Os Cavaleiros do Zodíaco

Notes Bugs Lua Scripts

Unused Music#

This page or section needs more audio. There's a whole lotta words here, but not enough audio. Please fix this. Specifically: Needs rips. IDs 23-24 are unused, unique compositions.

Music ID $20 is a repeat of the Stage 1 theme; since it is also the opening song but in 2A03 form, and it comes before the Prologue and Ending themes respectively, it could have been meant as the title screen theme originally.

Music ID $21 is a 2A03 form of the Prologue song, which is never used by the final game.

Interrupted Voiceline Snippets#

Every playable character gets interrupted at the last second of their voiceline. As a result, the audio data for that last second of all of their voicelines goes unused.

Language Selection#

The game loads to name tables #2 and #3 a language selection screen with the wrong tilemap half.

Changing the address $004B in CPU RAM to 07 forces the tilemap to be the screen's and shows a mockup of how it would look like.

See the Notes page for information on how to perform it.

There are also graphics for an "ARABIC" option unused by this screen at 333C10.

English Version#

The English version that was originally planned for the game is still in the ROM. It is unknown exactly why it was cut, but it is speculated that the game was deemed too small to contain voice samples in both languages.
See the Notes page for information on how to enable it.

This version was not built with the current sampled audio in mind.
The text is typed out much slower, like the one in Titenic's cutscenes, instead of simply displaying every line at once like with the released Arabic version, causing the more hopeful part at the end of the intro music to keep playing while the first text is still rolling out (which is not very connected with the intro showing Abel's towering figure).
The game also keeps pointing to voice samples that do not exist in the ROM anymore; half the time they point to machine code and other miscellaneous data which comes out as garbage noise, but on the other half, uninterrupted audio data is used from The Dragon in the console's ROM. (If playing the game standalone, the instances from The Dragon will produce no sound.)
Voice samples for end of battle and their special attacks remained in the ROM and are used by the English version, but are swapped.
Characters' victory lines are swapped with their defeat ones, and none remained for Pegasus Seiya.
The special attack ones are swapped between Dragon Shiryu and Pisces Aphrodite, Cygnus Hyoga and Coma Bellenike, and Carina Atlas and Phoebus Abel, despite internally being ordered in the same way that the game's cast is (Seiya, Shiryu, Hyoga, Aphrodite, Bellenike, Atlas, and Abel). Pegasus Seiya is the only one to use the English voice sample for his special attack normally.

On the game's final Arabic version, these graphics at 3068B0 go unused.

However, there are two uppercase letters, Q and X, which are never used at all either by UI (options menu and credits screen) or the cut English version, due to coincidentally not being used by many English words. Can be found at 3069B0 and 306A70.

Other Unused Graphics#

301810:

Unused Portuguese font arranged by order of the characters' names. In order: Abel, Shiryu, Hyoga (though interestingly the G and O are swapped), Aphrodite (or Afrodite since the Portuguese spelling is used), and Bellenike (Berengue as he's known in Brazil, presumably from a mishearing of the Japanese pronunciation, Berenike).
Unknown exactly where it would've been used, but since it's near the Arabic font, it can be presumed that it was used to display opponent names in battle like in the final game.

306E30:

In addition to the graphics used for English text in the game, there are graphics for accented Portuguese letters, but no Portuguese text remained in the game. Notably, there is an E with a grave, è, which is not a letter that exists in the Portuguese alphabet.

306FB0:

There are also these strange letters with dots, seemingly early attempts at drawing Portuguese letters with the acute accent. Notably, one of them is N, the closest letter of which is ñ, which is a Spanish letter instead of a Portuguese one.
It is unknown what the Roman-esque I was for.

30EB90:

Zodiac signs that are never seen in the game; Capricorn, Aries, Cancer, and Pisces. Aphrodite, the first enemy you fight in the game, is of Pisces, but none of the other signs correspond to the other enemies.
The used zodiac signs are arranged in VS Mode in a way that implies the game is meant to use all of them, but there's only 10 spots and not 12, indicating that it uses them to represent Pandora's boxes.

(Assembled into a viewable form:)

30F070:

Part of the font for opponent names, never used due to not appearing in the names of the game's small cast.
The letter K is normally never used by the game, but it is used by Gouki if you enable him and play as him.

30F1F0:

Leftover graphics from Street Fighter Zero 2 '97, Hummer Team's port of (bizarrely) Street Fighter Alpha. Originally used on the HUD for denoting the player's number and their amount of lives.

30F440 has the same leftover graphics as Street Fighter Zero 2 '97.

30F750:

30F9A0:

These graphics were used in SFZ2'97's Continue and Game Over screens, respectively.

331610:

Unused letters in the bank with tiles for the Copyright screen that is actually used.

333010:

Graphics for what seems to be a Continue screen of an unknown game, with many interesting points.
Different colors are used here even though the palette is unknown.
Space Shuttle Exploration, another Hummer Team game, uses the same font as the one used by the numbers, 2 pixels wider and for letters as well.
3334F0 is an unused version of the same graphic as 37EC00, used for in-game cutscenes.
333500 has a smaller version not seen in the game at all.

