fablog1.txt

The source code of Doom legacy for windows

=======================================================================
faB log of changes on DOOM LEGACY, since the Linux doom sources release
(23-Dec-1997) until the release of Legacy v1.20 (continued next file).
-----------------------------------------------------------------------
THE DOOM LEGACY PROJECT		          Contact: <[e-mail]>
=======================================================================

Project:  DOOM Legacy
Web site:  http://www.newdoom.com/doomlegacy
Log author: Denis Fabrice



Fri 26-12-1997: "First overview"
===============


*   Downloaded 'DOSDOOM' by Chi Hoang, compiled, runnable under Win95,
    crash under true DOS.

    No sound, no network, keyboard config doesnt work.

    REAL slow! -devparm shows 2 to 3 tics per frame on a P166MMX !

*   Corrected bugs in IdentifyVersion(), FindResponseFile();
    Now runs under DOS.

*   Added nice titlebar like original Doom, compile only with 'PC_DOS'
    defined (uses conio.h).

*   PC_DOS define : enable DOS/textmode specific stuff.

*   asm_defs.h : defines all the stuff needed so that asm routines
    know about vars and structs defined in C code.

*   tmap.S : converted the x86 code to AT&T syntax for GAS, NASM could
    be used but AT&T syntax is much better imho.

    R_DrawColumn, R_DrawSpan restored.

    Should be optimised for Pentium in the future.

*   USEASM define : compile with ASM routines instead of C ones.

    Actually, USEASM means i386/Pentium machine code.

*   Compiled with ASM routines, runs the first time, but it's still
    very slow... there's a lot of cpu time lost somewhere else than
    drawing the screen.



Sat 27-12-1997: - "Boris and Fab in search of the lost frame"
===============


*   tmap.S : FixedDiv2, FixedMul : translated to AT&T.

    After some profiling with 'gprof', we've noticed that FixedDiv
    and FixedMul where used a lot. We've restored the ASM ones.

    A lot of cpu time freed, but still a minimum of two tics/frame...

*   i_system.c : we've restored a TimerISR interrupt for counting the
    gametics, like it was in the original doom.

    I_GetTime was not supposed to count the tics itself, but just
    return it.

    I_TimerISR()
    I_StartupTimer()
    I_ShutdownTimer()
    I_SetTimer0()

    Result: IT S FAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAST !!!

    Problem: we have to call the old Timer int., but not at the same
    speed that our new TimerISR... but we didn't find a clean way to
    call the old interrupt from the new TimerISR.

    So, the old timer interrupt is called faster than 18.2/sec.
    Maybe we'll have to code an IRQ wrapper in ASM like in Allegro3.0



Sun 28-12-1997: - "Boris works alone"
===============

*   d_main.c  : no 'HOME' environment variable under DOS, now uses
    'DOOMWADDIR' for development.

*   tmap.S : removed use of %ebp in FixedDiv2 and FixedMul.

*   wi_stuff.c : line 1660 : object "WIOBJ" lump not found, should
    be used only with 'french' version, we'll have to check why the
    french version is active? Replaced by "WIOSTI" for now.

*   Network restored. Seems to work fine!

*   m_misc.c : M_SaveDefaults, M_LoadDefaults : now handles conversion
    of scancodes/ascii when reading/writing the config file.

*   Downloaded 'dosdoom2' by Chi Hoang.

    He goes so fast that we don't have the time to tell him the problem
    with 'I_GetTime'... we'll do a release of our port soon and we'll
    send him a mail. This guy got us started on dos doom sources.


Mon 29-12-1997: - "Doom2 in 320x400 looks weird!"
===============

*   i_video.c : arranged the keyhandler from DosDoom, which didn't
    PostEvent fast enough. Didn't respond when typing fast on the
    keyboard.

*   MULTIPLE VESA resolutions : added code to handle bank/switching
    and setting of VESA modes.

    - 320x400 runs 35fps on my P166mmx,
 640x480 needs 2 tics per frame, which is SLOW !
 (duke nukem runs at least 20fps 640x480 on my old P75...)

    - works only in full screen, crashes when Statusbar displayed,
 still lots of positioning/resizing to check out.

    - visplane 'top' and 'bottom' tables used byte for y-clipping
 coordinates, so of course it didn't support more than 255
 lines.

    - r_defs.h : visplane 'top' and 'bottom' tables changed to 'short',

 r_planes.c
 r_segs.c   : some compares used '0xff' changed to '0xffff'.

 This is the main stuff you have to change if you want to support
 multiple resolutions.

    - asm_defs.h : SCREENWIDTH must be changed to test different
 resolutions.

    - aspect ratio is not good for 320x400 and so...

    NOTE ! These modifications will not be in the 'clean' source code
    until it runs ok with status bar.

*   ALLEGRO3.0 : we'll definitely use a lot of code from this GREAT
    library by Shawn Hargreaves.

    We'll use the sound code from Allegro3 for sound support, like
    DosDoom2 does.


Tue 30-12-1997: - "Preparing the release for our doom dos port"
===============

*   Boris succesfully added the sound support with Allegro3.
    (as DosDoom2 does)

*   A lot of work to cleanup our release:

    DJGPPDOS/     contains the source code specific to DOS/W95.
	     i_xxxx.c
    LINUX/	     contains the source code specific to LINUX.
	     i_xxxx.c

    Implementation-specific code in the main directory has to be
    enclosed by #ifdef PC_DOS or #ifdef LINUX, or such.

    Linux code was removed from the DJGPPDOS/ implementation.

*   makefile : somewhat improved to support ASM routines with -DUSEASM,
	  added support for PGCC like Allegro3 makefile


Wed 31-12-97: - "Fixed that timer stuff!"
=============

*   DJGPPDOS/i_irq.c     : new timer routines, plus ticcount
	i_irq.h
	irqwrap.S   : asm wrapper for the irq handler

	i_system.c  : removed our dirty timer code, I_GetTime
		      now returns ticcount.

    The timer handler is now fixed, but then the Allegro sound doesnt
    work anymore, got to fix that.


Thu 01-01-1998: - "
===============

*   3:49 - Our doom version succesfully started with 5 players, but
    then it crashes after some seconds...

    We used DOOMATIC (which replaces Ipxsetup) to launch the game with
    5 net nodes. We used a map with 6 dmstarts.


*   Yeehaa! erreur ridicule avec le define ST_Y sans parentheses...
    Je peux enfin resizer en 320x240 !!! La status bar n'est pas affich閑
    en entier.

*   Le makefile ne marche toujours pas et ca m'enerve!

*   Tests transparence avec TINTTAB, super cool.

*   Dor閚avant il faut ajouter -file tinttab.wad avec un lump 'TINTTAB'.

*   Dans G_Initnew(), ajout