Faxanadu Retranslation

[Ascended Mermaid]

Here's my current checklist.

SPOILER: Click to show

- Check bank switch #$10; may or may not cause irregularity.

- Figure out something with the king, to eliminate the 1500 gold exploit.

- Check death routine. LDA "last guru visited", BEQ "alternate death", Player HP to starting HP, Player MP to starting MP, EXP to last level, don't change Gold/Ducats/Inventory, JMP "spawn at last guru"... or just reset everything to base.

- Check door data outside Eolis. LDA "current screen", CMP "outside Eolis", BNE "end routine", LDA #$##, STA "door destination", LDA #$##, STA "door YX", LDA #$##, STA "exit YX"

- Test magic damage with the Battle Helm + Ointment. Does the Battle Helm block magic? What kinds of magic? Check magic defense.

- Figure out whether or not the Ointment should prevent magic damage. Possibly add a separate routine for shieldless magic damage.

- Fix dialogue crash bug.

- Add "how many letters until next script code" check. LDX "letter position", INX, LDA X "current dialogue", CMP #$FA, BCC "LDA X", AND "letter position", CMP #$20, BCC "end routine", LDA "line position", CMP #$02, BEQ "button prompt -> new page routine", JMP "new line routine"

- Edit "new page routine".

- Rewrite dialogue routine, keep in mind to clear the negative flag.

- Check and correct save routine to fix dialogue spill bug.

Resume looking for a jump/branch to $C970-C97F from 0F:C600 onward.

You can see I've played around with sortof an incomplete mockup of new code in writing my checklist. Just kinda preparing. I should also look for unused RAM. Maybe.

[toolshopguy]

Hello all! this is mike halphen, I used to post in the gamespot forums under the topics for unsavory's remake:

hey SkyeWelse!! hey buddy hope you are doing well! as I am reading this about the new material that you recently found on yahoo auction, I am shaking with excitement - but the link or image that you posted isnt there/doesn't work! could you please repost?


i lurk occasionally on this topic and am always reading with a huge smile on my face - unsavory, thank you for all the work you put into this remake - i don't care if it would take you 10 years to finish this project, I love that there are others out there that are as crazy about this game as I am, and it is always exciting to read through your progress and (personally knowing nothing about programming) see the game that I've loved for all these years from that point of view/perspective

[Xalphenos]

I only just realized this would be pronounced FaZanadu the other day.

[Ascended Mermaid]

Good to hear, Mike! I've definitely kept Faxanadu in mind all this time. I don't have much time to work on it, but will if I can find time. Otherwise, I'm willing to pass the torch to someone else. Should anyone be interested in taking this off my hands, let me know.

[Ascended Mermaid]

In theory, cleaning up the continue/load routine will eliminate one cause of the crash bug.

// I'm thinking about checking for JSRs that lead to other JSRs, and are not followed by an equal or greater amount of RTSs. In theory, this could cause stack overflow, which can possibly crash the game... am I right?

[SkyeWelse]

Hey Mike,

Good to see you around still. Yep, I did get that thin booklet on Faxanadu that Famitsu put out but I don't think there was anything too new in there. Kind of a disappointment. There "may" have been some new artwork for the items/spells that might not have been in the guide, but sadly I have all of my Falcom stuff packed up in a box right now since we're moving into a new home so I cannot check it.

Glad to hear that the is project is still making some progress, Unsavory. : )

-Thomas

[RyuKisargi]

I still hope to see this. I love Faxanadu.

A buddy got his old OLD Xbox modded and a bunch of ROMs on it. What was the NES one I put on there?

Faxanadu.

What do I want on 3DS VC? Faxanadu.

[Ascended Mermaid]

Well, it's mostly just testing and theory at the moment. I haven't altered any data since January. I hope to squeeze a little debugging into my schedule soon.

// In the future...

LDA #$005B ; Load name screen check
BNE ; If not on name screen, skip 3 JSRs
(That spot with the 3 NOP'd JSRs)
Big letters.

