logboris.txt

The source code of Doom legacy for windows

3-4-99
- installed vc6.0, but haven't compiled because afxxxx.h not found, must ask
  to fab
- make the dependency for the dos version using gcc -MMD
- finaly found why lmp 1.9 get out of sync, stupid if(!fastparm) transformed in
  if(fastparm) (switching to cvar :()

4-4-99
- have compiled the join with Bell kin on my brother slackware 3.5
  (egcs 1.0.3), it work but i have sigsegv in map01 :(

5-4-99
- finaly installed mfc and now it compild under vc6.0, but i am not at end of
  deboging now with vc6 the demo get out of synch directely !! damned
  i will try to don't flame MS VC maybe is our fault (ambiguity in c language)
  Anyway for now compiling with debug release it work fine.
- first test with cross platform network test : IT WORK ! until the server
  (linux) hang with a sigsegv :(
- with gdb i have seen that the fault come from vquadloop (this is a label
  in tmap.s) must check this....
- well 4 hours latter after a real finding what wrong with vc6 i found it
  here it is : when i do p_pspr.c,p_gunshot() line 607 :
    angle += (P_Random()-P_Random())<<18; /* original 1.9 doom code */
  the compiler do : /* i see the desassemble code */
    rnd1=P_Random();
    rnd2=P_Random();
    angle += (rnd1<<15-rnd1+rnd2)<<18;
    /* in short : the compiler have use the first call for the negative value */
    /* in debug mode it use the seconde one for negative value */

  Well i have fixed this putting :
        angle += (P_Random()<<18);
        angle -= (P_Random()<<18);
  it work fine, but i don't remember if i am wrong or if the compiler have
  wrong but i will check....
- to fix the problem of the demoplayback out of synch if when record we go
  in the menu :
  the problem is cause by paused variable : this one is set by extradataticker
  ce tic est ecrit dans la demo puis executer puis la demo est pauser
  puit on va pas ticker le jeux tant que paused=true de nouveux pour le metter
  a false on passe par writedemoticcmd qui ne l'ecrit pas (paused=false)
  mais il va etre executer et la demo est a present ecrit a nouveau

6-4-99
- found why it crash under linux : the doom2.wad have been corupted by the
  transfer (i suppose a bug of tcp/ip stack of MS-windoz)
- forget about yesterday demo problem (french stuff (BP very tired))
  it is a doom 1.9 original bug caused by menuactive variable that don't
  enable p_ticket to tic the game but it continue to record tics
  the solution is not to don't write tic because we have extra tics pushed
  at eatch cvar change in the menu, we must record in pause mode, maybe
  when replaying the demo we can run fast the pause ones. (will cause some
  little user interfase problem (chat text comming too fast) but is far from
  the actual problem.
- added pause command
- push the pause command when menu is enabled
- demo continue to record while paused=true but when playing back demo
  when pause it will run tics until pause go false

7-4-99
- little note about vc problem : it is not the vc fault, the problem there
  because the evaluation order in not define in the ansi-c therefore
  gcc and vcc have just do what he want...
- fixed +respawnmonster 1, map command set this to 0 if !nightmare
- fixed canot save spaced savegame
- fixed seconde wipe after a demo played
- readded code to fix the multiplayer menu quit don't change the name
- now it is splitscreen change function that set secondarydisplayplayer
  using playeringame[1], this allow to set secondarydisplayplayer after
  a splitscreen savegame
- fixed no more binding during chat
- fixed load splitscreen saved game
- fixed some load saved game don't unpause
- added COM_CheckParm
- use it to change a little the map command :
    map <mapname[.wad]> [-skill <1..5>] [-monsters <0\1>]
- added timelimit cvar and remove leveltimerinmin
- remove respawnparm since there is respawnmonsters cvar
- with all this change there is no more game parameters in servercfg packet

8-4-99
- fixed some stuff from yesterday (BP tired at 3am :))
- begin water physics