Interestingly, the graphics for menu options (on the title screen and options menu) are ordered in a very specific way. "PUSH FTAR" is written, presumably for "PUSH START"; it is unknown why there is an F. However, the final game simply has a "START" option for starting the game, and an additional "OPTIONS" for the options menu. U, H and F at 37EC20, 37EC40 and 37EC60 respectively are never used.

37F010:

37FC10:

These are from a bank with generic graphics for the Rinco logo, but they go unused because the game uses the ones at 331010 instead.

Leftover Characters#

Watch on YouTubeAdon and Gouki (Akuma) from Street Fighter Zero 2 '97, character IDs 07 and 08 are still present in the game.
See the Notes page for information on how to enable them.
Adon merely glitches the characters' sprites and crashes the game. Gouki is still playable as a copy of Hyoga with his palette from Street Fighter Zero 2 '97. His special attack also crashes the game via triggering illegal opcodes, but he has a second special attack (Down, Down, A) which plays the opponent's animation for a special attack and does nothing else.
It is possible to make the game glitch with him by crouching to the left of the stage and pressing A.

Unused Text#

Menu Character Name Data#

00F6B2~00F73D, 05F6B2~05F73D and 06F6B2~06F73D have copies of character names for the English version's descriptions that are not used whatsoever.

Similarly, 00F748~00F806, 05F748~05F806 and 06F748~06F806 have copies of character names for the final Arabic version's descriptions.

00F80C~00F82D, 05F80C~05F82D and 06F80C~06F82D have unused copies of the "Street fighter II" [sic] disclaimer that appears after the copyright screen, right after the menu character name copies; this is because it merely uses tiles for the English menu character names.

Early English Text#

04B99E has a differently formatted, seemingly earlier version of the text in the prologue cutscene.
For starters, the two scenes are not broken into two chunks of text data, and are instead part of the same chunk.
It also ends halfway through before data for the text in the intro cutscene starts, although it ends with 0xFF characters like every other text data chunk, indicating that it was perhaps already cut before the final text was written.

Final English Text Copy#

At 04D640, there is a copy of English text data for the narration starting at the "at home in dealing with" line.

Copies of Miscellaneous Text#

065B5E + 065C22 and 067B5E + 067C22 are two copies of text data for the credits and options menu, respectively.

Copies of In-game Fighter Names#

The fighter name data for the HUD in-game has copies that are not used by the game at 05C408 and 06C408. Each name is a sized string, which means one byte indicates the length of the name, and the name itself X bytes long (X being the aforementioned byte) comes right after.

Unused Fighter Name#

ADON is still in the PRG at 07C438.

Uncompiled Source Code#

The game contains snippets of uncompiled source code scattered across the PRG. A lot of the data is corrupted, so guesswork has been taken in order to restore it to a readable form, comparing it to similar places in the code that aren't corrupted. Replacement characters indicate data whose original form is unknown.

960D:

;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
play_snd_down:
                txa
                and     #$07
                sta     a0              ;;_Mselect index

                lda     language
                and     #$01
                asl     a

                asl     a
                asl     a
                asl     a
                asl     a
                asl     a               ;;x����
                clc
                adc     a0
                sta     a0              ;a0:=language*64+x+
;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
        ldx     _TalkCnt
        dex
        stx     r0
        ldx     _TalkCnt+1
        dex
        stx     r0+1

        lda     r0
        ���     r0+1
        and     #$03                 

There are also 18 spaces after the last instruction instead of a newline, implying that there would have been a comment after. It is unknown how comments looked at that part of the source code, however.
At 041D06, 047D06 and 049D06 there are exact copies of this source code.
04BDCB and 04DDCB have copies that start at da in the lda language instruction instead, located right after English text data for character descriptions and the introduction cutscene, and its copy respectively.
04FF37, 051F37, 053F37, 071F37 and 073F37 have copies that start 561 characters in, at the second comment line. 055FF0 has a tiny copy starting two spaces before and #$03.

8EA0:

������������     SR
;                lda     dhi
;                sta     $2006
;                lda     dlo
;                sta     $2006
;?2
;                lda     _DBuffer,x
;                sta     $2007
;                iny
;                inx
;                cpy     dwide
;                ���     ?2
;                lda     dlo
;             

AD22:

        ldy     #$00
        sty     x1              ;;; x1 pos x index.
        ;;;<XPOS>;;;;;;;
?x000   lda     xpos,y
        sta     $303,x          ;;; set x pos.
        lda     y1
        sta     $300,x          ;;; set y pos.
        ���     a2
        lda     (r0),y
        bne     ?s00            ;;; if pat exit?
        lda     #$ff
        sta     $300,x                  ��;     at =0 clear pat.
        jmp     ?next_x
        ;;;
?s00    ora     Bit76           ;;; SET PAT BIT 7,6 FOR BANK 3,4.
        sta     $301,��         ;;; set pat name.
        lda     (c0),y          ;;; GET BANK PAT MAP COLOR.
        and     #$01
        ora     Sprface            ������OR��������T.
        sta     $302,x          

Interestingly, the same corrupting pattern of FF FF ?? ?? 00 00 00 00 that appears in Wait and See!'s source code snippets starts appearing here. At the end of these code snippets, a pattern of YY YY YY YY 00 00 ?? ?? 0X 00 00 00 ZZ ZZ ZZ ZZ is present, which also exists in the previous snippets; it could be a variant of the same pattern. YY YY YY YY and ZZ ZZ ZZ ZZ are two 32-bit addresses each.