Hmmm... of course, the game crashes because my line break interrupt code is incomplete, I think.

---

0F:F601:AD 16 02 LDA LetterPos = #$1F
0F:F604:2D 17 02 AND LinePos = #$00
0F:F607:C9 03 CMP #$03
0F:F609:90 01 BCC MessagePrinter
0F:F60B:60 RTS

Right. I need to figure out why A + M = #00. I got this far. This is all the time I have for today, so until next time.

Memo: JSR to $BFDF happens after switching to bank 10, my new bank.

[Xalphenos]

Nov 21, 2013 15:37:40 GMT -5 Ascended Mermaid said:

Right. I need to figure out why A + M = #00. I got this far. This is all the time I have for today, so until next time.

I'm not sure quite what you mean.  There is no addition going on in the code you posted.

If you bitwise AND any number with 0x00 the result will always be zero.

Bassed on the code you posted and your notes beside it:
It looks like you load the value from $0216, which must be 0x1F, into the accumulator
Then you AND it with the value from $0217, which must be 0x00. Which will yield 0x00.
Then you compare the accumulator, 0x00 now, with 0x03. Since the accumulator is less then 0x03 the carry flag will be cleared.
Since the carry flag is clear the branch will occur.

Edit: Oops. The value of $0217 doesn't have to be 0x00. Other numbers would result in 0 as well. 0xa0 0xc0 and 0xe0 would all result in 0x00.

[Ascended Mermaid]

Nov 21, 2013 17:04:04 GMT -5 Xalphenos said:

Nov 21, 2013 15:37:40 GMT -5 Ascended Mermaid said:

Right. I need to figure out why A + M = #00. I got this far. This is all the time I have for today, so until next time.

I'm not sure quite what you mean.  There is no addition going on in the code you posted.

If you bitwise AND any number with 0x00 the result will always be zero.

Bassed on the code you posted and your notes beside it:
It looks like you load the value from $0216, which must be 0x1F, into the accumulator
Then you AND it with the value from $0217, which must be 0x00. Which will yield 0x00.
Then you compare the accumulator, 0x00 now, with 0x03. Since the accumulator is less then 0x03 the carry flag will be cleared.
Since the carry flag is clear the branch will occur.

Edit: Oops. The value of $0217 doesn't have to be 0x00. Other numbers would result in 0 as well. 0xa0 0xc0 and 0xe0 would all result in 0x00.

Ah! I wish fceux didn't crash before I could save my solution. I'm thinking ADC might be better. The RTS should return to the routine after, assuming I don't have to change too much.

I basically want it to prevent starting over with the message routine, so I wrote that checksum.

[Xalphenos]

Nov 21, 2013 19:28:06 GMT -5 Ascended Mermaid said:

Ah! I wish fceux didn't crash before I could save my solution. I'm thinking ADC might be better. The RTS should return to the routine after, assuming I don't have to change too much.

I basically want it to prevent starting over with the message routine, so I wrote that checksum.

Yeah ADC is the only addition opcode on 6502. Posting ASM gets my interest piqued but you can feel free to ignore me. I'm curious though, wouldn't you hit the RTS regardless of the the compare. 6502 is still pretty foreign to me but if you are branching to the very next operation you will land there if you branch or not right? If you don't branch you just go to the next line if you do branch you're only branching to the very next line. am I right or did I miss some fundamental aspect of 6502?

EDIT: Oh. I didn't see the 01 in your branch. So you are branching past the RTS if the Accumulator is greater than or equal to three?

[Ascended Mermaid]

Nov 21, 2013 21:43:45 GMT -5 Xalphenos said:

Nov 21, 2013 19:28:06 GMT -5 Ascended Mermaid said:

Ah! I wish fceux didn't crash before I could save my solution. I'm thinking ADC might be better. The RTS should return to the routine after, assuming I don't have to change too much.

I basically want it to prevent starting over with the message routine, so I wrote that checksum.