9-4-99
- continue water physics
- do some stuff for the chasecam

10-4-99
- use con_keymap for the chat
- fix the default key for next and prev weapons
- more chasecam stuffs
  added cam_height,cam_dist, cam_speed to configure camera
  added new mobj flag : MF_NOCLIPTHING to allow camera wolk on thinks
  TODO : create a path when player change sector this path will
         be used to find a path to use for camera to join up the player
         when camera is blocked
- added playdemospeed to change the speed of the playback of the demo
  (very funy)
- fixed problem of splitscreen that the multiplayer variable is false
  this fix also demoplayback of splitscreen demos
- now "-notime" can be used on the server side to exclude timecheck for
  all client

11-4-99
- fix pause with load games
- added check of heap malloc in the linux vresion
- added check of data field in I_FreeSfx (the sound is not sure to be
  loaded)
  Anyway Fab have remobe the lock of sound in dos version this will cause
  a sigsegv under OS/2 i will probably lock the wall heap in OS/2.... :(

12-4-99
- added a clipaiming to the r_setupframe to use allow playback of a 3dfx
  demo under sofware rendering
- added ylookup1 and ylookup2 to precalulate ylookup in splitscreen
- fixed crosshair display in 3dfx

13-4-99
- fixed cls usage in hwr_main (remove -1)
- fixed the full autoaim, maybe there is some more autoaim problem
- fixed floor and ceilling puff position (line side problem)
* little note about client prediction :
  add x and y offset (spirit.x-position.x) too all missile spawned from player
  and use dead reconning algorithme for all other mobj (pos=pos(t)+(t-oldt)*mom)

15-4-99
- release 1.28

17-4-99
- fixed can take the chainsaw in doom1 or udoom
- fixed can't save when use -warp or +map

19-4-99
- read all mail from the 1.28 release and do a full bug list to fix.
- fixed spaced lmp don't play

20-4-99
- make join with Thierry V. modification for OpenGL driver
- added some basic stuff for "join in game" (numplayers in server
  packet)
- now map check file and lump _beffore_ send and execute

+- 7-5-99
- put all the code for join in game, work in singleplayer
  must check on multiplayer
- added some code in doomatic... unfinished :(

11-5-99
- fix many bugs in the texture cache 3dfx the bug is in the
PU_3DFX stuff, should ask to fab for more
- fixed timedemo reset timer at levelbegin !!

15-5-99
- continue join in game code

19-5-99
- fix a lot of join in game problem, it is neer to be finish...
- fix cdvolume cvar bounds
- now when a level is complite the next is started with map command
- added -noresetplayer to map command but can't be used when player join
  the game because the new player don't know the old config !
  after savegame is implemented no more problem....
- the join in game is now working ...

20-5-99
- fix the last bug founds....
- now the engine wrape around level in deathmatch no more final screen etc...

21-5-99
- added allowturbo and allowexitlevel cvar
  when allowexitelevel is "no" the only way to exit the level is timer or
  fratglimit or exitlevel command, switch, line and monster death don't exit
  the level

22-5-99
- added proper servershutdown message and clientquit message
- and fix many bugs

18-6-99
- reading mail from Rich Dawn (libsocket programer) i foudnthat the local
  ip is not used and then do be asked to the gethostbyname !!! it is mush
  more simple now
- fixed little problem with new protocol connecting with direct ip number

23-6-99
- found and fixed 3 big network bugs :
  - the size was not used for resend a acknoledged packet
  - since i added BASEPACKETSIZE to HSendpacket the sv_sendtics don't
    reinitialize datalength
  - there was textcmd cleared with cleartic when client is very slow caused
    by extratics parameters
- fixed also join in game with splitsreen

24-6-99
- added check in recsearch , chdir can fail if we have't the right
- added checkfile in map command when loading wad
- support lfn for wad file trouth network

26-6-99
- use "xor"(^) insteed of "or"||" for mouselook and autorun

19-7-99
- i am working on the code since last word on this log unfortunately
  i am only fixing the new code, maybe i have done some little feature
  i have also release the 1.29 beta (1 to 4) last days
- have put a more interesting error when client don't have the file
- fixed the minor bug of 1.29 beta 4, can't start a game while
  demo is runing, it was caused by the addplayer code since the demo
  added a player to the game, and nodewaiting[0]==1 then the player
  added was 2 and since he isn't in the game he can't do map command
  (only player 0 can execute command without being in the game)

20-7-99
- if have got my mouse handler to work in dos version !
  i have added mouse 2 support for the splitscreen game
  TODO : + use_mouse2, mouse_sens2, mouselook2, mouse2 config (COM ?)...

