doom / src / _logs / kin_progress.txt

important:All sources must be LF format, no CR and DOS EOF(^Z)

all changes to platform independent code
are marked with "199xxxxx by Kin"
(or use "diff -w -B" to compare)

about unix makefile:
makefile is for pmake
Makefile is for make
so the case must be reserved

RedHat must use "egcs" for -mpentium,
but "gcc" in Slackware 3.5 is already "egcs"

The Makefile* and #include are dirty, sorry.

CD music module is dummy now, 1. I have no music CD, 2. I hate music.

Use I_RegisterSong(music->name) with musserver, musserver will search
cmdline of doom for wad files, no modification is necessary

read() in a pipe will not always fill the whole
buffer, in Linux, read will only get 4096 bytes at once, now I use
sfx->data to store the sfx id, it's really a good idea.

There are 3 types of sound: SNDINTR, non-SNDINTR(define nothing),
SNDSERV, only SNDSERV is used by 1.28.

legacy uses I_StopCD() to pause cd playing, but xdoom uses CD_pause,
and CD_stop is to stop whole track.

I_ShutdownGraphics() in xdoom will just unhook the shared memory
doom needn't to know sound data length, sfxinfo_t only have pointers
to sound data, but not length, it is the job of sound module,
I_GetSfx(legacy) or getsfx(xdoom) will remember this, in dos legacy,
the sound data length is stored within sound data, in xdoom, it's
stored in a separate array lengths[](internal sound).

in Got_NameAndcolor(d_netcmd.c), there is a segment fault, why?
gcc will "optimize" some array into "const array", and ...

The code segment is read only, but it is not protected in djgpp,
so I removed the .text in tmap.s, the asm code is in data segment now.
The music lump loading/unloading is unnecessary with musserver,
but it is still there to minimize the possible bugs, check s_sound.c

old style "-net 1 ooxx" is unusable, why?
some transparent stuff in icarus and eternal III
is now solid black since 1.27(1.25 is OK), why?

use XResizeWindow(display,window,w,h),

minor problem:
eliminate -lnsl in Slackware(3.5)
and RH(5.2)

log:(if some weekends are not listed, I was playing doom)
added VID_X11 in v_video.c, MUSSERV in s_sound.c, i_sound.h
now I use "-res" to control resolution, vid.bpp is always 1 for now
#include <time.h> in cd_ioctl.c should be sys/time.h in RH 5.2.
check the SFX, where is the length?

SFX length(s) is in lengths[](SNDINTR/non SNDINTR internal sound)
Let Legacy use xdoom style sound.
.align 5 -> .align 1

All code successfully "compiled"
convert all sources into LF format,
and 'Makefile'...

Yes! My code works without SFX, the SFX value range is incorrect.
.code -> .data

Yes! SFX works, shift vol to 4bits, memcpy in Got_NameAndcolor(d_netcmd.c)
will cause a seg fault writing "readonly" play_names.

Yes, gcc sucks... it incorrectly thinking that player_names is untouched,
changed it into an 2d array, or you can put a dummy function with memcpy,
but not 'for(ooxx) player_names[ooxx][ii] = 0;' why!?

added xdoom990125 new event handling, now process mouse in I_StartTic(),
sleep if event queue is empty.

testing new code with 1.25, for moving to new legacy code.
SNDSERV is working!

reset in VoxWare 3 drivers(and many others) are broken,
all further ioctl on this FD will fail, don't reset /dev/dsp.
1.25 finished.

1.28 is running.

making i_video_ggi.c with 1.28, use -res to specify screen resolution,
only 256 colors are supported.

bug in d_clisrv.c if(VERSION!=....) are fixed,
client-server TCP/IP link is usable.

support VID_SetMode() in i_video_ggi.c, with 8/15 bits.

oops, don't forget G_CheckDemoStatus() in I_Quit()...

buffersize fixed(vid.bpp), 15 bits is runable, but ....

some keys added,
now svgalib ggi crash with icarus menu lump, why?

search modules in "."
I_Error() in I_ZoneBase()
quit flag in I_Quit() to prevent error loop

VID_BlitLinearScreen() was replaced with memcpy,
it does not work with Linux gcc

now team_names have the same problem as player_names,
so it was changed to an 2D array

I forgot recsearch(), added now.

Icarus big menu crash bug fixed

Use r_data.c from Legacy 1.25, no R_Render2sidedMultiPatchColumn()

bug using MAP command fixed