[TI ASM] Application Header Format

Post by **benryves** » Wed 07 Jun, 2006 10:37 am

I'm adding app header generation and app signing (via Wappsign) to Brass, and have hit upon the problem that I can't work out what format the app header is meant to follow.

The app header program gives me a header that has a different format to the app header in the example that comes with the SDK (and, as far as I can tell, neither work as expected, though that's probably more a case of my mixing up the two different forms).

Is there an app header reference anywhere?

NanoWar · Post by **NanoWar** » Thu 08 Jun, 2006 5:53 am

For Wizards I use the following header with Wabbit:

Code: Select all

   .org $4000
   .db $80,$0F      ;Field: Program length(MUST BE 1ST 2 BYTES)
   .db 0,0,0,0      ;don't worry about this
   .db $80,$12      ;Field: Program type
   .db $01,$04      ;leave as is
   .db $80,$21      ;Field: App ID
   .db $01
   .db $80,$31      ;Field: App Build
   .db $01
   .db $80,$48      ;Field: App Name
   .db "Wizards "      ;MUST BE 8 BYTES
   .db $80,$81      ;Field: App Pages
   .db $02         ;Number of pages(don't worry wabbit corrects this)
   .db $80,$90      ;No default splash screen (below is just data, leave it)
   .db $03,$26,$09,$04,$04,$6f,$1b,$80,$02,$0d,$40,$a1
   .db $6b,$99,$f6,$59,$bc,$67,$f5,$85,$9c,$09,$6c,$0f
   .db $b4,$03,$9b,$c9,$03,$32,$2c,$e0,$03,$20,$e3,$2c
   .db $f4,$2d,$73,$b4,$27,$c4,$a0,$72,$54,$b9,$ea,$7c
   .db $3b,$aa,$16,$f6,$77,$83,$7a,$ee,$1a,$d4,$42,$4c
   .db $6b,$8b,$13,$1f,$bb,$93,$8b,$fc,$19,$1c,$3c,$ec
   .db $4d,$e5,$75
   .db $80,$7F
   .dw 0,0,0,0,0,0,0,0,0,0 ;(Padding is important!!!!)

Post by **benryves** » Thu 08 Jun, 2006 9:24 am

I switched to using Andy J's header (for single page apps) and that now works fine for single page apps. Multiple page apps still don't work, so I'll try slotting yours in as well. It's all a bit trial-and-error

Andy_J · Post by **Andy_J** » Thu 08 Jun, 2006 11:36 am

Yeah, and you might want to not force the minimum base code into it. I only did that as I use the hook for _getKey, and that was introduced in 1.13. Of course, the more elegant way would be to check it yourself and put a proper error message up instead of ERR:VERSION.

Post by **benryves** » Thu 08 Jun, 2006 11:46 am

None of the optional fields (hardware revision, base code version, & al) are added unless a value is specified apart from the 'No splash screen' one (which has a sort of reverse logic).

Jim e · Post by **Jim e** » Thu 08 Jun, 2006 12:50 pm

Whats wrong with multipaged apps? theres no difference in the header other than number of pages field.

You should also becareful on the appname, I remember some old TI apps were using 0 instead of spaces and that resulted a ti connect incapatibilty, and even though logic would say changing $80,$4* to the length of the name TI has decided that its better if their programming departments don't understand what each are doing.

Post by **benryves** » Thu 08 Jun, 2006 1:55 pm

I'm currently padding the name with spaces, truncating to 8 characters if it's too long.

As for what's wrong with multipage apps - I have no idea, otherwise I would have fixed it

The .hex files load into Flash debugger (no error message displayed) but the app doesn't appear in the Apps menu (and 'erases' the old app if one was installed with the same name).
Wappsign signs it (and appears to fill the unused $4000->$8000 range with $FF), but loading it into Flash Debugger ends up throwing lots of 'invalid instruction' errors. Sending it to PTI/real calculator displays a 'Receiving *<app name>' message, but gives up half way through (no validation).