Yeah ADC is the only addition opcode on 6502. Posting ASM gets my interest piqued but you can feel free to ignore me. I'm curious though, wouldn't you hit the RTS regardless of the the compare. 6502 is still pretty foreign to me but if you are branching to the very next operation you will land there if you branch or not right? If you don't branch you just go to the next line if you do branch you're only branching to the very next line. am I right or did I miss some fundamental aspect of 6502?

EDIT: Oh. I didn't see the 01 in your branch. So you are branching past the RTS if the Accumulator is greater than or equal to three?

That's right! I may rewrite the code as:

LDA letterpos
ADC linepos
BEQ writemessage
RTS
writemessage

// Otherwise, the reason I picked 03:
linepos is added to twice if letterpos is >= 1F, and letterpos is -1F of current letterpos. I may have to alter the 1F rule.

[Xalphenos]

BEQ branches if the zero flag is set. ADC sets the zero flag if the result of the addition is 0x00. So this code will branch only if your accumulator ends up being 0 otherwise you will RTS.

I'm still not 100% on what you're doing.

[Ascended Mermaid]

Nov 22, 2013 9:06:38 GMT -5 Xalphenos said:

BEQ branches if the zero flag is set. ADC sets the zero flag if the result of the addition is 0x00. So this code will branch only if your accumulator ends up being 0 otherwise you will RTS.

I'm still not 100% on what you're doing.

That's perfect. Basically I'm preventing the message code from restarting. Every message begins with both letterpos and linepos at 0. The message code recalls information from before the checksum, and the RTS will return to the message code currently in progress.

---

The message dialog is handled accordingly:
Even LetterPos
Odd LetterPos
Even LetterPos+LinePos
Odd LetterPos+LinePos

This is what LetterPos is doing:
0,2,4,6,8,A,C,E,10,12,14,16,18,1A,1C,1E
1,3,5,7,9,B,D,F,11,13,15,17,19,1B,1D,1F
0,2,4,6,8,A,C,E,10,12,14,16,18,1A,1C,1E
1,3,5,7,9,B,D,F,11,13,15,17,19,1B,1D,1F

This is what LinePos is doing:
0
0
2 ; Always increments of 2. Otherwise, it's only half a LinePos.
2
4 ; End Message

0F:F601:AD 16 02 LDA LetterPos = #$00 ; Load LetterPos
0F:F604:6D 17 02 ADC LinePos = #$02 ; + LinePos
0F:F607:F0 01 BEQ $F60A ; Start message routine if new message
0F:F609:60 RTS ; Otherwise, return to message routine
0F:F60A:EE 16 02 INC LetterPos = #$00
0F:F60D:EE 16 02 INC LetterPos = #$00 ; LetterPos+2
0F:F610:20 6E F5 JSR SwitchToBank10 ; Check if letter is special character, print Letter and play sound
0F:F613:AD 16 02 LDA LetterPos = #$00 ; LetterPos to A
0F:F616:C9 1F CMP #$1F ; Is LetterPos >= 1F? (16 characters per line)
0F:F618:90 F0 BCC $F60A ; If not, go back to F60A and print another letter.
0F:F61A:AC 16 02 LDY LetterPos = #$00 ; LetterPos to Y
0F:F61D:C0 1F CPY #$1F ; Is LetterPos >= 1F? (16 characters per line)
0F:F61F:B0 06 BCS $F627 ; If so, go to F627, LetterPos-15
0F:F621:C8 INY
0F:F622:C8 INY ; If not, LetterPos+2
0F:F623:C0 1F CPY #$1F ; Until >= 1F
0F:F625:90 FA BCC $F621
0F:F627:98 TYA ; LetterPos to A
0F:F628:E9 1F SBC #$1F ; A-15 to alternate odd/even
0F:F62A: 16 02 STA LetterPos = #$00 ; A to LetterPos
0F:F62D:D0 E1 BNE $F610 ; Print this letter if LetterPos is odd
0F:F62F:EE 17 02 INC LinePos = #$02
0F:F632:EE 17 02 INC LinePos = #$02 ; Otherwise, add LinePos+2
0F:F635:AD 17 02 LDA LinePos = #$02 ; LinePos to A
0F:F638:C9 04 CMP #$04 ; Is LinePos >= 04? (4 lines per window)
0F:F63A:90 CD BCC $F609 ; If not... go to RTS?
0F:F63C:4C 10 F6 JMP $F610 ; Print this letter
0F:F63F:4C 85 F6 JMP $F685 ; ??I don't remember??
0F:F642:A9 00 LDA #$00 ; 0 to A
0F:F644: 13 02 STA MessageID = #$00 ; End message
0F:F647:60 RTS ; End of routine

