawn item and don't remove weapons
- give keycard when changing deathmatch from 0 to anything

xx-12-98
- release it as 1.27 beta 1

28-12-98 : Boris Bored at the Party 98
- fixed menu sigsegv with sound playing in the menu

xx-12-98
- release it as 1.27 beta 2

18-1-99 : End of exams

- working on heretic and hexen source

24-1-99

- put the code from fab's one, for the blood splat and bullet hole.

26-1-99

- fixed load game when there is demo playing back.

27-1-99
- fixed deathmatch value for old lmps

2-2-99

- first step for "join in game", for the moment when a new node is detected
  the server incresse the session number, and go to the D_ArbitrateNetStart.
  Client receive a server packet with a newsetion, this cause a call to
  D_ArbitrateNetStart, with this little modification it permit to synchronize
  all computer to have the same playeringame vector. But this cause many
  problem :
    - when returning to the game, player have changed there consoleplayer
      number, exept the server.
    - the new player don't have a "incarnation" in this game, all computer
      must spawn a new players at the same place.
    - the new player don't have the game in memory, it will kicked out with
      consystancy problems.
  This note is not realy to be read, it is for my a way to undertand what i do
  I propose to me this solution :
    don't do a real new synchronisation, just redo it for the new player
    the first problem will be fixed with this, the second will be fixed
    easly by doing a spawn at the end of the function. The third problem
    will not be fixed, since we have't a file copy feature. I will do a
    little hack someting like a variable newplkayerwaiting=true, and checked
    at intermision, to prevent the new player to do p_setuplevel...

12-2-99

- i remind one stuff of the lost beta4 implemented :
  the loadconfig, saveconfig, and changeconfig
- i have document all the change, but fab is a little lazy and don't whant to
  correct my bad english, to bad :(...

26-2-99

- debug a little bit the "join in game" code : before call d_arbitratenetstart
  increment doomcom->numnode it is nomal we have a new node. Also put acceptnew
  node always true. It work a little, but hang on the server side with a
  consistancy falure. Normal i don't have spwaned new player for the node ...
- ok fixed a little more stuffs, now it work since the server don't move, if he
  do "map xxxxx" at the console it work and both can play in the new level
  exept a little problem when the joined player die and respawn, it do a
  consistancy failure... should work for that...
- fixed some stuff for fab the splitscreen player did't have there proper
  localangle and localaiming, and fab don't understand this stuff :P
- added support for textcmd for the splitscreen player it use TEXTCMD2 the same of
  TEXTCMD but the server add 1 to the playernumber (it use the particulariti that
  the second player is the first player +1)

27-2-99

- finaly fixed config of player2 (register name2, color2, skin2 cvar)
- made a new status type for the menu : IT_DISABLE it will show the patch
  in gray and unusable, i use it for the multiplayermenu to desable
  the setup of the player 2
- fixed some more stuff put " between string in exec call because ':' and '\' are
  separator
- and finaly improuved splitscreen support

1-3-99
- improuve splitscreen support

3-3-99
- convert aiming to be a angle_t, work fine and it is much cleaner
- splitscreen aiming is now automatic no need for a special case

5-3-99
- change the way that ack packet work, better duplication detection
- fix the damned config of the splitscreen
- begin the savegame compression (save only the differance)

6-3-99
- finish savegame modification, but need more work maybe just zip it....
- save only usefull field of mobj, i have learn that save game save also
  static mobj, that not a good idea but i haven't change that
- save game save also taget and tracer pointer with a little trick...
- reenabled the loadskin in addfile command work well....
- now you can use -connect witout parameter it will use broadcast packet
  and autodetect the server in the neighbourhood

8-3-99

- begin file transfer protocol (ftp :))

9-3-99

- fixed some stuff about splitscreen : cl_maketic is no more a table of
  player but a table of node, new table : playerpernode
- added cv_notinnet flag, to don't permit change of this variable in netgame

10-3-99
- fixed player name don't save when leaving menu
- now the leave procedure is reliable with a timout of 5 sec
- fixed some free call in djgpp/i_sound.c (don't free not allowed memory)
- added support for joystick from allegro 3.11 see djgpp/i_system.c

11-3-99
- fixed some problem in the protocol caused by too much textcommand in a packet
  can happen when double splitscreen or just because there is buch of extracommands
  is't realy a fix just detect it better and reduisse the probablility that happens

14-3-99
- fixed -file getting parameters begining by '+'
- finaly fix the textcmds overhead problem put textcmds in futur tic if this happen
  and when is no more place (32*255 byte received in on tic=8Kb*35/sec=280Kb/s ) quit
  with i_error
- do a join with fab, this include win32 midi player and the new little
  window at begining, and some other stuff
- we have discover a differance between optimization and non-optimized code with VC5
  but anyway it is fixed and is not the fault of VC5

15-3-99
- fixed the spawn of coop player when more than 4, who play in coop with
  4 players ? Anyway this bug appear quickely if someone forget to put -deathmatch

18-3-99
- fixed the fix of 15-3-99
- found a strange think : sometime when stated from a fullscreen dos, legacy don't
  detect any video mode. Why ? anyway : put it in the doc as "know bug"

19-3-99
- finaly fixed the textcmd stuff, now is deleted only when acknoledged

20-3-99
- fixed a little bug in new savegame code
- splitscreen player have now there sound played

21-3-99
-  begin next and preview weapons

22-3-99
- fixed and finish next and preview weapons

23-3-99
- fixed some splitscreen issues
- show key in the statusbar overlay don't forget to add "k" in the overlayxxx

25-3-99
- implement recursive search for wad files

26-3-99
- added fastmonsters cvar
- added file name sent and loaded at runtime if not found in network

27-3-99
- fixed "+" pushing issues
- now demo can be recorded by only one computer

30-3-99
- i lost my wall journey to try to inderstand why demo get out of sync
and especially it get out of sync differantely using vcc and gcc with
the same source file !!!
- finaly using printf __file__,__line__ in #define all have disapear
  damned i don't understand anything all MUST be explicable....
  I will assume that Vcc don't rebuild all when i say rebuild all :(
  but i am not sure.... a case for F.Mulder and D.Skully  :)
- i have remove my #define with the __file__ and the other stuff and the
  lmp out of sync reappear, now i'am sure vcc is bugger to hell, must get
  VC 6.0 or get the 5.0 service pack... :(
- well finily i have change p_random and m_random to returning a byte
  then vcc no more generate buged code, fine but how many bug like this
  one vcc have generated, i will probably use more pgcc to test first :)