BD0E:

         ;
?2      lda     pad1
        beq     ?1
        inc     sub_flow
?1      rts
        rts
end12
        lda     #$06            ;;ending
        jsr     next_pagex
        lda     #END_TIMER
        sta     Delay           ;;; vs delay.
        rts
end13
        jsr     TimeOut
        bne     ?ed00           ;;; is delay time out?
        lda     #END2_TIMER
        sta     Delay           ;;; ���������
        inc     sub_flow
?ed00   rts
end14

        jsr     EndColor
        SET_SPRITE    #$05,#$50,#$a0,#$00,#$00
        SET_SPRITE    #$��,#$��,#$a0,#$00,#$00
        WAIT_NMI
        ;;;----------------     < TITLE COUNT DOWN >
        jsr     TimeOut
        bne     ?ed00 ���������������� delay time out?
        ld

043F37, 045F37, 059F37, 05BF37, 069F37, 06BF37, 079F37, and 07BF37 have copies of this same source code snippet, starting at the last #$00 on the first SET_SPRITE line.

Debug Symbols#

Some of the data for these debug symbols is corrupted much like the source code.

CCBE:
The third debug symbol was corrupted beyond recognition by the aforementioned pattern with a 32-bit address, 0x5D3B7627.

ow_pow_x
show_vs_sel
smt
v;]␀
snm
sound_ndx
sound_reg
split
spr_ndx
sptmode
start_addr
start_fg
␀␀Rx;]opt
stl
story_0
story_1
story_2
story_3
story_4
story_5
Ìx;]␀␀cu␂␀␀␀
sub_flow
swd
sxh
sxl
syh
sylt0
t0093
t0094
t0123
t0124
talk_0
talk_1
talk_2
talk_3
talk_4
talk_5
talk_6
talk␁␀␀␀
talk_box_scr
tall
temp_cnt
tempxx
testxx
ti_flow
tit_word0
tit_word1
tit_word2

DC70:

class
clr_demo_man
clr_speechon
col_msk
col_ndx
color
color_len
comp2
control
credit
ctu_0
ctu_2
cur_flash
d0
d1
d2
d3
d4
data1_hi
data1_lo
data_hi
data_lo
dbuffer_tbl
dead_fg
delay
delay1
delay_2
demo_fg
demo_ptr
dhi
dhv
dlo
dtall
dwide
end0
end0x
end1
end10
end11
end12
end13
end14
end2
end3
end4
end5
end6
end7
end8
end9
endcolor
english
f_color
f_msk1
f_msk2
fade_cnt
fadef
fades
fill_talk_word
fill_word_fg
go_show_man
head_3456
head_apper_tbl
head_area_tbl
head_area_tbl2
head_tbl
ini_language
init_head
init_head2
init_para
init_sound0
init_sound1
init_sound2
init_to_fill
init_vs_name
initial
irq_clk
irq_cnt
irq_deep
irq_wide
irqx_cntn
k0005
k0022
k0037
k0040
k0042
k0043
k0066
k0083
k0090
k0095
k0099
k0117
k0125
language
language_tbl
lib2_flow
lib2_subx
m4_flag
main_flow
man_bank_tbl
message2_00
message2_01
message2_02
message2_03
message2_04
message2xx_01
message_00
message_01
message_02
message_022
message_022d
message_03
message_032
message_04
message_042
message_05
message_cnt
message_flow
move_cnt
next_case
next_pagex
no_pat
no_use
nul
ocounter
on_clk
on_off_y_tbl
only_allar
overBlk
p2000
pad1
pad2
pad3
padlast1
padlast2
patbank
patbank0
path
pause
photo_tbl
port
pro_no
probank
push_s_fg
r0
r1
r2
r2000
r2001
random
random1
re_color
read_linex2
read_linex2_2

05DFE1, 06DFE1, and 07DFE1 have exact copies of these debug symbols, starting at f_color. However, about 47 bytes in, 1283 bytes of them were replaced by miscellaneous data; after that, the copies continue for 77 bytes before getting interrupted by machine code.

Oddities#

Icon Placement#

The game uses the Libra icon to replace every character that has been defeated, but for some reason the Scorpio icon comes first in the ROM at 30E5A0. The other icons only appear all the way over at 30EAD0.

RLE Redundancy#

The game's own RLE is used to encode strings of text with repeated characters that, if not using it, would only be two bytes long instead of three, most evident in the text for the Options menu and the credits (such as the "FF" in the Options' "EFFECT" or "MM" in the credits' "PROGRAMMER"), though it is also used for the 9s in the 1997 for Rinco's copyright screen at the beginning. The game's cut English version does not use this RLE despite having multiple strings of repeated two characters.

References#