I'm guessing a lot here, but confirmation would help - I can see that page 0 is at $4000, size $4000, and run at $4000. I'm guessing that page 1 is at $8000, also size $4000, and is either run from $4000 (paged in) or $8000 (?)

Jim e · Post by **Jim e** » Thu 08 Jun, 2006 2:31 pm

The first execute opcode should be at $4080(atelast thats the important one), where the app chooses to run after that fact is entirely up to it.

You are leaveing room on the last page for the signature right? It needs 96 bytes, so even if it is a 1 page app it needs 96 bytes on the end. I'm pretty sure Wappsign doesn't do any checking to verify if the number of pages is correct (it will simply tack on the signature even if it leaves the page bounds).

I can't think of much more that would mess with multipaged apps and not single paged apps.

Post by **benryves** » Thu 08 Jun, 2006 2:54 pm

Yes, sorry, I should have said that - if you take a look at this there's a .block 128 after the start of page 0 to allow space for the header.

I didn't know about the final 96 bytes (not that that's an issue at the moment) - I'll make sure to add a check to the assembler to make sure you're not overwriting it with your own code.

Here's a test app - doesn't contain any code, just the sequence $33,$66,$99 on both pages:

Code: Select all

.binarymode ti8xapp                 ; TI-83+ Application

.defpage 0, $4000, 16*1024, $4000
.defpage 1, $8000, 16*1024, $4000

.page 0
    .block 128                      ; Advance 128 bytes for header   
    .db $33,$66,$99

.page 1
    .db $33,$66,$99

... produces:

Code: Select all

:20400000800F000000008012010480210180310180484150502020202020808102809003C7
:20402000260904046F1B80020D40A16B99F659BC67F5859C096C0FB4039BC903322CE003E0
:2040400020E32CF42D73B427C4A07254B9EA7C3BAA16F677837AEE1AD4424C6B8B131FBBC7
:20406000938BFC191C3CEC4DE575807F000000000000000000000000000000000000000023
:04408000336699FF0B
:038000006699FF7F
:00000001FF

Erm, actually, that does look wrong. The record at $8000 is missing $33.
I think I need to check the hex output again

I'll get back to you on this one.

CoBB · Post by **CoBB** » Thu 08 Jun, 2006 3:12 pm

Where do you mark the start of the next page? Using the address $8xxx seems quite unorthodox to me instead of :0200000200rpcs (rp: relative page, cs: checksum).

Jim e · Post by **Jim e** » Thu 08 Jun, 2006 3:23 pm

Yeah thats definately missing the page record in ti hex format. If its 1 page then wappsign will add it, but multiple pages the page record is needed on all pages.

Post by **benryves** » Thu 08 Jun, 2006 3:32 pm

You can tell I use hex formats a lot...

I would just insert a ":0200000200rpcs" record between the main data chunks?

I assume, therefore, that '$8000' is never used, and the address field would be $4000 again when the second page starts again?

In which case, the .defpage directive is almost entirely redundant (and confusing) in Brass.

Jim e · Post by **Jim e** » Thu 08 Jun, 2006 3:57 pm

I would just insert a ":0200000200rpcs" record between the main data chunks?

I assume, therefore, that '$8000' is never used, and the address field would be $4000 again when the second page starts again?

Yup

In which case, the .defpage directive is almost entirely redundant (and confusing) in Brass.

Still good for rom generation.

Post by **benryves** » Thu 08 Jun, 2006 4:07 pm

All you really need to know for a page is the size (to stop you from going over the bounds) and, optionally, an origin (to save you from typing in .org). I can't really see any situation where you'd need to add an offset into the file, as you'd assume the pages would be consecutive in the binary.

Code: Select all

.defpage number, size, [org]

That seems a lot more sensible, in my mind - can you see any problems with it?

CoBB · Post by **CoBB** » Thu 08 Jun, 2006 4:29 pm

benryves wrote:I would just insert a ":0200000200rpcs" record between the main data chunks?

Where rpcs is 00FC (before the first page), 01FB, 02FA etc., and of course you don't need it at the end, only the 0...01FF.