Unfortunately, the game has other plans:

f ; LetterPos gets stuck
offence does not ; If LetterPos is even, it's not stuck anymore
. ; LetterPos gets stuck again

[Xalphenos]

Is this the code in the game or your edited code?

So the game writes half of a character at a time? first even halves then odd? Is this still trying to fix the letters from being written to the box all funky?

EDIT: I don't think that ADC is going to do what you want. The z flag will only be set if linepos = -leterpos. Which seems to be impossible. If I'm understanding you correctly the only possible values for linepos are 0, 2, and 4.

[Ascended Mermaid]

Nov 22, 2013 15:39:18 GMT -5 Xalphenos said:

Is this the code in the game or your edited code?

So the game writes half of a character at a time? first even halves then odd? Is this still trying to fix the letters from being written to the box all funky?

Pretty much. The original characters were twice the size, so now it prints a quarter of a character. I did fix this code to work 100% a long time ago, but I lost my progress.

[Xalphenos]

That's a bummer. Do you still have the original code or a tracelog?

[Ascended Mermaid]

Also, the line at 0F:F621:C8 that always gets skipped? That's the special line break character code. A code in Bank 10 jumps there.

---

Nov 22, 2013 16:19:17 GMT -5 Xalphenos said:

That's a bummer. Do you still have the original code or a tracelog?

Unfortunately, I don't think so. Unless it's in this thread or in one of my notes (which is where I got this version of my code) it's lost.

// I believe I see the problem and part of the solution.

// Two things stand out. The last BCC goes to an RTS. I honestly don't remember why. The JMP after starts the whole message routine over again. Also, if the message routine is already in progress, it needs to return properly. It all loops back to the check at F601, so making sure that works is important. It needs to RTS if either the letterpos or linepos aren't equal to 0. I can move more code to bank 10 and include separate checks for both if ADC won't work.

// I could probably use BIT to check if letterpos + linepos = 0.

I also see that BCC can execute the JMP at the same time. In this instance, that seems impractical.

// Also, that first RTS can be changed to a JMP to the line after the bank 10 JSR.

Still odd though, that odd letters print in this instance. I think the aforementioned RTS may be preventing even letters from printing correctly.

// It makes sense that the bank 10 JSR becomes a bank 10 JMP when I get rid of that RTS. The RTS should've worked... but something kept getting in the way.

// Wait. The RTS is working, but the ADC is where everything gets thrown for a loop. So yeah, I'll see where BIT gets me first.

[Xalphenos]

I've been trying to think of the easiest way to check both values but I'm not coming up with anything. AND works most of the time but not always.

Edit: for instance line pos 2 and letter position 4 would yield 0. BIT is the same as AND it just throws away the result.

[Ascended Mermaid]

Nov 22, 2013 20:23:16 GMT -5 Xalphenos said:

I've been trying to think of the easiest way to check both values but I'm not coming up with anything. AND works most of the time but not always.

Edit: for instance line pos 2 and letter position 4 would yield 0. BIT is the same as AND it just throws away the result.

I forget if LDA requires CMP if followed by BEQ. I think not IIRC. I'll check if LDA alters the zero flag. I can move virtually everything between Bank 10 and F601 to Bank 10, and check both values. LDA BEQ LDA BEQ RTS.