21-7-99
- added all cvar for the mouse2.

24-7-99
- After days of trying to get two mouse with enumdevices in directinput
  Mircosoft informe me that there is a bug in the direct input !
  Since i am not blocked by Microsoft i have do it by myself, configure the
  mouse port and at last it work !
- added mouse2port

26-7-99
- added beautiful menu for the seconde mouse in the splitscreen menu

27 & 28-7-99
- implement that a will call clientprediction 2 it is on'other methode
  to do client prediction it send x, y in absolute value, on the
  receiver side it do a p_trymove to cross the special lines and take
  object, but it is inefisiante, because the diferance can have a big
  z differance and the trymove will refuse to move there, also special
  line are not crossed if the diferance is too big. And finaly object
  in the path is not taken (well maybe in futur with linear
  interpolation). With this point i have found a little original bug
  of doom : you can take a object even if trymove return false :)
  (plasma in map 1) . Well all this "bugs" hare found using and random
  lost of packt of 9/10 ! that why it is so bugy but it CAN happens

29 & 30-7-99
- well at this time i have fixed soem problem with clientprediction :
  - z checking is just desabled when we do p_trymove with player when
    folowing the spirit (using MF_NOZCHECKING)
  - when server copy the ticcmd when he lose tics from client it put a
    bit to 0 (set 1 on buildticcmd by clientpredition2 code) this allow
    to follow the player like in a true "dead rockoning", well this fix
    a little the "get thing" and "cross line" problem, but player
    movement are strange the best should be send momx and momy but this
    will take more bandwith, anyway i don't know what to do ?
  now there is some stuff to fix too : the spirit can't get teleport,
  the spirit don't get momx and momy like the players when there is
  a rocket explosion !

6-8-99
- fixed that we can reconect after deconnect (d_freenodenum)
- fixed exitlevel command when there is playerwaiting map change

14-8-99
- try to fix the problem : when a client respawn there is a consistancy fail
  in clientprediction (fail to fix the problem)

17-8-99
- fixed little issue in seconde mouse support in dos version

18-8-99
- fixed freez when using idclev
- fixed playback of demo 1.29 !

19-8-99
- added wad version check for main wad file and doom3.wad

20-8-99
- continue ftp

21-8-99
- added toggle command also added cv_showmodif ans cv_showmodifonetime flags

23-8-99
- added a "bad code" to fix splats clipping, also investigate on problem with
  splats, found two problem, should ask to fab for more :
  - in some littel texture and particarely in texture having same x for v1
    and v2 splat are bad disigned, "one line splat" problem
  - when player is neer from a splat and there view is is the same direction
    of the texture the splat do a line !
  - when messing with splat code i have try to remove x checking line
    165-171 of r_segs.c i am very suprice that this line don't do what i thouth

24-8-99
- sleep is a good thing !
- fixed splat doing line on walls ! caused by a oveflow using viewangletox
- try to fix the problem of "one line splat", doing some printf see that the
  lenght of the colum is 1 ?! why ???

25-8-99
- using lost packet simulation found and fixed little problem in network
  code, i hope is the lasts

26-8-99
- continue ftp, have transfered a piece of file, the end of file detection was
  wrong caused by wrong duplication detection

30-8-99
- added ipx support under win32 never thougt is was soo easy, the most
dificult part will be to addapt doomatic to the new system, also the launcher
should show the option easely

31-8-99
- inprove duplication detection
- implement addaptive timout using jackobson/Karn algorithme
- finilize ftp, it is working now but i should optimize a little more
  packet size usage also i should take in consideration the bandwidth
- now the tcp/ip code use 1450 maxpacketsize, also i have used the 1400
  byte of doomatic (TODO : check doomatic version beffor)
- TODO : + send one file by node, + bandwidth, + directory wadcache

1-9-99
- use net_bandwidth for file transfer optimisation

2&3-9-99
- searching for a damnid bug in the latest network code.
  It is now working perfectely : there is now way to hang with
  duplication or packet miss.
- this improuvement is for the ftp, this work now at perfection.
- TODO: ftp requester (-nodownloadfile), + fix playerdeadview, aiming
  in 3dfx in 90